0% found this document useful (0 votes)

268 views158 pages

Reiner, A Linguistic Analysis of Akkadian

Uploaded by

Antonio Filippin

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

268 views158 pages

Reiner, A Linguistic Analysis of Akkadian

Uploaded by

Antonio Filippin

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

A LINGUISTIC ANALYSIS OF A K K A D I A N

JANUA LINGUARUM
STUDIA MEMORIAE

N I C O L A I VAN WIJK D E D I C A T A

edenda curat

C O R N E L I S H. VAN S C H O O N E V E L D
STANFORD UNIVERSITY

SERIES PRACTICA
XXI

1966

M O U T O N & CO.

LONDON · THE H A G U E · PARIS

A LINGUISTIC ANALYSIS
OF AKKADIAN
by

ERICA REINER
UNIVERSITY OF CHICAGO

1966

M O U T O N & CO.

LONDON · T H E H A G U E · PARIS
© Copyright 1965 Mouton & Co., Publishers, The Hague, The Netherlands.

No part of this book may be translated or reproduced in any form, by print, photoprint, microfilm,
or any other means, without written permission from the publishers.

LIBRARY OF CONGRESS CATALOG CARD NUMBER : 65-24780

Printed in The Netherlands by Mouton & Co., Printers, The Hague.

ACKNOWLEDGMENTS

This grammar, my first attempt toward a formulation of the structure of Akkadian

which has intrigued me for many years, owes its existence to a set of fortunate cir-
cumstances. As to the subject matter, discussions with my colleagues at the Oriental
Institute, as well as work with the vast range of material for the Assyrian Dictionary
Project over the years, provided stimulation and challenge; as to the presentation as
a linguistic study, it was the encouragement of my colleagues in linguistics that led me
to attempt it. To all the friends, students, and colleagues who read various drafts or
discussed sundry points with me go my sincere thanks. I am particularly indebted to
my colleagues Eric P. Hamp, Professor of Linguistics, and A. Leo Oppenheim, Orien-
tal Institute Professor of Assyriology, who so generously gave of their time to read the
manuscript at various stages, and sustained me in my effort to bring this grammar to
completion.
I have been especially fortunate to profit from the constructive criticisms and sug-
gestions of my colleague Bulcsú László (University of Zagreb and University of
Chicago), whose help in the final redaction, in particular with the formalization of the
phonological and transformational statements, has been invaluable.
It is my pleasant duty to acknowledge the financial contribution of the Division of
the Humanities of the University of Chicago toward secretarial expenses.
TABLE O F C O N T E N T S

ACKNOWLEDGMENTS 5

ABBREVIATIONS 13

INTRODUCTION 15
0.1 Purpose and scope 15
0.2 Organization 17

1. T H E C O R P U S 20
1.1 The language 20
1.2 Dialects 20

2. THE WRITING SYSTEM 23

2.0 Preliminary statement 23
2.1 The script 23
2.1.1 Graph and referent 23
2.1.2 Graphotactics 24
2.2 Elements of the transliteration system 25
2.2.1. Determinatives 25
2.2.2 Word signs 26
2.2.3 Phonetic complements 26
2.2.4 Syllabic signs 27
[Link] Types of syllabic signs 28
[Link] V signs 28
[Link] CV signs 28
[Link] VC signs 29
[Link] CVC signs 29
[Link] Sequences of signs 29
2.3 Changes in writing habits 29
2.4 Problems caused by the writing system 30

3. PHONOLOGY 33
8 TABLE OF CONTENTS

3.0 Basic assumptions 33

3.1 Phonemes 34
3.1.1 Prosodies 35
3.1.2 Configurative feature 35
3.1.3 Phones without phonemic status 35
[Link] /s/ 35
[Link] [w] 35
[Link] [j] 37
[Link] Distribution of glides 37
3.2 Stress 38
3.3 Distinctive feature analysis 39

4 . PHONOTACTICS 42
4.0 Questions to be treated 42
4.1 Vocalic and consonantal length 42
4.1.1 Cluster of identical consonants 43
[Link] Evidence of the writing 43
[Link] Evidence of morphophonemic alternations 43
4.1.2 Clusters with length 43
[Link] Occurring clusters 43
[Link] Members of clusters 44
[Link] Inventory of clusters 44
[Link] Advantages of the notation/:/ 45
[Link] Free variation o f / : C/ and /C :/ 45
4.2 Sequence of two vowels 46
4.2.1 Boundary between two vowels 47
[Link] Indication of boundary by writing conventions 47
[Link] Indication of boundary by a syllabic sign 47
[Link].1 Between/u/and/a/ 47
[Link].2 Between /i/ and /a/ 48
[Link].3 Between/a/and/a/ 48
[Link].4 Between/a/and/i/(or/e/) 48
[Link].5 Between/a/and/u/ 48
[Link].6 Between /u/ and /i/ (or /e/) 48
[Link].7 Between /u/ and /u/ 48
[Link].8 Between /i/ and /i/ 48
[Link].9 Between/i/(or/e/) and/u/ 48
[Link].10 Between/e/and/i/ 49
[Link].11 Between /e/ and /a/ 49
[Link].12 Between /a/ and /e/ 49
[Link].13 Between /i/ and /e/ 49
[Link] Notation of syllable boundary 49
TABLE OF CONTENTS 9

4.3 Consonant clusters 49

4.3.1 Non-occurring clusters 50
[Link] Notes on non-occurrence of certain clusters 50
[Link] Morphological significance 51
4.3.2 Distinctive feature analysis 51
4.4 Resolving of clusters 51
4.4.1 Two consonants 52
[Link] Consonant and length 52
4.4.2 Three consonants 52
[Link] Two consonants and length 52
[Link] Consonant and two lengths 52
[Link] Positions of the epenthetic vowel 53

5. MORPHOLOGY 54
5.0 Morphemes 54
5.1 Morphophonemic alternations 55
5.2 Inflectional classes 56
5.3 The noun 56
5.3.1 Declensions 57
[Link] Inflectional and syntactic distribution of the declensions . 57
5.3.2 Nominal subclasses 58
5.3.3 Gender 59
5.3.4 Number 59
[Link] Endings of the singular 60
[Link].1 Historical note 60
[Link] Endings of the plural 60
[Link].1 {u: ~ i:/e:> 60
[Link].1.1 Writing of {:} 60
[Link].2 {a:nu ~ a:ni} 61
[Link].3 {a:ne:} 61
[Link].4 {a:n ~ e:n} 61
[Link].5 {u:tu ~ u:ti} 62
[Link].6 {a:tu ~ a:ti} 62
[Link] Distribution 62
[Link].1 {u: ~ i:/e:} and {a:nu ~ a:ni} 62
[Link].2 {a:nu ~ a:ni} and {a:tu ~ a:ti} 63
[Link].3 {u:tu ~ u:ti} 63
[Link] Infixed plural 64
[Link] Plural morphemes 65
[Link] Generative statement of the plural 65
5.3.5 Case inflection 66
[Link] Declension (1) 66
10 TABLE OF CONTENTS

6. MORPHOPHONEMIC ALTERNATIONS 105

6.0 Isolating morphophonemic change 105
6.1 Consonants for which morphophonemic change may be documented . 106
6.1.1 /m/ 106
[Link] C+m 106
[Link] m+C 107
6.1.2 /n/ 107
[Link] m+n 107
[Link] ( + ) « + C 108
6.1.3 β / 109
6.1.4 /t/ 110
[Link] C+t 110
[Link] + / + C Ill
[Link] (C)+Vf Ill
6.1.5 /k/ 112
12 TABLE OF CONTENTS

6.2 Non-morphophonemic changes 112

6.2.1 Historical change 113
[Link] Changes from OB to MB 113
[Link] Changes from OA to MA 113
[Link] Changes from MA to NA 114
[Link] Other changes 114
6.2.2 Dialectal variation 114
[Link] Old Babylonian and Mari dialect 114
[Link] Assyrian and Babylonian 114
[Link] Babylonian (OB, MB, SB) and Neo-Babylonian . . . . 115
6.2.3 Free variation 115
[Link] Free variation of /a/ and /e/ 115
[Link] Free variation between voiced and voiceless consonant . 115
[Link] Free variation b e t w e e n / m / a n d / n / 116
[Link] Free variation between /a/ and /i/ in Assyrian 116
[Link] Free variation between is and us before consonant . . . 116
[Link] Consonantal metathesis 116

7. PHONOTACTIC ALTERNATIONS 117

7.0 Environments that determine alternants 117
7.1 Before the gender suffix {t} 117
7.1.1 Pattern 1 117
7.1.2 Pattern 2 117
7.1.3 Pattern 3 118
7.1.4 Pattern 4 119
7.1.5 Pattern 5 120
7.1.6 Pattern 6 120
7.1.7 Before feminine plural 121
[Link] Traditional explanation 122
7.2 Before terminal juncture 123
7.2.1 Other attestations of terminal forms 124
7.3 Bound form 125
7.3.1 An alternant of the bound form of adjectives 125

APPENDIX 128
1. A generative statement of the case inflection 128
2. A generative statement of the inflected verb 135

GLOSSARY OF AKKADIAN FORMS 137

GLOSSARY OF SELECTED LINGUISTIC TERMS 153

BIBLIOGRAPHY 154
ABBREVIATIONS

ABL R. F. Harper, Assyrian and Babylonian Letters.

AfO Archiv für Orientforschung.
AHw. W. von Soden, Akkadisches Handwörterbuch.
Baghd. Mitt. Baghdader Mitteilungen.
BiOr Bibliotheca Orientalis
Boyer, Contribution G. Boyer, Contribution à Γ histoire juridique de la Ire dynastie babylonienne.
CAD The Assyrian Dictionary of the Oriental Institute of the University of Chicago.
CT Cuneiform Texts from Babylonian Tablets in the British Museum.
GAG W. von Soden, Grundriß der akkadischen Grammatik.
Idrimi S. Smith, The Statue of Idri-mi.
JCS Journal of Cuneiform Studies.
OIP Oriental Institute Publications.
OLZ Orientalistische Literaturzeitung.
Or. N.S. Orientalia, Nova Series.
RA Revue d'assyriologie et d'archéologie orientale.
Syria Syria.
Szlechter, Tablettes E. Szlechter, Tablettes juridiques de la Ire Dynastie de Babylone.
TCL Textes cunéiformes du Louvre.
Tukulti-Ninurta Epic E. Ebeling, Bruchstücke eines politischen Propagandagedichtes aus einer
assyrischen Kanzlei (= Mitteilungen der Altorientalischen Gesellschaft, XII/2).
VAB Vorderasiatische Bibliothek.
Wiseman, Treaties D. J. Wiseman, "The Vassal Treaties of Esarhaddon", Iraq, 20, Part 1.
WZJ Wissenschaftliche Zeitschrift der Friedrich-Schiller-Universität Jena.
YOS Yale Oriental Series, Babylonian Texts.
INTRODUCTION

0.1. PURPOSE AND SCOPE

The purpose of this study is to present a substantial part of the corpus of the Akkadian
language in the form of a grammar. This grammar has roughly the subdivisions cus-
tomary in the field of Assyriology, but its organization has profited from the gram-
matical models and descriptions evolved in current linguistic research.
I have attempted to give a description of the elements of the language - phonemes
and morphemes - and the occurring combinations of these elements in terms of cate-
gories established from the analysis of Akkadian itself, and not in terms of some gram-
matical model - such as one or another language of the Semitic family - nor in terms
of the grammatical categories of the Assyriologist's native tongue or of some second
language acquired by him, such as Latin or Greek, in the humanistic gymnasium.
However, I have not attempted to adhere to any particular linguistic theory, and
this, in the eyes of many, may constitute a major shortcoming of this study. My rea-
sons for this stem in part from the nature of the corpus of Akkadian texts and of the
particular section that I have chosen for this grammar, and in part from the compo-
sition of the public to whom such a grammar should speak.
By the nature of its content and purpose, this grammar addresses itself to a twofold
audience. For the Assyriologist, it offers a series of restatements of language data.
Most of the data are well known and, moreover, have been arranged and organized
in patterns which are sufficiently contrastive and articulated, but they became over-
grown with a tangle of exceptions, diachronic considerations, and descriptions of
graphic or phonetic alternations unjustifiedly raised to a phonological level. Some of
the data, I believe, are new both in their statement of occurrence and in their interpre-
tation.
For the linguist, it tries to present the language data, stated, sometimes alternatively
in descriptive and generative terms, in such a way that material hitherto only available
in an incomplete collection, or, on the contrary, in an over-atomized form becomes
clearly articulated.
The grammar does not address itself to the beginner, unless the student be linguisti-
cally trained, since it offers neither hard-and-fast rules, nor extensive paradigms, nor
the customary handy categories to assimilate the new language to one already known
16 INTRODUCTION

to him, usually Hebrew. Nevertheless, while language teaching is not the primary pur-
pose of this description, its aim is that distributional facts and correlations be stated
in such a way that test sentences must be correctly generated; this serves, in the case of
Akkadian, not only a pedagogical purpose aimed at the student, but also serves the
scholar in need of providing a restoration for a fragmentary passage.
Some basic principles, by now elementary in most fields of linguistic study, have
been observed in this presentation, with the consequence that secondary or even ir-
relevant aspects of language description, to which the Assyriologist reader is accustom-
ed, have been deliberately neglected. Thus, while the title specifies the language as
"Akkadian" and not as one or another dialect or historical period that this term sub-
sumes, the grammar is essentially synchronic. In that, it adheres to the principles en-
unciated by F. de Saussure, and followed in the various structural schools of Europe
and America, according to which linguistic description should be synchronical and
never mix the synchronic and diachronic levels. Indeed, the prerequisite for a dia-
chronic study (historical grammar) is a set of studies on the linguistic systems of the
various periods of time and/or the areas (geographical dialects) which are genetically
related to the language.
There is in fact a greater need for studies on individual dialects and periods of Ak-
kadian than for a comprehensive grammar. 1 Such a grammar actually exists (W. von
Soden, Grundriss der Akkadischen Grammatik, 1952, to be cited henceforth as GAG) ;
based on a vast collection of material, it cites the regular and irregular forms of each
dialect and period, for each grammatical category; it also takes cognizance of excep-
tions and deviating forms, often however through a set of rules devised to take care
of precisely such forms. On the other hand, descriptive studies of dialects are non-
existent or inadequate by linguistic standards. The best dialect study, that of I. J. Gelb
on Old Akkadian, is of necessity based on a very limited corpus and has to resort to
etymological reconstructions to fill in missing paradigmatic material. All other studies
duly note what is divergent from an ideal dialect - Old Babylonian is usually arbitrarily
selected to serve - and fail to give a complete list of occurring forms. They illustrate
the "Barmecide approach", the apt term of B. Bloch (Language, 21, 112f.) for des-
cribing a language by telling what it lacks.
In keeping with the Saussurian principle, this study keeps apart the two levels of
synchronic description and diachronic analysis, although a purely synchronic ap-
proach is made difficult by the nature of the corpus, about which more will be said
presently. In the face of the difficulty involved in selecting a dialect that would offer
adequate coverage, the corpus on which the study is based is the literary dialect of Ak-
kadian, which will be further defined in Chapter 1. This has been selected as the one
which provides the widest differentiation among the most material. Contrasting dia-
lects - the most diverging, or the less well attested, or the less easily elicited ones - are
1
"Als erstes möchte ich da Einzeluntersuchungen nennen, die der Schreibweise, der Grammatik
und dem Wortschatz innerhalb der einzelnen Sprachperioden und Dialekte gewidmet sind" (W. von
Soden, 1961, p. 47, referring to the main tasks in the field of Akkadian).
INTRODUCTION 17
occasionally adduced; none of the peripheral dialects is included in the primary
presentation, but will be used occasionally as illustration when helpful.
To have selected a body of texts uniform enough to exclude dialect variation and
historic change would have meant to restrict the data to the point that the grammar
would have ceased to be useful to anyone but a specialist. My decision to include a
large and varied corpus has extended the range of variations beyond that accountable
for by distributional or similar criteria. Inevitably then, material which a finer seg-
mentation of the corpus into dialects, genres, and even idiolects, would have sorted
out further, appears here lumped together. In such cases, and further, in order to
deal with features and contrasts that may be of interest to the reader and may also
serve to delimit the dialect described against other dialects of Akkadian, I have added
a number of "footnotes" and "asides" ; these are set off in the layout by reduced
type. These asides take the place of perhaps desirable short sketches on the other
dialects, which, however, it would at this stage be premature to try to present.
Etymological considerations are naturally not taken into account, although some-
times they are referred to in comparing the traditional analysis of some forms with
the analysis preferred here.

0.2. ORGANIZATION

By way of general introduction, the first two chapters deal with the corpus of Akka-
dian texts and with the system of writing. The main body of the grammatical treat-
ment that follows is divided into Phonology, Phonotactics, and Morphology. Since
these represent convenient chapter headings rather than a hierarchy in the Bloom-
fieldian sense, morphophonemic alternations are treated after the morphology. The
contrast between morphophonemic alternation and other types of phonetic change is
illustrated at the end of that chapter. An interesting prospect is to eventually follow
up morphophonemics with a generative phonology.
Higher units of syntax have not been treated in the framework of this study - apart
from a tentative analysis of partial morpheme structure in the Appendix - because I
considered the statement of the distribution of the lower rank units as the more urgent
task. Eventually of course, and ideally after some agreement has been reached by As-
syriologists on the distributional categories here defined, phrase structure should re-
ceive similar synchronic treatment, preferably according to a generative model, not
because of any inherent virtue of the generative grammars, but mainly because such
an approach, more in harmony with the viewpoints prevalent up to now in Assyrio-
logical circles, is more likely to enlist the cooperation of philologists in our field. For,
as suggested by Chomsky, "It would not be inaccurate to regard the transformational
model as a formalization of features implicit in traditional grammars, and to regard
these grammars as inexplicit transformational generative grammars" (Chomsky, 1964,
p. 918).
18 INTRODUCTION

The chapter headed Phonology is my attempt to classify the information that is

available on the basis of the written texts and comparative Semitics, the two factors
which permitted the decipherment of Akkadian. Hence, rather than phonemics, it
contains the minimum morphophonemics and phonetics gained by combined consid-
eration of the graphic system and Semitic etymology. The chapter headed Morphe-
mics treats more extensively some morpheme classes (nouns, verbs), and rather cur-
sorily only, or not at all, some others which belong rather to the lexicon or to syn-
tax, such as prepositions, adverbs and adverbials, and numerals.
In this sense, this study is not really a grammar, but rather a set of statements and
restatements. I believe that making these restatements is a necessary first step before
a grammar of any dialect can be attempted.
Every Assyriologist will be able to quote references to constructions that have not
been treated in this grammar. I myself have a collection of such examples, both of
morphological and of syntactical oddities, which I have not included because a gram-
mar should not be a collection of oddities. Once a satisfactory distribution of these
seeming oddities is found, they have to find their place in the grammar. For the time
being, it has seemed to me more important to give a basic framework for the operation
of the Akkadian language.
The above mentioned comprehensive grammar of Akkadian (GAG), with the many
dialect grammars it comprises, and the recent dictionaries (AHw. and CAD) have been
extensively consulted as an insurance against significant omissions of both "normal"
and "irregular" forms. Hence much of the study takes the shape of restatements of the
distributional categories apparent in the GAG and in the dictionaries.
New terminology has been avoided in order to permit the use of this book by any
Assyriologist who may be inclined to consult it. The current and accepted terms are,
however, in many cases more strictly defined. When it was found necessary to intro-
duce new terminology, the current term is cited in parentheses after the new one.
Akkadian words - cited in their transcription forms as they appear in the diction-
aries, or in transliteration - are normally not accompanied by references when these
can be found in the published volumes of the AHw. and the CAD. These two diction-
aries now cover the letters A, B, D, E, G, {J, I, J, Κ (partly), Ç, and Z. References are
given only for words beginning with letters that have not yet appeared in these two
dictionaries, as well as for forms or spellings not cited in the dictionaries. Words not
glossed in the text are found in the Glossary on p. 137ff.
The categories, lists, and paradigms here given do not claim to be the only possible
categories, lists, paradigms, or other divisions applicable to Akkadian ; the solutions,
as I have often tried to indicate explicitly, have no claim to uniqueness ; in short, I do
not claim to know the truth. I have presented the linguistic facts, trying to take care
of all the facts and to show what is significant about them. There might be more eco-
nomical and more elegant ways of presenting them, and certainly alternate ones that
would also do justice to the facts; in fact, I have indicated some of the alternate pos-
sibilities in the relevant paragraphs.
INTRODUCTION 19

The structural linguist is asked to bear with polemical or narrative sections intended
to rectify or refute views held by a great majority of philologists in the field, even if
such polemics seem to him pounding at open doors. Such doors may indeed be still
closed to the philologist, not entirely without the fault of the linguist. The aim of this
study will be reached if it helps to exorcise some ghosts and, at the same time, serves
as usable material to the general linguist who, no doubt, will find many opportunities
to refine and further restate the restatements of the present work.

Symbols and notations: capital letters are used for conventional, or conventionally
identified, signs of the cuneiform syllabary; / / enclose phonemic transcriptions; [ ]
enclose phonetic transcriptions, or rather, phonetic approximations; { } enclose
morphs or morphemes. Akkadian words or segments not enclosed in one of these
pairs normally represent the citation forms currently used by Assyriologists, or a
preliminary transcription, and are mostly in italics. Whenever special attention is
drawn to the written image of a cited form, the latter is enclosed in < > ; for the same
reason values of cuneiform signs are also sometimes enclosed in < ).
1. THE CORPUS

1.1. THE LANGUAGE

Akkadian belongs to the Semitic family of languages and was written in Mesopotamia
and the literate countries of the ancient Near East from roughly 2400 B.C. to 100 A.D.
It was spoken in Mesopotamia for a considerably shorter period of history, since lan-
guages belonging to the same Semitic family were concurrent or competing with it, to
an extent that cannot be determined, for precisely the reason that the extant written
records are in Akkadian. For this reason, I prefer the term "written language" to that
of "dead language" when referring to Akkadian; the former term is not only histori-
cally more correct but also indicates the nature of the difficulties that face the linguist,
since problems arise not so much from the fact that the language is dead, i.e., has no
more living speakers, and hence no more new material can be elicited nor the analyzed
material submitted to checks by such speakers, but from the fact that it is known to us
only through written records. While a "dead" language may be known, e.g., from
tape recordings, and thus subject to phonetic analysis, but have a finite corpus, a writ-
ten language cannot be subjected to phonetic analysis but its corpus may be expanded
by the discovery of more - even if usually of the same kind of - written records. For
the peculiar problems confronting the linguist dealing with a written language see, e.g.,
Mcintosh, 1956, and Chao, 1961.

1.2. DIALECTS

The written records of Akkadian form no continuous stream but fall into isolated
groups of texts from areas geographically distant from each other through a period
of, as has been mentioned above, over 2,000 years. The compulsion of tripartite di-
vision has quite naturally led Assyriologists to divide such groups of texts according
to the postulated two main dialects of Akkadian, into Babylonian and Assyrian, Old,
Middle, and Neo. Since these divisions soon proved insufficient, while conserving the
terms of Old, Middle and Neo-Assyrian (OA, MA, NA), and Old, Middle and Neo-
Babylonian (OB, MB, NB), in recent years the dialects Old Akkadian (OAkk.) - for
the oldest periods - and Late Babylonian (LB) - for the Persian and Seleucid periods -
have been added to the dialect classification, and research has started on the dialect
areas of Old Babylonian.
THE CORPUS 21

Obviously, these dialect groups need to be further refined; ideally, material from
one or more neighboring sites should be described each in its own right, and when lim-
iting such a dialect in time, one must not leap over gaps of a hundred years or more.
Peripheral dialects, and documents written outside the Akkadian language area, must
be investigated separately; in fact, the only pertinent linguistic analyses up to the pre-
sent have been performed on such documents only. The dialect texts include letters,
legal and administrative texts, about 30-40 thousand in number, of which, however, no
more than 4-5,000 are letters, while the rest contain, apart from recurrent formulas,
very little language material.
The definition of dialects and their relation to each other present difficulties which
should at least be pointed out here.
The first question is whether our subdivisions as mentioned above delimit dialects or
rather separate languages. I have here included under the name "dialect" not only
geographically different forms of speech but also the historical stages of the language
or languages considered.
Taking the terms dialect and language in their rough definition,11 would be inclined
to consider Old Akkadian and Neo-Babylonian as distinct languages. For the chrono-
logically intermediate periods (i.e., roughly 2000 - 600 B.C.), I assume the existence of
two dialects, Assyrian and Babylonian.
This assumption may be considered the statement of negative evidence, i.e., of the
fact that there seems to be no convincing way of deriving the earliest attested Assyrian
or Babylonian texts from Old Akkadian, nor deriving Neo-Babylonian from the pre-
ceding stages of Babylonian without admitting at least considerable interference from
other Semitic languages.
Outside the requirements for a definition of a geographical dialect area fall those
texts that are customarily called "literary" which originated at an earlier, the Old Ba-
bylonian and post-Old Babylonian, period, and have survived either in both Old Baby-
lonian and late copies, or, for the great majority, in late copies only. While the writing
habits changed greatly in the interval (see 2.3), the late copies usually preserve rather
exactly the tenor of the original text; the slight differences must reflect phonetic, per-
haps even phonemic change, and some change in morphology which, of course, may
conceal much more far-reaching historical change. Since, however, the wording of such
texts is changed in very rare and exceptional cases only, we have no answer to the
question whether such "literary" texts still represent a language intelligible to persons
other than the ancient scholars who copied them, nor a clue to the relation they may
have had to the language spoken. The dialect of these "literary" texts has been, and
will here be termed, Standard Babylonian (SB); whatever the relation of such texts may
have been to the spoken language at any given period, they represent a legitimate cor-
pus for linguistic analysis. The number of lines of SB texts may be estimated at about
1
"The most spectacular and best known (linguistic variation) is that between geographical forms
of speech. When the differences are small, these are known as dialects. When larger, they are known
as languages. However, no exact definition of these two terms is feasible" (Gleason, 1961, 398).
22 THE CORPUS

40 thousand, or twice that of the Iliad and the Odyssey. A great many of these exist
in several duplicates. This corpus should be kept separate from other dialects that can
be circumscribed in time and geographically, which reflect the language of a scribal
center and which themselves may or may not be integrated into another kind of rigid
transmission, such as legal, economic, and administrative texts. It is only in private
letters - predominantly from the OA, OB, NB, and NA periods - that some evidence
can be found for the spoken language of an area at a given time.
Another group of texts, customarily included in "Standard Babylonian", have again
to be analyzed in a different way : these are the reports of named and known Assyrian
and Babylonian kings that can thus easily be dated but that use the literary dialect and
are clearly modeled after those of their predecessors in style, without being just slavish
copies of them. Reports on the military achievements and building activities of the
Assyrian and Babylonian kings written at their courts have to be grouped according to
their literary models - sometimes a Sumerian one - with due consideration given to the
efforts of the sometimes very ingenious authors to introduce novelty and variety in
this "court-style" by a new use or combination of an old phrase. Whether such creati-
vity of individual scribes at royal courts should be taken as antiquarian interest, lit-
erary flair, or conscious language reform has to await further investigation based on
the above-sketched lines. Again, as any of the distinct parts of the corpus enumer-
ated above, these written records can likewise be adduced for language description if
their specific character is constantly borne in mind.
2. THE WRITING SYSTEM

2.0. PRELIMINARY STATEMENT

The following aperçu deals with the elements of the writing system, their preferred
combinations and limitations. It is intended partly as preliminaries to some statements
made in the section on phonology, and partly as a general statement of the principles
which I follow in my linguistic interpretation of the texts. It does not constitute a
"graphology" or "graphemics", the theoretical foundations of which are very scant
yet and have not been applied to the writing of Akkadian. At this stage it is, however,
important to stress the difference between graphs, the elements of the writing system,
and their referents; the two are often loosely used one for the other. Ideally, we should
be speaking only in terms of allographs, morphographs, etc. ; if we wish to avoid the
tedious encumbrance of such a terminology and use the terms customary in grammati-
cal analysis, we must at least be aware that we are using a shortcut.

2.1. THE SCRIPT

The written records of Akkadian make use of the cuneiform writing system, the graphs
of which, when Akkadian texts begin to be written in it, are arranged in horizontal
lines progressing from left to right (with the notable exception of the stela with the
Code of Hammurapi).

2.1.1. Graph and referent

A graph in the cuneiform writing system is a cluster of wedges imprinted in clay, or

imitations of such imprints in other materials, which are usually set off by a space at
the left and the right. Such a graph is called, in Assyriological parlance, a "sign" (as
graphs in the Chinese system are called "characters"). The referent of the sign in the
language for the writing of which it is used is called its "value". (Thus there are "Su-
merian values", "Akkadian values", etc.) The referent or "value" of a graph is a text
segment statable as a sequence of one or more phones. A text stated as a sequence of
values is called "transliteration". This definition - which is not necessarily cotermin-
24 THE WRITING SYSTEM

ous with other definitions of transliteration used by Assyriologists1 - is of importance

for the following, wherein we shall speak of the elements of the system called transliter-
ation and not of the system of writing proper.
A configuration of wedges which is the graph is not normally a compound of
constituent configurations. E.g., the graph PA plus the graph IB do not equal a graph
PA + IB but the graph SAB. This statement is necessary because some graphs (those
corresponding to some of our numerals) are indeed compounds of their constituents,
and also because some graphs have been considered by the ancient scribes as com-
pounds of some other graphs; this is evidenced partly by the scribes' analysis of such
graphs into two or more components in the sign lists, and partly by the names there
given to such graphs (see Christian, 1913).
Since a graph is not normally a compound of constituents, obviously the constitu-
ents of a graph have no relation to the constituents of the referent ("value") ; this may
be proven by the fact that allographs - or historical diagraphs - of the same graph are
not necessarily decomposed into similar constituents.
Such ancient learned constituent analysis is different from the case of the so-called
ligatures, which are rare instances of graphs whose constituents have a relation to
the constituents of the value; this relation is sometimes indicated in the translitera-
tion practiced by a +-sign instead of a hyphen in connecting the constituents of the
value. Such ligatures are the sign IA (ligature of I + A with non-syllabic i, i.e., a
monosyllabic value), and such ligatures as i+na, as+sum, as+sur, as well as liga-
tures with the numeral 1: 1 +en, and 1+ei, and with determinatives (see 2.2.1), such
as the Sumerogram DINGIR+EN.

2.1.2. Graphotactics

The arrangement and the distribution of the graphs (graphotactics) express linguistic
information. The distribution of the graphs segments the written utterance into
smaller units, in a way somewhat comparable to punctuation marks. The units thus
gained we identify with linguistic units, with the word as the lowest level; what is
important here is the fact that units below the rank of word are not segmentable from
the writing system as it has been analyzed so far.
The largest unit is the line which is normally coextensive with a clause or a group
of clauses ; a line may or may not be coextensive with a sentence. Specifically, a line
in a literary text does not contain a group less than a clause; if there is no room on the
line for the whole clause, the next line is indented. This holds for the literary dialect ;
there are exceptions, even in SB royal inscriptions, but mostly in letters, and in texts
written in peripheral areas. A line on a tablet inscribed with a letter often contains
1
For instance: "Transliteration is a form of graphic transfer wherein one sign (or a combination
of alphabetic signs and artificial symbols) stands for each character of the writing we are recording"
(Gelb, 1948, 2).
THE WRITING SYSTEM 25
less than a clause, and a line on a peripheral tablet inscribed with a literary text may
contain part of one clause and part of the next clause ("enjambement").
The smallest unit is the word. A graphemic word is one or more graphs the first
and last of which at times co-occur with the beginning and the end of a line respective-
ly. In other words, a word as defined by graph-arrangement is not divided between
two lines, and in this definition stands for the grammatical unit: base plus affixes or
compound. Again, exceptions occur, mostly in letters, when the suffix morph or
morphs consisting of several graphs may run on into the next line ("rejet").
Selection in the occurrence of graphs also determines words. Thus, word final is oft-
en marked by occurrence of a graph which normally occurs in word final only, e.g.,
su, and some others which rarely occur in other than word final, e.g., turn, kám, and
conversely, word initial is marked by the absence of such graphs or of other graphs
that do not occur in word initial, i.e., u.
A word consisting of one graph is sometimes so identified by the occurrence of a
special graph, or graph sequence, e.g., ù {u}, or lu-u {lu} (however, the graph la
which corresponds to the word {la} is identified in some dialects only by the sequence
la-a).
Additional information is sometimes provided by the properties of a written
document (writing material, size and shape) ; these normally provide a rough indica-
tion of the type of its content (letter, royal annals, literary text, and so forth), of its
geographical situation, and of the time when it was written. In marginal cases, the
configuration of a given graph may carry a direct semantic reference. Thus, numerals
written with horizontal wedges refer to particular units of volume, and numerals
written with vertical wedges refer either to other units of volume or to things counted.
These considerations however shall not concern us here.

2.2. ELEMENTS OF THE TRANSLITERATION SYSTEM

The system of transliteration as defined above is a composite system in two respects:

through the elements' two-fold relation to the written utterance, and through the di-
versity of the elements themselves.

2.2.1. Determinatives

First, some elements correspond to text segments which have a reference in the phono-
logical system, i.e., we assume that they formed part of the utterance, were pronounced,
while other elements correspond to text segments which have no reference in the pho-
nological system, i.e., were not pronounced. The referential status of the latter is that
of an "ideogram", inasmuch as they directly represent meaning. Their property is that
of a graphic affix - pre- or suffix - which marks form-classes. Hence, the usual names
"determinative", which more properly refers to their graphic affix property, and "se-
26 THE WRITING SYSTEM

mantic indicator", which more properly refers to their ideogram status, each describe
one of their aspects. I would also suggest the term "class indicator" as describing their
property of marking form-classes, in the manner of a "written inflection". It has not
yet been determined what kind of inflection is manifested by this "written inflection",
but such an investigation should form the basis of a segmentation of the corpus in the
analysis of the writing system itself.
Secondly, we may divide the elements of the system by the ways in which they match
text segments. Their usual division is into syllabic signs (syllabograms) and word signs
(logograms). These may be considered two sub-systems.

2.2.2. Word signs

The word signs (logograms) occurring in Akkadian texts are mono- or polysyllabic
words or phrases borrowed from texts written in Sumerian, the language for which
this writing system is first attested. Logograms are signs which have Sumerian graphic
etymologies and also represent text segments, but they represent Akkadian phonic
material only insofar as such is substituted for them, and in consequence have to be
excluded from the phonological and morphophonemic analysis. Note that the class in-
dicators are also word signs which have Sumerian graphic etymologies, but, although
they may be homographous with logograms, they do not have the further morphemic
references of logograms.

2.2.3. Phonetic complements

The phonetic complements or phonetic indicators are a type of graphic affixes whose
reference may be identified as morphophonemic. They occur with logograms in the
following way: When a word sign (logogram), i.e., a Sumerian word, occurs in an Ak-
kadian context, it is often followed by a sign whose value usually consists of the last
consonant of the corresponding Akkadian word (i.e., the translation word) and of a
vowel, sometimes of yet another consonant, i.e., by a CV or CVC sign; this sign nor-
mally represents the paradigmatic ending of the word required by the Akkadian con-
struction. This syllabic sign, in contrast to the preceding logogram which is customar-
ily romanized in Roman capitals, is customarily romanized in italics just as other Ak-
kadian text-segments, and connected to the preceding word sign by a hyphen, or, if
according to another custom of transliteration the logogram is transcribed in Akka-
dian and thus romanized in lower case italics, the "phonetic complement" is set after
it raised one half line: e.g., KI-tim or er?etirntim, 'earth'; some transliterate even
erpeti" by giving the sign TIM the syllabic value "ti". However, two diachronic stages
have to be distinguished : In the earlier periods, including OB, the sign groups A.SÀ-
lum, A.SÀ-Zaw and A.ÈÀ-lim for instance are to be read respectively eqlum, eqlam and
THE WRITING SYSTEM 27
eqlim, and the syllabic sign after A.SÀ indicates the grammatical ending of the word,
as well as the fact that the preceding sign or sign complex is to be read in Akkadian,
not in Sumerian. In later periods, the syllabic sign appears in a form - usually in one
or two only - that does not correspond morphophonemically to the grammatical func-
tion of the word. Thus, the signs KI-tim (more rarely Kl-tum) stand for any of the fol-
lowing forms of the word {ersetu} : ersetu, er$eta, erseti·, the signs ΑΝ-ιί or AN-e stand
for the forms Samû or samé, interchangeably; here the phonetic complement indicates
that the graph adjacent to the left is to be read in Akkadian, but no longer indicates the
grammatical ending of the word. However, it still indicates which lexeme is to be
selected from among the several Akkadian correspondences (translations) of the
Sumerian word, and thus such writings as KI-tim, AN-« indicate that {ersetu} is to
be selected rather than {itti} or {asar} for KI, {samû} rather than {ilu} for AN.
Thus, the phonetic complement or phonetic indicator (Gelb, 1948, 104f.) gives in all
periods (1) morph information, namely directions (a) to select an Akkadian morph and
(b) which morph to select, i.e., the morphophonemics of the adjacent graph, and (2)
suffix information, namely (a) it fixes a suffix position, i.e., the occurrence of a suffix,
(b) it indicates the suffix set, and (c) in the earlier periods, it specifies the morphopho-
nemics of the suffix. Hence it is a graphic affix with morphophonemic reference.
This clarification is necessary in order to avoid drawing phonological or morpholo-
gical conclusions from a feature that is peculiar to the writing system. For example,
one is not permitted to say that the sign TIM has the reading -ti because it is appended
to the word to be read erpeti at a period when the final /m/ of the nominal case-inflec-
tion had disappeared (see [Link]); nor is one permitted to say, on this basis alone, that,
because the sign complex KI-//m also stands for ersetu, after a certain date the case
endings all coalesced in a "schwa" (which for the Orientalist often is a synonym of
"indistinct vowel"), or that final vowels were "dropped". For a discussion of this
problem see below [Link].

2.2.4. Syllabic signs

Syllabic sign here, as in comparable writing systems, is used in a special technical sense
(see Gleason, 1961 p. 414); syllabic signs match, that is, their values are expressed in
terms of, syllables or syllable segments. A border case is that when a syllable consists
of a single vowel, and in this case only is the sign "alphabetic". There is no sign whose
referent is less than a phoneme, although some graphs represent the configurative
mark "syllable boundary" (see 3.1.2).
The values of syllabic signs, i.e., statements of text segments in terms of phones,
have been established, in the process of the decipherment of the cuneiform writing, by
combined consideration of the graphic system - the first values of which were given
with the help of bilingual texts (including proper names found both in cuneiform writ-
ing and other known writing systems, e.g., Old Persian, Greek, Hebrew) - and Semitic
28 THE W R I T I N G SYSTEM

etymology. Since Akkadian was recognized to be a Semitic language, phonemes

known in other Semitic languages, Hebrew, Arabic, etc., were expected to be found in
Akkadian too. For the implications of this assumption, see below, 3.0.
The retracing of the procedure which led to the assignment of values to signs over
more than a hundred years of decipherment cannot be attempted here; what seems to
have been a trial-and-error method continues to be applied as various refinements are
brought into the transliteration system.

[Link]. Types of syllabic signs.

The elements of the syllabic sub-system are signs representing vowel (V) alone, or
vowel and consonant (VC, CV, and CVC). (VCV and CVCV values given to some
signs are precisely the result of applying morphological - or morphophonemic - rules
to the writing system.2)

[Link]. V signs.
Of V signs, there is one each for a, e, and i, but three for u. The homophonous u-
signs have been taken over from Sumerian, and apart from the OAkk. period when
one of these may have represented the syllable /ju/ in word-initial, these are in fact
homophonous though not always interchangeably used, as they became stabilized in
certain writing habits.

[Link]. CV signs.
The distribution of CV signs is uneven. For some consonants (t, s), there is a sign
each for the syllables Ca, Ce, Ci and Cu ; for others (ρ, d, t, k, g, q, s, z, s, h, 1, r) there
is only one sign for the syllables Ce/i (e.g., left, re/i, ge/i, etc.); in still other cases,
although there exist two different signs for Ce and Ci syllables, only one of them is
used in one dialect, or the use of one is restricted (ne/ni, me¡mi, be/bi). There is only
one sign for any of the vowels that may occur with the two "semi-vowels" [w, j] and
the glottal stop [?], (i.e., wV may be theoretically read [wa, we, wi, wu], etc.); for these
see below, 3.1.3.
Conversely, as to CV signs in which the vowel is the same, there are sometimes less,
sometimes more signs than correspond to the phoneme inventory. In the case of ho-
mophonous signs - differentiated in the transliteration by diacritics - such as su and
sú, the distribution is conditioned by scribal habits, such as, in the mentioned case, the
second is normally used in word final, and not (although exceptions occur) in word
initial. On the other hand, while some signs can be grouped into pairs which differ in
one phonological feature only, such as voice, other CV signs are polyvalent in this
respect: For example, there is a different sign for each of the syllables /ba/ and /pa/,
/bi/ and /pi/, but one sign only for both /bu/ and /pu/. Another asymmetry resulted
in the dental and velar row: In Old Babylonian there were two signs, ka and ga, for
« Such values have been suggested for the NA dialect by Deller, Or., N.S., 31 (1962), 7-26 and 186-
196.
THE WRITING SYSTEM 29

writing the three k-row consonants followed by the vowel a, i.e., the syllables /ka/,
/ga/, and /qa/; similarly, there were two signs, ta and da, for writing the three /-row
consonants followed by the vowel a, i.e., the syllables /ta/, /da/, and /ta/. To the first
pair a third sign, qa, was added in the course of the centuries, thus providing one sign
each for the syllables /ka/, /ga/, and /qa/, but no third sign was added to the pair ta
and da ; nevertheless we continue to interpret the sign ta as representing the phonemic
sequence /ta/, or /ta/, transliterated as ta or tá respectively, and the sign da as re-
presenting the phonemic sequence /da/ or /ta/, transliterated as da or ta respectively,
depending on etymological or morphophonemic considerations.

[Link]. VC signs.
Among the VC signs, most iC signs can also be read eC, with the exception of is,
il and in, since there exist the signs es, el, en. Moreover, there is only one VC sign - in-
cluding homophones having a traditionally restricted distribution, such as as and ás,
ar and ár, at and at - for any syllable whether the final consonant is voiced, voiceless,
or "emphatic". In other words, the oppositions voiced vs. voiceless vs. emphatic are
neutralized in VC signs, e.g., in the signs ad/t/t (also ádltjt), ab/p, ig¡k¡q, iz/s/s,
uz/s/?, etc.

[Link]. C VC signs.
CVC signs are less evenly distributed; the oppositions voiced/voiceless or voiced/
emphatic, or all three, are again neutralized for the final C, and in some signs also for
the initial C, as in d/tan, d/t/tal, b/pan, d/t/tim, b/pal, djtir, z/sum, g/k/qir, etc.

[Link]. Sequences of signs.

A sign sequence CV^VjC may be equivalent to a sign CVXC when such a sign is a-
vailable in the graphic system, and the sign sequences CV 1 -V 1 and Vj-VjC may be
equivalent to one sign CVX and VjC. No other sign sequence (e.g., VC-CV(C), CV-
CV(C), CVi-VüQ is equivalent to a sign CVC, i.e., represents less than two syllables,
and the sequences first mentioned may also represent two syllables, see below 4.2ff.
The writing habits are such that within one word, sequences of (C)V-CV-CV signs
or of (C)V-VC-CV-VC signs occur, but not normally sequences of (C)VC-VC-VC
signs. The latter restriction is of importance for phonology and morphology, see
below, 3.1.2 and [Link].

2.3. CHANGES IN WRITING HABITS

The changes in the writing system between early texts (partly including OB) and later
texts (OB and later) mentioned in 1.2 manifest themselves not only in such material
features as the sign forms, or the introduction of new signs in the form of ligatures or
by differentiation by means of diacritics (V? and V£f), or of new values - such as ad-
dition to or restriction of the polysemy of a particular sign - or the increasing use of
30 THE WRITING SYSTEM

CVC signs as well as a more frequent use of word signs, but also in important innova-
tions in writing habits which affect the "fit" of the system. Innovations that can be
dated to the post-Old Babylonian period, or that became current then, are :
(a) Long consonants are indicated by (C)VC 1 -C 1 V graphs (and not as before by
(C)V-CjV graphs) ; it is to be remembered that the doubling of the consonant element
is not visual, as the above symbolization might suggest;
(b) A tendency of treating components of groups - compounds or other stress-
units - as free forms (i.e., decrease of sandhi-writings) ;
(c) Lack of indication of the morphophonemic alternant of the "head" or free lexi-
cal item element of a paradigmatic form, more observable in nouns than in verbs;
(d) Technical texts (medical, pharmaceutical, divination texts) often use a group of
graphs that represent either the free form or an invariable bound form of a word, most
likely for quick visual identification, with presumably only secondary reference to the
paradigmatic form; e.g., they write mar-ha?, tak-fir, tar-kas, tal-pap, etc., for items
whose full phonological form is /marhasV/ i.e., marhafu, marhasa, marhaçi, /takçitrV/,
/tarak:as/, /talap:ap/, etc.

2.4. PROBLEMS CAUSED BY THE WRITING SYSTEM

The asymmetry of the writing system is of course the legacy of the Sumerian, or pre-
sumably even of the Pre-Sumerian language which was first reduced to writing in cu-
neiform. Signs homophonous in Akkadian may have been distinctive in Sumerian;
polyvalence often resulted from the adoption for the same sign of both its Sumerian
word-value and its Akkadian translation value.
Since the writing system used by the Sumerians was not adequate to express dis-
tinctions significant in their own language - which makes one suspect that this system
was not originally invented to write Sumerian, see e.g., Dávid, 1945, 13ff. - the inade-
quacy, or lack of "fit", remained and increased when it was borrowed by the Akka-
dians, both in the direction of overdifferentiation and of underdifferentiation. Pallia-
tives such as scribal conventions and later secondary differentiations of signs to intro-
duce into the writing differences pertinent in the language never reached a point where
it could be said that every combination of phonemes found expression in the writing.
While it is well known that the writing system was not designed for Akkadian, it is
not generally accepted that it was not designed for Sumerian either. Statements
found in several articles on the subject which, after reconstructing the Sumerian pho-
nology from the writing system, and the Akkadian phonology from the same writing
system, find that the two phonological systems are indeed very close, only point up
the inadequacy of the writing system for either language. It seems to me that the more
one is able to match the phonemes - and a fortiori, the phones - of the two languages
on the basis of the common writing system, the less likely it becomes that there was an
original fit between Sumerian writing and Sumerian language. Hence the assumption
THE WRITING SYSTEM 31
- in my mind unwarranted - of a quasi-perfect fit involving two languages as different
as Sumerian and Akkadian can only indicate that there is a fit with neither.
When, at the time of the decipherment, many values assigned to the syllabic signs
were based on the Semitic etymology of the words, a certain arbitrariness was intro-
duced into the system of transcription from which Assyriology, in spite of many ef-
forts to remedy this, cannot be completely freed. The so often alleged polyphony and
homophony of the syllabary is to a large extent due to lack of rigor in the system of
transliteration. In recent years it has been recognized and advocated - especially by
I. J. Gelb, see e.g., Gelb, 1952,108f. - that there is little, if any, polyphony in the syllab-
ary of any given period. The syllabary of W. von Soden (von Soden, 1948) is the first
systematic attempt to limit certain syllabic values to a given dialect. However, the
inherent falsification that results from applying etymological correspondences and a
priori established, normative grammatical rules, i.e., interpretation before the facts
have been collected, in assigning syllabic - and hence phonemic - values to the signs
of the syllabary has not only not been corrected, but has not even been admitted.
In truth, the restrictions and refinements introduced by the above-mentioned scholars
and others are only palliative in effect, and Assyriologists still seem to move in a vi-
cious circle from grammar to transcription to grammar and so forth.
The origin of all this confusion seems to be the historical accident that the first
transliterated - or deciphered - texts were Assyrian royal inscriptions written around
1000 B.C. and later; the values ascertained from these texts were then transferred to
texts written about 1000 years earlier, in a less differentiated system of writing.
Since these earlier, Old Babylonian, texts soon were taken to represent an "unspoil-
ed" - often called "classical" - period of the language, syllabic values for which no
special sign existed in the OB system of writing were assigned to some Old Babylonian
signs, and these secondarily posited values could in turn conveniently be substituted
in the later texts whenever etymology necessitated the positing of a consonant or vowel
which had not been included in the reading of the particular sign when originally as-
signed its syllabic value. As peripheral texts - written by Hittites, Hurrians, or
other non-Akkadian speakers - were discovered, further new values came to be as-
signed to the signs, when the use of diverging sets of signs only indicated the inability
of the foreign speakers to differentiate Akkadian phonemes, or their adherence to a
scribal tradition transmitted through foreign channels.
Such are the reasons for which a great number of homophones have been intro-
duced into the syllabary by the decipherers, and for which the unaware observer thinks
that the Assyrian texts give a more exact transcription of the language than the Old
Babylonian ones. No systematic attempt has as yet been made to revise the entire sys-
tem of transcription; the efforts of some scholars to adhere rather to the principles
of the writing system than to a normative grammar have been greeted with derision
and often construed as a lack of knowledge of the grammar.
Two roads are open to the Assyriologist: to observe in his transliteration only the
differences overtly expressed by the writing system, or to expand the values of signs by
32 THE WRITING SYSTEM

introducing the maximum number of differentiations for one sign within the limits set
by the system. As an example : he may choose to transliterate the sign BU always bu,
whether the etymology requires bu or pu. On the other hand, he may give any of the
following values to the sign DI (with the exception of OA texts) : di, de, ti, (e; or to the
sign IB the values ib, ip, eb, ep ; even to the sign GI the values gi, ge, and (the less fre-
quent values) qì, qè. All these remain within the possibilities of the system (see above,
2.4.2 and 2.4.3). However, violence is done to the system itself if the syllabic value se20
is assigned to the sign Si when there is, and the texts in fact use, the available sign SE
to write this syllable. To a ridiculous extreme go those who go ad absurdum in as-
signing "phonetically required" values to signs used in a peripheral, non-Akkadian
scribal center, in order to change this particular, peripheral, system, conditioned by
the language of the area, into a system consistent with Akkadian phonology.
The system of transliteration presently current seems rather adequate to its users,
but is harmful in obscuring some points of phonology and morphology. These points
are : distinction between the phonemes3 /e/ and /i/ and in consequence the morphemes
into which one or the other enters; the distribution of voiced, voiceless, and emphatic
sibilants and of the palatal groove fricative /§/; the distribution of vocalic and conso-
nantal length; the morphemic shape of endings (verbal or nominal suffixes).
This rather short list does not look unduly dangerous to affect the description of the
language; it will look different, however, if we state the positive versus the negative:
not affected by our present system of transliteration are: the vowels /a/ and /u/; the
consonants /m, n, 1, r, ? and h/; consonantal prefixes; most, but not all, /s/ and sibilants
in syllable initial (sa, si/e, su; zi, zu; ?u; but ambiguous za — ça, zum = sum, zìi =
sar = sar, sib = sib, etc.) ; some stops in syllable initial.
There are thus grave uncertainties, or rather a large scope for subjective interpre-
tation of ambiguities ; not to mention in this connection the fact that it is customary
to blame the ancient scribe for any dissonant form that shocks the sense of symmetry
of the grammarian and contradicts the norms he has established. Therefore, every
new look at the structure of Akkadian ought to start with a reconsideration of the ac-
cepted system of transliteration. Nevertheless, this study must forego a re-investi-
gation of the writing system, in spite of the fact that when the foregoing has been said,
such an investigation must seem as necessary for a grammatical study as phonetics
before establishing the phonemic system of a language. Work with the material has
shown that the errors resulting from the arbitrariness of the system of transcription
correct themselves; in case of doubt, a minimal range of distinctiveness has been pre-
ferred over the maximal distinctiveness current in the field.

For the phonemes here cited in phonemic notation, see 3.1.

3. PHONOLOGY

3.0. BASIC ASSUMPTIONS

As mentioned in 2.2.4, the syllabic values by which Akkadian text segments are repre-
sented in the transliteration are composed of phonemes whose general classification
was established by Semitic etymology, although, of course, their phonetic identities
cannot be recaptured.
Further refinements among members of a class with a similar component, as, e.g.,
dentals, sibilants, high vowels, depend on the considerations applied in eliciting them
from the contrasts supplied by the writing system. As stated in 2.4, I follow the
principle of preferring a minimal range of distinctiveness. This principle contrasts
with current practice in Akkadian phonology, which often moves on the allophonic
level. This is evidenced not by the absence of the term "phoneme", because the terms
used ("sounds", "Laute", "Sprachlaute"), following the accepted usage of the neo-
grammarians, are to be understood as standing approximately for "phoneme" ; rather,
the allophonic level of work on Akkadian is apparent from the increasing number of
studies in recent years which "discover" another and yet another "sound" in Akka-
dian. These may be "intermediate" vowels such as ö, ü, o, ä (i.e., ae) or consonants
such as ?, g (the voiced counterparts of the voiceless emphatics ç, q), interdental spi-
rants, laryngeals, and other consonants usually regarded as part of the Proto-Semitic
consonantal stock.
The reasons for this practice lie in two tacit assumptions. One regards the writing
system, the other general Semitics. With regard to Semitics, it is assumed that since
Akkadian is the oldest recorded Semitic language, it must contain, or have once con-
tained, all sounds that comparative linguistics has posited for Proto-Semitic. An ex-
ception is made for the common Semitic laryngeals, it being accepted that the number
of these has been reduced due to the influence of the substratum language, in this case,
Sumerian.1 The hitherto accepted theory that Proto-Semitic interdentals were equally
1
" . . . im Akkadischen [sind] infolge der Substratwirkung des Sumerischen die allgemeinsemitischen
Laryngallaute [h], ßi], ['] und [g] geschwunden, einer Substratwirkung, der auch noch das Aramäische
auf babylonischem Boden unterlegen ist, der sich erst das Arabische, das als die Sprache einer Herren-
schicht nach Babylonien gekommen ist, hat entziehen können" (Falkenstein, Genova, N.S., VIII,
303f.).—"Das Akkadische macht abgesehen von dem schon erwähnten Laryngalschwund und der Ver-
schiebung der dentalen Spiranten zu Zischlauten den Eindruck einer verhältnismässig 'konservativen'
semitischen Sprache" (ibid., p. 305).
34 PHONOLOGY

"lost" in Akkadian is nowadays being modified by the positing of these sounds for at
least some dialects of Akkadian.2
The other assumption, with regard to the writing system, is, paradoxically, that,
while the system of writing is permitted to operate with homophonous signs, the gram-
marians seem to exclude the possibility of homophonous morphemes. Where such
seemed to emerge through the customary system of transliteration, it was found ex-
pedient to substitute for one of the values a value which differs only in one phonetically
similar consonant. This phonetically similar consonant, which is then usually consid-
ered a different phoneme, serves then as the minimal difference between the two homo-
phonous - or "phonetically similar" - morphs, and the road is open to prove the ex-
istence of the particular phoneme by the "minimal pair" thus generated. Another
device substitutes /e/ for /i/ in the polyvalent Ce/i syllable to arrive at two morphs.
On the other hand, wherever there appears a spelling that would seem to indicate a
phonetic difference (b for p, e.g.) the transcription of the morpheme is not changed,
but instead the unusual sign, e.g., BA is assigned a value homophonous with that of
the usual one, PA (i.e., ρά for ba), so that the number of homophonous signs keeps
increasing as new texts from yet little known regions and dialects are being published.
In extreme cases, where the phonetic similarity seems to be insufficient, the spelling is
attributed to an error of the scribe. Instead of citing examples, I can simply refer to
the sign lists (the standard, and comparatively sober, sign list is von Soden, 1948), to
illustrate this overgrowth of homophonous signs.
The above sketched circular methods are also used either to describe, or to obscure,
phonetic change, as the case may be. This question will be further dealt with in the
section of morphophonemic alternations (Chapter 6).

3.1. P H O N E M E S

The literary (also the OB) dialect of Akkadian has the following phonemes: the stops
/b, p, d, t, t, g, k, q/, the spirants /z, s, s, S, 1)/, the sonants /m, n, r, 1/, the vowels
/a, e, u, i/.
labial dental groove-alveolar velar palatal tongue-base

voiceless Ρ t s k s h
voiced b d ζ g
pharyngeal- t s q
ized*
nasal m η

2
Cf., e.g.: "Die neuen Beispiele lassen sich ohne die Annahme eines gesprochenen / in bestimmten
Fällen im späten Sumerischen wie im Akkadischen schwer erklären" (W. von Soden, AfO, 18, 120;
cf. id. BiOr, 19, 150f.).
PHONOLOGY 35

Vowels : i u
e a
For a distinctive feature analysis see 3.3.
* The phonemes transcribed t, q, and s are the so-called "emphatics", phonetically either velarized
or pharyngealized stops and spirants.

3.1.1. Prosodies

The prosodies include length, which is distinctive (phonemic), hereafter noted /:/ after
V(owel) or C(onsonant), i.e., /V:/ and /C:/, see 4.1.1, and stress, which is non-distinc-
tive, see 3.2.

3.1.2. Configurative feature

On the basis of writing habits, a configurative feature similar to plus juncture e.g.,
in English an aim vs. a name can be established. This feature is expressed in the
writing either by a sequence of syllabic signs VC VC (see [Link]), or by the so-called
aleph-sign ?V (glottal stop). The plus juncture, similarly as in English, demarcates
the word boundary but sometimes, very rarely, is incorporated in some lexical items
as an internal juncture, cf., e.g., English /pléy+tòw/ for written <Plato>. The differ-
ence in comparison to English is that internally juncture occurs only before a vowel
not preceded by another one, e.g. /sum+ud/, /is+al/, etc., see Reiner, 1964, 169.

3.1.3. Phones without phonemic status

At the OB period the language had already lost two phonemes that are to be assumed
for previous periods : /s/ and /w/. The glottal stop [?], formerly possibly also a pho-
neme, has merged in SB with juncture in such forms as /§um+ud/, i.e., <§um?ud>,
/i§+al/, i.e., <iS?al>; in other positions, the same way as the glides [w] and [j], it is an
allophone of the zero phoneme. That [w, j, ?] are non-phonemic and omissible is also
apparent from the fact that the signs used for their writing <wV>, <jV>, <?V> are
either altogether omissible or occur in free variation with each other.

[Link]. I s / .
/s/, postulated for Proto-Semitic, can be reconstructed for Old Akkadian from writ-
ing habits of the Old Akkadian period surviving into OB, and from different morpho-
phonemic alternations in the immediate environment of etymological /á/ (see 6.1.3 c).

[Link]. [w].
Before the OB period, Akkadian dialects (Old Akkadian, Old Assyrian) had a pho-
36 PHONOLOGY

neme /w/ with a limited distribution : initial, intervocalic, and post-consonantal. The
phonemic status of [w] in the earlier periods may be seen from the existence of such
minimal pairs as /warka:tim/ 'future (fem. pl.)' and /arka:tim/ 'long (fem. pl.)',
/wir:u:/ 'intestines' and /ir:u:/ 'colocynth'; contrastive writing of such pairs is still
often preserved by OB writing practice. The fact that the phonemic status of [w] was
lost in OB already is evident however from such writings as <warka:tim> for <arka:
t i m ) 'long (fem. pl.)', <warhi§> for <arhi§> 'quickly', (waäpuram) for ( a s p u r a m )
Ί wrote', as graphic hypercorrections.
These examples also illustrate that initial [w] which contrasts with all other con-
sonants as well as with 0 (zero) before the O B period, does not contrast with zero in
the OB period any more. Graphic replacement of the initial <wV> sign by a <bV> sign
occurs in older periods and sporadically in some OB dialects; the phonetic inter-
pretation of this equivalence is of course ambiguous.
In intervocalic position, the earlier <wV> sign has as graphic replacement (in OB
and later) either nothing (i.e., zero graph) or an <mV> sign.
In postconsonantal position, the <wV> sign has as graphic replacement an <mV>
sign.
In preconsonantal position, graphic <w> does not occur because the writing system
does not include a Vw sign. In certain dialects of OB and in OA one finds however
the sign wV alternating with a sign Vm or Vb, in words that are stabilized either as
personal or other proper names. This may indicate an older preconsonantal [w],
which may or may not have formed a diphthong with the preceding vowel ; see Gelb,
1961b, 194ff.
The non-occurrence of preconsonantal [w] may be deduced also from the fact that
foreign words, which have a wC cluster, appear in Akkadian transcriptions with a Cw
cluster, e.g., Hurrian ewri appears in the Akkadian texts from Hurrian-speaking Nuzi
as erwi.
On the basis of the above enumerated graphic replacements, the sign <wV> (and
phone [w]) could be (disregarding the graphic replacements by <bV> or <Vb>) re-
garded as an allograph (and historical allophone) of <mV>, or, initially and inter-
vocalically, of <V>. This corresponds to the assumed historical development of
intervocalic and postconsonantal /w/ > /m/, and initial /w/ > 0 ; later use of graphic
<w> in these positions being historical spelling.
Such an analysis is not affected by the fact that phonetic [w] occurs as a speech-
sound throughout the life of the language. Occurrence of [w] as a glide between vow-
els, especially after /u/, is treated in 4.2.1 if. Apart from the graphic notation of such
a glide in OA (a dialect which chronologically precedes OB, but follows a different
development), as in kuatim ~ kuwatim, ituar ~ ituwar, which show the graphic alter-
nation with and without the <wV> sign, it can be seen that the graphic sequence <u-a>
contains the glide [w] from the writings <ú-a-ar> ~ <ú-wa-ar> in Hittite for the ending
{-uwar}, and the like.
Occurrence of [w] as an allophone of /m/ in intervocalic position may also be ac-
PHONOLOGY 37
counted for by assuming a spirantization of all intervocalic [m]-s; the single sign avail-
able for writing wV (the sign PI) having become restricted for the segment <pi> (and
<pe>), except in copies of older texts or in archaizing writings, its place in the system
could have been taken by mV-signs which in intervocalic position, after the spiranti-
zation of intervocalic [m] had taken place, may well have acquired the values wV, to
render the sound-sequence [wV] < [mV] in intervocalic position. However, since mV
signs replace the wV sign in postconsonantal position too (although the list of forms
in which these occur is extremely limited), and since intervocalic long m [m:] is
sometimes written with the sign sequence Vm-mV, the question of the phonetic reali-
zation of /m/ is best left open. We may state only that the writing ceases to make dis-
tinctions between graphic wV and mV after the OB period.
The diachronic assumption of initial /w/ > 0 and intervocalic or post-consonantal
/w/ > /m/ can be expressed in graphemic terms such as initially the <wV> sign ceased
to be written, and in other positions was replaced by M-signs (i.e., signs whose trans-
literation contains the letter <m>). That some of the M-signs were interpreted as
phonetic [w] in the LB period is shown by their rendering in transliterations with
Greek letters by upsilon (cf. Ungnad, O LZ, 1926, 426 η. 1).
For a description of divergent dialects, the statements on the occurrence and alter-
nations of <w> may easily be modified. For the Assyrian (post-OA) dialects, e.g., it
can be stated that a <wV> sign is replaced by a <bV(C)> sign in intervocalic position
(e.g., OB a-wa-tam ~ MA, NA a-ba-ta; OB la-wi ~ MA, NA la-bi)·, for MA a-P-i-lu,
a-P-i-lu-ut-tu < OB /awi:lu/ and /awi:lu:tu/, see [Link]. In other positions, sufficient
evidence is not available; I have noted Ν A lu-Sal-bu-ku-nu (Wiseman, Treaties, 621),
and the writing tal-bi-tu, which may represent OB *tal-wi-tu. In initial position, the
<wV> sign is replaced by a V sign; this general statement also includes, though
does not specify, the fact that the V sign <u> replaces the signs for which a trans-
literation <wa> and <a> is given in other dialects, i.e., the historical statement that
OA wet > MA u, OA wV (where V is not a) > MA V, see [Link].

[Link]. [j].
The glide [j] mostly appears after /i / and is written with the sign I + A = IA (with
values [ja], [ji], [ju]).

[Link]. Distribution of glides.

The two just discussed glides, [w] and [j], are usually regarded as semi-vowels. How-
ever, in Akkadian they do not enter into diphthongs but rather serve to indicate syl-
lable boundary between two (identical or dissimilar) vowels. The signs used to write
these glides and the glottal stop[?]seem to be predictable; the <wV> sign occurring
after /u/, the <jV> sign (transliterated as ia) after /i/, and the <?V> sign after /a/;
in the phonological analysis these glides are best grouped as allophones of the zero
phoneme, or of morphophonemic length {:}. For examples see 4.2.1ÍT. and [Link].
38 PHONOLOGY

Note that the signs wV and IA belong to a sub-system with three members ; I have
discussed their interpretation in another context, see Reiner, 1964.

3.2. STRESS

Stress is not phonemic; this is indicated by the following considerations: While

length is phonemic, and serves to differentiate morphemes (words), there are no cases
where stress alone would differentiate two morphemes (words).
In all grammars of Akkadian, stress is assumed to fall on the long syllable (i.e., the
syllable which contains a vowel plus length, a vowel plus consonant plus length, or a
vowel plus two consonants); this coincidence of stress and long syllable may be cha-
racteristic of the native language of many Semitists and Assyriologists, or assumed
for Latin and Greek as taught in European schools, but has never been shown to be
characteristic of Akkadian. This stress pattern is assumed for Akkadian partly on the
basis of the place of stress in Arabic (see e.g., Moscati, 1960, §§ 166f.), but here the
same objections are valid, see Fleisch, 1956, p. 139 n. 8, and Birkeland, 1954, pp. 9ff.,
and passim. Even if it were true that stress and long syllable coincide, this would only
mean that stress is predictable in terms of syllable length, i.e., it is not distinctive but
culminative, in that it signals the word peak. If, however, stress was to be shown to be
fixed on the first syllable of the word, it would again mean only that it is not distinctive
but demarcative, in that it demarcates the word.
I cannot make the general assertion that primary - non-phonemic - stress falls on
the first syllable of the word, because at the present this can be shown with a great
probability for the Assyrian dialect only. The following considerations may, however,
be mentioned: in a language where the stress is at the end of the word, there is usually
a profusion of proclitics (e.g., Hebrew); this is not the case in Akkadian. In a lan-
guage where the stress is not on the first syllable, loss of the initial vowel often occurs
in the course of language history; this is not the case in Akkadian (with the excep-
tion of some words of four or more syllables, usually loanwords, which exist in a
shape with and without an initial /a/, mostly in the distribution that the Babylonian
dialect has the form with /a/, the Assyrian without /a/, such as (a)gadibbu, (a)lahhinu,
(a)gurratu, (a)maluktu, etc. ; for words of less than four syllables, there is one instance
du-ú-du for /atu:du/, a-ka-ju-ú for /kaiu/ and a few occurrences of a-za-pi, a-zap-pi for
/zap:u/, /zap:i/). It is generally observed that short vowels often drop in unstressed
open syllables; in Akkadian, such loss of short vowels is frequent, and the syllable
that is preserved, the one that precedes the lost unstressed vowel, usually is the first
syllable of the word, hence, in these words at least, it is the initial syllable that was
stressed. By-forms such as litmudujlitamdu, damiqtu/damqatu (see [Link]) are best
explained by assuming stress on the first syllable. On the other hand, the reconstruc-
tion damiqtu from *dámiqátu usually accepted has to assume, (a) shift of stress from
*dàmiqâtu to *datniqatu, then (b) loss of the unstressed penultimate vowel [a], but
PHONOLOGY 39
still cannot account for the by-form damqatu, unless assuming that both stress-
patterns, the original *dámiqátu and the shift *damiqatu coexisted; the assumption of
stress on the first syllable: *dâmiqàtu, explains both forms damiqtu and damqatu.
According to the same theory (put forth by Poebel), stressed and unstressed vowels
alternate in a word; so while damiqtu is said to result form *damiqatùm, the plural
/damqu:tum/ (damqütum) is said to result from *dámiqútum. Only in NB does stress
tend toward the penultimate, under Aramaic influence (Poebel, 1939, 60).

3.3. DISTINCTIVE FEATURE ANALYSIS

The following matrix of distinctive features and the branching diagram of the phone-
mes is based on the distinctive features established by Roman Jakobson, see, e.g.,
Jakobson and Halle, "Phonology and Phonetics", in Roman Jakobson, Selected
Writings, I: Phonological Studies ('s-Gravenhage, 1962), pp. 464-504, esp. pp. 484ff.

ρ b t d t k g q s ζ s S h m η r l a e u i
obstruent/sonorant + + + + + + + + + + + + H
consonantal/non-consonantal + + + + —
vocalic/non-vocalic l· +
continuant/interrupted _ _ _ _ _ —
1 —
—I —
—
f — —f — +
backward-flanged/forward-flanged + + + + H l· + H h +
peripheral/medial + + h + — H 1
flat/non-flat 1 1 h
tense/lax H 1 H H

The inventory of phonemes can be most economically divided as follows :

All phonemes are divided into obstruents (all consonants except the nasals and the
liquids) and sonorants (nasals, liquids, and vowels).
The obstruents are divided into continuants (spirants) and interrupted (stops). The
stops are divided into backward-flanged (or front), i.e., labials and dentals, and
forward-flanged (or back), i.e., velars. The backward-flanged (front) stops are divided
into the peripherals (labials) and the mediáis (dentals); both the backward-flanged
mediáis (dentals) and the forward-flanged (velars) are first divided into flat (pharyn-
gealized, "emphatic") q/, and non-flat, and the non-flat ones /t, d, k, g/, into tense
(tenuis) /t, k/ and lax (media) /d, g/; similarly the backward-flanged peripherals
(labials) are divided into tense /p/ and lax /b/.
The continuants are divided into backward-flanged (dental spirants, or sibilants)
and forward-flanged (palatal and velar spirants). The backward-flanged (sibilants)
are divided into flat (pharyngealized, "emphatic") /§/ and non-flat /s, z/; the non-flat
ones are divided into tense /s/ and lax /z/. The forward-flanged continuants are
divided into the peripheral (velar) /h/ and the medial (palatal) /§/.
The sonorants are divided into consonantals (nasals and liquids) and non-conso-
nantals (vowels). The consonantal sonorants are divided, by the feature vocalic VÍ.
non-vocalic, into non-vocalic (nasals) and vocalic (liquids). The nasals are divided
40 PHONOLOGY
PHONOLOGY 41
into the peripheral (or grave) nasal /m/ and the medial (or acute) nasal /n/. The
liquids are divided into the interrupted liquid /r/ and the continuant liquid /l/.
The vowels are divided into forward-flanged (low vowels) /a, e/, and backward-
flanged (high vowels) /u, i/. Both groups are divided, by the feature peripheral vs.
medial, into the forward-flanged peripheral /a/, the forward-flanged medial /e/, the
backward-flanged peripheral /u/, and the backward-flanged medial /i/.
The "incompatibility rules", i.e. the rules on the occurrence or non-occurrence of
certain consonant clusters (4.3.1), can be rephrased on the basis of the distinctive
feature analysis as follows :
(a) clusters of identical members occur;
(b) the distinctive feature flat (pharyngealized) occurs in a cluster only once, i.e.,
*ts, *st, *tq, *qt, *qs, *sq do not occur (the so-called Geers law) ;
(c) phonemes that are both plus-consonantal and plus-vocalic (liquids) do not
cluster, i.e., *rl, *lr do not occur; plus-vocalic sonorants (liquids) may be followed
by minus-vocalic sonorants (nasals), i.e., rm, lm, rn, In occur; in the reverse order
only /m/ may be followed by a liquid, i.e., mr, ml occur, but not *nr, *nl;
(d) in clusters in which one member differs only in the distinctive feature continuant
vs. interrupted, the member which contains the feature continuant precedes the
member which contains the feature interrupted, i.e., st, sd, st, zt, zd, z\, çt, sd, §t, §d,
St, Ijk occur (hg and hq are not attested), but not *ts, *tz, *ts, *ds, *dz, *ds, *ts, *tz,
*t§ (for the last see also above (b), *tS, *d§, *t§, *k& *gh, *qlj;
(e) consonants (obstruents and plus-consonantals) in which the only opposition is
in the distinctive feature tense vs. lax or flat (pharyngealized) vs. non-flat do not
cluster, i.e., *pb, *bp, *td, *dt, *kg, *gk, *sz, *zs, and *tt, *tt, *dt, *td, *kq, *qk, *gq,
*qg, *ss, *ss, *zç, *sz do not occur;
(f) of the non-peripheral continuants, the forward-flanged (palatals) precede the
backward-flanged (dentals), but not the reverse, i.e., *ss, *z§, *çs do not occur;
(g) peripheral interrupted obstruents (p, b) do not cluster with the peripheral
sonorant (m), i.e., *pm, *mp, *bm, *mb do not occur.
4. PHONOTACTICS

4.0. QUESTIONS TO BE TREATED

Besides giving an inventory of the phonemes, a grammar has to treat the various ways
in which they combine, e.g., which phonemes occur in clusters with which others and
in what environment, which can be long as well as short - if long phonemes have not
been listed as separate phonemes in the phoneme chart - etc. Some of these points are
treated in the traditional grammars under such headings as "The syllable", "Stress",
etc. In this presentation these and similar questions, all relating to the order and ar-
rangement - taxis - of phonemes, are subsumed in this chapter under the general title
"Phonotactics".

4.1. VOCALIC AND CONSONANTAL LENGTH

At the start, a justification of the omission of long vowels from the phoneme inven-
tory has to be given. Up to now, it was customary to list, beside the short vowels
/a, i, u, e/ the long counterpart of these same vowels: /â, ï, ü, ë/. The macron over the
vowel is the customary assyriological notation for a long vowel, whether considered
originally long or as having resulted from the contraction of a short vowel and a
consonant that was lost during the historical development, while the circumflex (in
/â, î, û, ê/) is the notation for a long vowel having resulted from a contraction of two
vowels (two syllables) ; some Assyriologists use the tilde to mark long vowels that were
shortened through processes in historical development. This threefold (or the more
usual twofold) notation of the length of the long vowel has never been said to imply
that Akkadian had more than two degrees of vowel length : short and long. The writ-
ing itself rarely indicates length by "inserting a vowel sign after the sign for open syl-
lable (formula CV+V), usually called piene writing" (Aro, 1953, 3).
On the other hand, it is customary to consider consonants usually written double
(i.e., VQ-QV), as well as consonants which, although never written double [?, w, j],
occur in the same grammatical forms where usually double consonants are written,
as phonologically double consonants. It has recently been suggested that there may
have been a phonetic difference between long consonants and double consonants
(Goetze, 1958, 149).
PHONOTACTICS 43

Since it is preferable, wherever possible, to assume a symmetrical phonological

structure, i.e., either both long vowels and long consonants, or both double (geminate)
vowels and double consonants (to be transcribed /aa, bb/, etc., as it has been done for
Classical Arabic; Cantineau, 1946, 1281Γ.), I propose to use such a unified system. In
this system, the prosodeme length is noted by the symbol /:/. If the symbol /:/ follows
a vowel, it indicates that the vowel is long; if it follows a consonant, it indicates that
the consonant is long. Accordingly, long consonants will henceforth be written /C:/
and not /CC/, the latter notation being reserved for a cluster of two identical con-
sonants, for which see below 4.1.2ff. and [Link].3. E.g., /br/ is symbolized /CC/ or
/ Q Q / , /bb/ is symbolized /CC/ or /C 1 C 1 /, but double, i.e. long b, is symbolized /C:/.
Of course, by definition, the two notations (/CC/ and /C:/) are interchangeable and
in fact do so interchange in the Akkadian writing system. Note also the statement:
"Die auf Grund des Schriftbildes als Konsonantenverdoppelung bezeichnete Erschei-
nung stellt in Wirklichkeit meist eine betonte Längung der Konsonanten d a r . . . "
{GAG, § 20a).

4.1.1. Cluster of identical consonants

The distinction between long consonants and a cluster of two identical consonants is
based on two criteria: (1) the writing, (2) morphophonemic alternation.

[Link]. Evidence of the writing.

Although the writing does not always indicate long consonants - older periods, such
as the OA dialect, never do, OB only rarely - in certain words there is always a double
writing of the consonants, with the syllabograms VC-CV. For example, always du-ub-
ba, never du-ba.

[Link]. Evidence of morphophonemic alternations.

A cluster of /C:/ in final position or preceding another consonant behaves differently
from a cluster of CjCj, for example \/dan: = /da:n/, \/dan:-t = /dan:at/, but ^/lazz
= /laziz/, \/lazz-t = /lazizt-/. For further examples see 7.1.4, 7.1.5, and 7.2.

4.1.2. Clusters with length

If we examine the morphs which contain a long vowel /V:/ or a long consonant /C:/,
it becomes apparent that the distribution of length, i.e., of the symbol /:/, is that of a
consonant. In other words, /:/ behaves like a consonant, and hence forms a cluster
with a preceding or following consonant. This will be exemplified in the following.

[Link]. Occurring clusters.

On the evidence of the writing system as described above (2.4), it can be stated that
44 PHONOTACTICS

no consonant cluster occurs in either initial or final position, and that no cluster of
more than two consonants occurs intervocalically. Or, to state it differently, clusters
of not more than two consonants occur, and these always intervocalically.

[Link]. Members of clusters.

With reference to clusters of a consonant and length, it can be stated that the
clusters /QC 2 / (or / Q Q / , short /CC/), /:C/, and /C:/ all occur, but /CCC/, /:C:/,
/C:C/, /:CC/, /CC:/, do not occur.

[Link]. Inventory of clusters.

List of occurring intervocalic clusters into which /C/ and /:/ enter:

short vowel : VCV long vowel: V:CV

VC:V or VCCV V:CV:
VC:V: or VCCV:
VCV:
These can be summarized as
(i) pattern with single consonant: V(:)CV(:),

(ii) pattern with consonant cluster : VC

both formulas may
ay be united as :
Π-
e
(iii) V
C
I}
(0 (C)
The combinations of the same elements into sequences *V:C:V or *V:CCV, and
*V:C:V: or *V:CCV: do not occur, these being clusters of three consonants. This
statement affects many hitherto usual transcriptions (entries in dictionaries) that will
have to be revised. For example, there is no item *bärtu */ba:rtu/ since this repre-
sents a cluster /:CC/, only an item bartu /bartu/; no *bïltu */bi:ltu/, for the same
reason, only /bi:iltu/ (see below, [Link].8).
The sequences extracted above, having been based on the number of mathematically
possible combinations of these elements, correspond to the groups of noun-forms lis-
ted in GAG, § 55 as "A. Formen mit nur kurzen Vokalen zwischen den Konsonanten"
(type CVCC, CVCV(C)), "B. Formen mit einem langen und einem kurzen Vokal"
(type CVCV:C, CV:CVC(C)), and "C. Formen mit Verdoppelung eines Radikals''
(type CVCCVC(C), CVCCV:C, CVCVCC-, i.e., /CVC:VC(C), CVC:V:C, CVCVC:-/),
the forms CVCV:CC cited sub Β, and CVCCV:CC cited sub C, i.e. /*CVCV:CC/
and /*CVC:V:CC/, must however be excluded. This is in fact done in the mentioned
§ 55 sub 9b, 10b, l i b , 12b, 14b, 15b, 16b, 17b, 23b, when the type is cited as e.g. (sub
9b) paräst, but the examples are cited as anantum, Sapartum, etc., with the exception
of (sub l i b III) rûmtu, (sub 23b) tajjärtum; note also that sub 15b the type is cited as
puräst, but the examples are usätum /usa:tum/ and rubätum /ruba:tum/.
PHONOTACTICS 45
[Link]. Advantages of the notation /:/.
The statement on the occurrences of combinations of vowels and consonants per-
mits clearer statements on the morphs and their combinations in the morphology.
The extraction of / :/ as a separate symbol also greatly facilitates the description of
derivational morphs both in the verb inflection and in the noun-formation (see 5.0).

[Link]. Free variation of /:C/ and /C:/.

The phonotactics of / :/ include the characteristic and peculiar feature that the place
of length before or after a consonant is non-distinctive. This can be stated as there is
free variation between /:C/ and /C:/. Such free variation can be profusely illustrated
with alternant spelling practices. The only situation where free variation is not opera-
tive is the case when the place of length has a functional load and serves to differen-
tiate two morphs ; this obtains in the inflection of the weak verb, where the place of
length differentiates preterit and present, see [Link] and [Link].
The main advantage of considering the alternation :C ~ C: as phono tactic free
variation can be gained in morphology where divergent forms may be grouped together
when the only phonological difference between them is the presence of a long vowel in
one, of a double consonant in the other. For example, not all nouns with "Verdoppe-
lung eines Radikals" (see [Link]) and those with "einem langen und einem kurzen Vo-
kal" represent different derivational types (see the remarks in GAG, § 55 passim), but
they rather represent the free variation of the infix /:/, i.e., the length which is part of
the vocalic morph (see 5.0) may occur before the root consonant (i.e., after the vowel
of the vocalic morph) or after the root consonant. For such pairs see e.g., from among
the list given in GAG, § 55, the types päris and parris, parass- and paras, paruss- and
partis, parussat and parûsat, etc., which may be regrouped as {pa:ris ~ partis},
{parus: ~ parurs}, etc., and written in some abbreviated way such as {paris}, {parus},
etc. (see below).
A further advantage of this statement of free variation is that it eliminates from the
"phonetics" the treatment of a phenomenon usually called "compensatory lengthen-
ing", occurring for example in such pairs as ku$$u ~ kü$u 'cold' and described in
terms of a process as the lengthening of the vowel to compensate for the reduction of
two consonants, or doubling the consonant to compensate for the shortening of
the vowel. Such pairs, if rewritten as /ku?:u/ and /ku:$u/ can again be grouped as one
morpheme {kusu} on the basis of the free variation of length. This not only accounts
for the "compensatory lengthening", of course, but also for the reverse process, the
"doubling" of a consonant to replace the length of the vowel and the phonologically
equivalent phenomenon described as the shortening of the vowel before a double
consonant, that which occurs in such pairs as Mrruttu ~ sarrütu and may be re-
written as {sar:ut:u ~ sar:u:tu} or {sanutu}. Such an analysis of the pair not only
does not obscure what is often taken as dialect variation (Assyrian -uttu versus Baby-
lonian -ûtu) but even permits statement of dialect variation - observable in a great
number of similar cases - as the preference of one dialect for a particular place of the
46 PHONOTACTICS

length, in this instance the preference of the Assyrian dialect for the variant in which
/:/ occurs after, and not before the consonant.
A convenient notation for a cluster of consonant and length would be either some
such formula as VC, to be interpreted as /V:C/ or /VC:/, or one could choose one of
the latter, e.g., the notation V:C, and include a transformation rule "move /:/ after
C " for the dialects in which the pattern is VC: and for the paradigmatic forms in which
length appears after C.
As several phonotactic statements, the present one on free variation of length not
only permits us to group morphs into morphemes (see e.g. 7.2, [Link].1), but also to
consider together lexical items separately cited in the dictionaries as variants, e.g.,
zäbilu and zabbilu, däbibu and dabbibu.

4.2. SEQUENCE OF TWO VOWELS

Akkadian has no diphthongs; moreover, as is the case in most Semitic languages,

there occurs no cluster of two vowels. Since we now note a long vowel by /V:/ and not
by /V X Vj/, it needs to be shown only that the combinations /V 1 V 1 / (i.e., two identical
vowels in contact) and /VjV 2 / (i.e., two dissimilar vowels in contact) do not occur.
This can easily be shown by variant spellings for CV1-VeC types of syllables, for which
the writing alternates between ( Q V ^ V ^ C ) , (C)V r ?V-V 2 (C), (C)V1-wV2-V2(C), and
(C)V r jV 2 -V 2 (C). Since the last three are usually transcribed as C V ^ V ^ , CV^wVaC,
and CVjjVaC, i.e., as disyllabic, the group written (CyVj-V^C) also has to be interpre-
ted as disyllabic, and not as a diphthong. Only a border case of the above spellings is
that in which V 1 = V 2 , i.e., two identical vowels follow each other. In the cuneiform
syllabic writing system they can be taken either as forming only one syllable with the
preceding and following consonants - for example, the syllable /mak/ can be written
only as <ma-ak>, since no single syllabic sign *<mak> exists in the system - or as form-
ing two syllables. The latter case, two syllables, has to be assumed whenever the writ-
ting CV 1 -V 1 C alternates with one of the above three patterns that insert a CV sign
between the two syllables, in which CV sign the consonant is [?] (aleph or glottal stop),
[w], or [j]. This fact, that has not been sufficiently noted up to now, will be of basic
importance in the description of the morphology. To use nonsense syllables as an
example, the writing <ba-ab>, if at any time it interchanges with <ba-?a-ab> or with
(ba-ia-ab) 1 has to be interpreted as phonologically two syllables, /baab/. (No
example with wV syllable is here given, for reasons see below.)
The same type of syllabic writing also often interchanges with <ba-a-ab>. This may
be transcribed (i.e., phonemicized) accordingly as two syllables, namely /ba:ab/.
This transcription is in conformity with (a) the writing system, which sometimes -
though rarely - indicates length by the so-called "piene" writing (see above, 4.0) and
1
This syllable combination is not actually attested, the same patterning occurs however with e.g.
<ba-al>, cf. simply AHw. s.v. ba'älu.
PHONOTACTICS 47
(b) our interpretation of the behavior of length as if it were a consonant /:/, since in
the transcription /ba:ab/ the symbol /:/ is replaceable by a consonant and thus
obviously indicates a syllable boundary between the two identical vowels.

4.2.1. Boundary between two vowels

We may now summarize the foregoing in stating that the three so-called "weak" con-
sonants w, j, Ρ or, as they are also called, the semi-vowels w, j and the aleph (or
glottal stop) P, (see above [Link] and [Link]), and length /:/ are in complementary
distribution or in free variation in intervocalic position, and that there always occurs
one of these consonants between two vowels when they belong to two separate syl-
lables, whether the vowels are identical or not. The item to which we assign these con-
sonants may be called the morphophoneme {length} or "syllable boundary".

[Link]. Indication of boundary by writing conventions.

As indicated in [Link], within one word the sign sequence (C)VC-VC-VC does not
normally occur. Consequently, apart from the just discussed signs which serve to
indicate syllable boundary (the three CV-signs where C = w, j, or ?), divergence from
the customary syllabification pattern can also be used to indicate syllable boundary.
While the three CV-signs indicate syllable boundary between two vowels, a sign se-
quence VC-VC indicates either word boundary between the two signs, or, when it
interchanges with VC-CVC or VC-?-VC, syllable boundary between two consonants.
In the latter case, this syllabification pattern indicates that the first C ends a closed syl-
lable, i.e., is followed either by /:/ (length) or by [?] (glottal stop). Thus, a spelling
VC-VC stands for a spelling VC-PVC or for a spelling VC-: VC, since the system has no
?VC or :VC signs available (presumably because these sound-sequences do not occur
in initial position). As an example from SB texts, the spellings bur-ru-um-tú and bur-
um-tú may be cited, both of which are to be phonemicized as /bunumtu/. For such
writing practice in OAkk., see Gelb, 1961a, p. 42f., where the reverse practice, that of
a syllabification (CyVCx-QV to indicate a word division -VC V-, or a syllable division
(C)VC-?V, is documented for OAkk.

[Link]. Indication of boundary by a syllabic sign.

Although, as said above, some free variation occurs, at least in the writing, between
the allophones - or allographs - of/:/ between vowels, normally the following comple-
mentary distribution may be observed: <w> occurs after /u/, <j> occurs after /i/, and
/ :/ or <?> occurs after /a/ and /e/. Some examples to illustrate this distribution, from
actually attested spellings:

[Link].1. Between/u/and/a/:
e-pis-ú-a, e-pis-tu-u-a /e:pisu:a, e:pistu:a/
pa-nu-ú-a /panu:a/
48 PHONOTACTICS

ku-mu-u-a ~ ku-mu-ia /kumu:a/

[Link].2. Between/i/and/a/:
li-?-ab-bit ~ li-ia-ab-bit /li:ab:it/
i-?-dar ~ ?-a-dar ~ ia-ad-da-ru /i:ad:ar(u)/
A-hi-ia-qar /ahiraqar/ (compound)
a-bi-ia-bi-ia ( B a g h d . Mitt., II,
56ff. iii 27) /abi/+/abi:a/
[Link].3. Between /a/ and /a/:
ba-a-lu ~ ba-?-a-lum ~ ba-?-lu /ba:alu(m)/
ma-a-du ~ ma-?-du /m[Link]u/
ta-ba-? ~ i-ba-?-a /taba:a, iba:a/
ta-sa-ah ~ a-ça-ia-ah /tasa:ah, asa:ah/ (see [Link].1)
[Link].4. Between /a/ and /i/ (or /e/):
da-i-ku ~ da-a-a-i-ku ~ da-?-
i-ku /da:iku/
da-is ~ da-a-is ~ da-i-is /da:is/
a-ha-is ~ a-ha-IA-si /aha:is/
ba-?-it ~ ba-i-it ~ ba-it /ba:it/
ba-?-il ~ ba-i-il ~ ba-il /ba:il/
[Link].5. Between /a/ and /u/:
ba-a-u ~ ba-?-u /ba:u/
ba-?-ú-la-a-tu ~ ba-?-u-la-ti ~
ba-ú-la-a-ti /ba:ula:ti/
u-ba-?-u ~ u-ba-wu4/MU-ú /uba:u/
[Link].6. Between /u/ and /i/ (or /e/>:
mu-ir-rum ~ mu-?-i-ru ~ mu-u-e-ru
(Hitt. orthography) /mu:ir:u/
da-bu-ú-es /dabu:es/
[Link].7. Between /u/ and /u/:
bu-ú-ra ~ bu-?-ra /bu:ura/
zu-uz-ti ~ zu-?-uz-ti /zu:uzti/
ru-ú-a ~ ru-?ú-a /ru:u:a/
du-um ~ du-?-um ~ du-?-ú-m(u) /du:um/
[Link].8. Between /i/ and /i/:
di-i-? ~ di-?-i [Link]
bi-il-ti ~ bi-?-il-ti /bi:ilti/
[Link].9. Between /i/ (or /e/) and /u/:
le-?-ú ~ le-e-um /le:u/
ti-ú-tú ~ ti-ú-?-ti ~ ti-?-ú-tu /ti:u:tu/
lu-sab-bi-?-[u] ~ lu-sab-bi-ju-u /lusab:i:u/ (ABL 157:7 and
115:13)
PHONOTACTICS 49
[Link].10. Between /e/ and /i/: (Ass. only)
a-he-is ~ a-he-?-is (in is-sa-he-
?-is) ~ a-he-IA-si /ahe:is/
de-iq ~ de-e-iq ~ de-IA-qi /de:iq/
[Link].11. Between /e/ and /a/:
te-?-a-ti ~ te-? /te:a(ti)/
e-a ~ e-?-u ~ i-a-ú /e:a(:u)/
[Link].12. Between/a/and/e/:
No certain attestations.
[Link].13. Between/i/and/e/:
No certain attestations.
[Link]. Notation of syllable boundary.
Since we have established that the morphophoneme {:} or "syllable boundary"
is always present between two dissimilar vowels, there should be no need for indicating
this in our transcription (and especially not after the haphazard and inconsistent fash-
ion of Assyriologists who sometimes do and sometimes do not insert an ? between
contiguous vowels) ; a notation liabiit, e.g., would suffice to account for the morpho-
graphemic spellings li-ia-ab-bit, li-P-ab-bit, and li-a-ab-bit. To keep in conformity with
widespread practice in Assyriology, I would nevertheless prefer to use a symbol be-
tween the two vowels, and since the symbol ? is the one used by Assyriologists when-
ever they choose to indicate the syllable boundary, this could be substituted (liPab:it)
for the symbol /:/ adopted here for reasons of neatness.
Since there are many examples in which two different or identical vowels are separa-
ted by a consonant - w, j, Ρ or .· - and no examples where writings with CVC-signs
would prove that the two vowels were contracted into one, there is no reason to as-
sume such contraction for the historical periods of Akkadian, except where they are
actually attested as a historical development. Thus, many words up to now cited in
two variant forms, one "uncontracted" and one "contracted", have to be rewritten as
always containing the two vowels as two syllables. This affects, to cite from the recent
dictionaries (AHw. and CAD) such forms as baPälu (not *bâlu), ahePis (not *ahîs,
*ahês), biPiltu (not *bïltu), ziPiru (not *zîru), etc., and leads to a reconsideration of the
conjugation of the so-called "weak" verbs, for which see [Link].

4.3. CONSONANT CLUSTERS

According to statistically established rules (see Greenberg, 1950, 162-181), also called
"incompatibility" rules (Landsberger, 1935, 450-52), no "root", with another term,
consonantal morph (see 5.0 and 5.4.3), in Akkadian (and in the Semitic languages in
general) contains two homorganic consonants as either the first and second, or the
second and third consonants. This rule can be expressed in terms of phonotactics, in
stating what clusters do not occur, since both the first and second, and the second and
third consonants of a root come into contact position in some inflectional forms.
50 PHONOTACTICS

4.3.1. Non-occurring clusters

Instead of listing all occurring clusters, we list here the clusters that do not occur, to
conform to the procedure customary in describing the "incompatibility", although
this list goes beyond occurrences limited to "root-incompatibility". Non-occurring
clusters are:
(1) "reversible" clusters:
two dissimilar labials (*pb, *bp, *pm, *mp, *bm, *mb)
two dissimilar dentals (*td, *dt, *tt, *tt, *dt, *fd)
two dissimilar velars (*kg, *gk, *kq, *qk, *gq, *qg)
two dissimilar groove-alveolars (sibilants) : (*sz, *zs, *ss, *ss, *zs, *sz)
two dissimilar liquids (*lr, *rl)
two pharyngealized consonants (emphatics) *{s, *s{, *tq, *qt, *sq, *qs)
Referring to the chart of phonemes (3.1), it may be seen that the phonemes standing
in a vertical row and those standing in the third horizontal row, do not occur in
clusters with each other. For an alternate statement, see 3.3.
(2) "Non-reversible" clusters:
dental and groove-alveolar (sibilant) (*ts, *tz, *ts, *ds, *dz, *ds, *ts, *tz,
*ts) [for the last see also above sub (1)]
dental and palatal (§) (*tS, *dS, *tS)
groove-alveolar (sibilant) and palatal (s) (*ss, *zs, *ss)
η and liquid (for /n/ classed as a liquid in Semitic, see Cantineau, 1946, 101,
see also Greenberg, 1950, 162f.) (*nl, *nr)
moreover the clusters *S$, *kh ;
and note that the "reverse" of the above clusters do occur, such as groove-alveolar
and dental (sd, st, zt, çt [but see [Link].2]; not attested: *zd, *sd, *st, zt); palatal (s)
and dental (sd, st, i>t) ; liquid and /n/ (rn, In).

[Link]. Notes on non-occurrence of certain clusters.

A few limitations and explanatory notes must be added to the above list.
First, some of the clusters which the list gives as non-occurring, never occur; how-
ever, some other clusters which appear in the list of non-occurring clusters because
they are excluded from the dictionary - or rather the root lexicon - may occur in for-
eign words. Whether a word written with such a cluster actually represents a phonetic
realization of the cluster or is a graphic approximation of a foreign language pattern,
cannot be established.
Second, the cluster dental and palatal (s) does indeed not occur phonologically, but
it occurs morphophonemically. When such a cluster results from a morphemic pat-
tern it must be rewritten in either of two ways: /§:/ or /sD/ (i.e., sd, st, st). This state-
ment is necessary to account for the two forms es-Su (/es:u/) and qasdu, beside their
PHONOTACTICS 51

corresponding feminine forms edistu and qadistu. For, if we are going to extract the
masculine from the feminine form, we will arrive at V e dsu and V q ^ s u . Neither of
these occur, but one morph occurs as /es:u/ and the other as /qasdu/ (and not conver-
sely as *esdu, *qaS\u). The argument is not changed by the existence of the feminine
forms essetu (/es:etu/) and qassatu (/qas:atu/) beside edistu and qadistu, since by a sim-
ilar reasoning they would predict the masculine forms essu and *qassu, whereas the
actually occurring eSSu and qaSdu are not predictable. Eventual transformation rules
for generating these pairs will be determined by the dictionary.
Third, the cluster of the two labials /mb/ occurs only phonologically, as the dis-
similation of /b:/ (see [Link].5), i.e., it does not occur morphophonemically with /m/
and /b/ as successive consonants of a root.

[Link]. Morphological significance.

The advantage of stating the non-occurrence of certain clusters is that it relieves the
descriptive grammar of the necessity of including in the "phonetics" a great number
of assimilation, dissimilation, and metathesis rules established by historical reconstruc-
tion of the Proto-Semitic root and its resulting Akkadian morph.
Note that while in other Semitic languages the incompatibility rules only apply with-
in the morpheme (see Greenberg, 1950, 178f.), in Akkadian they apply within the
morpheme only if the affix is a grammatical morph, such as a personal (see 5.4.1), or
a morph which occurs with derived stems (see [Link]), but straddle the morpheme
boundary when the affix is derivational; for example, the derivation prefixes {ma ~
me} occur as /na ~ ne/ before a root that contains a labial; on the other hand, the
derivational prefix {mu} when it occurs as derivational morph of a derived verb stem,
always occurs in the shape /mu/.
Note that the "incompatibility" rules pertaining to emphatics are not operative in
NA (e.g. iqfibi occurs), which seems to indicate that the NA dialect did not have
three contrasting rows of stops, only two, such as voiced vs. non-voiced, or emphatic
vs. non-emphatic.

4.3.2. Distinctive feature analysis

If, besides the conventional phonological chart given in 3.1, we analyze the consonants
of Akkadian in terms of distinctive features, the statement on clusters can be made
more general, see 3.3 above.

4.4. RESOLVING OF CLUSTERS

It has been stated in [Link] that no initial or final consonant clusters occur, nor clus-
ters of more than two consonants intervocalically. Wherever the morphemic pattern
produces such clusters, they are resolved by means of an epenthetic vowel.
52 PHONOTACTICS

4.4.1. Two consonants

Initial CC > CVC
Final CC > CVC

[Link]. Consonant and length.

If one of the consonants is /:/, the phonotactic pattern is modified as follows: Ini-
tial */C:/ and */:C/ do not occur. Final /:C/ remains /:C/ (ex.: /iku:n/).
Before terminal juncture, in the Assyrian dialect, /:C/ > /:VC/ (ex. : /iku:an/, /zaku:
at/, /ibi:ad/), in the OB dialect, only when the vocalic morph is /i/, /i:C/ > /i:VC/ (ex. :
/iri:ab/, /rabi:at/), for further examples see [Link] and 7.2.
Final /C:/ > /C/ or, according to 7.2, /:C/ (ex.: \/kunuk:- > /kunuk/, \Alan: >
/darn/).

4.4.2. Three consonants

When in the inflection or derivation, a third consonant comes into contact with a
cluster of two consonants, the resulting cluster of three consonants is resolved by
means of an epenthetic vowel, as follows (since there is a morpheme boundary (see
5.1) between these consonants, it is here indicated by a +-sign):
C + C C > CVCC { + , # } , i.e., before either morpheme boundary ( + ) or terminal
juncture ( # ) ; (ex.: uS+pras > uSapras)
C C + C + > C V C C + (ex.: d a m q + t > damiqt-)
C C + C # > C C V C # (ex.: p a r s + t > parsat, before terminal juncture; see 7.2).

[Link]. Two consonants and length.

If one of the consonants is /:/, the phonotactic pattern is modified as follows:

Final C : + C # > C : V C # (dan:+t > dan:at); final C + : C # does not occur.

Final : C + C # > : C V C # fta:b+t > ta:bat); final : + C C # does not occur.
Intervocalic : C + C > CC (ta:b+tu = tabtu); intervocalic : + C C does not occur.
Intervocalic C C + : > CVC: (i-pr+:-as = ipar:as); intervocalic C + C : does not
occur.
Intervocalic C + : + C > C:VC ( d a m + : + q u > dam:aqu; d a q + : + q u > daq:aqu);
for further examples see [Link].
Note that initial cluster of two consonants and /:/ does not occur because initial
/C:/ and /:C/ do not occur, as stated above, [Link].

[Link]. Consonant and two lengths.

If two of the consonants are / :/, the phonotactic pattern is :
Intervocalic : C + : > C: ( i + : b + : + a l > ib:al);
Intervocalic : + : + C > :C ( i + k u : + : + n u > *iku:nu) or > C: (ikun:u), but the
former variant is not attested, see [Link].
PHONOTACTICS 53

[Link]. Positions of the epenthetic vowel.

The epenthetic vowel which resolves a cluster of three consonants does not always
appear between the first and second consonants of the cluster. In some instances whose
distribution has not been established as yet, but which seems to be dialectal, it appears
between the second and third consonants. In either case, the epenthetic vowel forms
the second syllable of the word.
One interpretation of this feature may be purely in terms of the writing system which
has no means to indicate clusters of three consonants ; the epenthetic vowel would be
then assumed to be purely graphic and phonologically a three-consonant cluster could
be recorded.2 Such an interpretation is made unlikely by the fixed place of this vowel
in most cases; one would expect fluctuations were it only due to writing.
The other interpretation, phonological, must be in terms of stress. Indeed, in a
word of the presumed syllabic structure CVCCCV- (with or without following syl-
lables), the fluctuation of a phonetic realization either as CVCVCCV- or as
CVCCVCV- has to be due to stress placed on the first syllable, with presumably secon-
dary stress on the third, and weakest stress on the second syllable; the second syllabic
vowel in this case would be a morphophonemic vowel occurring only to facilitate the
pronunciation of the cluster and realized probably as a schwa. In favor of the inter-
pretation of this epenthetic vowel as morphophonemic is also the fact that it is almost
invariably written as /a/ and does not have the quality expected if the writing were
morphographemic. The syllabic structure of such words would be, in my proposal,
CVCVCCV-, and as a consequence, we have to assume that the alternant also has the
stress pattern CVCCVCV-, that is, stress lies on the first syllable and not on the "long"
syllable.
Examples :
morphological or morphographemic: morphophonemic:
litmudu litamdu (OA)
litbusu litabsu (OA)
atlakam atalkam (OB)
sa libbi+ka+ni (/lib:+ka-ni/) sa libbakni (NA)
Du:r-Humid Du:ruhmid (OA)
sitnunu sitannu (OB)
Both forms morphographemic
regular: rare:
karaski /kars+ki/ karsaki (SB)
damiqtu /damq+tu/ damqatu (SB)
The dialect distribution indicated in parenthesis is, apart from the OA dialect, a rough
approximation only, since in later texts other dialect influences may have played a part
in a text predominantly in the standard language of the period.
* This seems to be actually the case in the rendering of foreign words in the cuneiform writing system.
5. MORPHOLOGY

5.0. MORPHEMES

Most of the words of the Akkadian lexicon can be analyzed as being a combination
of three - rarely of four - consonants, and of one or more vowels. According to more
recent grammatical analyses, two consonants alone can also appear in a word, but in
that case usually one of the vowels is long. As stated in 4.1.1, length - including vowel
length - behaves like a consonant and thus these words too can be viewed as a
combination of three consonants - the two apparent ones and length - and of one or
more vowels. These consonants, called radicals, form the root which determines the
lexical meaning of the word, while the vowel or vowels and their position in relation
to the radicals determine a typical modification of the meaning, such as the grammati-
cal category - adjective, noun, infinitive, passive participle, etc. - or some other dis-
tributional class, such as abstract, diminutive, and the like. The vocalic element which
imparts this typical modification to the root may be called, with Cantineau, 1950,
"scheme", or, with Harris, 1941, "pattern". In Assyriological textbooks it usually
goes under the name "noun-form".
Semitists and linguists have found it useful to consider the consonantal "root" and
the vocalic "scheme" or "pattern" of the word as two discontinuous morphemes, and
the word - or word-base - as the result of the interdigitation of these two. Since this
terminology involving discontinuous morphemes is consonant with previous termin-
ology and does not contradict the historically determined concept of the root in the
Semitic languages, and since, moreover, it may lead to useful distinctions and simpli-
fied descriptions on the morphological level, I see no reason not to adopt it beside the
more common terms of "root" and "noun-form", and to resort to it when expedient.
A residual number of items in the lexicon cannot be broken down into roots and
patterns; into this group belong first of all the foreign words and loanwords, also some
proper names; secondly, words whose pattern - vocalic discontinuous morph - does
not recur in other words of similar distribution although they may have an etymology,
i.e., a root, or whose root even does not recur in other words. Such words are most
prepositions, conjunctions, interjections, also pronouns, some numerals, etc. These
words are best considered single morphs, not the interdigitation of discontinuous
morphs, and we may give them the name "continuous" or "linear" morphs.
MORPHOLOGY 55

The items of the lexicon we have considered up to now were words, i.e., free forms ;
moreover, the bound forms of the lexicon, i.e., the derivational and inflectional affixes,
must also be considered morphs; such morphs belong to the group of "continuous"
or "linear" morphs. These affixes are either consonantal, or vocalic, or consist of both
vowel and consonant.
We may now redefine the items of the lexicon consisting of a root and a pattern :
they consist of a discontinuous consonantal morph, of a discontinuous vocalic morph,
and may or may not include one or more linear affixes : one or more prefixes, one or
more linear infixes, and one or more linear suffixes. A "pattern" or "noun-form"
may be redefined as the vocalic discontinuous morph plus the linear affix. Since vo-
calic morphs contain both short and long vowels, the definition of the vocalic morph
has to be enlarged to that of a morph containing one or more vowels and length.
The distribution of consonants in affixes is very limited: only the consonants /m, n,
s, t, k /and length {:} enter into affixes.
While it is useful to analyze a word into discontinuous morphs and thus establish
its root, i.e., its etymological affiliation, and its "noun-form", i.e., its typical distribu-
tion class (grammatical function or meaning), it is not necessary to break a word-base
into its discontinuous constituents when we consider its inflection. I would thus like
to make the distinction that the interdigitated morphs build up the lexicon, while on
the inflectional level only the combinations of linear morphs, both bases and affixes,
operate. Some of these bases, which on the grammatical level, i.e., in the inflection,
are considered as linear morphs, may be broken down, in the lexical analysis, into two
interdigitating discontinuous morphs.
At the boundary between the word-base and the affix - prefix, infix, or suffix - vari-
ous rules of assimilation, dissimilation, and replacement obtain; hence it is necessary
to describe these rules. Such rules do not affect any two phonemes in contact, only
those that come into contact at these morpheme boundaries; hence these rules are
not phonemic but morphophonemic rules, and are described in a special section,
customarily termed morphophonemic alternations.

5.1. MORPHOPHONEMIC ALTERNATIONS

Whenever two morphs come into contact, in inflection, derivation, or in compounds,

a set of morphophonemic rules obtains. These rules are rather easy to elicit, since the
inflectional and derivational linear morphs have an extremely limited roster of
morphophonemes: V (any vowel), {m, n, §, t, k, :}. The morphophonemic rules, i.e.,
assimilation, dissimilation, metathesis, elision, and the resolving of consonantal
clusters (also called prothesis and anaptyxis), once established for these seven classes,
enable prediction of all forms.
It is hence not necessary to list for each phoneme of the language the so-called
"changes" that it can undergo, since these changes affect a phoneme only in contact
56 MORPHOLOGY

with one of the above listed phonemes. Neither must morphophonemic changes, i.e.,
the changes that occur at morpheme boundaries, be confused with other so-called
"phonetic changes" ; the latter, usually collected and listed at random under the head-
ing "Phonetic changes", or the like, need to be sorted out into morphophonemics,
historical changes, dialect variations, or whatever category they prove to belong to.
The introduction of the concept of morpheme boundaries is not arbitrary, since it
may be said, conversely, that a word-boundary is certain only at the juncture where
the morphophonemic rules do not obtain. Word boundaries are rather easily found
in such positions as the beginning and the end of lines of a cuneiform inscription (see
2.1.2) and at boundaries of graphs that do not conform to the normal syllabification
pattern (see [Link]).
All morphophonemic rules must be established on the basis of the oldest dialects,
since these rules eventually became obscured through a peculiar scribal practice, to
which the name "morphographeme" has been given (by Gelb, although he used this
term in a different context in Gelb, 1961b, 194).
This scribal practice consists in restoring in the writing as much as possible of the
free form of the morph which is the base of the word, indicating in the writing the
morphophonemic alternant of the affix only, and this in itself must be sufficient to in-
dicate the morphophonemic alternant of the base. For example, the bound morph
/ma:t/ (substantive nominative singular /ma:tu/), when combined with a suffix be-
ginning with /§/ has the morphophonemic alternant /ma:s/ and the suffix the alternant
with initial /s/. The inflected form /ma:t/+/Su/ yields \ / m a : s : u > /massu/ or /mas:u/.
This is written in the earlier periods as <ma-su> or <ma-as-su> ; later, however, it is
written <mat-su> or <ma-at-su>. In such a case, the morpheme alternant of the suf-
fix sufficiently indicates the shape of the whole inflected form. Writings of the type
*<ma-at-§u> occur in peripheral scribal areas ; elsewhere they are extremely rare and
insufficiently attested for drawing conclusions of a phonetic nature.
However, the discussion of the morphophonemic rules is best taken up after the
description of the morphology, since it will constantly draw on combinations of
morphs that remain to be isolated. At that time the different treatment of consonants
in contact position within one morph and of consonants in contact belonging to
different morphs will also be illustrated.

5.2. INFLECTIONAL CLASSES

On the basis of the categories for which certain classes of words inflect, three in-
flectional classes may be distinguished : noun, verb, and pronoun. Uninflected free or
bound forms - words and clitics - may be grouped together as indeclinables.

5.3. THE N O U N

A noun is a member of a class of words which is subject to nominal inflection, i.e.,

which participates in the categories of gender, number, and case. This class comprises
MORPHOLOGY 57

substantives, adjectives, and participles, distinguished according to their degree of

participation in these inflectional categories, see 5.3.2. These three subclasses of the
noun are usually distinguishable also through their "pattern", see 5.0.
Nominal inflection is marked by suffixes affixed to the noun which appears in var-
ious allomorphs before these suffixes. A list of these suffixes and allomorphs alone -
usually presented in paradigmatic form - is however obviously not sufficient to des-
cribe the morphology of the noun, especially since in Akkadian their distribution in
various syntactic constructions is of a great complexity. I will here deviate from the
organization of the existing grammars of Akkadian, which treat the distribution of
noun-allomorphs partly in the morphology and partly in the syntax, and endeavor to
show their relation to various syntactic constructions through the application of phrase
structure and morpheme structure rules. This treatment, in the Appendix (pp. 128ff.),
will be preceded by a conventional description of the contextual allomorphs (5.3.1),
inflectional categories, and suffix sets (5.3.3-5.3.5).

5.3.1. Declensions

The inflection of the noun falls into three types of declension, depending on the envi-
ronment. These three contextual "situations" of a noun are called in the traditional
terminology, adapted from that of other Semitic languages, the three status of the
noun: (1) the independent declension ("status rectus"), (2) the bound declension
("status constructus"), and (3) the dependent declension1 ("status absolutus"). In
terms of morpheme analysis, these three status can be called: (1) free form, (2)
bound form, when the noun is followed by some morph or morphs with which it
forms a compound (presumably a stress-unit), and (3) terminal form, followed by
terminal juncture (in the terminology taken from other Semitic languages, "Pausa"),
also definable syntactically as predicative. These three forms can be regarded as
allomorphs of the same morpheme, determined by their environment described in
terms of juncture.

[Link]. Inflectional and syntactic distribution of the declensions.

The difference between the three declensions of the noun, stated above in environ-
mental terms, is also reflected by their selection of the nominal inflectional categories:
Declension (1) inflects for gender, number, and three cases; declension (2) inflects for
gender, number, and two cases; declension (3) inflects for gender and number, and
both the gender and the number categories are restricted to certain nominal subclasses,
see 7.2.1.
Contextual (syntactic) distribution of the three declensions may be summarized as
follows:
1
"Declension (3)" is in reality no declension, since nouns in this status are not inflected for case,
see [Link].
58 MORPHOLOGY

Declension (3) (dependent declension) is used when the noun is the predicate of a
clause. A limited number of nouns occur, partly with different allomorphs, in this
declension as a member of idiomatic constructions for which see below, 7.2.1.
Declension (2) (bound declension) is used when the noun is followed (a) by a per-
sonal suffix (see [Link]), (b) by a noun in the subjoined case (genitive), (c) by a subord-
inate clause, i.e., a verb phrase with its predicate ending in a subjunctive suffix or the
mood suffix which takes precedence over the subjunctive suffix (see [Link]).
Declension (1) (independent declension) is used when none of the criteria for declen-
sion (2) or (3) obtains.

5.3.2. Nominal subclasses

Substantives, adjectives, and participles share the same inflectional categories (gender,
number, and case), but participate in them in varying degrees, and also do not equally
occur in all three declensions.
The adjective and the participle are the marked subclasses ; they are identifiable by
the following marks:

(a) adjectival or participial "pattern" (noun-form), i.e., the vocalic pattern or vocalic
pattern plus a prefix, see 5.0;
(b) obligatory mark of gender (an adjective or a participle is either overtly marked
for feminine gender or not so marked and is then a masculine);
(c) the selection of the allomorph {u:tu ~ u:ti} as ending of the masculine plural, see
[Link].5;
(d) inflection for number (as well as for gender) in declension (3).

The substantive is characterized by the absence of mark (d), and marks (a), (b), and
(c) are facultative, i.e., the substantive may or may not have a "pattern" or vocalic
morph, it may or may not be marked for gender, and it may or may not occur with
the ending {u:tu ~ u:ti}. As to the last mentioned criterion, the fact that a small group
of substantives, mainly designating professions, social classes, and the like, occurs now
with the {u:tu ~ u:ti} allomorph of the plural, now with another, does not affect the
distinction between the nominal subclasses, but is a matter of lexicon and syntax, see
[Link].3.
Both substantives and participles occur in declension (2); adjectives do not occur in
declension (2), i.e., they do not form a compound (stress unit) with a following morph
or morphs.
Specifically, (a) an adjective is never followed by a personal suffix, (b) it is never
followed by a subordinate clause, (c) when it is followed by a noun in the subjoined
case, an adjective does not take the bound form but occurs in one of two forms
depending on the gender concord required : masc. -a(ro), fem. -at, see 7.3.1.
MORPHOLOGY 59

5.3.3. Gender

The noun has two genders, masculine and feminine. Of these, the feminine gender is
the "marked member" of the pair, the gender suffix being {t}. In the singular, the gen-
der is overtly indicated for all adjectives, but not for all substantives ; hence the absence
of the gender suffix does not necessarily indicate masculine gender, except for adjec-
tives, and since an adjective is not always recognizable by its form as such, in the ab-
sence of gender suffix it is the syntactic concord alone that indicates whether a noun is
masculine or feminine. In the plural, the masculine and feminine endings are different
morphs and are syncretic, i.e., carry the gender distinction as well as the number
specification (gender-with-number). Since the gender suffix is consonantal, the pho-
notactic alternants of the base have to be stated. For these see 7.1.

5.3.4. Number

The maximum differentiation in the category of number in Akkadian is three-way:

singular, dual, and plural. Both dual and plural contrast with singular by different
sets of endings in the periods preceding SB, namely in OAkk., OA, and possibly still
in OB.
The distinction between the numbers dual and plural is abandoned in OB, in two
ways:
(1) dual endings are replaced by (fem.) plural endings in concord with a head-noun
even if this noun has the dual ending (i.e., adjectives are not inflected for the dual);
(2) dual endings that occur in the (OB and) SB dialect with nouns that denote a
pair of objects - mostly parts of the body occurring in pairs - do not contrast any
longer with plural endings; they have become allomorphs of the plural morpheme.
Their distribution with such words as /qa:tu/ 'hand', /se:pu/ 'foot', /i:nu/ 'eye',
/nak:aptu/ 'temple(s)', etc., is such that when their number is more than one - nor-
mally two, but also when more than two - these words select the set of endings that
have functioned as dual endings in previous periods; when the same base occurs
with the regular plural ending, the noun has a different meaning, i.e., it is the two
bases that are homophonous and in contrast, but the two plural allomorphs do not
contrast grammatically. For example two or more eyes are /i:na:n/, while /i:na:ti/
means 'two or more beads (or springs)'; two or more hands are /qa:ta:(n)/, while
/qa:ta:ti/ means 'two or more shares'. Even a noun which refers to a singular item,
such as /re:Su/ 'head', occurs with the old dual ending, in a specialized meaning 'top'
in the two case forms /re:si:/ and /re:sa:/. The contrast is on the lexical level.
Since, as it will be discussed under case inflection in [Link], to the three cases of the
singular correspond only two cases in the plural (and formerly also in the dual), the
singular is marked solely by the case ending while the plural morpheme is a syn-
cretic "number-with-case" ending. For details see [Link] and [Link].
60 MORPHOLOGY

[Link]. Endings of the singular.

Three vocalic endings mark the substantive and adjective as singular: -u, -a,
and -i/e. -u marks the subject case (nominative), -a the object and referential case
(accusative), and -i¡e the subjoined case (genitive).
It is to be noted that of the three singular endings, the -a ending is the first to be re-
placed by the -u "nominative" ending, thus equalizing the pattern of case inflection
between singular and plural into two cases each.

[Link].1. Historical note. — The older form of these endings (OAkk., OA, partially OB) is -um, -am,
-im ; in the OB period and sporadically earlier, in OA, the loss of the final /m/ can be noted; the literary
dialect however still often writes final /m/, though usually only by selecting a CVC sign whose last con-
sonant is /m/. It need not be necessarily assumed that these signs have to be read as a CV sign, e.g.,
tut and not turn, lut and not lum, ti and not tim, rù and not rum, etc., since in the literary dialect these
CVC signs are used only for forms in which the ¡ml ending is historically correct, in contrast to ar-
chaizing texts of the NB period which use these signs even when historically incorrect, i.e., when in the
earlier period there would be no such /m/ ending, for example <ni-sim> = /niäi:/, <at-ma-nim>
= /atma:ni/ VAB 4,178:23 and 28 ;2 such late texts even use CV-VC combinations such as tu-um,
/«-«TO,etc., to write words in which the final /m/ is not historical,® and so these writings can best be ex-
plained as a graphic hypercorrection, and the CVC signs of this period too are best transliterated
with their full syllabic value.
The literary dialect, with few exceptions, uses the three cases correctly, and this seems to indicate that
the literary or learned language has maintained the three cases of the singular. In the spoken language
the final vowels seem to have fallen together, thus bringing an end to the case inflection which was re-
placed by fixed word order. Even in the literary language, occasional interchanges of the vocalic end-
ings can be noted, indicating that the process that was taking place in the spoken language encroached
on the speech-forms considered "correct".

[Link]. Endings of the plural.

The endings of nouns in the plural are {u: ~ i:/e:}, {a:nu ~ a:ni}, {a:ne:}, {a:n ~
e:n}, {u:tu ~ u:ti}, {a:tu ~ a:ti}.
[Link].1. {u: ~ i:/e:}. - This ending occurs with substantives, and the distribution
of the two allomorphs is that of nominative and non-nominative, or, as is frequently
the case in the literary dialect, of free variation. Free variation between the two allo-
morphs seems to indicate that case distinction was first lost in the plural.
If we consider that the "accusative" is the least stable case of the inflection (see
[Link]), we may define this morpheme as an ending /:/ added to the case ending of the
singular. In earlier texts, and in the literary texts in which the final /m/ of the singular
is still written (see [Link].1), the plural is contrasted with the singular by the ending
/ :/ instead of /m/.
Since this ending does not occur with feminine substantives, it is assigned to a mor-
pheme "masculine plural", see [Link].
[Link].1.1. Writing of {:}. - The plural suffix {:} is sometimes indicated by "piene"
writing (see 4.1) especially in OB, and it is always so indicated for certain substantives.
It can at present not be decided whether this writing is merely "piene" writing of

' Note, however, in OB already, Sa ana lawi-Ju ta-al-la-kum, for /tal:aku/ (RA, 27, 142:8).
3
E.g., <ul-lu-um> for ul-lu-u = /ul:u:/, 3. pl. prêt, of elû 'to be high', D-stem.
MORPHOLOGY 61

length /:/, or is to be considered phonetically and even phonologically different from

/V:/ as for instance "overlong vowel" or even disyllabic /V:V/, the case being similar
to the ending <Ca-ni-e> discussed in [Link].3.
Examples :

nominative : (genitive :)
a-wi-lu-ú a-wi-li-e /awi:lu:/ or /awi:lu:u/
su-ha-ru-ú su-ha-ri-e /suha:ru:/ or /suha:ru:u/

Note that Gelb, 1955, 107, § 61, considers the spellings <Ci-e> as graphic device only,
"in which -e functions regularly as the phonetic indicator of the oblique case of the
Pl.". This interpretation can only be accepted for the OA examples cited loc. cit., since
also the nominative case of the two substantives cited above is written "piene".
[Link].2. {a:nu ~ a:ni}. - This ending also occurs with substantives, in the same
distribution as noted above. Again, it does not occur with feminine substantives, and
thus is to be assigned to a morpheme "masculine plural".
[Link].3. {a:ne:}. - An allomorph, or possibly only a spelling variant or phonetic
variant of {a:nu ~ a:ni} described in [Link].2, has the graphic shape <(C)a-ni-e>
which might equally be transcribed <Ca-né-e>, i.e., can be phonemicized as /ani:e/,
/a:ne:/ or /ane:/. It is the last of these phonological transcriptions that I prefer, since
it permits the statement that the ending /a:ni/ has a variant /ane:/ with a shift in the
position of / :/. This ending occurs with substantives that end in a long vowel in their
singular inflected forms; it can thus be stated that the singular ending /V:/ plus the
plural ending /a:ni/ result, instead of */CV:+a:ni/, in /(C)ane:/ (or /(C)a:ne:/) by a
shift of the stem-final length to word-final position.
Examples :

writing: sg.: writing: pl.:

bu-ru (genitive:) bu-re-e bu-ra-ni-e /burane:/
ha-ru-u /haru:/ ha-ra-ni-e /harane:/
i-tu-ú /itu :/ i-ta-ni-e /itane:/
as-ma-ru-ú /asmaru:/ as-ma-ra-ni-e /asmarane:/
as-ma-ra-a-ni-e or /asmara:ne:/
nu-dun-nu-u /nudun:u:/ nu-dun-na-ni-e /nudun:ane:/
çi-ru-u /siru:/ si-ra-ni-e /çirane:/
gab(a)-ru-u /gabaru:/ gab(a)-ra-ni-e /gabarane:/
gu-(uq)-qu-u /guq:u:/ gu-(uq)-qa-ni-e /guq:ane:/
mu-su-ú /mu:su:/ mu-sa-ni-e /mu:sane:/

Most of these forms appear in the NB period - although the form /asmarane:/ occurs
in the literary dialect of this period - and thus may be ascribed to the NB dialect.
[Link].4. {a:n ~ e:n}. - This ending occurs with substantives, with the same distri-
bution of the two allomorphs as the preceding plural endings. The ending {a:n ~ e:n}
62 MORPHOLOGY

is in complementary distribution with {:} and with {a:nu ~ a:ni}, and occurs with
names of the parts of the body and other substantives denoting items that come in
pairs, since it represents the dual number suffix which contrasted with the plural at
earlier periods (see 5.3.4). This ending co-occurs with the gender concord "feminine
plural", and thus is assigned to a morpheme "feminine plural".
[Link].5. {u:tu ~ u:ti}. - This ending occurs with adjectives and participles in the
masculine, and is, as noted in 5.3.2, the only formal differentiation between substan-
tives on the one hand, adjectives and participles on the other. In OB and later, when-
ever the writing uses the signs TUM and TIM in word final, the endings are {u:tum ~
u:tim}, their distribution in SB is, as for the preceding endings, that of the two cases
of the plural or free variation. It is also assigned to the morpheme "masculine plural".
For a homophonous ending, see [Link].3.
[Link].6. {a:tu ~ a:ti}. - This ending occurs with nouns - substantives, adjectives,
and participles - in the feminine. In OB and later, whenever the writing uses the signs
TUM and TIM in word final, the endings are {a:tum ~ a:tim}, their distribution is the
same as the previously enumerated plural endings. They are assigned to a morpheme
"feminine plural"; for the selection of either {a:n ~ e:n} or {a:tu ~ a:ti}, see 5.3.4.
For the spread of the "feminine plural", see below [Link].2. This ending occurs with
nouns that are marked for the feminine by the gender morpheme {t} as well as with
substantives that are unmarked for gender. When it is affixed to the former, it re-
places the gender suffix {t}. According to the phonotactic alternations that result
from affixing a /t/ to the noun-base for feminine singular, and suffixing {a:tu ~ a:ti}
to it for feminine plural, the two forms differ in structure, see 7.1.

[Link]. Distribution.
The distribution of these endings which we have assigned to a morpheme "plural",
or to two morphemes "masculine plural" and "feminine plural", has not yet been suf-
ficiently investigated. Some of these endings are in complementary distribution, i.e.,
one of the endings occurs with some nouns, and another with other nouns, some seem
to be in free variation, and some seem to contrast. Whether such contrast justifies
assigning them to different morphemes, or whether the co-occurrence of two or more
endings with the same noun reflects dialect difference or whether it is the two nouns
that contrast in such cases, is difficult to determine, again because of the nature of the
corpus.
A more complete study will become feasible once the dictionaries currently in
process of publication are excerpted for plural forms.
[Link].1. {u: ~ i:/e:} and {a:nu ~ a:ni}. - In the earlier periods, in OA and OB, it
seems that these two endings belonged to two different morphemes, with different
meanings such as "pluralization" and "pluralization plus... ." The ending {a:nu ~
a:ni} occurred with three or four substantives only, and these presumably contrasted
with the same substantives plus the {u: ~ i:/e:} ending, such as /sar:a:nu/ 'kings',
contrasting with /sar:u:/, /du:ra:ni/ 'walls' occurring in the same text - and contrast-
MORPHOLOGY 63
ing? - with /du:ri:/,4 and /a:la:ni/ 'towns'. For a suggestion about the meaning dif-
ference between these two morphemes, see Goetze, 1946, 121ff.
In the literary dialect, a few more substantives occur with the ending {a:nu ~ a:ni}.
These are, apart from the mentioned /sar:a:ni/, /du:ra:ni/ and /a:la:ni/, also /ila:ni/,
/sada:ni/, /salma:ni/, /zaqi:pa:ni/, /kak:aba:ni/, /taksi:ra:ni/, and in the NB dialect the
plurals /dap:a:nu/, /ebu:ra:nu/, /pagra:nu/ and the compound /rab-bu:la:nu/.
These substantives also occur with the ending {u: ~ i:}, and this variation may be
due to the survival in certain texts o f / : / plurals for substantives that have otherwise
generalized the {a:nu ~ a:ni} ending.
[Link].2. {a:nu ~ a:ni} and {a:tu ~ a:ti}. - Also in the literary dialect, some sub-
stantives occur both with the ending {a:nu ~ a:ni} and {a:tu ~ a:ti}, e.g. :
/di:na:ni/ and /di:na:tu/ /[Link]ni/ and /[Link]ti/
/bibla:ni/ and /bibla:ti/ /nasi:ka:ni/ and /nasi:ka:ti/
/tup:a:ni/ and /tup:a:tu/ /papa:ha:ni/ and /papa:ha:ti/
There seems to be no contrast between the two sets of forms that can be determined
at the present. It may be noted that, although plurals in {a:nu ~ a:ni} become more
frequent in later dialects, the most common plural ending, both in the early and in the
late dialects, is the ending {a:tu ~ a:ti} which is suffixed to a great number of substan-
tives which do not have the overt mark of the feminine {t} in the singular; such plurals
require, nevertheless, the formal gender concord "feminine plural".
[Link].3. {u:tu ~ u:ti}. - The "adjective masculine plural" morpheme {u:tu} is
homophonous with a derivational suffix {u:tu} that forms abstract nouns. Hence it
is often difficult to decide whether the ending /u:tu/ that occurs with masculine sub-
stantives is to be considered the plural morpheme, indicating that these substantives
have shifted to the category of adjectives, or the derivational morpheme, which
would shift these nouns into a category "substantive singulare tantum" with a gram-
matical meaning "collective". (For the derivational suffix {u:tu}, see 5.6.3.)
Since the substantives with which these endings occur are usually names of pro-
fessions, age classes and various other social, occupational, etc., classes, both views
are equally possible. In the utterance however the question can be decided on the
basis of grammatical concord (see 5.9).
Substantives that occur with the ending /u:tu/ are, e.g. :
pl. sg. pi.
ame:lu 'man' ame:lu:tu 'slaves' 'mankind' ame:lu:
ardu 'slave' ardu:tu 'slaves' 'slavery' ardami
ibru 'friend' ibru:tu 'friends' 'alliance' ibru:
(OA ebru " eb:aru:tu 'collegium')
abu 'father' ab:u:tu 'chiefs' 'fatherly attitude' ab:u:
agru 'hired man' agru:tu 'hired men' 'hire' agru:
4
Possibly singular, wr. du-ri (Syria, 32, 12:21, Mari); note also that in OA the form /sar:a:ni/ alone
occurs, without a contrasting form.
64 MORPHOLOGY

The plural forms cited in the last column are intended to show that regular plurals also
occur, their dialect distribution has however to be stated. Normally only one or the
other plural occurs in the same dialect. The distribution can bejjchecked by consulting
the dictionaries. However, for the reasons mentioned above, the forms may have to
be looked up under two entries, one without, the other with the ending -u:tu.
The above-cited nouns - all substantives, with the possible exception of the last
cited agru which, according to its pattern, may be considered an adjective and has been
cited above because its plural /agru:/ puts it in the category of substantives - which
have two plurals according to the dialect in which they are attested and belong to two
different semantic and grammatical categories in their suffixed form, may be com-
pared to a group of nouns which, in the plural, always take the /u:tu/ ending and are
to be classed as adjectives on this basis as well as on the basis of their pattern, but
whose syntactical class is that of substantives, both in the singular and in the plural.
Such nouns are: nakru, kabtu, dan:u, eb:u, taklu, halqu, abku, etc.
The question whether the /u:tu/ ending is plural or derivational is to be decided by
the grammatical concord it governs: a noun with derivational ending /u:tu/ is singular
and governs concord for the feminine singular. A noun with derivational suffix {u:tu}
does not occur in the plural, presumably because the derivational suffix {u:tu} is mu-
tually exclusive with the homophonous plural suffix {u:tu} and with the latter's allo-
morphs.

[Link]. Infixed plural.

A restricted number of nouns - mainly adjectives, but also a few substantives (abu,
ahu, bakru, zikaru) form their plural, in addition to the ending {:} or {u:tu}, by an in-
fix {:}. The infix appears after the second consonant of the stem, schematically: sg.
CVCC-, pl. VCVC:C-, with the morphophonemic vowel resolving the cluster of three
consonants (see [Link]) as /CVC:VC-/.
Examples:
sg. pi. (adj. plurals here cited ending in -a and not -a:tu,
•u and not -u:tu, are predicatives, i.e., declension
(3), see [Link] and 7.2).
/ arku/ /ar:aka/
/rabu:/ /rab:u:tu/
/sehru/ /seh:eru:tu/
/kuru:/ /kur.a:/ (OB /kur:i:a/, see [Link])
/se:mu:m/ /sem:u:/ (OB)
/damqu/ /dam:aqu:tu/
/daqqu/ /daq:aqu:tu/
/bakru/ */bak:aru:/, /bak:ari:/
/zikaru/ /zik:aru:/
/ebru/ /eb:aru:tu/
/hanbu/(?) /han:ama:tu/ (OB), or /ljan:aba:tu/
MORPHOLOGY 65
[Link]. Plural morphemes.
Two plural morphemes can be isolated according to the gender concord they gov-
ern: a "feminine plural" morpheme, {a:n ~ e:n ~ a:tu ~ a:ti}, and a "masculine
plural" morpheme, {: ~ a:nu ~ a:ni ~ u:tu ~ u:ti}, or, disregarding the distribution
of the final vowel, the morphemes {a/e:n ~ a:tu/i} and {: ~ a:nu/i ~ u:tu/i}. The
plural allomorph /:/ occurs either as suffix or as infix. Infixed {:} co-occurs with
(a) the suffix {u:tu ~ a:tu};
(b) the suffix {:} only in the forms ba-ak-ka-ri-i/e (/bak:are:/, pi. to /bakru/ cited
[Link].), in the SB dialect, in ah-hu-ú (/ah:u:/, pi. of /ahu/), ab-bu-ύ (/ab:u:/, pi. of
/abu/), both in the earlier dialects and in SB;5 the plural /zik:aru:/ cited in [Link] be-
ing written < zi-ik-ka-ri > , may have to be taken as /zik:aru/, with infix /:/ only,
without any suffix of the plural.

[Link]. Generative statement of the plural.

A generative ordering of the plural could be made in the following steps (where we
restrict ourselves to the word level) :

A d j . pi. -»• A d j . base χ ( : ) + Ju:tul (i.e., the base may in-

|a:tu| terdigitate with an in-
fix {:})
Subst. pi. Subst. base +
fei
pli - (0 + u:tu
a:tu
pl 2 Ja:n
-*•

ja:nu

These rules express the statement that the adjective plural is formed by an optional
infix {:} and the suffix {u:tu} or {a:tu}, which are selected depending on the gender
concord. The substantival plural is formed by adding to the substantive base one of two
sets of plural morphs. One set (plx) includes the suffixes {:}, {u:tu} and {a:tu}, and
these may occur alone or together with the infix {:}; the second set (pl2) occurs only if
the infix {:} is not present, and includes the allomorphs {a:n} and {a:nu}.

5.3.5. Case inflection

Case inflection is best described according to the three declensions, see 5.3.1 and
[Link]. (For an alternate presentation, see the Appendix).
6
The plural /i?:u:/ cannot be taken with certainty to belong to this group, since it may be the plural
not only to /i$u/, but also to its by-form /i$:u/, with suffixed ending only.
66 MORPHOLOGY

[Link]. Declension (I).

Substantive and adjective inflect in declension (1) for three cases in the singular, for
two cases in the plural (and in the periods where dual number is distinguished from
the other two, in the dual). Previously cited inflected forms of the noun, unless hy-
phenated to indicate syllabic transcription, as also all citation forms, unless otherwise
specified, were taken from declension (1) and represent the free allomorph ("status
rectus") of the noun.
The subject case ("nominative") ending is /u/. This case is used for the subject of
the clause and the words that are in concord with it: adjective, appositive substantive,
pronoun. (For the pronoun inflection see 5.5.3.)
The object and referential case ("accusative") ending is /a/. This case is used for
the object of a transitive verb and the words that are in concord with it, and the term
of reference of an intransitive verb, adjective, adverb, etc., that cannot be further speci-
fied here.
The subjoined case ("genitive") ending is /i/ or /e/. The possibly allophonic distri-
bution of /i/ versus /e/ seems impossible to state at the present time and conceivably
never will be stated, owing to the ambiguities of the writing system as outlined in 2.4.2.
The name subjoined case is preferable to that of "genitive" because this case is deter-
mined by its antecedent. The noun is in the subjoined case if its antecedent is (1) a
preposition (see 5.7), (2) a determinative (see 5.5.4), (3) another noun in the bound
form (see [Link]).
Case inflection is in full use up to and including the OB period; subsequently, the
endings occurring most frequently are /u/ and /i ~ e/, both in their proper case func-
tions subject and subjoined case ("nominative" and "genitive") and - usually /u/ -
for the "accusative". In the NB dialect all three endings, /u/, /a/ and /i ~ e/ occur
without reference to their syntactical function. This indicates that the vocalic endings
had been dropped or had coalesced into one vowel ; possibly the final vowel, when it
is written, is either only the graphic device for writing the word-final consonant when
it forms the second member of a cluster, or an attempt to write the "correct" gram-
matical form. In NB texts, and most likely in the literary dialect likewise, the syn-
tactical function of case endings is taken over by word order, with the following
restriction: object precedes the verb, subject precedes the object; the subjoined case is
sufficiently indicated as such by its antecedent.
In the plural, only two cases are distinguished ; the endings are /u/ for the subject
case ("nominative") and /i ~ e/ for the non-subject case ("genitive-accusative").
These endings co-occur with the endings of the plural : they appear before the plural
suffix /:/ and after the plural suffixes /a:n-/, /a:t-/, and /u:t-/, that above in [Link],
were cited in their forms with these case endings (as a:nu/a:ni, a:tu/a:ti, u:tu/u:ti).
For example, subject case /kalbu:/ non-subject case /kalbi:/
/a:la:nu/ /a:la:ni/
/sar:a:tu/ /sar:a:ti/
/dan:u:tu/ /dan:u:ti/
MORPHOLOGY 67
Since the plural suffix was extracted in [Link].1 from {u: ~ i/e:} as {:}, it may be
useful to apply this segmentation to any of the allomorphs of the plural and rewrite
them as {a:nu/i:}, {a:tu/i:}, {u:tu/i:}.e It has been preferred here to consider all plural
endings as allomorphs of two plural morphemes (see [Link]), or even - if we include
into a morpheme "plural" the feminine plurals - as allomorphs of one morpheme
plural; this enables us to describe the case inflection of the plural as affecting the final
vowel of the plural morpheme, whether that vowel is followed by / :/ or not.
Note that in the periods when final /m/ is written, the only nouns that end in a vowel and not in /m/
are the plurals of the masculine substantives, for which length {:} or {a:nu ~ a:ni} contrasts with
/m/, while the other plurals (feminine /a:tum ~ a:tim/ and adjective /u:tum ~ u:tim/) also end in /m/.

The analysis of the case endings as attested by the final vowel of the plural morpheme
equally applies to the allomorph described in [Link].4 (the former dual number). In
this morph the vowel that precedes the final /n/ is subject to alternation of case, and
equally is to be described as the final vowel of the plural morph; the two case endings
are /a:n/ and /i:n ~ e:n/.
The variation of the final vowel noted for the singular case endings also applies to
the plural case endings. The conditions and environments of the variation have not
yet been determined.

[Link]. Declension (2).

As stated in [Link], declension (2) (the bound declension) inflects for two cases. In
the singular, the distribution of the two cases is that of subjoined case ("genitive")
and non-subjoined case ("nominative-accusative"). The singular subjoined case has
the ending -z, the non-subjoined case is the noun base without any inflectional ending,
i.e., zero ending. For phonotactic alternants see 7.3.
In the plural, the distribution of the two cases is subject case ("nominative") and
non-subject case ("genitive-accusative"). The non-subject case has the ending -i co-
occurring with the plural suffix, i.e., occurs in the shape {i: ~ a:ti ~ u:ti ~ a:ni};
note that the plural allomorph {a:nu ~ a:ni} does not occur in declension (2) in word
final, only when followed by a personal suffix. The subject case is represented by the
plural suffix minus the final vowel for the suffixes {a:ti} and {u:ti}, i.e. occurs in the
shape {a:t ~ u:t}, by zero or by /u/ when the plural suffix is not {a:ti} or {urti}, the
first of which - zero - may be subsumed under the category "minus the final vowel"
just established, and the second - /u/ - which is homophonous with the case-ending
of declension (1) may be interpreted as the bound form having shifted to declension
(I)·

[Link]. Declension (3).

Declension (3) does not inflect for case. For phonotactic alternants, see 7.2.

" In that case, the case endings of the plural would be segmented in all instances into case ending plus
plural suffix {:}.
68 MORPHOLOGY

[Link]. "Adverbial" cases.

Apart from the three cases that are differentiated in the singular, two more cases are
listed in the grammars: one with an ending {um ~ u} (the distribution being the same
as for other endings that occur with and without final /m/, see [Link].1 and [Link]),
the so-called "locative-adverbial"; the other with an ending {is}, the so-called "ter-
minative-adverbial". The name "adverbial" is given to the cases for which these end-
ings function because the same endings form adverbs and prepositions when affixed,
as derivational endings, to nouns which then, alone or in combination with a prepo-
sition (see 5.7), serve as prepositional locutions. Most of the nouns to which the end-
ings {um} and {is} are suffixed denote parts of the body (for semantic parallels cf.,
e.g., French lez < Latin latus, à côté, vis-à-vis; Hungarian hátra, mellett, szemben, etc.).
Akkadian prepositions with suffix {um ~ u} are: lib.u 'within', qerbum 'within',
balum 'without', with suffix {i§}: se:ris 'towards', mahris 'towards, in front of',
qudmis 'in front of', [Link] 'in front o f , etc.
The two cases, here for brevity's sake called simply "locative" and "terminative",
are very rarely contrastive, and it seems that they are so only in certain dialects. In
most dialects, they are in complementary distribution. The terminative (ending {is})
occurs in positions where the free form occurs, and the locative (ending {um ~ u})
occurs in positions where the bound form occurs, i.e., primarily before a personal
suffix or a noun in the subjoined case.
Examples :
terminative : era:siS (OAkk.), mu:a:tis (OA), ama:ris (OB), ilis (in Ilis-
tikal, OAkk., OA, OB, MB).
locative :
a) with noun in the subjoined case: sil:um Suen (OB), qabaltu $e:ri,
ik:ul:atu ila:tim (<in(a) kul:atu ila:tim).
b) with personal suffix: napla:sus:a, qibi:tuS:a, Se:pu§:u, $e:ru§:u,
al:a:nu§:u.
The above examples sub b) are the result of the assimilation of final /m/ of the suffix
/um/ to the initial /s/ of the suffix at morpheme boundary, see 6.1.3. Id. However, from
the re-analysis of this ending /us:u/ as /us+su/, a new locative ending/us/ is abstract-
ed, and hence the personal suffix, even one that does not have initial /s/, is affixed to
this ending /us/, or a noun joined to it, as, e.g., in
a) with suffix: pa:nuska (besides pa:nuk:a < pa:num-ka), bi:tuska;
b) with noun in the subjoined case : idus ti:amat.
The secondarily abstracted ending /us/ may even occur as locative without follow-
ing noun or suffix, as in /irtus/, or replace the bound form in some words, as in
/mislus:a/ (accusative, for /miSilsa/).
While a noun with ending {is} normally occurs as a free form, there are instances
where the ending {i§} precedes a personal suffix or a noun in the subjoined case.
MORPHOLOGY 69
a) with noun in the subjoined case: da:ris u:mi: (OB), da:ri:a:tis u:mi: (OB),
bi:tis emu:tim (OB), *sil:is Dagan (OB).7
b) with personal suffix: rigmiska, qa:tis:u (OAkk.), puhris:un (OB), pantis:a (OB),
qablis:a (OB).
Such "hybrid" forms may either attest to a case-system where the two cases loca-
tive and terminative contrast, or may simply be phonetic variants of the /us/ ending
before consonant (for this see [Link] and n. 7). It is to be noted that the ending
{i§} is productive (e.g., Idiglat-es, OIP 2,114, viii 24), while the {um} ending is not.
No doubt a false archaism is the NB form (in Nbk.) [Link] sama:mi, in which a
derivational (not case ending) {is} is linked as a bound form to a subjoined case.
With some nouns, the ending {is} or {um} is preceded by a morph {a:n} ; this morph
should not be taken as the plural allomorph {a:n} or {a:nu} but may be connected
with the ending {a:nis} which forms comparatives and which occurs e.g. in ara:nis,
anza:nis, asmara:nis, etc.
Examples for {a:n} plus {is}: sulma:nis, ba:ma:nis.
Examples for {a:n} plus {um}: supra:nus:u, kis[Link]nus5:u.

A residual number of forms are interpreted in the grammars as locatives; their meaning and concord
is plural (including the former dual); the ending is /u/ which is taken as the allomorph of {um} and is
suffixed to the plural morph before personal suffix or subjoined case: bina:tus:u, ana si:ma:tu+
[Link]'.tim, or are taken as locatives without an overt mark, as birka-Ja, [Link], and the "locative" is
posited only on account of their meaning.
Moreover, one OA form with the exceptional order personal suffix-case ending, qaq:arüum ([Link]
+Sa+um for *qaq:ar+um+sa) seems to indicate that these "cases" were no more than relics of an
older system. For {iä} as still productive, see above.
On the basis of these considerations, I would extract one morpheme "locative-terminative" with
the allomorphs {um ~ iä}.

5.4. VERB

The verb is the grammatical category subject to verbal inflection, i.e., which inflects
for actor, for tense, and for mood.
The morphs for actor (actor-affixes or personals), for tense and for mood are affixes
(prefixes, suffixes, and infixes) ; they are linear morphs consisting of a vowel, a conso-
nant, or a combination of these (see 5.0) ; the stem to which they are affixed is, for the
purposes of the description of the inflection, also a linear morph, although it may be
broken down further into two interdigitating morphs (see 5.4.3).
An inflected verb form may occur with personal suffixes, stating the person acted
upon (direct or indirect object; see [Link]) and with clitics (see 5.8).

5.4.1. Personals

The actor affixes (personals) specify person, gender, and number. They appear as both
7
wr. B A U . SA-jf and }i-lu-uS; see Goetze, JCS, 17, 34.
70 MORPHOLOGY

prefixes and suffixes. Two paradigmatic sets, in complementary distribution, deter-

mine two conjugations. For their distribution, see below [Link].
Set 1 Set 2
Sg. PI. Sg. Pl.
masc. fem. masc. fem. masc. fem. masc. fem.
3. i- i- i- -u i- -a u- u- u- -u u- -a
2. ta- ta- -i ta- -a ta- -a tu- tu- -i tu- -a tu- -a
1. a- a- ni- ni- u- u- nu- nu-

An initial analysis of the prefix and suffix elements seems to indicate that person is
designated by prefix morphs, gender and number by suffix morphs.
Gender has an overt mark for the second person singular feminine only, a suffix {i}.
In the third person plural the feminine suffix {a} contrasts with the masculine suffix
{u}.
Number is marked by a suffix in the second and third persons; for the second per-
son plural, the ending is {a}, for the third person plural, the masculine ending {u}
contrasts with the feminine ending {a}. Number for the first person is marked by the
prefix.
Some dialects (OAkk., Assyrian, NB) differentiate gender in the third person by
means of the prefix {ta} in Set 1, {tu} in Set 2 for the feminine, and {i} in Set 1, {u} in
Set 2 (as above) for the masculine.
Prefixes {a} and {ta} have the alternants {e} and {te} before verbs with a stem deter-
minant /e/ (see [Link]).
In paradigmatic form, the inflection of the verb-stem /prus/ is, with Set 1 personals,
as follows :
Sg. PI.
masc. fem. mase. fem.
3. i-prus i-prus i-prus-u i-prus-a
2. ta-prus ta-prus-i ta-prus-a ta-prus-a
1. a-prus a-prus ni-prus ni-prus
The paradigm shows that the prefix determines the actor not only as to person, but
also as to number in the first person, and the suffix determines the gender in the second
person singular, the number in the second person plural, and gender-with-number in
the third person plural.

[Link]. Paradigmatic illustration of personal Set 2, with verb-stem /par:is/:

Sg. PI.
masc. fem. masc. fem.
3. u-par:is u-par:is u-par:is-u u-par:is-a
2. tu-par:is tu-par:is-i tu-par:is-a tu-par:is-a
1. u-par:is u-par:is nu-par:is nu-par:is
MORPHOLOGY 71
This paradigm shows that for Set 2, the first and third person singular are not differ-
entiated, as they are in Set 1. Otherwise, the same distribution of gender and number
affixes obtains.

5.4.2. Mood

There are three moods: indicative (unmarked); subjunctive (suffixed); and ventive,
also called allative (also suffixed).
Subjunctive and ventive suffixes are mutually exclusive; a verb with a ventive suffix
does not take the subjunctive suffix.

[Link]. Ventive.
The ventive suffixes are:
singular and prefixed plural : -(a)m {am after C, m after V) ;
other: -ram (/m/ occurs when final /m/ occurs in a dialect (i.e., -nim and (a)m when
final /m/ occurs in a dialect, elsewhere -ni and -a/V).
Note that the distribution is not singular-plural : the prefixed plural selects the "sing-
ular" ventive suffix.
Examples : aprus-a(m), purs-a(m), niprus-a(m), taprusi-(m) ;
pursa-ni(m), taprusa-ni(m), iprusa-ni(m), etc.

[Link]. Subjunctive.
The subjunctive suffix is -u. A personal actor-suffix ending in a vowel is mutually
exclusive with the subjunctive suffix.

[Link]. Sequence of suffixes.

The relative preference sequence of verbal suffixes is as follows : Rank 1 : personal
suffix; Rank 2: ventive suffix; Rank 3: subjunctive suffix.
Rank 1 has preference over Rank 3. Rank 2 has preference over Rank 3; Ranks
2 and 3 are mutually exclusive.
Examples: 3. pi. masc. i-prus-u; subjunctive: same; ventive: i-prus-u-nim; subjunc-
tive of ventive: does not occur (alternately: is the same as ventive).

Note that the so-called "subjunctive" ending -ni occurring in the Assyrian dialects is not a mood end-
ing, but a clitic added to (1) the subjunctive; (2) the ventive; (3) to nomináis; it is followed only by
pause, i.e., it occurs in final position of the string of affixes. For this ending, see 5.8.

For the imperative, see [Link].

[Link]. Optative and vetitive.

It is possible to consider as a mood the paradigm that comprises cohortative, op-
tative (or "precative"), and imperative forms, and similarly the one which comprises
the vetitive forms. These, however, are characterized by a prefix, and hence structur-
72 MORPHOLOGY

ally differ from the other moods that are overtly marked by a suffix. For the "preca-
tive" paradigm, see GAG, Paradigm 14 and n. 5, also Paradigms 16, 18, 23, 24, and 26.

Sg. PL
masc. fem. masc. fem.
3. liprus liprus liprusu liprusa
vetitive ajiprus ajiprus ajiprusu ajiprusa
2. purus pursi pursa pursa
vetitive etaprus etaprusi etaprusa etaprusa
1. luprus luprus (i) niprus (i) niprus
vetitive ajaprus ajaprus eniprus eniprus
Mood prefixes : lu- "cohortative" (1. person only)
li- "optative" (3. person only)
aj "vetitive" (1. and 3. person)
e "vetitive" (2. person)

5.4.3. Root and stem

The morph to which the inflectional actor and modal affixes described in 5.4.1 and
5.4.2 are affixed is the verb stem. The stem may be further segmented into two discon-
tinuous morphs (see 5.0) plus affixes. The consonantal discontinuous morph is usually
called, in reference to Akkadian and the related Semitic languages, the root; the other
discontinuous morph is vocalic. Affixes are pre- or infixes which form derived stems
(see below, [Link]).
The vocalic morph of a simple (non-derived) stem is not predici able and must enter
the definition of each stem as stem-vowel. All other vowels are morphophonemically
conditioned and predictable as to occurrence from morphophonemics, as to quality
from the stem-determinant (see [Link]).
The constituent analysis into consonantal and vocalic morphs has importance for
derivation - both for nominal derivation and for derivation of verb-stems - and will
here be examined briefly; for the description of the verb inflection proper, however,
the interdigitated consonantal and vocalic morphs may be considered simply as the
stem to which the inflectional morphs are affixed.

[Link]. The root.

The root almost always consists of three consonants. For limitations of their oc-
currence see 4.3.1. There is a small group of roots that consist of four consonants, so
patterned that the second and third consonants - R(adical) 2 R 3 - are a liquid (/, r)
plus a stop (ρ, b, t, d, (, k, q), a sibilant (s, z, s or m. Although this type of cluster
seems to indicate that the root originally had only three consonants, the liquid being
syllabic, morphophonemically all four radicals of the so-called quadriliterals behave
MORPHOLOGY 73
like consonants and do so cluster, i.e., the phonotactic rules for clusters equally apply
to the second radical. Attested clusters of R 2 R 3 in quadrilaterals :
lp It lk
rk
rb rd
It
rq
Is
rz rs
1?
lm
rm
From the chart appears that I occurs with non-voiced, i.e., voiceless and emphatic
consonants; r with voiced consonants, except for the velars (with k, q and not g);
both occur with m ; only r with s.
Other so-called quadriliteral verbs have no quadriliteral roots, only form a separate
conjugation; their root consists of three radicals, the second of which is either length
or identical with the third radical; the third radical is always a sonorant (a liquid or a
nasal) ; they occur only in derived stems Ν and S.
Some roots appear to consist only of two consonants and a long vowel or of
"strong" and "weak" consonants; in fact, they all can be said to consist of three con-
sonants if we include length among the radicals. For fuller discussion see below 5.4.5
and ff.
Most roots occur not only in the simple stem - root plus stem-vowel - but also in a
varying number of derived stems.

[Link]. Derived stems.

Derived stems are formed with a prefix or infix from another stem or from a root.
Stems are numbered with roman numerals or given a letter symbol. The simple (non-
derived) stem is called Stem I or B(asic), in grammars written in German, G(rund-
stamm).
Stem II or D (for "Doppelungstamm"), is formed with an infix {:} after the second
radical (in the customary terminology, with reduplication of the second radical).
Stem III or § is formed with a prefix {§}; Stem IV or Ν with a prefix {n}.
Every stem, the simple as well as each of these derived stems, may serve as the base
of other derived stems, one type formed with an infix {t}, symbolized either as 1/2,
II/2, III/2, IV/2, or as Bt (in German Gt), Dt, St, Nt, and another type formed with
an infix {tan} (or {ta+:}, see [Link].2) and symbolized either as 1/3, II/3, III/3, IV/3,
or as Btn (in German Gtn), Dtn, Stn, Ntn.
The derivational infixes {:} of Stem II and {t} of Stems 1/2, II/2, III/2, IV/2 are
homophonous with the tense infixes {:} and {t}; see [Link].
74 MORPHOLOGY

To each stem belong finite verb forms inflected for actor and mood in three tenses;
nominal forms - infinitive inflected for case only and participle inflected for gender,
number and case; and forms sharing in both inflections - imperative and stative - , see
5.4.6. Inflected forms divergent from the regular paradigm are often said to belong
to further stems, such as the stem SD or III/II, a reduplicated R-stem (Kienast, 1961,
59ff.), and a "second St" stem; in Assyrian also a Dtt-stem.

Synoptic chart of Stems

I (G or Β) II (D) III (§) IV (N) R
infix {t} 1/2 (Gt) II/2 (Dt) III/2 (St) Rt
infix {ta(n)} 1/3 (Gtn) II/3 (Dtn) III/3 (Stn) IV/3 (Ntn)
infix {ta+ta} II/4 (Dtt)
III/II (§D)
The four primary or primarily derived stems I, II, III and IV fall into two groups : I
(G or B) and IV (N) take actor-affixes (personals) of Set 1, all others (II (D), III (S)
and the residual forms) take actor-affixes of Set 2. Secondarily derived, i.e., {t} and
{ta(n)} infixed stems follow, for the selection of personals, the stem they have been
derived from.

[Link]. Stem determinant.

The predictability of occurrence of a vowel /a/ or /e/ in certain positions in a stem
(other than the stem vowel) is treated as stem determinant, see [Link].

5.4.4. Tenses

Three tenses are formally distinguished; they are usually called present (or present-
future), preterit, and perfect. Since this is a study in morphology, considerations
whether these are really tenses or aspects are here left aside.
The tense inflection of the verb, just as the person inflection, is here segmented into
a sequence of linear morphs; since the verb stem is considered as a phonetically
existing, pronounceable, segment, similarly the tense affixes can be segmented as
infixes, which form neither part of the consonantal morph, nor part of the vocalic
morph (pattern).
The tense morph is either an infix or a vowel replacive of the type called "Ablaut"
in grammars of Akkadian, and of other types. Present tense is formed by means of an
infix {length}, or of a vowel replacive, or of both; perfect tense is formed by means of
an infix {t}. The preterit is characterized by the lack of infixes, and in certain groups
of verbs contrasts with the present only by means of the stem vowel.
In the derived stems II and III the contrast between the present and the preterit is
provided by the discontinuous vocalic morph. This is {a-a} for the present, {a-i} for
the preterit. The preferred way of analyzing the derived stems type /parras/ and
MORPHOLOGY 75
/par:is/ (II) and /sapras/ and /sapris/ (III) is to extract first the consonantal and vocalic
morphs: /p-r:-s/ and /s-pr-s/, and /a-a/ and /a-i/, since other forms belonging to the
same derived stems, e.g., the nomináis, again differ only in their vocalism, e.g., in
their vocalic morphs {u-u} and {u-i}, in /pur:us/, /suprus/, /pur:is/, /supris/, etc.

It may be noted here in anticipation to a description of the Assyrian dialects that, while the cited vocal-
ic morphs describe some Babylonian dialects, they furnish a convenient way of describing Assyrian
dialects as well: in Assyrian, the first vowel of the vocalic morphs, if it is an /u/, is replaced by /a/, i.e.,
the above cited morphs are attested as /a-u/ and /a-i/ ; in other cases, the vocalic morphs remain un-
changed. Note that the replacement of /u/ by /a/ in such morphs obtains for nomináis as well, see,
e.g., Assyrian sahurtu, falu:lu, ahusiu, akus:u, habuliu, kaduriu, etc., and Babylonian fuhurtu, fuluilu,
uhultu, ukultu, hubuhu, kudur.u, etc. The MB dialect can also be described by a replacement of {a-i}
by {e-i}, see [Link].

The consonantal morph of these derived stems can be further analyzed into the root
/prs/ and the pre- or infix. The alternate analysis, into a tense morph {a} (present),
and {i} (preterite), and morphophonemic -\/pr:-s and V^pr-s, which, according to
4.4.2, is realized as phonological /par:-s/ and /sapr-s/, pays for the advantage of ex-
tracting a simple tense morph by a need for additional statements regarding other
derived forms.
Derived stem IV has, with regard to its preterit and present tense morphs, an inter-
mediate position between the simple stem I and the other two derived stems. As ob-
served above, it shares the actor-affixes of Set 1 with the simple stem; the present and
preterit contrast, as the tenses of the simple stem, by an infix length, and the stem
vowel is that of the simple stem. However, due to the analogical influence of the tense
contrast a/i of derived stems II and III, as well as, as we shall see in [Link], of a sub-
group of the simple stems, stem IV tenses tend to take, first, a preterit morph {a-i},
then, a present morph {a-a}. This inflection will be discussed in [Link] as a sequel to
the tense inflection of the simple stem from which, as mentioned, it depends.
Another distinction between derived stems II and III on one hand, and the simple
stem, stem IV, and all secondary derived stems in {t} and {ta(n)} on the other is the
relation of the number of syllables for each tense. Verbs of the first group are isosyl-
labic in all tenses (present, preterit, perfect), while those of the second group are heter-
osyllabic, the preterit usually having one syllable less than the other tenses. The dis-
tribution is such that either the preterit has three syllables while the present and per-
fect have four (Gtn, Dtn, Stn, St2, Ntn), or the preterit has two syllables (before
vocalic suffix, although it has three with zero suffix), while the present and perfect
have three syllables (Gt, N), or the preterit has two syllables, the perfect likewise two
(before vocalic suffix, although it has three with zero suffix), while the present has
three (G).
Heterosyllabic tense forms :
76 MORPHOLOGY

present perfect preterit (pi.)

Gtn 4 4 3
Dtn 4 4 3
Stn 4 4 3
St, 4 4 3
Ntn 4 4(?) 3
Gt 3 3 2
Ν 3 3 2
G 3 2 (pl.) 2
Dt 3 4 3
St! 3 4 3

Two stems - last listed in the chart - differ in their syllable distribution. Their present
and preterit are isosyllabic, while their perfect has one syllable more. The two stems
also share a meaning "passive", as well as the syllable structure, and it is on the basis
of syllable structure, and not solely on the basis of meaning, that Stem III/2 or St
proves to comprise two stems, Stx which is parallel in structure and in meaning to Dt,
and St2 which is a derived stem with heterosyllabic tenses and an independent gram-
matical meaning.
Heterosyllabic are also the quadriliterals with a cluster of liquid + obstruent or m.
Isosyllabic are the stems D, § and the residual stems §D and probably R.

[Link]. Stem allomorphs.

In order to present the tense inflection of the simple stem, a more detailed analysis
of this stem is necessary. Such analysis will at the same time provide a pattern for
many of the subgroups of the Akkadian verb which up to now have been treated as so
many separate phonetic developments (see 5.4.5).
The simple stem (Stem I, G or B) is defined by its root - the consonantal morph
usually consisting of three radicals - and the stem vowel. The analysis of the Akka-
dian verb as a stem in which the vowel has to be specified has long been advocated by
Landsberger. It is practiced in GAG where for each stem not only the radicals but
also the stem vowel is given. For certain stems, forming a subgroup usually called
"Ablaut-class" two vowels usually are given, since the vowel of the preterit and that
of the present differ. The "Ablaut-class" stems are those which form the present tense
by vowel replacement, and since, as it will be seen, the replacive is always {a}, there
is in fact no need to specify both vowels; see below [Link]. For example, the root
/pqd/ is cited as the "Verbalwurzel" pqid in GAG, § 73c.
Instead of this citation form alone, I posit for a simple stem two forms, e.g., {-prus}
and {purs-}, these being allomorphs in complementary distribution: the allomorph
/-prus/ occurs after vowel (e.g., in all prefix-inflected forms, as i-prus, ta-prus, etc.),
and the allomorph /purs-/ after zero, i.e., initially. The latter allomorph, /purs-/, is
MORPHOLOGY 77

convenient to describe the imperative which forms the feminine with a suffix /i/, the
plural with a suffix /a/ (see [Link]), and the ventive with a suffix /(a)m/ (see [Link]),
as pursi, pursa, pursam; the imperative singular, which has a zero ending, can equally
conveniently be described as the simple stem, or the stem with zero ending: morpho-
phonemic y/ purs, phonological /purus/, where the already stated morphophonemic
rule applies that a final consonant cluster is resolved by an epenthetic vowel; in the
imperative, this vowel is of the same quality as that of the postvocalic allomorph
(for the need of this restriction, see [Link].1).
The stem and its allomorphs can be stated for each verb, whatever the stem vowel;
for example, to take the usual paradigm-illustrating verbs, the stem of /pqd/ is
{-pqid ~ piqd-}, the stem of /sbt/ is {-sbat ~ sabt-}. The stem allomorphs of the so-
called "weak verbs" to which we shall revert in [Link] are also sufficiently stated in
the above way. There are exceptions : for two groups of verbs, those that have {length}
as last radical, and the "primae w" group (see [Link].1), the initial allomorph, of the
pattern /CVCC-/ cannot be set up. For those that have {length} as middle radi-
cal, the postvocalic allomorph is missing, see [Link]. Since these stems have other
divergent features, the non-occurrence of one of the allomorphs is not unduly worry-
ing, the more so since the imperative forms of verbs that have {length} as last radical
and those of the "primae w" and "primae η" cannot be adequately explained by the
traditional grammars either without resorting to ad hoc rules. In spite of these
exceptions, which will be discussed in [Link],1 find that positing two stem allomorphs
provides a simpler description for not only the "regular" but also for other verb stems
which, according to the traditional grammars, are "irregular".
[Link].1. A subclass of stem allomorphs. - Note that there is a small group of verbs
for which the stem-allomorphs stated in the way just indicated show a difference in the
stem-vowel too. For instance, for the verb which has the preterit iplah and the im-
perative pilaff, the allomorphs {pilh- ~ -piai}} have to be given, and similarly for the
verbs {limd- ~ -lmad}, {rikb- ~ -rkab}, {pish- ~ -psah}, {tiki- ~ -tkal} and, for some
"weak verbs" (see [Link]), {kil kla}, { s i m — -sma}. (For verbs with a "weak"
first radical, see [Link].) The customary analysis of the stem as /prus/ without posit-
ing an initial allomorph avoids this problem but also has to account in some fashion
for the /i/ occurring in the first syllable instead of the more common stem-vowel. Since
the divergent forms (/i/ in the first syllable of the imperative) are paralleled by other
imperatives (of derived stems or irregular verbs) which likewise have an /i/ in this po-
sition, or rather, an /i-a/ vocalic pattern: such as tizkar, tisab, ital (and not *tizkur,
*tasab, *itil), the question would bear further examination.

[Link]. Infixed tense morphs.

The present tense infixed morph is {length}. This length (or "gemination of the
second radical") has not yet been given its proper status as tense morph, and has been
noted only as a concomitant feature of the present going back to an aspectual inter-
pretation of the present "(Durativ; daher wohl auch die Verdoppelung des 2. Radikals
78 MORPHOLOGY

im Prs. G und Ν!)" (GAG, § 78a); "Das Prs. wird von der durch Verdoppelung des 2.
Radikals erweiterten Wurzel gebildet"(italics mine) {GAG, § 87j).
There is^nothing'unusual in the assumption that a tense in Akkadian is formed by
means of an infix; .the perfect tense (see [Link]) is formed by means of an infix {t}
(cf. "Das Akkadische ist|die einzige sem. Sprache, die ein 'Tempus' mit infigiertem
-ta- bildet." GAG, § 80a). Moreover, the use of an infix {length} is^by no means unique
in Akkadian : it is an infix {length} that is the derivational affix of stem II (D). It may
of course be difficult at first to accept the notion that the same item should func-
tion as morpheme of the present tense and as a derivational morpheme; however, it is
not the extraction of the present tense morph as {:} but the linguistic facts that require
this definition, and the current definition that stem II (D) is formed with the redupli-
cation of the second radical likewise defines the present tense and the stem-formation
by the same morphological phenomenon. It has been pointed out by Greenberg,
1952, Iff., that it is precisely this "lack of logic" that made Semitists reluctant to ad-
mit that the Akkadian present tense contains a reduplicated second radical. Hence,
when I posit an infix {:} for both these forms I do not attempt to change or to obscure
the facts. Neither my notation, nor that current in the field ("reduplication") in itself
can solve the problems generated by the cumulation of functions of this particular infix.
Just as there is a pair of homophonous infixes {:}, there is another pair of homopho-
nous infixes which similarly cumulate the function of tense and of derivation : these are
realized as /t/, which, when tense infix, forms the perfect, and when derivational infix,
forms the derived stems Gt, Dt, St and Rt (see [Link]), in such meanings as reciprocal,
reflexive, passive. A difference between these two pairs of infixes is that the two /t/
infixes can occur side by side, while the infix {:} of tense never occurs concomitantly
with the infix {:} of the derived stem. For the homophony of /t/ see "Da das ία-Infix
gleichzeitig auch bei der Bildung abgeleiteter Verbalstämme verwendet wird ...", and
"Wir wissen auch noch nicht ob das ta- des Pf. mit dem stammbildenden ta- her-
kunftsgleich ist . . . " (GAG, § 80a).
The present tense morph {:} is infixed after the second consonant of the post-vocalic
allomorph, which yields morphophonemic \/-pq:id, \/-§b:at, etc. The resulting clus-
ter of three consonants is resolved, according to [Link], by an epenthetic vowel /a/ or
/e/, according to the stem-determinant (see [Link]), yielding phonological /-paq:id/,
/-§ab:at/, etc. Due to the non-occurrence of clusters of more than two consonants,
this present base of the "strong" verb differs from the preterit base not only in the infix
{:} but also by having one syllable more, see 5.4.4.
The perfect tense morph infix can be segmented as {ta} or {t} ; it is infixed after the
first consonant of the post-vocalic allomorph. For examples and the discussion of
segmentation see [Link] and ff.

[Link]. Vocalic replacement.

For some groups of verbs, it is not sufficient to state the two allomorphs of the
stem, because these verbs have one stem vowel in the preterit and imperative, and an-
MORPHOLOGY 79
other in the present and perfect. Hence, either two stems have to be given, or the
present and perfect tense morphs have to include an additional morph, "vocalic re-
placement". This morph can be stated as {a <- V} since the vowel of the present tense
is, in case of a replacive, always /a/, and hence, if the formula for the replacive is used,
there is no need to specify two "stem vowels". To one of the groups of verbs which
form the present with vocalic replacement belongs the stem /-prus/ used in the illus-
trations above; its present stem, with infix {:} and with vocalic replacement {a u},
is /par:as/. This particular case of vocalic replacement is called "Ablaut" in the
grammars. However, since other vocalic replacements also occur, the more general
term "vocalic replacement" or "replacive" is preferable.
The perfect of verbs which form their present with vocalic replacement is likewise
formed with an infix {ta} and vocalic replacement : /-p-ta-ras/.
[Link].1. "Irregular" verbs. - Another subgroup of verbs which form their present
tense by replacive - however, using the replacive only as tense morph, without infix
{:} - have been classed as "weak" or "irregular" verbs. These verbs have a stem vowel
/i/, and once the preterit is marked by an /i/ vowel, the present can be effectively
contrasted with it by a replacive /a/, just as in the derived stems. These verbs are:
preterit il:ik, iz\iz, id'.in, present [Link], iz:az, [Link] (the last of the three is an Assyrian
dialect form). For the last two, cf. now Kienast, 1963, 141 and n. 8.

[Link]. Perfect.
The simple stem remains unchanged throughout its inflection in any of the moods
and tenses, i.e., the occurrence of pre- or suflixes does not condition any further mor-
phophonemic changes. Not so in the perfect tense, where the occurrence of a vocalic
suffix morphophonemically conditions the shape of the perfect stem. For this reason,
two alternate segmentations are possible:
[Link].1. Perfect tense infix. - The tense infix is segmented as {-ta-}. The stem with
which this infix occurs is {-p-ras} before consonant and in final position, and {-p-rs-}
before vowel, e.g., 3. pi. iptarsu, iptarsa, 2. sg. fem. taptarsi, 3. sg. ventive iptarsam,
etc. That is, the perfect stem is the same as the present stem, and the stem-vowel or
the replacive is elided before vowel. The disadvantage of this segmentation is that it
introduces vowel elision which otherwise proves unnecessary as a morphophonemic
rule, and also that it cannot be used efficiently to describe the perfect of some "weak"
verbs.
[Link].2. Perfect tense infix: alternate segmentation. - The tense infix is segmented
as {-t-}. The stem with which this infix occurs is a special stem, derived from the root
with a vowel /a/. This stem R^aR^R^ recurs among derived forms, as a nominal
(adjective). Note that this stem has no post-vocalic allomorph *R 1 R 2 aR 3 . The per-
fect accordingly is realized as -yAp-t-ars, which is the morph occurring before vowel;
before consonant and in final position, the final cluster is resolved by an epenthetic
vowel, according to 4.4.1. The disadvantage of this segmentation is that (1) it requires
the additional statement that the epenthetic vowel is the same as the replacive, or if
80 MORPHOLOGY

there is no replacive, as the stem vowel; (2) while with this corollary the description
is adequate for the literary dialect, it is not usable without further statements for the
description of other dialects.

[Link]. Recapitulation of tense morphs.

The present tense morpheme can be stated as {-:-(+a<-V)}+v , R 1 R 2 -VR 3 .
The perfect tense morpheme can be stated as {-ta-(+a*-V)}+\/Ri-R2VR3.
Alternately, the perfect tense morpheme can be stated as {-t-}+VR-i-aR^R-a·
Explanation of symbols: R l5 2, 3 stand for R(adical) 1, 2, 3, i.e., the first, second,
third consonant of the root.
Hyphen stands for structurally conditioned empty slot for the infix.
V stands for stem vowel (not predictable).
V stands for morphophoneme, i.e., to obtain the phonological form, the phono-
tactic and morphophonemic rules have to be applied.
Examples:

present tense
a) root /pqd/, stem vowel /if: {->}+Vpq-'d== Vpq:id=/paq:¡d/
b) root /prs/, stem vowel /u/: { - : - + a < - V } + V p i " - u s = V P r : a s =
/par: as/
perfect tense
a) root/pqd/: {-ta-}+Vp-qid=/-ptaqid/
b) root/prs/: {-ta-+a<-V}+<v/p- r u s = /-ptaras/
perfect tense, alternate analysis
a) root /pqd/: {-t-}+Vp-aqd=/-ptaqdu/ (pl.), /-ptaqid/ (sg.)
b) root /prs/: {-t-}+\/p-ars=/-ptarsu/ (pl.), /-ptaras/ (sg.)

[Link]. Tense morphs of Stem IV (N).

As observed in 5.4.4, the tense morphs of Stem IV (N) present different contrasts
in different periods. The present morph is an infix {:} which is infixed in the stem de-
rived by a prefix /n/, in which the stem vowel of Stem I is present. However, with very
few exceptions, the vowel in the preterit of Stem IV is /i/, as in the preterit of the other
derived stems, II, III, and in the cited subgroup of Stem I. Evidence for this /i/ vowel
in the preterit, as well as for eventual exceptions, is scant, because most attested pre-
terits of Stem IV occur with vocalic suffixes, in which case no vowel occurs between
the two last consonants of the stem.
After the mark of the preterit of Stem IV came to be /i/, the present likewise came
to be formed on the analogy of stems which express tense contrast by the contrast /i/—
/a/, by a replacive {a«-V}. This development is not quite clear from the paradigms
given in GAG (No. 13); in spite of an indication along these lines in § 90b, c, it is evi-
MORPHOLOGY 81

dent that the author considers that in both tenses of Stem IV the same vowels as in the
tenses of Stem I occur. However, since in the paradigm No. 13 preterits with vowels
other than /i/ are cited in parentheses, i.e., as rare forms, it is evident that this is no
accident, and that all preterits in /i/ must be taken not as illustrating the occurrence
of the stem vowel /i/, but that of the preterit vowel /i/. This view can be supported by
many examples of Stem IV preterits in /i/ where the stem vowel of Stem I is /a/ or /u/,
and by examples of Stem IV presents with a replacive /a/. Similarly, the present forms
cited in the GAG paradigm do not illustrate a "stem vowel" /a/ but the present tense
replacive /a/.
Examples :
stem vowel Stem IV preterit Stem IV present
u ib:atiq ib:at:aq
u in:amir in:am:ar
u ah:abit ih:ab:at
u ih:ari§ ih:ar:aç
u ih:asis ih:as:as
u in:adir in:ad:ar
u in:ahiz in:ah:az
u in:apil in:ap:al
u [Link] ik:an:ak
u in:ak:al
u is:ah:ar
a is:abit is:ab:at

5.4.5. Weak verbs

The verb stems examined up to now have as radicals three consonants that recur in
all inflected forms and in all derived stems - exception made of course of the morpho-
phonemic changes that occur when the first or last radical comes into contact at a
morpheme boundary with a consonant that conditions such a change, in which case
the rules described in Chapter 6 (Morphophonemic alternations) apply. Such verbs
are customarily called "strong verbs".
Other verb stems exhibit in most of their inflected forms only two consonants, or
even only one consonant, although, especially in the periods earlier than the dialect
under investigation, some forms are sometimes written with a syllabic sign whose
consonant component is a [w], a [?] (glottal stop), or a velar fricative /h/. The fact
that in the related languages these verbs correspond to stems that are written in purely
consonantal script with the consonants w, j, glottal stop, or a laryngeal, as well as di-
vergences in the inflection of these stems, have brought about the classification of
these stems into "weak verbs", "doubly weak verbs", and "strong verbs with phonetic
peculiarities"; a "weak verb" contains in its root one "weak" consonant, i.e., a con-
sonant that either was lost between Proto-Semitic and Akkadian, or was lost during
the historical evolution of Akkadian, or which simply "drops" in the inflection; a
"doubly weak verb" contains two such consonants in its root; a "strong verb with
phonetic peculiarities" is a verb which contains a glottal stop in some of its forms, or
has once contained one, and which exhibits differences in vocalism from other stems.
While, at least until recently, in the field of Semitic all "weak verbs" were considered
as containing such a weak consonant, partly on the basis of the consonantal writing,
partly from a desire to posit three radicals for every root, in Akkadian the classifica-
tion of "weak" verbs made progress through an innovation introduced through GAG,
the view that some of the "weak" verbs never had more than two consonants.
It is not my intention here to reopen the question of "triliterality" versus "biliter-
ality" of the Semitic root, nor to investigate the historical question of the origin and
etymological connections of the "weak" stems. The discussion of these verbs in
current grammars is valid from the point of view of comparative or historical
grammar.
Since the purpose of this restatement is to describe actually occurring morphs, I
am not allowed to speculate on the possible origin of these morphs, even if these
morphs look irregular because, instead of having three radicals, i.e., three consonant
phonemes, in their consonantal morph, they seem to have only two or even only one.
What has to be stressed here is the fact that the weak consonants that are used to sym-
bolize the non-occurring radical never occur in any of the forms of the verbs to be
considered. Verbs which are attested in some inflected forms which exhibit a conso-
nant that is not present in other forms must receive separate treatment, see [Link].1.
My purpose is to attempt to describe these "weak" verbs with the same methods as
those used for the description of the "strong" verb, without recourse to a proto-form
MORPHOLOGY 83
never attested in any dialect of Akkadian, or inventing ad hoc "sound laws" devised
uniquely to take care of the divergent forms.
Since this study is not intended as an exhaustive treatment, there will remain a resi-
due of "irregular" verbs, i.e., verbs whose inflected forms are not accounted for, or
predictable by the statements that follow. This residue is however not larger than the
"irregular verbs" so treated in the traditional grammars.

[Link]. Roots which contain length.

Here I wish to remind the Assyriologist that in the section on weak verbs in GAG
there is a group called "hollow roots" which belongs to the type mentioned that are
said never to have contained more than two consonants ; according to GAG, the hollow
roots contain two consonants and a long vowel. I also wish to remind the reader that
in 4.1.2 length was said to behave like a consonant.
Accordingly, any hollow root may, by definition, be rewritten as consisting of three
consonants and a short vowel; one of the consonants (in this case the second) is the
morphophoneme length. Further, it is also possible to rewrite all the weak verbs as
consisting of three consonants, one of which is length. The inflection of weak stems
will usually be parallel to that of stems which do not contain length, and the divergent
forms merely - and automatically - result from the phonotactic rules governing length
and clusters with length, as stated in the chapter on Phonology and Phonotactics.
Additional morphophonemic rules which prove necessary for these stems are not more
complicated than the rules required by the traditional analysis, and moreover have
the advantage of having a more general application throughout the morphology.

[Link]. Stem-determinant.
Before analyzing the stems which include length, and illustrating their inflection, a
further element, the stem-determinant has to be introduced. The feature here treated
as stem-determinant is normally treated in the grammars in relation to the class
established below as Class (1), the "primae aleph", since etymologically their vocalism
depends on the etymological laryngeal: if the laryngeal was ?, or h, their personal
(actor-prefix) vowel, the first vowel of the imperative, of the infinitive, etc., is /a/, other-
wise, if the first consonant was a voiced pharyngeal stop ['ayin], a pharyngeal or
laryngeal fricative (h, g), this vowel is /e/. However, the distinction in the vocalism is
not restricted to this particular class; there are other verbs which have no length as
radical and also have either /a/ or /e/ in similar positions, and this fact which is usually
mentioned in the phonetics, has not been given a satisfactory explanation, i.e., its dis-
tribution cannot be established.
Consequently, I prefer to treat the occurrence of /a/ or /e/ in certain positions - to
be listed presently - as a stem-determinant, and to indicate such stem-determinant for
all verbs. If we disregard etymological reasoning, the difference in occurrence of a
stem-determinant /e/ with verbs of the above class (1) and others is only that /e/ occurs
more frequently with verbs of class (1) and (3), i.e., with verbs that have length as one
84 MORPHOLOGY

of their radicals than with verbs that have no length among their radicals. Since the
stem-determinant /e/ is less frequent than /a/, it would suffice to mark only those stems
for which /e/ occurs ; all verbs not so marked are assumed to have a stem-determinant
/a/. As a practical matter, since the common citation form both in the ancient word
lists and in modern dictionaries is the infinitive which, as will be seen presently, is also
determined by the stem-determinant, any dictionary entry in itself automatically in-
dicates the stem-determinant.
In Babylonian, the stem-determinant occurs in the following forms :

e a
infinitive sebe:ru sapa:ru
stative sebir sapir
adjective sebru sapru
personals prefixed teseb:ir tasap:ar
eseb:ir asapiar
suffixed sebre:ku sapra:ku
sebret saprat
infixed forms -tan- isteneb:ir istanap:ar
(perfect) -ta- istebir istapar
morphophonemic vowel (present) iseb:ir isap:ar
morphophonemic vowel (N-stem) nenduru nanduru
imperative (for verbs with /:/ as erub alik
radical) be:l sa:l

In summary, the stem-determinant is seen to occur

a) in the nominal forms (infinitive, stative, adjective, imperative), i.e., in the vocalic
morph of the nominal, and as the morphophonemic vowel of the imperative;
b) as the morphophonemically conditioned vowel of the cluster in the infixed forms
present, perfect, etc.;
c) in the personals (prefixed or suffixed, see 5.4.1 and [Link]).

In contradistinction to this statement of occurrence in Babylonian, no such stem-determinant occurs

in the Assyrian dialect in the above stated positions; /e/ occurs in Assyrian only with stems that in-
clude / :/ among their radicals. Hence, the occurrence of /e/ can be predicted for both dialects most
economically from its occurrence in the infinitive.

Note that the alternation a ~ e in the first syllable of the derived stems II (D) and III
(S) is not considered stem-determinant, but is dialectally conditioned, see 5.4.4 and
[Link].
Some further near-homophones which differ in one feature as well as in their stem-
determinant are here cited in order to illustrate the difficulty of establishing its distri-
bution on the basis of the feature that differs. The feature cannot be a contrast
voiced-voiceless, because besides the already cited pair Sebe:ru ~ §apa:ru, there are
the pairs
MORPHOLOGY 85

meke:ru maga:ru
[Link] zaka:ru
keze:ru kasa:ru
qebe:ru kaba:ru and kapa:ru
qererbu kara:bu

and moreover the doublets qere:bu ~ qara:bu and hepe:ru ~ hapa:ru.

Although all cited verbs have /r/ as one of their root-consonants, an /r/ in the root
alone does not determine variation, e.g., there is neither a doublet nor a pair for the
verbs pa{a:ru, maha:ru, qata:ru, kada:ru, [Link], etc.; moreover, verbs without /r/
have occasional variants with /e/, e.g., lebe:nu ~ labœ.nu.
For free variation a ~ e see also [Link].

[Link]. Stems with length among their radicals.

Let us first consider the stems that have /:/ in place of one of their radicals. Since
a stem consists of three consonants and one vowel, there are twelve mathematically
possible combinations of the three elements vowel, consonant (other than length),
length, into items of four members, when we limit the occurrence of length to one in
a stem, i.e., we only deal with the "simple weak" verbs and not with the "doubly
weak" verbs, in which length occurs twice. Of these twelve combinations, six do not
occur ( * R 1 R A : V , * V R J R 2 : , * R 1 : R 2 V , * V R 1 : R 3 , * : R A R S V , * V : R 2 R S ) because they
would be arranged to form clusters of three consonants (as discussed in 4.1.2, length
forms cluster with other consonants). The other six combinations all occur, but not all
belong to different morphemes : similarly to the distribution of stem-allomorphs dis-
cussed above in [Link], these six morphs combine into three morphemes:

1. { : V R 2 R 3 :R2VR3 ~ -RA:VR3}
2. { - R I R 2 V : }
3. { R x Y : R 3 ~ RIVR3:-}

In the first morpheme, the first two allomorphs are in complementary distribution,
one occurring initially before vowel, the other occurring postvocalically, just as the
allomorphs {purs- ~ -prus}. The third allomorph belongs to the same morpheme on
the postulate that the place of length before or after consonant is non-distinctive
(see [Link]); the distribution of this third allomorph however is limited to verbs which
form their present with a replacive (see [Link].1) to eliminate homophony between
the two tenses; see [Link]. In the third morpheme, the two allomorphs belong to one
morpheme on the same grounds of the non-distinctiveness of the place of length be-
fore or after a consonant; their distribution will be discussed presently. This mor-
pheme has no special postvocalic allomorph, since both CV:C and CVC: can occur
after vowel and initially.
As preliminary example, we can illustrate the three morphemes by the following
verb stems:
86 MORPHOLOGY

1. {:abl- ~ -:bal ~ -b:al} citation form (infinitive): abdlu (/aba:lu/) 'to dry up'
2. {-bni:} citation form (infinitive): banû (/banu:/) 'to build'
3. {ku:n ~ kun:-} citation form (infinitive): kânu (/ka:nu/) 'to be stable'
The three morphemes extracted correspond, in the traditional grammars, to the three
classes (1) "primae aleph", i.e., verbs which in other Semitic languages have as first
consonant one of the five laryngeals which have been lost in Akkadian (see [Link]);
also the "primae j", i.e., verbs which in related languages are written with a consonan-
tal /j/ as first radical; (2) the "ultimae aleph" and the "ultimae infirmae", i.e., verbs
which in the related languages have one of above mentioned "laryngeals" or a /w/
or /j/ as third radical; (3) the "mediae aleph" and the hollow roots, i.e., verbs which
in related languages have a "laryngeal" as middle radical, and verbs which in related
languages are written with a consonantal /w/ or /j/ as middle radical and in Akkadian
have a long vowel and two radicals.
In a recent treatment (Kienast, 1963) on a similar subgroup of the verb, with his-
torical reconstruction of the "biliteral" paradigm, i.e., the paradigm based on the
root without a consonant length, such verbs are typologically classified as PIS, PSI
(here morpheme 2), and PÜS (here morpheme 3).
[Link].1. Behavior of the radical /:/ in clusters. - In the majority of the stems with
/:/ as one of the radicals, this /:/ follows the phonotactic rules of Chapter 4 established
for clusters of consonant and /:/ (4.4). There is however a subgroup of these stems in
which /:/ follows the phonotactic rules for clusters of two consonants (CjCg or C^Cj)
other than length, i.e., with another affix /:/, of three consonants of which only one is /:/.
We may term these stems, adopting the traditional terminology, as forming a
"strong conjugation"; a "strong conjugation", in phonological terms, is predicated
on the phonotactic behavior of /:/ as if it were some other consonant. In the writing,
this /:/ that behaves like a consonant usually appears written with the <?V> sign, and
in intervocalic position this sign indicates that there is syllable boundary between the
two vowels (see 4.2).
Such subgroups are attested for Class (1), often as by-forms of Class (1) of the weak
verbs; e.g., the stem V:rur/ is attested in the present of Stem I as /ir:ur/ (conforming
to the phonotactics of /:/), and as /i:ar:ur/ (not conforming to the phonotactics of /:/).
The distribution of these by-forms remains to be determined. For Class (2), there are
the contrasting forms (present pi.) /imas:u:/ and /imas:a:u/, etc., see [Link].2. For
Class (3) there are the contrasting forms (present pi.) /irub:u/ and /ira:ubu/, etc., see
[Link].2. See also [Link].2 and [Link].4.

[Link]. Inflection of Class (1).

The inflection of a stem of Class (1) (length as first radical) is, according to the pho-
notactic rules: stem y/:abl/, inf. aba:lu
preterit i-:bal = /iibal/
present i-:b:al = = /ib:al/ (with reduction of cluster, see [Link].)
MORPHOLOGY 87

perfect sg. i-:-t-abl = Vi:tabl = /i:tabal/

pi. i-:-t-ablu = Vi:tablu = /i:tablu/
or: sg. i-:-ta-bal = /i:tabal/
pi. i-:-ta-blu = /i:tablu/
:
imp. sg. V abl = /abal/
pi. Λ/:ΪΛΛ-Ά. = /abla/

stem \ / : k u l / ; replacive /a/; stem-determinant /a/:

prêt. /i:kul/ pres. /ik:al/ pf. /i:takal/, /i:taklu/, imp. /akul/, /akla/

stem V : m i d / ; stem-determinant /e/:

prêt. /i:mid/ pres. /im:id/ pf. /i:temid/, /i:temdu/, imp. /emid/, /emda/

stem V:rub/; stem-determinant /e/ :

prêt. /i:rub/ pres. /ir:ub/ pf. /i:terub/, /i:terbu/, imp. /erub/, /erba/

Note that no phonotactic vowel intervenes when the first radical of the stem is {length}
(similarly when the second radical is {length}, see [Link]), because the infix {:} of the
present forms with the middle radical a cluster of only two consonants, which is per-
missible. The feature which distinguishes the present from the preterit in these stems,
unless there is a replacive as present tense morph, is solely the presence and position
of length, in such pairs as /i:rub/ ~ /ir:ub/, /i:mid/ ~ /im:id/, /i:bal/ ~ /ib:al/. It has
been stated in [Link] that the position of length is non-distinctive in a cluster formed
of length and consonant. The preterit/present contrast of the "weak" verb is the only
case in which this position is distinctive, since it is precisely the place of length before
or after the radical that differentiates the two tenses. Hence the allomorphs /:CVC/
and /C:VC/ are not in free variation in stems which do not form the present tense with
vowel replacive. See also [Link].

That another phonetic feature may have once accompanied this distinction based later solely on the
place of the length is evidenced from the spelling common in the OB period for the present forms of
such stems. In this practice, the present is written either i-ru-ub, i.e., not differentiated from the pre-
terit, or written i-ir-ru-ub, i-im-mi-id, etc. Since the writing habits of the cuneiform system never use
an initial vowel syllable to indicate length of the vowel, some other phonetic feature must have been
indicated by these spellings, which are paralleled by the writing V-VC before consonant in monosyl-
lables, e.g., α-α/for /a:l/, e-edfor /e:d/, etc., and other words, e.g., i-ib-bu-ú-um, etc., see Reiner, 1964.
A contrast ir-ru-ub ~ i-ru-ub of course would have been sufficient to indicate the presence of length
in the present tense; however, no OB spelling ir-ru-ub or the like occurs. In stems with vocalic re-
placement, where even a spelling such as i-ka-al would differentiate the present from the preterit i-ku-
ul, the V-VC initial spelling is likewise frequent, such as i-ik-ka-al, i-ip-pa-al, etc., and so it has to be
assumed that the writing not only strives to exclude ambiguity but also to indicate some phonetic
feature. Some verbs belonging to Class (1) which form the present with infix length in Babylonian,
form their present in the Assyrian dialect with infix length plus vocalic replacement: epe:su, [Link]
(prêt. e:pus, e:rub, pres. ep:as, er:ab).

[Link]. Class (2).

Class (2) has length as third radical and corresponds to the class "ultimae aleph"
88 MORPHOLOGY

and "ultimae inftrmae" of the traditional terminology. A subgroup usually set apart
from this class is composed of verbs with a "strong aleph" as third radical ; these can be
described in the terms here established as stems in which the third radical, /:/, always
functions as syllable boundary ?, see [Link].1. The stem of class (2), of the shape
{-RiRjjV:}, has no initial allomorph CVCC; such an allomorph, with /:/ in the position
of the final consonant, has already been assigned to the stem-morpheme of class (3),
see above [Link] ; in place of such initial allomorph, the imperative masc. has a pattern
CVCV, and a pattern CVCV:, instead of the morphophonemically expected *CVC:V
(which pattern is again reserved for Class (3)), in the plural, in the feminine and in the
ventive, before vocalic suffix as in the attestations:

sg. bi-ni fem., pl., ventive: mu-na-a

hi-di, hu-ú-du hu-di-i
bi-ri
hi-pi hi-pa-a
hi-ri
ti-bi ti-ba-a
§i-ti si-ti-i

It appears from the examples that length occurs only with vocalic endings, to be pho-
nemicized presumably as /muna:/, /ljudi:/, etc., while the imperative singular is to be
phonemicized as /hidi/, etc., either on the analogy of the imperative pattern /pursa/ ~
/purus/, i.e., disyllabic singular with length dropped in absolute final position (see be-
low), or taken as formed from a shortened stem /bin/. To be noted is that no impera-
tive *bin:i or the like occurs as feminine, nor a singular *bi:n or the like (since both pat-
terns occur with Class (3)).
The inflection of stem /-bni:/ is as follows:

preterit \/i-bni: = /ibni/

present \/i-bn:i: = /iban:i/
perfect pi. i-b-t-an-u = /ibtanu:/
sg. i-b-t-an-i = /ibtani/
imp. pi. [stem /-tbi:/] = /tiba/ or /tiba:/
sg. /bini/

Stem /-mnu:/
preterit Vi-mnu: /imnu/
present /iman:u/
perfect pi. /imtanu:/
sg. /imtanu/
imp. pi. /muna/ or /muna:/
sg· /munu/
MORPHOLOGY 89

Stem /-mia:/
preterit -\/i-mla.: /imla/
present \Z\-mV.a. /imal:a/
perfect pi. /imtalu:/
sg. /imtala/

To be noted is the phonotactic feature that length drops after vowel, in all above cited
forms; the perfect, since the stem has no CVCC allomorph, is again formed either from
a shortened stem /ban/, or after the analogy of other stems (iptarsu ~ iptaras). In
Assyrian and OB, the plural of the perfect is -\/ibtaniu, i.e., vocalic endings are suffixed
to a stem ending in -i (as is also the case with other vocalic endings, as in yliqia,
\/ired:iam), and it is only in this position that the morphophoneme /:/ of the stem is
overtly marked as syllable boundary (/ibtani:u/, while with a suffix /i/ (2 fem. sg.),
/:/ is to be deduced from writings such as si-ti-i, tab-ni-i, etc. (/siti:/, /tabni:/).
[Link].1. A suggested alternate analysis. - The typology that has been followed here
is that of the GAG (§ 105), where the verbs "ultimae infirmae" are considered having
two radicals and a long final vowel. The recent proposal of Kienast, 1963, 148f., to
omit final length of the stem and consider these verbs of the type PSI, i.e., two conso-
nants and a final short vowel, eliminates the need to assume that length drops after
vowel in the inflection (whatever the historical antecedents of this type may be), and
would permit us to assign the stem type CCV? - and its inflected forms - unequivocally
to the type CCVC and thus raise no conflict in finding a contrast [:] and [?] in the verb
class (2) since there is no phonological contrast elsewhere between [:] and [?].
[Link].2. "Strong conjugation". - While the spelling with ?V-sign is optional in
Class (2) stems which do not have vocalic replacement as tense-morph, there is a sub-
group of Class (2) stems in which the radical /:/ is always written in its allographic
form with <?). This subgroup must be described by a list. The argument might be
brought that verbs which regularly select the allograph <?) and those which select /:/
or one of its allographs except <?) form minimal pairs, and that thus the procedure
which resulted in the grouping of [?] as allophone of {:} must be reconsidered. How-
ever, the procedure is insofar justified as the contrast between the allophones in the
subgroup is restricted to the root, i.e., to a morph which does not occur in the utter-
ance by itself, while inflected forms (i.e., the root plus a vocalic morph) contrast not
only by means of the consonantal allophone, but by their syllable structure. Where
[?] forms syllable boundary, the inflected - or derived - forms have one syllable more
than those of stems in which the allophone does not occur; in other cases, the vocalic
morph supplies the contrast. For a general survey of stems in which /:/ forms syllable
boundary, see [Link].1. Example:

Class (2) Subgroup

inf. masu: masa?u /masa:u/
prêt. sg. imsi imsu? /imsu:/
pi. imsu: imsu?u /imsu:u/
90 MORPHOLOGY

pres. sg. imas:i imas:a? /imas:a:/

pi. imas:u imas:a?u /imas:a:u/
Due to the small number of stems of this subgroup (nine are cited in GAG, the tenth
there cited is not attested since the cited forms must be given another reading), it is
impossible to establish which forms are contrasting in one and the same dialect.

[Link]. Class (3).

Class (3) has length as second radical. As this length appears in one of the stem
allomorphs (see [Link]) after the stem vowel, this class has been described as "hollow
roots", i.e., "roots" which have a long vowel as middle radical. Since the cluster/V:C/
is permissible in any but preconsonantal position (see [Link]), this stem has no posi-
tionally determined allomorph, only one which is due to the free variation of the place
of length, {CV:C ~ CVC:}. Both of these can occur either as initial or as postvocalic
morphs; it is the second which occurs as the prevocalic allomorph. Since there is no
phonological reason for such a distribution in prevocalic position, the distribution is
morphemically conditioned, and carries the functional load of differentiating tenses.
The inflection of Class (3) is as follows :

Stem /ku:n/ preterit sg. i-ku:n pi. i-ku:n-u

present i-ka:n i-kun:-u
perfect i-k-t-u:n i-k-t-u:n-u
imperative ku:n kun:a

Two functionally determined patterns occur that are to be noted :

(1) the present singular is formed with a replacive (which is, as in all replacive
morphs, /a/), because, if it were formed with an infix, the resulting present form would
be *\/i-ku::n, which would yield phonological /iku:n/ and thus would be homopho-
nous with the preterit.
(2) In the plural of the present and in all vowel-suffixed forms of the present, a pre-
vocalic allomorph {kun:-} occurs and thus these forms are in contrast with preterit
forms where the prevocalic allomorph does not occur, thus excluding the free varia-
tion of the place of length (see [Link]), which here carries a functional load. Since
vowel-suffixed forms of the present tense thus form sufficient contrast with similar
forms of the preterit, no replacive occurs in these forms.
In the Assyrian dialect, the present singular is distinguished from the vowel-suffixed forms by means of
a vowel /a/ appearing between the stem vowel and the last consonant, as in /iku:an/, etc., for a parallel
see 7.1.3.

The prevocalic allomorph /CVC:/ is a productive pattern for the differentiation sing-
ular-plural in the present, as can be seen from its frequent occurrence with stems for
which this distinction need not carry the functional load of singular-plural or preterit-
present differentiation; note NB spellings such as i-ru-ub-bu for /ir:ubu/, NA e-kul-lu
for /ek:ulu/ (prêt. /e:kulu/), and several more. This is possible because once a contrast
MORPHOLOGY 91
prêt, (pl.) /iku:nu/ ~ pres. /ikun:u/ operates, the new present-tense morpheme for
verbs with two consonants can spread from the /ku:n/-type to other types as well, as
in i-ru-ub-bu, i.e., /irub:u/ where the present infix appears, instead of after the second
stem-consonant /r/, after the second consonant of this particular inflected form, /b/,
just as in /ikun:u/. The present /irub:u/ seems to contrast with a prêt. /i:rubu/ and not
with a - hitherto not attested - prêt. */iru:bu/.
Inflection of stem /si:m/
preterit sg. i-si:m pi. i-si:m-u
present i-sa:m i-sim:-u
perfect i-s-t-i:m i-s-t-i:m-u
imperative si:m si:m-a (the free variant *sim:a is not attested)
[Link].1. Class (3) stems with stem vowel /a/. - The inflection of all "hollow roots"
with stem vowel /u/ or /i/ as described above shows that the present formation with
vocalic replacive provides sufficient tense contrast. However, there are stems of Class
(3) with a stem vowel /a/, and for these the replacive would not effect any change be-
tween present and preterit singular. This numerically small group of "mediae ä" is
best treated with the group of "mediae aleph".
The verbs "mediae aleph" are usually considered as having a consonantal middle
radical, namely a glottal stop. Only in a few verbs however is the glottal stop written,
and then only in a limited number of forms. Mostly, no consonant is written apart
from the first and third radical; the etymological explanation is that the glottal stop
assimilates to the preceding or following vowel and combines with it to yield a long
vowel. The stems in which the middle consonant is an etymological laryngeal more-
over have, as the result of this environment, a long vowel [è] in the "contracted" forms;
descriptively of course the quality of the vowel, [ä] or [è], can be considered the stem
determinant, see [Link]. Any stem which exhibits a long vowel is of course suspect
of having length as one of the radicals, and in fact these verbs with glottal stop as
middle radical can be described as having length as middle radical, and their inflection
as parallel to that of Class (3) verbs with stem vowel /u/ or /i/. Such a description will
nevertheless fail to disclose the tense contrast preterit-present overtly marked in all
verbs. The traditional grammars indicate a tense contrast in the reconstructed forms
(e.g. *ib?el versus present *ibe??el) and in the attested forms by the purely graphic
means of contrasting preterit ibël «*ib?el) with present ibêl {(*ibe??el). The recon-
structed present form indicates a present tense infix {:}, since *ibe??el can be rewritten
as/ibe:el/or*ibe?:el,but the notation ibël versus ibêl is purely graphic, since the macron
and circumflex are only meant to be historical notations (see 4.1.) and it has never been
assumed that these represent two degrees of length realized differently phonetically.
I submit that the notations ibël and ibêl have to be rewritten so as to indicate an
actual difference between the two tenses. The notation ibël is to be rewritten /ibe:l/),
and the notation ibêl which is etymological in that it indicates that the vowel length
resulted from the contraction of two syllables, is in fact transposable into a phonolo-
92 MORPHOLOGY

gical notation /ibe:el/, which thus indicates that the contraction of two syllables did not
in effect take place. What occurs in the present singular of these verbs is not a long
vowel, but two contiguous identical vowels which are realized as two syllables. The
phonological notation with /:/ separating these two identical vowels may be phoneti-
cally realized as a glottal stop or as a glide, which, according to 4.2.1 are the allophones
of morphophonemic {length}.
In the verbs "mediae aleph" the phoneme realized between the two vowels can be
considered etymologically the glottal stop or the laryngeal attested in related lan-
guages, but in the verbs "mediae ά", where no such etymological consonant was ever
present, this phoneme must be considered morphophonemic, functioning as a con-
sonant which forms syllable boundary between the two vowels. In these verbs, the
preterit and the present singular also effectively contrast as /isa:m/ and /isa:am/, etc.
The phonological proof for this purely structurally elicited contrast lies in the writ-
ing. While the preterit singular forms are often written with a CVC syllable as e.g.,
i-bel, i-sam, i-ram, the present singular is never written with such a CYC syllable, only
with CV-VC syllables. (Rare exceptions, in which moreover it is not always certain
that a present tense has to be assumed, are i-bar, ta-bar ; for references see AHw., s.v.
bâru(m) I.)
That the syllables CV-VC have to be read as two syllables and not one is shown not
only by the frequent "piene" writing, i.e., a V-sign occurring between them, but also
by spellings which tend to indicate a consonantal syllable boundary by the use of a
syllabic sign which has as consonant element one of the allographs of length : glottal
stop <?) or <j> (not <w> because the latter does not occur between the vowel cluster
a-a or e-e, see [Link]).

Examples : a) with V-sign : i-da-a-as

a-ba-a-sú (for /aba:as/ rather than /abas:u/)
ta-pa-(a)-as
b) with ?V-sign : ta-na-?-ad
c) with jV-sign: i-sa-ia-ad
a-sa-ia-ah
[Link].2. "Strong conjugation". - In the subgroup of Class (3) stems which regularly
selects the allograph <?>, the <?) forms syllable boundary and thus the inflected forms
pattern like those of any stem in which there is a consonant other than /:/ in this posi-
tion; in the present plural, the stem is disyllabic as in [Link]-u and not monosyllabic
as in i-kun -u, and consequently the present singular is formed by infix /:/ and not by
vocalic replacement, as in [Link] and not as i-ka:n\ for example, present sg. ira?ub,
pi. iraiubu, present sg. alaiut, etc. For a similar subgroup see [Link].2; for a general
survey see [Link].1.

[Link]. Divergent verb stems.

Previously treated verb stems with divergent inflection could be successfully treated
MORPHOLOGY 93
parallel to the regular inflectional pattern on the basis of phonotactics applicable to a
radical /:/. There are three further groups of verb stems with divergent inflection which
cannot be satisfactorily described by phonotactic alternations of the stem in its in-
flected forms. These stems are those which have as first radical /w/ in the citation
form (infinitive) in the OA and OB periods, i.e., have an etymological /w/ as first
radical, those which have as first radical /n/, and those which have identical consonants
as second and third radicals. The two first type stems are related to the previously
discussed "weak" stems with /:/ as one of their radicals, since in many of their in-
flected forms length occurs instead of the etymological or phonological first radical
/w/ or /n/. However, the phonotactic distribution of length is different from that of the
discussed "weak" verbs.
[Link].1. "Primae w". - Verbs with an etymological /w/ are given the name "primae
w". In the derived stems, they inflect like the stems with first radical /:/, for example
derived stem III (§) usœ.bal (usabial, see [Link]) and prêt, usœ.bil. In the simple stem,
in the preterit and the present the stem takes the personals of Set 2 (with vowel u, see
[Link]) but in the perfect the personals of Set 1 (with vowel i); the perfect seems to
have a suppletive perfect stem with first radical /t/ which also often occurs in the im-
perative and infinitive.
The preterit and present of the simple stem are inflected according to a pattern typ-
ical of this stem class. Apart from the personals of Set 2, this pattern involves a re-
placive in the present, which contrasts with the stem vowel /i/ of the preterit; it also
involves the optional elision of this stem vowel before a vocalic suffix, a feature unique
for the preterit tense. This elision is optional in the SB dialect, but regular in some
other dialects.
Example: stem V : bl, infinitive aba:lu (OB baba:lu) and taba:lu.
prêt. sg. prêt. pl. pres. sg. pres. pi. perfect sg. perfect pi. imp. sing.
ubil ublu ~ ubilu ub:al ub:alu bil
suppletive stem: itbal itbalu tabal
or it:abal it:ablu
Verbs which in earlier dialects have an initial /w/ in their nominal forms (infinitive,
participle, stative) but do not inflect according to the above described pattern, are not
considered here. Since their inflection is according to the regular pattern of Class (1)
as described in [Link], a fact which their SB citation form - infinitive - indicates, they
need not be singled out here.
[Link].2. "Primae «".-The name "primae η" was given to a stem class with irregular
inflection on the basis of cognates in other Semitic languages and of some of its nom-
inal and inflected forms which have an /n/ in the position of the first radical. However,
other inflected forms of this stem class have no /n/ in this position, and, in order to
account for the absence or the assimilation of /n/ in these forms, special rules, affecting
this stem class only, have to be established.
In this classification of the verb stems, I have preferred to assign this class too to
94 MORPHOLOGY

the stems which have /:/ as one of their radicals. The presence or absence of ja/ in
some of the inflected forms may be interpreted as the dissimilation of the morpho-
phoneme /:/ of the stem in the environment of another /:/, namely in the forms with
an infix /:/ (Stem II and present of Stem I).
The reasons for classifying this stem class as having /:/ and not /n/ as first radical
are the possibility of simpler morphophonemic statements. If the first radical is taken
as /n/, we have to devise morphophonemic rules for /n/ which differ from the morpho-
phonemic rules applicable to /n/ (see 6.1.2.2d). If, on the other hand, the first radical
is taken as /:/, we can use the properties of /:/ discussed above ([Link].1) to account
for forms that are seemingly irregular, and we have to account for the presence of
/n/ only in some of the inflected and derived forms.
Inflection of "primae η" :
If we assume that the first consonant of the stem is /n/, "irregular" forms will be the
imperative, in which no /n/ appears, and the preterit of Stem I and inflected forms of
Stems III and IV, in which the first consonant of the stem is in contact with the second
consonant. Normally, as described in [Link], /n/ does not assimilate to a following
consonant within one morph, only at morpheme boundary ; the latter case is present
in the inflection of the perfect, where /n/ is followed by the tense morph {t}, to which
it assimilates as /t:/. In the preterit and in the derived stems, the /n/ of the stem like-
wise assimilates to the following radical.
Examples:
infinitive naka:su, stem vowel /if ;
infinitive naqa:ru, stem vowel /u/; replacive /a/;
infinitive naba:hu, stem vowel /u/.
prêt. present perfect [Link]. imp. pi.
ik:is inak:is it:akis ikis iksa
iq:ur inaq:ar it:aqar uqur uqra
ib:uh inabruh it:abuh ubuh ubha
Derived stems : prêt. pres. imp.
II unak:is unak:as nuk:is
III usaqñr usaq:ar suq:ir
IV/3 ittankis [?] [?]
Note, on the one hand, the "assimilation", i.e., non-occurrence of /n/ in the preterit of
Stem I and in Stem III, where the forms correspond to those of verb stems with first
radical /:/, the non-occurrence of /n/ in the imperative which thus is likewise parallel
to that of stems with first radical /:/ except for the quality of the first vowel, which is
not that of the stem-determinant, but that of the stem-vowel, and on the other hand
the presence of /n/ before consonant in contact in the derived stem IV/3.
With /n/ in first position of the verb-stem, it would be very difficult to set up the
stem-allomorphs without resorting to extremely complicated morphophonemics. Of
a stem morph *{nkis ~ niks}, e.g., the imperative needs separate morphophonemic
MORPHOLOGY 95

treatment; of a stem morph *{k:is ~ :iks} which is suitable for the preterit and the im-
perative, the forms in which there is an /n/, in initial position as in the infinitive
(naka:su) and the derived nominal forms (adjective naks-, etc.) or in intervocalic po-
sition as in the present of Stem I, and in Stem II, cannot be predicted. In the latter
forms with /n/, the /n/ can be regarded of course as some type of augment or supple-
tism, as initial /t/ in the stems with /w/ (see [Link].1).
If we assume that the first consonant of the stem is /:/, the presence of /n/ in the
present of Stem I (e.g., inak:is) and in Stem II (e.g., [Link]) can be accounted for by
assuming a dissimilation of /:/ when another /:/ is infixed:

-\/i-:ak:is -» inak:is \/u-:ak:is unak:is

The presence of /n/ in the prêt, of Stem IV/3 on the other hand can be ascribed to the
dissimilation of long consonant, see [Link] sub (c).
Consequently, the /:/ of this class of stems, and its realization as /n/, behaves like
the /:/ which forms syllable boundary, see [Link].1. By forming syllable boundary, this
/:/ > /n/ preserves the heterosyllabic pattern of the preterit - present contrast discussed
in 5.4.4.
The descriptive analysis of /n/ as realization of /:/ in certain environments parallels
the historical view regarding this stem class, according to which the initial /n/ of these
stems is an augment.
The presence of an /n/ in intervocalic position to preserve a heterosyllabic pattern
may also be illustrated by the derived stems of the type 1/3, II/3, or Gtn, Dtn, etc.
It has been noted above in [Link] that the derivational infix of these stems can be seg-
mented either as {tan} or as { t a + : } . If a segmentation {tan} is accepted, the same
problem of assimilation of /n/ as in the case of /n/ as first radical is posed, although
in forms with "assimilation" the assimilation occurs at morpheme boundary as
in /iptar:as/ < V i-p-tan-ras, since, when the {tan} infix occurs before a consonant
cluster, the cluster is not resolved as /naCC/ as for other clusters of three conso-
nants (see 4.4.2), but by the loss of /n/, as /CC/, as in /uStapris/ < V u-S-tan-pris.
The segmentation { t a + : } on the other hand accounts for all inflected forms, except
for the presents [Link], [Link], ustanapras, [Link], in which /n/ appears
intervocalically. If this segmentation is accepted, the presence of /n/ can likewise be
accounted for as due to the dissimilation of /:/ in the presence of another infix /:/, and
functioning as a syllable boundary consonant to preserve the heterosyllabicity of the
contrasting tenses.
[Link].3. " M e d i a e geminatae". - Stems in which the second and third radical are
identical consonants are traditionally called "mediae geminatae". When the two iden-
tical consonants come into contact in the inflection, they behave as a cluster of two
consonants, and not as a long consonant, as pointed out in 4.1.2ff. It is only in some
derived nominal forms that the sequence of two identical consonants does not behave
as a cluster of two consonants but as a long consonant; for these, see the examples
cited in 7.1.4 and 7.2.
96 MORPHOLOGY

Consequently, derived forms inflected for nominal categories (gender, number)

either do not occur, or they occur either as CVCC or as CVC: depending on the verb,
or the forms CVCC and CVC: occur in free variaton.
Example: stem {-dbub ~ dubb-}
prêt. pres. perfect sg. perfect pi. imp. sg. imp. pi.
i-dbub i-dab:ub id:abub id:abbu dubub dubba
(Note that the consonants phonemicized as double are always written so and not as C:
for which the doubling may be omitted in the writing.) Statives (see [Link]) belonging
to such stems are, of the type CVCC, garir, ?alil, melil, laziz, etc.; of the type CVC:,
da:n, etc. ; and in free variation e:d and edud, e:z and eziz, etc.
Imperative does not occur for the stem /dnin/ (stative da:n), and for all those stems
which only have CVC: type statives.
[Link].4. Recapitulation of the realizations of /:/. - The above analysis of subclasses
of the verbal inflection, namely of the "strong conjugation" of classes (1), (2), and (3),
see [Link].1, [Link].2, [Link].2, and of the "primae η", see [Link].2, necessitates in
the morpheme list a specification of the stem consonant /:/ as
(a) non-clustering (weak verbs),
(b) clustering and forming syllable boundary between vowels, often written with the
sign <?V>,
(c) clustering and realized as /n/ between two vowels and initially and when another
/:/ is affixed to the stem.
All three realizations of /:/ can be noted from the citation form (infinitive, see
[Link]), since the infinitive is disyllabic when the /:/ is non-clustering, trisyllabic and
includes the syllable boundary graph <?V> when /:/ is clustering, trisyllabic and ap-
pears with initial /n/ when the /:/ is clustering and realized as /n/.
[Link].5. Residue. - There is a residue of irregular verbs that cannot be here treated
in the framework of a structural restatement. They are not attested in all, or even in
a sufficient selection of, inflected forms in any one dialect, and while a combined ana-
lysis of the forms appearing in various dialects may lead to a historical reconstruction,
it will not help to establish the classes to which they belong within one dialect. More-
over, the scanty attestations of most of these irregular verbs prevent a reliable philo-
logical interpretation and hence their assignment to one or another root. Witness to
these difficulties are the statements presently found in current grammars which may
at times be shown to cite non-existent forms, which owe their existence in the gram-
mars or léxica to a misreading, wrong cutting or misinterpretation of the texts.
For some of the residual irregular verbs and the rarer stems (§D, R, etc.), and for
the quadrillerais, a paradigm with all variant forms may be given. Other verbs belong
to the class of "doubly weak" verbs, i.e., verbs which have length in two positions in
their root and thus only one consonant at the most can be unambiguously elicited;
their patterning greatly depends on the interpretation of the syllabic writing in deciding
where two identical vowels form two syllables and where only one.
MORPHOLOGY 97
Nevertheless, it may be posited that these verbs will be found to belong to one or the
other pattern discussed under verb inflection; the present is formed with infix or re-
placive, the morphophoneme length is realized as length, glottal stop, or another allo-
phone such as [w] or [ j] ; if the preterit and the present seem to be homophonous, the
present will prove to have one syllable more; etc., etc.
For each form, each attestation has to be checked and sorted out according to dialect
distribution, so that we do not fill in blank spaces in a paradigm by forms reconstruct-
ed on a normative basis.
Similarly, since quadriliteral verbs form a small subclass and not all of them occur
in a great enough variety of inflected forms, and even such inflected forms as do occur
are distributed among various dialects, I will not attempt to construct a synchronic
picture of their inflection (a monograph devoted to the quadriliteral verb is the well-
known Heidel, 1940, to which further evidence has been added in some articles by von
Soden, see the lit. cited in GAG, § 108b).

5.4.6. Stative and imperative

Two morphological categories share in both the nominal and the verbal inflection,
the stative and the imperative.

[Link]. Stative.
The stative is a noun (i.e., is derived from a root by a vocalic morph {a} typical of
nominal derivation) which inflects for person and mood. The personals (actor-affixes)
are all suffixed and occur in two forms according to the stem determinant with /a/ or
/e/ as first phoneme.
Sg. PI.
masc. fem. masc. fem.
3. {-0} {-a:t ~ e:t} {-u} {-a}
2. {-a:ta ~ e:ta} {-a:ti ~ e:ti} {-a:tunu ~ e:tunu} {-a:tina ~ e:tina}
1. {-a:ku ~ e:ku} {-a:ku ~ e:ku} {-a:nu ~ e:nu} {-a:nu ~ e:nu}
When the suffix is zero, the phonotactic rules of 7.2 obtain. The relative ranks and
distribution of mood suffixes is as described in [Link] ; for this reason, only the follow-
ing forms inflected for mood occur (the mutual exclusivity rules of 5.4.2 would tolerate
a ventive of 1. sg. and pl., but no such forms are attested):
Ventive Subjunctive
3 sg. pars-am pars-u
3 pi. masc. parsu-ni(m)
3 pi. fem. parsa-ni(m)
[Link]. Imperative.
The imperative is a verb stem inflected for the nominal categories gender and num-
ber (not for case). The stem is determined as the prevocalic allomorph of the verb
98 MORPHOLOGY

stem in Stem I, by a vocalic morph {a-i} in Stem IV, by a vocalic morph {u-i} in other
derived stems, with a morphophonemic vowel resolving consonant clusters for some
of the secondarily derived stems, so that the vocalic morph has the shape {u-a-i}. It
is preferable to analyze imperatives of stems derived from Stem I by an infix {t} or
{ta+:} (for the latter see [Link].2), as mithas (stem mahs ~ mhas), kitpad (stem
kupd ~ kpud) and ritap:ud (stem rupd ~ rpud), pitarías (stem purs ~ prus, replacive
a), as derived by means of an infix {it} which recurs in other nominal forms (see 5.6.2),
than as formed with a vocalic morph {i-V} (V = stem vowel or replacive). In the
latter analysis, the /i/ of the first syllable would have a parallel in the imperatives of
the subclass discussed in [Link].1, e.g. pilah, rikab, etc. For the imperative of
irregular verbs see the discussion under the respective subgroups.
The gender and number affixes are suffixes, {i} feminine singular, {a} (masculine
and feminine) plural; compare the gender suffix {i} and the second person plural
suffix {a} of the personals cited in 5.4.1. For a generative statement see the Appendix.

5.5. P R O N O U N S

The grammars of Akkadian distinguish three types of pronouns : independent, suffix-

ed, and prefixed pronouns. The "prefixed pronouns" are the actor affixes (personals)
of the verb inflection which have been discussed in connection with the verb in 5.4.1.
The "suffixed pronouns" will be treated under "personal suffixes" in 5.5.3.
The independent pronouns are subdivided into "personal pronouns", "anaphoric
pronouns", "possessive pronouns", "demonstrative pronouns", "determinative pro-
nouns", "interrogative pronouns", "indefinite pronouns", "relative pronouns", and
"number pronouns (Zahlpronomina)". Only the "personal pronouns" and the "ana-
phoric pronouns" are said to have a special inflection; the others are said to be in-
flected "if at all, like nouns". The personal and anaphoric pronouns form a supple-
tive paradigm for three cases which will be described in 5.5.3.

5.5.1. Inflectional classes to which "pronouns" belong

The independent pronouns, i.e., free forms, do not constitute a separate, pronominal,
inflectional category with the exception of the personal pronouns mentioned above.
Some belong to the inflectional class of substantives, some to that of adjectives, some
are indeclinables, and some, at least in vestigial form, preserve traces of a separate
inflectional category, which may be termed pronominal, and which thus will be here
treated.

[Link]. Substantives.
To the inflectional class of substantives belong the "interrogative pronouns"
mannu, minu (/man:u, mi:nu/), and in some dialects the "indefinite pronoun" mimmü
(/mim:u:/).
MORPHOLOGY 99

[Link]. Adjectives.
To the inflectional class of adjectives belong the "interrogative pronoun" ajjû
(/aiu:/) and the "demonstrative pronouns" ullû, ammiu (/ul:u:/, /am:i:u/); for the pos-
sessive and demonstrative pronouns see below, 5.5.2.

[Link]. Indeclinables.
To the indeclinables belong the "indefinite pronouns" mamman (mammà) and
mimma (/mam:a(n)/, /mim:a/), the "determinative pronoun" sa (of the OAkk. period),
the MA and NA "number pronoun" jamuttu/jamattu (/iamut:u ~ iamat:u/), and a few
pronouns that inflect only for concord: SB annanna, fem. annannïtu (/an:an:a,
an:an:i:tu/), the compound (see the personal pronouns in [Link]) attamannu, fem.
attimannu (/at:aman:u, at:iman:u/), and the NB pronouns *agâ, fem. agât(u) (/aga:a/,
/aga:t/); agâsû, fem. agâsija (/aga:su:, aga:si:a/); pl. agâsunu (/aga:sunu/), fem. not
attested; agannû, (fem. sg. not attested), pl. agannûtu, fem. pl. agannâtu/agannêtu
(/agan:u:, agan:u:tu, agan:a:tu ~ agan:e:tu/).

5.5.2. Pronominal inflection

The "possessive pronouns" and the two "demonstrative pronouns" annû and annummû
belong, in the SB literary dialect, to the class of adjectives; however, in the earlier pe-
riods, in OB, they form a separate inflectional class which is indicated by some forms
although the process of transfer into the adjectival category is seen to have begun.
This inflection presents a pattern of agreement marked -n "masc. sg."; -t "fem.
sg."; -n "pi." which has been analyzed by Greenberg, 1960, 317-21, as demonstrated
by him "particularly common with, and perhaps originally confined to demonstra-
tives and the numeral one." In the mentioned article Greenberg could adduce the
Akkadian numeral "one" istën "mase.", istët "fem." as containing the singular por-
tion of the pattern of agreement. In the pronominal class, the tripartite n/t/n pattern
can however still be observed, perhaps as an operative pattern of agreement, perhaps
only as surviving frozen forms, since in OB the shift to the class of adjectives was al-
ready taking place, and this shift was accomplished in Assyrian by the time of the OA
texts.
Forms attested with pronominal inflection in OB :
demonstrative annûm fem. annïtin pl. (masc.) annütun, annütin
(/an:u:m/) (/an:i:tin/) (/an:u:tun/, /an:u:tin/)
(fem.) anniätin (/an:i:a:tin/)
annummûm annimmûtin
(/an:um:u:m/) (/an:im:u:tin/)
possessive ja:um fem. juttun pl. (masc.) juttun (/iut:un/)
(/iut:un/)
jattan (acc.) (fem.) jattun (/iat:un/)
(/iat:an/)
100 MORPHOLOGY

kûm (/ku:m/) pl. kuttun (/kut:un/)

CTCL 18 145)
nûm (/nu:m/) (fem.) nu-ut-ti-in (/nut:in/) (gen.)
(A 3520:6, unpublished)
§ûm (/Su:m/) Suttun (ß[Link]/)
(AfO 18 66:10)
sâtun (/§a:tun/)
(see Goetze, Sumer, 14,48)
Compared to the n/tjn pattern observable in Afro-Asiatic languages, the traces of this
pattern of agreement in OB are restricted to the plural, while for the feminine the -t
element is followed either by an -n, or by the nominal ending -m.
Here may be cited, apart from the numerals masc. istën, fem. istët, 'one', mase.
iStënu, fem. isîëtu, 'first', also frozen forms like kilallun, fern, kilattun or kilattan,
'both'.
In NB, the demonstratives masc. /aga:a/, fem. /aga:t/ also seem to attest the fem.
-t, but this pronoun is no doubt a late borrowing from Aramaic.

5.5.3. Personal pronouns

The personal pronouns form a suppletive paradigm. The "anaphoric pronoun" men-
tioned in 5.5 supplies the 3rd person and is in the nom. homophonous with a demon-
strative pronoun sû ßu :/, see 5.5.4.
The personal pronouns are subject to verbal concord: one suppletive set functions
as subject of the verb or as predicate of the nominal clause, this is called the nomina-
tive case; when the pronoun is the direct or indirect object of the verb, a paradigm
which differentiates the direct and indirect object case by contrastive suffixes ("case
endings") obtains. In the three persons of the singular, the free form contrasts with
the bound form which occurs after the verb, and has been given the name "suffixed
pronouns" (Pronominalsuffixe) or personal suffixes; in the three persons of the plural,
free and bound forms do not contrast except by position. Free forms occur in the
direct object "case" when they are the direct object of a transitive verb and after a
preposition; in the indirect object "case" if they are the indirect object of a non-transi-
tive verb and after the preposition ana when this latter is a paraphrase of the object of
a non-transitive verb.
In the OA dialect only the subject and object paradigms are distinguished, i.e.,
there is no distinction between direct and indirect object. In the literary dialect both
forms occur but they are in free variation.
The bound forms of the singular direct object pronoun also occur after substantives
and prepositions as another paradigmatic set - differing in the three persons of the
plural from the object pronoun occurring as free form and after verbs - which may
be called personal suffixes. After a substantive, they function as possessives, after a
preposition, as referential object.
MORPHOLOGY 101

[Link]. Paradigm of personal pronoun, free form, subject case:

Sg. PI.
masc. fem. masc. fem.
1. ana:ku ana:ku ni:nu ni:nu
2. at:a at:i at:unu at:ina
3. su: si: Sunu sina
[Link]. Paradigm of personal pronoun, free form, object case:
Direct object Indirect object
Sg. PI. Sg. PI.
1. ia:ti ni:a:ti ia:si ni:a:§i(m)
2. ka:ti (ka:ta) kunu:ti ka:si kunu:si(m)
3. su:ati (sa:ti) sunu:ti su:asi(m) (sa:si(m)) [?]
NB: For the object case, the gender distinction is not attested, except for a fem. pl.
sina:ti. As noted above, the indirect object forms do not occur in OA, and occur in
free variation with the direct object form in the literary dialect. Note also that not all
attested forms are here cited, they differ a) in the quality of the final vowel (e.g., ka:ta,
ka:ti, sa:su, sa:si, ni:nu, ni:ni) or b) in the first vowel (e.g. su:ati, sa:ti).

[Link]. Paradigm of personal suffixes:

Sg. PI.
masc. fem. masc. fem.
1. i ~ ia i ~ ia -ni -ni
2. -ka -ki -kunu -kina
3. -su -sa -sunu -sina
This set of personal suffixes occurs after substantives and prepositions (see 5.7). The
two allomorphs of the 1. sg. are distributed acccording to non-subjoined case ("nom-
inative-accusative", see [Link]): /i/, and subjoined case ("genitive") and adverbial
case (see [Link]): /ia/. After prepositions, only the subjoined case occurs.

[Link]. Personal pronoun, bound form.

It is not necessary to give a separate paradigm for the bound form of the personal
pronoun since, as mentioned in 5.5.3, the three persons of the plural are identical with
the free form of the personal pronoun, and the three persons of the singular are iden-
tical with the personal suffixes, if we add to the first sg. the allomorphs /ni, in:i, nin:i,
an:i/ and to the 3. sg. fem. the allomorph /si/ which occur after a verb. In the SB dia-
lect the two object cases of the plural pronouns are in free variation after a verb.

In OB the two object cases are distributed according to direct and indirect object, in OA the same dis-
tribution applies but the forms differ: for direct object, the personal suffix forms occur, for indirect
object, the forms ending in -ti (i.e., those that in OB occur with transitive verbs) occur.
102 MORPHOLOGY

An illustrating comparative paradigm follows :

OB SB OA

dir. obj. indir, obj. dir. obj. indir, obj.

sg. pl. sg. pl. sg. pl. sg. pl. sg. pl.

1. ni ni:a:ti (a)m ~ nim ni:a:äim (an) (n) (in)ni ni:a:ti ~ ni:a:Si i ni:a:ti nin:i ni:a:ti
2. m. ka kunu:ti ku(m) kunu:Sim ka ~ ku kunu:ti ~ kunu:Si ka kunu ka kunu: ti
f. ki kina:ti ki(m) kina:5im ki kina:ti ~ kina:5i ki kina ki kina:ti
3. m. Su 5unu:ti Su(m) Sunu:Sim Su Sunu:ti ~ Sunu:Si Su Sunu Su Sunu:ti
f. Si sina:ti Si(m) Sina:Sim Si ~ Su Sina:ti ~ Sina:Si Si Sina Si Sina:ti

N o t e s relative t o distribution of allomorphs:

1. sg. allomorphs in SB: anni after consonant, inni and ninni after vowel.
1, sg. indirect object allomorphs are distributed like the ventive endings, see [Link].
Final / m / is preserved in dialects where it is preserved elsewhere in final position, see [Link].1. and
[Link].
SB allomorphs with ending /¡Si/ are rare.
3. fern. sg. in SB is Si only when referring to animate beings; elsewhere Su occurs.
The two object suffixes m a y co-occur, in the sequence indirect object — direct object.

5.5.4. Determinative pronoun and relative pronoun

A determinative pronoun, with inflection for the three cases, subject case ("nomina-
tive"), object case ("accusative"), and subjoined case ("genitive"), exists at the OAkk.
period, see Gelb, 1961a, 133 ff. In the literary dialect this pronoun belongs to the in-
declinables and inflects only for concord (see [Link]): sg. sa, mase. pl. su:t, fem. pl.
sa:t. The singular form, sa, occurs as relative pronoun without number or gender
concord.

5.6. D E R I V A T I O N

Derivation is either by means of vocalic morphs from a consonantal morph (root),

see 5.0, or by means of linear affixes from a base which is either simple or itself con-
sists of a root and a vocalic morph.
Vocalic morphs consist of one or more vowels with empty slots in a fixed position
where the consonantal morph interdigitates. For examples see [Link], 5.4.4, [Link].
Linear affixes are prefixes, infixes, and suffixes. Derived verb stems have been de-
scribed in [Link]. Some linear affixes only occur with interdigitated morphs.

5.6.1. Prefixes

Prefixes only occur with interdigitated morphs. They derive nouns (including par-
ticiples and infinitives). The prefixes are :
{i-}, {mu-}, {ma- ~ me-}, { n a — ne-}, {nu-}, {sa— se-}, {su-}, { t a — te-}
MORPHOLOGY 103
Distribution of prefixes with allomorphs ending in /a/ and /e/ is according to stem
determinant, see [Link].
{mu-} mostly occurs with participles; {su-} with infinitives and adjectives; the re-
mainder with substantives. The sole vocalic prefix /i/ may also be considered part of
a vocalic morph {i-i-} as in Vi-Pt-i-r-, V'-kr-i-b- and a few more.

5.6.2. Infixes

Derivational infixes are: {:}, {t}, {it}, {ut}.

The derivational infixes occur with derived verb stems ({ut} only with derived stem
Dt and Dtn in the infinitive, and hence may be resegmented as {t} plus the appro-
priate vocalic morph for derived infinitives), and nouns. For a homophonous infix {:}
as grammatical morph of the plural see [Link], for the homophonous tense infixes
{:} and {t} see [Link].
5.6.3. Suffixes

Derivational suffixes form nouns and adverbs.

Nominal suffixes :
(substantives) {t ~ at}, {u:t}, {a:n}.
(adjectives) {i}, {i:t}, {aia}, {u:} (older {a:u}).
Adverbial: {is}, {a:nis}, {tan}, {at:a ~ at:i}, {us:u}, {am ~ a}, {e:}.
Several of these suifixes combine, as e.g. adjectival i, aj, u: with (fem.) i:t; adverbial
is with am. The adverbial suffix us:u is in some constructions treated as a pronominal
and is subject to concord of gender and number, i.e., for the pi. has the endings (us~)unu
and (us-)ina from the personal suffix set (see [Link]).

5.7. INDECLINABLES

To the class of indeclinables belong certain pronouns (see [Link]), furthermore, ad-
verbs (for their derivational suffixes see 5.6.3), and the so-called particles: interjections,
sentence connectives, negations, conjunctions, and prepositions.
Indeclinables may be divided into two groups: into those which always occur alone,
and those which may take a clitic. For these see below, 5.8.
Prepositions form a special subclass of indeclinables. They may occur not only
with a following substantive in the subjoined case (see [Link]) or with a pronoun in
the direct object case (see 5.5.3), but also with a personal suffix.
Prepositions occurring with personal suffixes : gadu, eli, itti /it:i/, isti, arki, assumi
/as:umi/, kümu, mahar.
Prepositions not occurring with personal suffixes but with a noun or the free form
of the pronoun: ina, ana, kì (/ki:/), lam(a), em{a), adi, istu, eia, assu (/a§:u/), ezib.
104 MORPHOLOGY

For adverbials (substantives in the locative case in adverbial function) occurring

with personal suffixes, see [Link] and [Link].

5.8. CLITICS

All Akkadian clitics are enclitics, i.e., suffixed. The following clitics occur:
ma grammatical meaning: sentence connective or emphasis
mi grammatical meaning: indicates that the clause that pre-
cedes is direct citation. The clitic occurs also with the
indeclinable umma which in itself indicates that what fol-
lows is direct citation. Rarely the clitic is suffixed to any
member of the clause which constitutes the citation,
man grammatical meaning: "irrealis", i.e., transforms the
clause into an irreal clause,
maku occurs only in peripheral texts, from territories where
Akkadian is not the native language and once in an NA
text; grammatical meaning "irrealis" or the like.
The clitic -ni, which has the grammatical function of indicating the end of a dependent
clause, is restricted to the Assyrian dialect. Previous studies have seen in this clitic a
subjunctive ending; however, it occurs not only with verbs but with any part of the
sentence, and it co-occurs with the subjunctive ending for which see [Link].
Distribution of clitics : clitics may occur with any member of any inflectional cate-
gory, with the exception of sentence connectives (u 'and', 'or', lu 'or' 'indeed',
sa 'who, which' (introducing a relative clause); this distribution supports their in-
terpretation, including that of the Assyrian clitic -ni, as sentence marker, since the
only morphemic category with which a clitic is mutually exclusive is a sentence con-
nective which occurs at the beginning of a clause, while clitics usually occur after the
last member of the clause.

5.9. CONCORD

Two types of concord occur : concord between the subject and predicate (for person,
gender and number) which affect any part of speech that occurs as predicate, such as
verb, noun, pronoun; and concord between a substantive and the words associated
with it: adjective, pronoun, appositional substantive. (For concord of the adverbial
in us:u, see 5.6.3.)
For pronominal concord see 5.5.3.
Numerals have a particular concord. Since the discussion of numerals is not in-
cluded in this description, their concord is likewise disregarded.
6. MORPHOPHONEMIC ALTERNATIONS

6.0. ISOLATING MORPHOPHONEMIC CHANGE

Dissimilation, assimilation, metathesis, and similar environmentally conditioned re-

placement of a vowel or consonant by another, are treated in the grammars of Akka-
dian in the chapter "Phonetics". Similarly, elision or epenthesis of vowels and similar
changes in the syllabic structure conditioned by affixation in the derivation or by junc-
ture are also subsumed under "Phonetics". Such chapter on phonetics comprises three
distinct types of "sound-change" : historical development (the replacement is analyzed
in terms of the etymon of the word) ; phonological variation, and morphophonemic
change. In this descriptive statement, only the last two will be treated. Affixation and
environment stated in terms of juncture affect the phonotactic shape of a base, and
these will be described in the following section, Phonotactic alternations.
A preliminary note is in order. First, the morphophonemic changes will be ob-
served only insofar as they are reflected in the writing. As was pointed out above in 5.1,
the writing does not always reproduce morphophonemic change, but uses "morpho-
graphemes". Moreover, not every spelling, especially if it is unique and hence suspect
of being a mistake, and even if it is only very rare, can be used to illustrate morpho-
phonemic change; also hypercorrection, although it often sheds a welcome light on
phonological phenomena, cannot always be taken as indicative of actual phonological
features.
My reason in isolating morphophonemic change from other types of sound-change
lies in the observation that the changes to be studied in this chapter affect phonemes
only at morpheme boundaries. These morpheme boundaries lie between a stem and a
suffix, a stem and a prefix, and before and after an infix. Morphophonemic change
does not span the boundary between two words, with the exception of compounds
consisting of a bound form (see 5.3.1) followed by a subjoined case, or of compounds
representing a clause when they serve as a proper name. Scribal practice rarely notes
morphophonemic change at such word boundary, tradition demanding that each
word be written as an independent unit (cf. the hypercorrection cited below 6.1.4.2a)
and only in exceptional cases, in frozen forms such as proper names and the like, does
one find an indication that the morphophonemic rules obtained in this case too. Note
for example the compound geographical name Du:r-Humid written once (TCL, 14,
106 MORPHOPHONEMIC ALTERNATIONS

52:4) Du-ru-ùh-mì-id, i.e., /Du:ruhmid/, indicating that the cluster / : C + C / (where +

indicates word boundary) is resolved as /:CVC/, as in final position (see [Link]), i.e.,
as if the boundary was [Link]+mid. Most occurrences are found in older texts,
written before the scribal practice was frozen, or in late (NB) texts of private nature,
which for the reasons cited in [Link] cannot be adduced as evidence.

6.1. CONSONANTS FOR WHICH MORPHOPHONEMIC CHANGE

MAY BE DOCUMENTED

Since we are obliged to disregard word boundary attestations and restrict ourselves
to changes occurring at the boundary of a prefix, a suffix, an infix, and the stem, the
morphophonemic description becomes considerably simplified, as the roster of pho-
nemes that enter into affixes is limited to /m, n, s, t, k, :/. In the following we shall il-
lustrate the morphophonemic behaviour of each of these consonants, contrasting it
with contact position of the same consonants in non-boundary positions, where no
change occurs.
We limit the cases of morphophonemic change to contact position of consonants,
because the orthography does not furnish sufficient evidence of changes that may af-
fect vowels at morpheme boundaries. Exception will be made only for the morph /Vt/
when it follows certain consonants, since the change affects both the preceding con-
sonant and the /t/ of the /Vt/ morph, in spite of the fact that they are not in contact.
For each consonant that enters into affixes, two cases will be differentiated: (1) as
initial consonant of a suffix; morpheme boundary before this consonant. This is sym-
bolized by + , thus: + C ; (2) as infix; morpheme boundary both before and after this
consonant, symbolized + C + . The position of consonant as prefix (i.e., of the type
C + ) is not treated here because, as stated in phonotactics [Link] and 4.4.1, initially
only CVC, never CC- occurs, and hence any initial * C + C is always written CVC.
For the purposes of stating the morphophonemic changes in a more succinct form, it
is convenient to symbolize the sibilants (the groove-alveolars /s, z, s/) by Z, and the
dentals (/t, d, \j) by D. Note that the symbols Ζ and D are not chosen because of the
phonetic or phonological classification as D(ental) or sibilant, but that the members of
the class symbolized by D and Ζ are defined as belonging to this class because the
same morphophonemic changes affect all the member phonemes.

6.1.1. /m/

[Link]. C+m=/m:/, where C=p, b, n.

Definition : at morpheme boundary, a labial stop (ρ, b) or a nasal (ri) are assimilated
to the labial nasal /m/ of which they share one feature, either the peripheral (i.e.,
labial) articulation or the nasality. When the nasal preceding /m/ is itself /m/, there
MORPHOPHONEMIC ALTERNATIONS 107
is of course no graphic indication of the assimilation, except the fact that the resulting
long m /m:/ is often written only once.
Examples :
us:ab+ma=us:am:a (wr. us-sa-am-ma and us-sa-ab-ma)
i:rub+ma=i:rum:a (wr. i-ru-um-ma and i-ru-ub-ma)
iqa:p+ma=iqa:m:a (wr. i-qa-am-ma and i-qa-ap-ma)
iskip+ma=iskim:a (wr. is-ki-im-ma and is-kip-ma)
(also is-ki-in-ma, see [Link])
ik:ip+ma=ik:im:a (wr. ik-ki-im-ma and ik-ki-ip-ma)
luskun+ma=luskum:a (wr. lu-us-ku-um-ma and as-ku-un-ma)
Note, at word boundary between bound form and following word: hisib+ma:tim=
hisim:a:tim (wr. hi-si-im-ma-tim) YOS, 10, 22:24.
Contact position of these consonants in non-boundary position does not occur,
since, according to the phonotactic rule stated in 4.3.1, homorganic consonants do
not co-occur in one morph, if they might cluster; morphs in which /n/ and /m/ co-
occur so that they might cluster are very rare in Akkadian (as also in Arabic, see
Greenberg, 1950, 164 and 166), and are normally attested only in forms where the
cluster does not occur, except for the peripheral an-mu-us (Idrimi line 18), which, as
expected, does not show assimilation. Morphs in which /n/ and /m/ co-occur are at-
tested in older texts with /n/ and /w/; in these, *nw is written <nm> or (mm).

[Link]. m+C
+ m + does not occur (i.e., /m/ does not occur as infix).
For m-\-s see 6.1.3 d.
For m+k see 6.1.5.
For m+n see [Link].
For m+t see [Link].1.

6.1.2. /n/

[Link]. m+n=n:

Definition : at morpheme boundary, /m/ assimilates to a following /n/.

Examples :
ispuram+ni=ispuran:i (wr. is-pu-ra-an-ni)
asrum+ni=asrun:i (wr. as-ru-un-ni)
Morphographemic spelling ta-aS-pu-ra-am-ni-si (Boyer,
Contribution, 208:8)=taspuram-\-ne:si
In non-boundary position assimilation does not obtain : lumnu, not *lun:u, imna, not
*in:a, amnu, not *an:u, etc.
108 MORPHOPHONEMIC ALTERNATIONS

[Link]. (a) + M + C = C :
Definition : at morpheme boundary, /n/ as infix assimilates to the following conso-
nant.
Examples (taken from Stem IV) :
i+η+paris=ip:aris
i+n+t+apras=it:apras
If the "primae η" are regarded as containing /n/ as first consonant, the following ex-
amples also apply:
a+n+t+umus=at:umus
i+η+1+akis=it:akis
In non-boundary position, assimilation does not obtain: kanSu, not *kas:u, inbu, not
*ib:u, etc. (Exception: has.u < hanSu < hamSu, sa$:u < Sansu < samSu). Note also
the hypercorrection cited [Link], is-ki-in-ma, where from iskim:a < iskip+ma an
assumed iskin+ma was graphically restored, and the plurals be:lpiqine:ti, al:una:tum
(sg. piqif.u < piqid+tu, [Link].u).

(b) +n+Yt=Yt
Definition: at morpheme boundary, + n (restricted to word-initial position) follow-
ed by an infix Vt (/it/ or /ut/) drops.
Examples (taken from stem IV) :
η+it+aprusu=itaprusu (root PRS, \ / n + i t + p r u s + u )
n+it+apras=itapras (root PRS, V n + i t + p r a s )
If the "primae η" are regarded as containing /n/ as first consonant, the following ex-
amples also apply:
n+it+at:uku=itat:uku (root NTK, V n + i t + n + t u k + u )
n+ut+az:umu=utaz:umu (root NZM, \ / n + u t + n + z u m + u )
In non-boundary position, loss of initial /n/ does not obtain: nitbi (ni+tbi), not
*itbi, nutar:ak:um (nu+tar.+am+kum), not *[Link]:um.

(c) n+:C=nAnC (A stands for stem-determinant /a/ or /e/)

Definition: at morpheme boundary, when /n/ is followed by a cluster of /:/ and con-
sonant, the cluster of three consonants is resolved by a vowel /a/ or /e/ (the stem-de-
terminant) appearing after /n/ and the /:/ (length) of the cluster is realized as /n/.
Examples: n+{:mr}+{-u-u}=nanmwrM (prefix /n/ of derived stem IV (N), plus
consonantal morph {:mr}+interdigitating vocalic morph {-u-u} of
the infinitive)
η+{:dr}+{-u-u}—nenduru (same, with consonantal morph {:dr},
stem-determinant /e/)
n+{:dr}+{-u-u}=na/ii/Mrw (same, consonantal morph {:dr}, of a
stem "primae η," i.e., citation form is [Link].
MORPHOPHONEMIC ALTERNATIONS 109

In non-boundary position, this change does not occur, e.g., na+:bal=na:bal(u),

ne+:pis=ne:pi$(u), na+:mar=na:mar(u), etc. (derived forms from similarly consti-
tuted stems with prefix na— ne-).

(d) +n+t+:C=+t:AnC (A stands for stem-determinant /a/ or /e/)

Definition : An /n/ infix followed by a /t/ infix and by a cluster of /:/ and consonant
assimilates to the ft/ of the infix, and the cluster of four consonants is resolved by a
vowel /a/ or /e/ according to the stem determinant, and the /:/ of the cluster is realized
as /n/. (This applies with a segmentation of the infix of the perfect tense as {t}, see
[Link].2).
Examples :

i + n + t + : m a r = i t : a n m a r (stem \/: m Vr, citation form ama:ru)

i + n + t + : p u s = i t : e n p u s (stem \/:Pus, citation form epe:Su)
i + n + t + : d i n = i t : a n d i n (stem \/:din, citation form nada\nu)

In non-morpheme boundary case, i.e., with different segmentation, this rule does not
apply, see above [Link] (b).
If the "primae η" stems are regarded as containing /n/ as first consonant, another
morphophonemic rule must be introduced :

(e) + n C = C :
Definition : at morpheme boundary, morph-initial /n/ assimilates to the following
consonant such as *i-nkis=ik:is, etc., see [Link].2; note the "etymological" writing -
or hypercorrection - ta-an-zi-im-ti=¡taznmti/ ( 7 0 S , 10, 33 ν 7).

(f) For n-\-k=/k:¡ see 6.1.5.

6.1.3. ß l

(a) {Z, D, §} (i.e., Ζ or D or S)+S=/s:/

Definition: when preceded by a morpheme boundary, i.e., as suffix, /§/ and the pre-
ceding dental, sibilant, or palatal (/s/) becomes long s (/s:/).
Examples :
riks+su=rikis:u
ma:t+su=mas:u
us:is+§ina:ti=us:is:ina:ti

In non-boundary position, the assimilation does not obtain, e.g., *qadsu=qasdu (see
[Link]), not *qas:u; paSSu, not *pas:u, etc.

(b) For + V i + f = / S : / see 6.1.4.3a.

Definition: at morpheme boundary, when s < s (see [Link]), it assimilates to the s

(itself < s) after the boundary to long S (/s:/). This rule subsidiary to rule (a) is
operative in the OAkk., OA, and OB periods and is the origin of /§:/ in morphs oc-
curring in other dialects which, etymologically, are composed of two morphs {Vs} +
{§V} and thus would have, according to rule (a), resulted in */Vs:V/.
The different behavior of $ < s at morpheme boundary would necessitate positing a
phoneme /è/ for every dialect in which rule (c) is operative, or rather, two phonemes,
and s2. However, S and S < s do not contrast in any other environment and an alter-
nate solution might be to give the list of morphs for which rule (c) applies, e.g., the
stems VIbis, qi:S, etc.
For the literary dialect, the morph {is:u ~ us:u ~ is:a ~ us:a} (see [Link]), need not
be considered as consisting of {is}+{su}, etc. - especially since etymologically the allo-
morph with initial /u/ is {um}+{su} - but, as noted in [Link], as {us:u} or {us} (which
is not *us) ; since the sequence β:/ and /ss/ both occur, in non-boundary position, the
behavior of this morph need not be included in morphophonemics.

(d) +(CV)m+i=/s:/
Definition: at morpheme boundary, final /m/ of a suffix morph (i.e., of the morphs
{am}, {um}, {nim}, {kum}, {kim}, {sum}, {sim}) assimilates to the initial /s/ of the
following morph.
Examples :
asr+um+su=asrus:u
atrud+am+su=atrudas : u
ublu+nim+sum=ublunis:um (wr. ub-lu-ni-sum)
ubar:+am+kum+su=ubar:ak:u§:u (wr. ú-ba-ra-ku-su)

In non-boundary position, the assimilation does not obtain : amsi, not *as:i, emsu, not
*eS:u, etc. When /m/ preceding the morpheme boundary is not the final of a suffix
morph as listed above, the rules stated in [Link] obtain.
(e) For + « + £ = / § : / see [Link] (a).

6.1.4. /t/
[Link]. C + i
(a) Z+t=Z:
Definition : at morpheme boundary, /t/ assimilates to a preceding sibilant.
Examples:
is+t+abt-=is:abat
iz+tan+k:ar=iz:anak:ar
is+tan+h:ur=is:anah:ur
Exception: if /t/ is the last derivational morph of the word (even if other, inflectional,
morphs follow it), i.e., if /t/ is the derivational morph of the feminine (see 5.3.3), the
phonetic and phonotactic rules stated in [Link] and [Link]. apply.
MORPHOPHONEMIC ALTERNATIONS 111
Examples :
tahsis+t+u=tahsistu or tahsistu
manzaz+t+u=manzaztu, maz:astu
i s + t + u m = i s t u m and istum (wr. i-i$-tum, i-is-tum)

(b) D + i = D :
Definition: at morpheme boundary, /t/ assimilates to a preceding dental.
Examples :
id+t+abb=id:abub
at+t+ard+am+kum=at:ardak:um

Exception: if /t/ is the last derivational morph of the word (even if other, inflectional,
morphs follow it), i.e., if /t/ is the derivational morph of the feminine (see 5.3.3), the
phonotactic rules D + t = t : apply, and it is the dental that assimilates to /t/.
Examples :
mubal:it+t+u=mubal:it:u
qard+t+u=qarit:u

[Link]. + i + C
(a) +t+Z=Z:
Definition: at morpheme boundary, /t/ assimilates to a following sibilant.
Examples :
h+it+sas=his:as
k+it+sur+u=kis:uru

Note the compound himif [Link], wr. hi-mi-if-se-tim, i.e., /himis:e:tim/ (Szlechter,
Tablettes, II, 142:4).
In non-boundary position, the sequence t+Z does not occur, see 4.3.1. Since the
writing is usually morphemic, i.e., the morphographeme <it-ZV> occurs instead of a
sequence VZ-ZV, the following hypercorrection is attested: <lit-su-ru> for /lis:uru/,
for l i + n s u r + u or li+:sur+u (WZJ, 8, 567, HS, 109:6, MB).

(b) + f + D = D :
Definition : at morpheme boundary, /t/ assimilates to a following dental.
Example: m+it+dud+a=*mid:uda=minduda (see [Link].5)
In non-boundary position, the sequence f + D does not occur, see 4.3.1.

[Link]. (C)+Vi
(a) +Vt+s=/s:/ (see 6.1.3b)
Definition : /t/ of an infix Vt assimilates to a following /§/.
Example: p+it+sas=pis:as
In non-boundary position, the sequence ί + ί does not occur, see 4.3.1.
112 MORPHOPHONEMIC ALTERNATIONS

(b) Z+Yt=tVZ
Definition: a sibilant in initial position, followed by an infix Vi, undergoes metathe-
sis with /t/.
Examples:
z+it+qar=tizqar
s+it+but=tisbut
z+ut+q:ur+u=tuzaq:uru
In non-boundary position, ZVt remains ZVi, e.g., in zif.u, [Link], [Link], si\tu,
zitlunu., [Link].

(c) D + V i = f V D
Definition : a dental in initial position, followed by an infix Vf, undergoes metathesis
with /t/.
Examples :
d+it+kuäat=tidkusat
d+it+u:ku=tidu:ku
In non-boundary position, DV< remains DVf, e.g., in [Link], dital.u, du\tu, mi(i:tu.
The example titabbû cited GAG, § 96e, is attested in a copy containing numerous scribal errors, for the
duplicates' titabbû /titab:u:/. Note that rules (b) and (c) are formulated so as to exclude f + V/; for the
latter case, a metathesis as variant for the sequence SVt- is attested in the OA dialect: ti$amme:a and
sitamme:a, also in MB (or MA) tuSeSkun for suteskun (Tukulti-Ninurta Epic, iii, 38).

6.1.5. /k/
C+k=¡k:¡, where C=n, m
Definition: at morpheme boundary, η and m assimilate to a following /k/.
Examples:
id:in+kum=id:ik:um
lil:ik+am+kim=lil:ikak:im
il:iku+nim+kum=il:ikunik:um

6.2. NON-MORPHOPHONEMIC CHANGES

While these morphophonemic changes apply without exception, and are expressed in
the writing except for morphographemic spellings, various phonetic changes - his-
torical, dialectal, and perhaps even synchronic - seem to be attested in free variation.
This may be due to a large extent to the conservativeness of writing, which accounts
for copies of older texts without adjustment to contemporary language, or to an at-
tempt to write a "correct" form, or to an attempt to write the old form. Whatever
the reasons - the attempt at correctness is evident only in those cases where a hyper-
correction is made, but such cases rarely occur in the literary dialect under investiga-
MORPHOPHONEMIC ALTERNATIONS 113
ti on - the fact remains that phonetic development or phonetic variation can be judged
only through the fluctuations exemplified in the writing. This feature of the writing
practice provides an additional argument for separating historical phonology from
morphophonemics, for which the writing always noted the change. A few exceptions
may of course be noted, but not more than may be safely assigned to scribal error or
convincingly argued to be an intrusion from another, the scribe's, dialect.
In order to contrast morphophonemic with non-morphophonemic changes, I shall
illustrate phonetic change, divided in three groups: 1. historical, 2. dialectal, 3. syn-
chronic free variation. The last group is illustrated with the understanding that what
is here dubbed free variation may in fact equally represent historical or dialectal
change only sporadically reflected in the historical type of writing.
Since my contention for the separation of morphophonemics from phonological
change affects only phonemes in contact, the examples do not include every change or
free variation of single phonemes. These are either treated in the phonology, or have
to be relegated to a book on historical phonology.

6.2.1. Historical change

Changes to be illustrated do not include (a) changes from reconstructed Proto-Semitic

or Proto-Akkadian to Akkadian; (b) so-called "changes" from Old-Akkadian to later
dialects (OA or OB), since it is not proven that there has been unilinear development
from OAkk. to the other Akkadian dialects; for a theory that OAkk. represents a dia-
lect earlier than, but not in direct historical line with, Old Babylonian and Old Assy-
rian, see Chapter 1.
[Link]. Changes from OB to MB:
OB [ia] ([ija], /i:a/) > MB /a:/: bini:a:tum > bina:tu, ri:a:bum > ra:bu, an:i:am
> an:a:
OB initial wV > MB /V/ : wardum > ardu, wilit:um > ilit:u.
OB {a-i} > MB {e-i} in D and S - s t e m of the verb.
OB final /am, im, um/ > MB /a, i, u/: bi:tim > bi:ti, etc. (see also above examples).
Note that the last change applies only to final /am, im, um/, not to final V C m =
/CVm/; e.g., rigim remains rigim, Sulum remains Sulum, because they stand for ^rigm,
Vsulm; it does not apply either to final /V:m/ which also remains /V:m/, e.g., sara:m,
(e:m, etc.
OB í D > MB /D: i§tur > iltur, iädu > ildu, aStakan > altakan.
[Link]. Changes from OA to MA:
OA initial wa > MA /u/ : wardum > urdu.
OA St > MA /It/: iStakan > iltakan.
OA VvfV > MA VÒV: awutum > abutu, awat > abat.
OA awi, ami > MA /a:i/ : awi:lum > axlu, 1 damiq > da:iq, ahamis > aha:i§.
1
The two MA forms a-i-lu and a-i-lu-ut-tu (/a:i:lu/, /a:i:lut:u/) show the relative chronology of the
.114 MORPHOPHONEMIC ALTERNATIONS

[Link]. Changes from MA to ΝA (see above for the MA forms):

OA it > MA It > NA /s:/: istakan > iltakan > is:akan.
OA ami > MA a:i > NA /e:i/: damiq > da:iq > de:iq, ahamis > aha:i§ > ahe:is.

[Link]. Other changes, not clearly ¡imitable to period·.

1. Partial assimilation to homorganic consonant (post-OB): m > η before dental
and /§/ : (mD > nD) : imdud > indud, imti > inti, amsi > anSi.
2. Assimilation to homorganic consonant: mb > /bb/: imbaru > ib:aru, embu:bu
> eb:u:bu.
The examples cited under (2) can be also considered dissimilation of bb > mb, see
(5) below.
Zt > /st/ : m a r u s + t u > maruStu, piris+tu > piriStu, maz:az+tu > maz:a§tu
nt > /t:l: lemun-ftu > lemut:u.
3. Assimilation rn > /n:/ : arnu > an:u. Also at morpheme boundary (of a com-
pound): Sa-pi-in-na-ri-im for Sa:pir+na:rim {CT, 44, 53:1, OB).
4. Partial assimilation :
a) to voiced consonant: i g + t + a m a r = i g d a m a r .
b) to emphatic consonant (in MA and NA): i q + t + a b i > iqtibi.
5. Dissimilation of voiced long consonants (post-OB):

d: > /nd/: inad:i > inandi

g: > /ngI : im:ag:ar > im:angar
b: > /mb/: *inab:u > inambu; ib:i > imbi
z: > /nz/: inaz:iq > inanziq

For further examples and dissimilation of non-voiced consonants (it-tan-am-pa-f}u=

/it:anap:ahu/, a-nam-$ar — /anaç:ar/, < u§amsak > = /uäas:ak/) see Aro, 1955, 36f.

6.2.2. Dialectal variation

[Link]. Old Babylonian and Mari dialect :

OB i:a ~ Mari /e:/: iqbi:am ~ iqbe:m, an:i:am ~ an:e:m.

[Link]. Assyrian ~ Babylonian:

Assyrian sb, sp ~ Bab. /sb, sp/: usbat ~ wasbat.
Assyrian e ~ Bab. /i/(personal, see 5.4.1): ek:al ~ ik:al.
Assyrian e ~ Bab. i (as final vowel of the genitive case).
Assyrian a before r, h ~ Bab. e: sahru ~ sehru, iqar:ib ~ iqer:ib, garu: ~ geru:

two last changes: MA /a:i:lu/ goes back to a form /amirlu/, so that the change ami > a:i includes this
word, and this (perhaps Babylonian) form must have replaced OA /awi:lum/ before VwV > V6V,
otherwise the word would have occurred as *abi:lu in MA.
MORPHOPHONEMIC ALTERNATIONS 115
[Link]. Babylonian {OB, MB, SB) ~ Neo-Babylonian.
Variation between other Babylonian dialects and Neo-Babylonian are included here
and not under "historical development", because, as discussed in Chapter 1, the Neo-
Babylonian texts attempt to render contemporary language which is influenced to a de-
gree unknown to us by one or more local Aramaic dialects. This question is in need
of further investigation.
Babylonian -u:ni ~ NB -u:nu (see [Link])
Babylonian rC ~ NB §C: amartu ~ amastu, ahirtu ~ ahistu, pirku ~ pisku.

6.2.3. Free variation

It is mainly free variation, to be more exact, free variation in the writing, which makes
a phonological description of any dialect of Akkadian so difficult. One may proceed
if one wishes with a rigorous methodological approach, holding to the irreversibility
of phonetic change. If we follow this approach, every deviation from the norm - and
by "norm" we might mean the statistically more frequent or the etymologically cor-
rect forms - would be taken as evidence of phonetic change, and henceforth any wri-
ting that does not reflect such a change would be declared historical, normative, or
morphographemic. Another approach would take the fluctuations of the writing as
free variation in every case; the difficulty here lies in the time-span involved, and in
order to assume the coexistence of two forms in free variation for any given length of
time, the very rough time-divisions in Akkadian dialectology would have to be nar-
rowed down far more. Here I can only attempt to give a sample of variant render-
ings of phonemes in contact without opting for either method.
Examples are collected from GAG and from several dialect studies.

[Link]. Free variation of /a/ and /e/.

The most typical free variation is that between /a/ and /e/. Apart from the cases
where the stem-determinant (see [Link]) determines the morphophonemic vowels of
the verb, there are many cases in which the stem-determinant itself is fluctuating.
igan:ih ~ igen:ih
arah:i ~ irelj:i
uhap:i ~ uhep:i
iktananu ~ iktener:u
pasa:lu ~ pese:lu
haru: ~ heru:
{ahu: ~ {ehu:
[Link]lu ~ [Link]lu

[Link]. Free variation between voiced and voiceless consonant:

suqtu ~ zuqtu, gala:tu ~ gala:du.
116 MORPHOPHONEMIC ALTERNATIONS

[Link]. Free variation between /m/ and/n/:

baqa:mu ~ baqa:nu, pasa:mu ~ pasarnu.

[Link]. Free variation between /a/ and /i/ in Assyrian (or vowel dissimilation) :
isi:tu ~ asi:tu, ziqi:pu ~ zaqi:pu.

[Link]. Free variation between iS- and uS- before consonant:

uSta§ni ~ iätasni, usparu ~ isparu.

[Link]. Consonantal metathesis:

nashat ~ nahsat, askup:u ~ aksup:u, hutpu ~ huptu.
7. PHONOTACTIC ALTERNATIONS

7.0. ENVIRONMENTS THAT DETERMINE ALTERNANTS

As was repeatedly mentioned in the chapter on morphology, the affixation of suffixes

(including the 0 suffix) to a base may result in the occurrence of clusters which result
in phonotactic and morphophonemic alternants for both the base and the affix. Mor-
phophonemic alternants have been discussed in Chapter 6; here we need to specify
the phonotactic alternants of the base.
The alternants appear (1) before the gender suffix {t}, (2) before terminal juncture,
i.e., in declension 3), and (3) before a morph with which it forms a stress unit, i.e., in
the bound form, in declension 2).

7.1. BEFORE THE GENDER SUFFIX {t}

Six patterns have to be distinguished, according to the phonological shape of the noun
base.
7.1.1. Pattern 1

When the base ends in a single consonant, the feminine ending /t/ is directly affixed to
the base, i.e., before the case ending, without any morphophonemic vowel intervening,
since two-consonant clusters are permissible.
Examples :
adj. /pa:ris-u/ fem. /pa:ris-t-u/
/mupar:is-u/ /mupar:is-t-u/
subst. /im:er-u/ /im:er-t-u/
/pur§um-u/ /pur§um-t-u/

7.1.2. Pattern 2

When the adjective base ends in a consonant cluster where the first consonant is not
/:/, the phonotactic rule stated in 4.4.2 applies ( / C C + C / > /CVCC/), and the phono-
logical solution is -CiCt-, -CuCt- or -CaCt-. The quality of the vowel occurring be-
118 PHONOTACTIC ALTERNATIONS

tween the consonants of the base is not predictable; the vowel is, in most cases, /i/; in a
restricted number of cases /u/; rarest is /a/. The quality of the vowel which occurs be-
tween the two last consonants of the base when this is followed by the gender suffix
is the same as that of the vowel which occurs in the same position when the base is in
absolute final position, before terminal juncture, see 7.2.
Examples :
a) with /i/: /pars-u/ fem. /paris-t-u/
/sehr-u/ /seher-t-u/
/qasd-u/ /qadis-t-u/ (see [Link])
b) with /u/: /mars-u/ \/marus-t-u=/marus tu/ (see [Link])
/lemn-u/ V'lemun-t-u=/lemut:u/ (see [Link])
/arq-u/ /aruq-t-u/
c) with /a/: /raps-u/ /rapas-t-u/
/aqr-u/ /aqar-t-u/
For an alternant shape -CCa/et-, see 7.1.4.
Since the occurrence of the vowel is not predictable from the masculine form, the
léxica or grammars must list the bases where /u/ or /a/ occurs before /t/ (or in absolute
final) between the two final consonants of the base; all similarly patterned bases not
so listed have the vowel /i/. The alternate solution is to take as base the form occur-
ring before the gender suffix, such as paris, maruç, rapasele., and state different phono-
tactic rules for predicting the masculine form, such as CVCVC before (short) vowel)
CVCC. This procedure is the one preferred by the Semitist for comparative and his-
torical reasons. We may use it with profit because it requires a simple rule for deleting
the vowel regardless of its quality. Its disadvantages are that such a rule (elision) does
not seem to be needed for the description of other aspects of the morphology; also,
taking the feminine as base form is not helpful since, while the assimilation rules for
C + / can be neatly stated (see [Link] and [Link]), it is not possible to recapture, e.g.,
the base marus or lemun from the feminine, assimilated, form.
Bases with more than one syllable form their feminine in a similar way, but the vowel
appearing between the base-final consonants is always /i/.
Example:
/mus:abr-u/ fem. /mus:abir-t-u/

7.1.3. Pattern 3

When the base ends in a vowel, i.e., the inflected forms with case ending /V/ end in a
long vowel /V:/, the feminine gender ending is /V:t/, where the V is /i/ with adjectives,
/a/ with most substantives, rarely /u/. For adjectives, this can be stated by the formula
that /V:/ is replaced by /i:/ before the gender suffix {t}.
PHONOTACTIC ALTERNATIONS 119

Examples :
1) adj. masc. (nominative sg.) /rabu:/ fem. /rabi:t-u/
/elamu:/ /elami:t-u/
/zaku:/ /zaki:t-u/
/da:ru:/ /da:ri:t-u/
/saplu:/ /§apli:t-u/
/e:pu:/ /e:pi:t-u/
2) subst. : /rubu :/ /ruba:tu/
/mah:u:/ /mah:u:tu/
Again, it is possible to take as base the form with the vowel which occurs before the
gender suffix, such as rabi-, ruba-, mak.u-. Then the masculine (and terminal) forms
have to be predicted by a rule, and the long vowel before a consonantal suffix account-
ed for. Since in some forms the base-final vowel is followed by a vocalic suffix (in the
OB and Assyrian dialects), in such forms as rabiat, zakuat, which we have rewritten,
to indicate the syllable division, as [Link], zaku:at, we may rewrite the base as Vrabi:-,
yzaku:-, V ru ba:-, etc. (Since clusters of vowels do not occur (see 4.2) and two iden-
tical or dissimilar vowels in contact exist in the writing only but have to be analyzed
as either [V?V] or [VwV] or [VjV] or /V:V/, the four notations being phonologically
equivalent though possibly phonetically different, the older form rabiu is phonemicized
/rabi:u/, etc.) The rule for the feminine would be to add simply the gender suffix /t/,
which would yield /rabi:t-/, (Ass.) /zaku:t-/, /ruba:t-/, etc., and for the masculine base
a replacement rule of /V1:/->/V2:/, where Vj is the base-vowel, V2 the case-inflection
vowel. Dialect differences may then be described according to whether the base
vowel Vj occurs in a given form or is replaced by V2 (the case-inflection vowel or the
vowel of some other affix).

7.1.4. Pattern 4

When the base ends in a long consonant (/C:/), a morphophonemic vowel /a/ or /e/
occurs between the base and the suffix, see [Link]. /e/ occurs when the stem-determin-
ant (see [Link]) is /e/.
Examples :
/dan:-u/ fem. /dan:-a-t-u/
/mar:-u/ /mar:-a-t-u/
/gaz:-u/ /gaz:-a-t-u/
/daq:-u/ /daq:-a-t-u/
/raq:-u/ /raq:-a-t-u/
/qal:-u/ /qal:-a-t-u/
/rab:-u/ /rab:-a-t-u/
/ez:-u/ /ez:-e-t-u/
120 PHONOTACTIC ALTERNATIONS

/ed:-u/ /ed:-e-t-u/
/el:-u/ /el:-e-t-u/
/eb:-u/ /eb:-e-t-u/
/em:-u/ /em:-e-t-u/
Note that the bases ending in CXC2 sometimes occur, besides the shape /C 1 VC 2 t/ des-
cribed in 7.1.2, also in the alternant shape / C ^ V t / here described.
Examples:
/damq-/ fem. /damiq-t-u/ and /damq-a-t-u/
/sehr-/ /seher-t-u/ and /sehr-e-t-u/
/qa§d-/ /qadis-t-u/ and /qa§:-a-t-u/
/ess-/ /edis-t-u/ and /ess-e-t-u/ (see [Link])
subst.: /sar:-u/ fem. /sar:-a-t-u/
/lab:-u/ /lab:-a-t-u/
/abarak:-u/ /abarak:-a-t-u/
/apkal:-u/ /apkal:-a-t-u/
/ser:-u/ /ser:-e-t-u/
/zab:-u / /zab:-a-t-u/
It is to be noted that many substantives with a base QCg, described in 7.1.2, form
their feminine like those whose base ends in a long consonant. This preference for a
particular pattern may be an additional distinction in classing substantives and ad-
jectives (see 5.3 and 5.3.2).
Examples :
/malk-u/ fem. /malk-a-t-u/
/kalb-u/ /kalb-a-t-u/
This pattern is sometimes preferred over Pattern 6 (7.1.6) as well.

7.1.5. Pattern 5

When the base ends in two identical consonants, the solution is according to 7.1.2.
Examples :
/kazz-u/ fem. /kaziz-t-u/
/lazz-u/ /laziz-t-u/
Pattern 5 does not occur with substantives.

7.1.6. Pattern 6

When the base ends in /:C/, i.e., long vowel and consonant, the ending /t/ is affixed to
the base without any morphophonemic vowel intervening; thus seemingly the word
PHONOTACTIC ALTERNATIONS 121

includes a cluster of three consonants: /:Ct/. To conform to the phonotactic pattern

of clusters, we must write this final cluster as /Ct/, i.e., assume that length drops (see
[Link]). Since the writing never indicates a long vowel in this position, our assump-
tion is justified; the customary transcription with a vowel symbolized as long by a
macron is only etymologically and morphophonemically correct. The fact that /:C/
in final position in such bases indeed represents a cluster is shown by the alternate
pattern 7.1.3 applied to these bases. This incidentally confirms our conclusions in 4.1
that /:C/ is equivalent to /CC/.
Examples:

1) adj.: /ta:b-u/ fem. Vta:b-t-u=/tabtu/

/se:l-u/ -\/se:l-t-u=/seltu/
/ki:n-u/ Vki:n-t-u=/kit:u/ (see [Link].2)
/asi:r-u/ Vasi : r-t-u=/asirtu/
Alternants : /s[Link]-u/ /s[Link]i:tu/
/sar:a:q-u/ /sar:a:qi:tu/
2) subst. : /ma:r-u/ fem. \/ma:r-t-u=/martu/
/be:l-u/ \/be:l-t-u=/beltu/ and /be:latu/
/kas:a:p-u/ v / kas:a:p-t-u=/kas:aptu/
/e:n-u/ Ve:n-t-u=/entu/ and /e:netu/
/si:b-u/ Vsi:b-t-u=/sibtu/
/bu:r-u/ V bu:r-t-u=/burtu/

7.1.7. Before feminine plural

The feminine plural ending {a:tu ~ a:ti} either replaces the gender suffix /t/ or is suf-
fixed to a noun not marked by the gender suffix.
Examples :
,/pa:ris-t-u/ pi. /pa:ris-a:tu/ (Pattern 1)
\/pars-t-u /pars-a:tu/ (Pattern 2)
/rab-i:tu/ /rab-a:tu/ (Pattern 3)
/sar:-atu/ /sar:-a:tu/ (Pattern 4)
•v/ma:r-tu/ /ma:r-a:tu/ (Pattern 6)
Veql-u /eql-a:tu/
V qa:t-u /qa:t-a:tu/

Note that Pattern 5 is not attested, since it would be homonymous with Pattern 4, ac-
cording to Pattern 2 exemplified above.
The examples illustrate that the feminine plural ending is suffixed to the singular
base without the gender suffix, and this accounts for differences in structure between
singular and plural, since the phonotactic changes arising from the addition of a con-
sonantal suffix /t/ do not obtain in the feminine plural because the ending begins with
122 PHONOTACTIC ALTERNATIONS

a vowel. Note that in Pattern 3 where in the singular the base ending /V:/ was realized
as /i:/ before the consonantal gender suffix, it is realized as zero before the vocalic
plural suffix.

[Link]. Traditional explanation.

The second solution proposed for Patterns 2 and 3 is very close to the traditional
explanation, except that of course the traditional explanation is based on etymology
and also has to account for dialect forms that are not our primary concern here. A
basic difference in approach however is that, traditionally, gender is not considered a
grammatical morpheme in substantives. Instead, gender is considered part of the deri-
vational scheme of the so-called "noun-forms". Often even for adjectives and parti-
ciples the masculine and feminine are considered as parallel but basically indepen-
dent formations, since for some only the feminine form, without a corresponding
masculine, is attested, and its meaning is not linked with a presupposed feminine gen-
der. Thus the grammars and dictionaries list as separate entries the "noun-forms"
parsu and paristu, pirsu and piristu, pa'.risu and pa:ristu, etc., e.g., bajjartu, baiertu,
ëpû and ëpîtu, etc.
As already indicated, I separate derivation from inflection, and hence consider as
pairs only forms that inflect for the feminine. Such forms as pirsu and piristu, two
substantives with no other relation to each other than the common sphere of relevance
indicated by the identical root, are analyzed as {p-r-s}+{-i-} or {-i-i-} respectively,
and for the latter the derivational suffix {t} added to the interdigitated base {piris}.
The case of parsu/paristu is different, since they represent an adjective that must take
one or the other form depending on the syntactical concord required, and the same is
the case with pa:risufpa:ristu, [Link]/[Link], etc.
The typology of the patterns listed by the traditional approach also differs from the
one analyzed in the preceding sections. Patterns 1 and 6 are analyzed similarly to the
analysis offered above, but they are not differentiated from each other, since the trad-
itional approach has no reason, such as the one required by our phonotactic state-
ments, to avoid a pattern "long vowel plus two consonants", because the long vowel
is not further analyzed into vowel plus length. Consequently, the morphophonemic
forms Vtäbtu, Vmärtu, etc., are cited on a par with pâristu, immertu, etc., see [Link].
The alternate Pattern 2 (=Pattern 4) (parsat) is reserved for substantives, and the
difference between Patterns 2 and 4 is attributed to a basic distinction between "pri-
mary nouns" (nouns not derived from a verbal or adjectival base) which conform to
Pattern 4, and "derived nouns" (such as adjectives specialized in a substantival func-
tion) which conform to Pattern 2. Exceptions are however admitted for the occurrence
of Pattern 4 in the case where the pattern of the "derived noun", i.e., Pattern 2, al-
ready exists in the language with a specialized meaning. In such a case, the "primary
noun" pattern - in my definition : Pattern 4 as alternate of Pattern 2 - serves to form
a word with a meaning different from that formed with Pattern 2, e.g., the pairs
fimittu and çimdatu.
PHONOTACTIC ALTERNATIONS 123
The origin of Pattern 3 is etymologically explained as the older dialect forms that
have a base ending in /V/ which, with the vocalic case ending, yields /νχν2/. The
feminine is formed with the simple addition of the gender ending /t/ to the base-final
/V/. Thus, to take the examples cited in 7.1.3,

base masc. nom. sg. fem.

rabi- rabi-u(m) rabïtu=/rabi:tu/
elami- elami-u(m) elamïtu=/elami:tu/
zaki- (or zaku-) zaki-u(m) or zaku-u(m) zakïtu=/zaki:tu/ or zakütu=/zaku:tu/
däri- däri-u(m) dârïtu=/da:ri:tu/
sapli- sapli-u(m) saplïtu=/sapli:tu/
ëpi- ëpi-u(m) ëpitu=/e:pi:tu/
ruba- ruba-u(m) rubâtu=/ruba:tu/.

This etymologically correct explanation is insufficient because it has to devise phonetic

explanations in order to account for the long vowel before the gender ending. An ex-
planation was proposed above, 7.1.3. To my mind, however, in the literary dialect
here analyzed, the rule of vowel replacement as sole criterion, without the necessity of
introducing "compensatory lengthening" is a sufficient descriptive statement.
Note furthermore that the traditional, etymological explanation can account for the
feminine rubâtu (/ruba:tu/) beside masculine rubû < rubau, and for the example
mah'.[Link]. However, the descriptive statement of 7.1.3, with the corollary that sub-
stantives have preference for the /-atu/ ending (after Pattern 4, see 7.1.4) over other
types of endings, also accounts for /ruba:tu/, and moreover accounts for such non-
Semitic substantives as /asa:tu/ (masc. /asu:/), for which the etymological explanation
would require either a masculine *as-, or *ass-, or *asa-, while asu (/asu:/) is borrowed
from Sumerian [Link], and hence the base in the masculine is asu-, and the masc. nom.
sg. with the case ending asu+u=asû, from which base a feminine *asûtu (/asu:tu/)
would be expected.

7.2. BEFORE TERMINAL JUNCTURE

The terminal form of declension (3) is the noun base without any inflectional ending,
i.e., zero ending. The terminal form of the feminine {t} appears as {at}. When the
base ends in a consonant cluster, the phonotactic alternants described in 4.4.1 and
4.4.2 obtain. The vowel that resolves a final cluster is /i/, /u/, or /a/ according to which-
ever appears before the feminine /t/ ending, as described in 7.1.2; when the base ends
in a vowel, the terminal form ends in /i/ - with the same replacive that occurs before
the feminine suffix /t/, as described in 7.1.3. Bases ending in /:C/ or in ¡C:/ (Pattern 4
and 6, see 7.1.4 and 7.1.6) do not end in a cluster, and thus their terminal form has no
vowel to resolve a cluster.
124 PHONOTACTIC ALTERNATIONS

Examples:
masc. fem. terminal terminal form fem.
form masc.
Pattern 1 /pa:ris/ /pa:ristu/ /pa:ris/ /pa:risat/
Pattern 2 /mars-/ /marustu/ /marus/ /marsat/
Pattern 2 /pars-/ /paristu / /paris/ /par sat/
Pattern 2 /raps-/ /rapastu/ /rapas/ /rapsat/
Pattern 3 /rabV-/ /rabi:tu/ /rabi/ /rabi:at/ (in OB and
sporadically later).
/raba:t/ (SB).
Pattern 4 /dan:-/ /dan:atu/ /dan:/ /dan:at/
Pattern 5 /kazz-/ /kaziztu/ /kaziz/ /kazzat/
Pattern 6 /ta:b-/ /tabtu/ /ta:b/ /ta:bat/
Pattern 4 shows in writing a short - or single - final consonant, e.g., <da-an>, <e-ez>.
These may be phonemicized as /da:n/, /e:z/, etc., with the length appearing before, in-
stead of after the consonant, see [Link]. Variant spellings such as <e-zi> and <e-bi>
beside <e-ez> and <e-eb>, indicate a phonological feature that the writing system can-
not unambiguously render, and since the same forms with vocalic suffix are written
<ez-zu>, <ed-da>, etc., they may indeed indicate the variation /:C ~ C:/ discussed in
[Link]. For spellings with initial V-sign (i.e., Vj-VjC) see also [Link] and Reiner,
1964, 172 and n. 10.
The plural terminal form ends in /u/ for nouns whose free form ends in {: ~ a:nu ~
u:tu} and in /a/ for nouns whose free form ends in {a:tu}.

7.2.1. Other attestations of terminal forms

Base-alternants similar to terminal (pre-terminal-juncture) forms (syntactically: pre-

dicative), also occur in a number of syntactic constructions - environments - that at
present cannot be defined but by giving a list of them. Such a list is given in GAG, § 62,
and may be here recapitulated, with some modifications, as follows :
a) numerals (cardinals) ;
b) units of measure and of the substance measured ;
c) reduplicated nouns, with a distributive meaning or as compounds;
d) some compounds formed with a proclitic negative particle /la/ or another noun
in a dvandva, or bahuvrîhi, type compound ;
e) in some dialects, nouns accompanied by some specific intonation contour, used
as vocatives.
The syntactic use of the terminal form cannot be analyzed here in detail, but there
are certain restrictions in its use determined not by syntactical but by morphological
categories. Thus, not all nouns occur in the terminal form; while adjectives in the
predicative ( = stative, see [Link]) are inflected both for gender and for number (i.e.,
PHONOTACTIC ALTERNATIONS 125
they occur with the endings 0 , -at, -u and -a), substantives occur in both genders but
only in the singular. This means that the plural nominal predicate of a sentence cannot
be the last element, but is followed by a copula (which may be an enclitic -ma, or a free
form of the personal pronoun), or has a subjoined (genitive) case ending as part of a
noun phrase preceded by a bound form or a preposition. Further restrictions and
solutions will have to be yet investigated.

7.3. B O U N D FORM

The bound form occurs, as stated in [Link], followed by a personal suffix, a noun in
the subjoined case, or a subordinate clause, i.e., a verb phrase with its predicate in the
subjunctive. Depending on these and whether the suffix begins with a vowel or a con-
sonant, various morphophonemic alternants obtain to resolve the cluster resulting
from the juncture. When the base ends in a consonant that is not /t/ (below I), the
alternants are normally differently stated than when the base ends in /t/ (below II) :
Alternants :
Base ends in +suffix V -j-suffix C(VC) +noun or clause
la -C Sum- CV §umi cc- sumka -C sum
b -cc sipr- CCV sipri cvcc- sipirsu -CVC/CCV sipir/sipri
c -C: tup:- C:V tup:i C:VC- tup:aka -C/C:V tup/tup:i
IIa -t ahat- tV ahati tc- ahatka -t ahat
b -Ct nidint- CtV nidinti CatC-/CtVC- nidnatka -Cat/CtV nidnat
nidintaka nidinti
c -t: irnit:- t:V irnit:i t:VC- irnit:aka -t:V irnit:i
d -t: §at:- t:V sat: i CatC- sanatka -Cat sanat
The difference in the alternants ending in CVC or CCV is partly morphologically
conditioned, as in the last cited example under Hd, where final /t:/ results from
assimilation of /Ct/, and in fact there are two alternants of the base, /Sant/=/sat:/,
and /san+t/=/sanat/, see also the examples cited under 7.1 and following, and part-
ly graphically conditioned, since the free variation in lb of, e.g., sipir and sipri may
be taken as a device to write the final cluster, phonologically /sipr/. The same may
apply in lie to final /t:/ when the base has no alternant, e.g., /ilit:u/ < *ild-t-u, and no
alternant *ilid- occurs; the bound form y/ì\ìt:/ is hence written <i-lit-ti> or <i-li-it-ti>
either in an attempt to indicate final /t:/ or to graphically differentiate this form from
/ilid/ from the base \/iId-/ which must be similarly written <i-lid/t> or <i-li-id/t>.

7.3.1. An alternant of the bound form of adjectives

A rare alternant of the bound form of adjectives before a following noun is the ending
/a(m)/ instead of the expected zero ending. The examples have been collected by von
Soden, 1960, to which a few additions and corrections may be added. I agree with
126 PHONOTACTIC ALTERNATIONS

von Soden's conclusion to see in the ending -am an alternant of the bound form before
genitive; while the origin of this form may or may not be sought along the lines sug-
gested by von Soden, an explanation of the occurrence of the -a(m) ending may be
given in purely formal terms : the homography of the bound form of the adjective
base (\/CaCC/, pars) with the bound form of the participle base (VCaCi:C/, parís, or
VCaiCiC/, päris), and the resulting ambiguity in syntactical construction.
The bound form of a participle /pa:ris/ or /pari:s/ (the derived participles /mupar:is/,
etc., are not relevant in this context) is written in the syllabic writing system <pa-ri-is>,
i.e., in this position length is normally not indicated. The corresponding adjective
(which syntactically may also be a passive) has a base /pars/, and its terminal form
(absolute state) would be /paris/, for morphophonemic Vpars/. If there were a
bound form of the adjective homophonous with its terminal form /paris/, it would be
written <pa-ri-is>, i.e., exactly as the same form of the participle /pa:ris/. In the vocalic
morph {a:-i} of /pa:ris/, length is distinctive, and a compound written pa-ri-is+
noun+genitive would be ambiguous. The ambiguity is avoided by the creation of
an allomorph which, with the vocalic morph {a}, preserves the final cluster of the
base /pars/ which is characteristic of its grammatical category, by the suffixation of a
vocalic ending. This then is the allomorph /parsa(m)/ which precedes a subjoined
case, although the homophonous -a(m) accusative ending cannot precede such a case.
In case of a feminine, there is anyway a distinction between the terminal form of the
participle, /pa:risat/, and that of the adjective, /parsat/ (see above, 7.2). For this
reason, feminine adjectives may occur in declension (2), as it has been stated in 5.3.2.
The compounds with bound form ending in -a(m) will have different transformations
into clauses (relative clauses) than compounds with the regular bound form ending.
A full treatment of the transformations cannot be given here, but an illustration will
indicate the lines along which they will proceed. For the transformations, we have
the good fortune of their having been made in antiquity; in the following illustrations
only those forms that are starred are not attested in the ancient vocabularies. Simi-
larly, contrasting participial formations are also well attested, and only some of them
need be expanded by a starred form to provide the parallel.

1 a. ia+substnom+verbtr+acc-> adj+am+substgen
b. ϋα+substacc + v e r b t r + 0 -»part +subst g e n

Rule (1): delete acc. suffix and transform verb into adj.+aw
General rule: move subst to end as substgen
Corollary : if verb has no acc. suffix, this transform is not possible.
Examples:

sa ilum iskipu+su sakpam ilim

but sa rag:am iskipu sa:kip rag:im

similar: mahsam u:rim, aklam isa:tim, lapta nablim.

PHONOTACTIC ALTERNATIONS 127

2a. sa+substpos8+verbintr-» adj+am+substg e n

sa+substpos8+verbtr part+substgen

Rule (2): delete possessive suffix and transform verbintr into a d j + a m (verbmtr=intr.
verb, or tr. verb in passive or stative)
Corollary : if there is no possessive suffix, this transform is not possible.
Examples: sa lib:asu ed:u ed:am qarni:n (for *ed:am lib:i)
sa i:na:su iz:aq:apa *zaqpa i:ni:
but *sa abasu ipahahu pa:lih abim
similar: palham zi:mi:, rapsam irtim, etc., see von Soden, 1960, sub "adj.".

3a. sa+substaec+verbmtr-> adj+am+substgen

Rule (3): transform verbintr into a d j + a w

Corollary: if substantive has no acc. or poss. ending, this transform is not possible, see
2a; if verb is transitive, this transform is not possible, see lb.
Examples : sa irtam malu: mali:a irtim
similar: sa epqam malu: (the example sa lib:a marsu is a scribal error for sa lib:asu
marsu), sa bustam halpu

4. sa+substaec+Hu:jidu: -> subst+subst g e n

(no adjectival transform)

Example: sa emu:qa isu: be:l emu:qi

Relative clauses with adverbials (preposition plus noun) may be transformed into either
construction :
rabsa elu:tim=*sa ina elu:tim rabsu
a:lik mahra=§a ina mahra [Link]

Two phrases do not fit into the above scheme: aklam anzikim and aklam asak:im.
Their transform is §a anzil:am/asak:am i:kulu, and thus we expect (according to lb)
a participial construction, a:kil anzil:im/asak:im.
Five phrases differ from the preceding constructions in that the final substantive is
not in the genitive but in the terminal form. No explanation for this form can be given
(for the occurrence of terminal forms in other than terminal position, see 7.2.1). The
phrases are
namra(m)-si:t
namra(m)-saru:r
atra(m)-hasi:s
apkal:am sipir
mas:u:tam Suka:m

They may be compared to compounds both elements of which are in the bound form,
as cited in 7.2.1 (la §ana:n; seher rabi; biri:t biri:t; se:p arik; emu:qa:n si:ra:t).
APPENDIX

1. A GENERATIVE STATEMENT OF THE CASE INFLECTION

The case inflection is a function of the structure of the Akkadian sentence and may
be arrived at through the analysis of this structure. For the present purpose, a partial
analysis of the sentence only is followed up; omitted are adverbial phrases and other
modifiers which do not participate in the case inflection, but included are those which
are affected by the case inflection.
In this partial analysis, the sentence can be specified as subject and predicate. The
predicate may be a predicative noun phrase or a verb phrase. The predicative noun
phrase is specified as noun plus predicative state. Predicative state has been character-
ized as declension (3) above, 5.3.1 and [Link].
The verb phrase is specified as an intransitive verb or as a transitive verb plus Object.

Abbreviations and symbols

Na = bound form of adjective Vs = stem vowel1

N' = bound form of substantive or parti- Ρ = pattern1
ciple χ = interdigitation symbol
Ν um = number # word boundary
Sg = singular -> rewrite as
PI = plural env. X in the environment before X
Vb = inflected verb env. X in the environment after X
AM = actor marker (personal) env. X _Y in the environment between X
1, 2, 3 = 1st, 2nd, 3rd person and Y
G = gender marker Opt optional rule
m = masculine A pair of ( ) enclose optional items.
f = femine A pair of { } enclose items among which there is
M = mood a free choice.
vet = vetitive A pair of [ ] enclose items which are connected
prec = precative by horizontal rows.
S = stem1 Symbols that have the same subscript (α, β, γ)
R root1 denote identical items.

1
For root, stem, stem vowel and pattern, see 5.4.3 and 5.4.4.
APPENDIX 129

(a) P/WÎOM/Jnon-nomlnatlve +
Pli
a(m)
1. o (b) N a +
PL
a(m)
(c) N' ß + ((Pl ) + (N'y + (Ply)) + i) + Personal suffix]
ß

2. -P/wioMMnon-nomiiiative + Sg -+ {ia:ti, ka:ti,...} (i.e., the object case of a personal

pronoun, free form, see [Link])
3. Pronomnoa-nominauve + PI -»• {ni:a:ti, kunu:ti, ...} (i.e., the object case of a
personal pronoun, free form, see [Link])
4. {β, γ} - {1,2, ..., 6}
5. Pl2 Plx env. N'
PI1 i:
Pit a:ni
6. Pis —• ane:
PI4 e:
Pi& a:ti
Ph. u:ti

N\ ma:r, ...
N\ 5ar:, ...
8. N\ -y bur,...
N\ i:n,...
N's 5ar:, i:n, §a:im, ...
N't §a:im,...
9. Na~* {rapä, ...}
10. α {5, 6}
Opt 11. Personal suffix 0 env. i
12. i - 0 env. V(:)

Opt 13. i -* im env

#
Opt 14. e: -»· e:n env. #
15. Personal suffix -* {i ~ ia, ka, §u, ...} (i.e., the non-subjoined case of a personal
suffix, see [Link])

Ad 1(a): Pronoun non-nominative cannot be further analyzed, and must be specified in

the form of a list, see [Link].
130 APPENDIX

Ad 1 : Ν is the base, i.e., the form without inflectional ending, of a noun, which is
further specified as substantive, participle, or adjective.
su
The object strings Na + b 1(b), and Ν'β + a(m), sub 1(c), select for the
singular the object or accusative ending -a(m).
Ad 1(c): Personal suffix must be specified in the form of a list, see [Link]; in the
examples it will be represented by one of its members, su.
Since N' is specified as the bound form, or "construct state", of the substantive
and participle, the object string sub 1(c) specifies the object as a bound form plus
personal suffix, or a bound form followed by another bound form with an affix -i
plus a personal suffix, i.e., generally, 1(c) specifies the object as a bound form followed
by a personal suffix where these two elements may be split by the insertion of another
bound form plus -/. Rules 11 and 13 specify the object as a bound form followed by
another bound form plus -/, or optionally (in older dialects obligatory), -im.
Expansion of Rule 1(c):
N' + a(m) sar: + a(m)
N' + su ma:r + Su 'his son'
N' + Pli + i + (Su) Sar: + a:ni + (Su) '(his) kings'
N' + Ν' + i + (Su) ma:r + ma:r + i + (Su) '(his) grandson'
N' + Pli + N' + i + (Su) sar: + i: + ma:t + i + (Su) 'the kings of
(his) land'
N' + Ν' + Ply + i + (Su) sar: + ma:t + a:t + i + (Su) 'the king
of (his) lands'
N' + Pli + Ν' + Ply + i + (Su) bi:t + a:t + il + a:ni + (Su) 'the houses of
(his) gods'
(or '(his) houses of the gods', i.e., '(his) tem-
ples')
Note that the "accusative" ending -a(m) occurs only in the singular, with the cate-
gory "noun". This is the only ending that is distinctive for the object; the other af-
fixes that occur in the above list occur also in the subject constructions, to which we
shall turn presently.

The subject of a sentence may be a single word or a phrase. (For the present analysis
we disregard the case when the subject is a relative clause, because such a relative
clause must be again reduced to subject, predicate, etc., and hence is underlaid by
sentence construction rules, and not "subject" rules.) Hence phrase in this context
can be more closely specified as "noun-phrase". The single word may be an (unin-
flected) proper name (inflected proper names are included in the category "noun-
phrase") or a pronoun in the nominative.
The noun phrase occurring as subject may be simple or complex, i.e., consisting of
many members. To include the entire complex, we may either start with the single
APPENDIX 131
element and keep adding on further members until we have exhausted all the members
of the string that occurs as subject, or we may start with the most complex string and
then omit one by one the elements until we arrive at the minimal subject. In this pro-
cess of omitting, we must be careful to proceed so that the remaining members should
not form a string which is ungrammatical.
A good test of our understanding of the categories with their grammatical marks and
their co-occurrences might be to follow the last outlined procedure. I would then note
a noun-phrase which may occur as subject as a consecutively numbered string, of
which some, but not all, members may be omitted.

(1) one of the following words: sa, sœ.t, su:t, rab(i), noted 1'; or a substantive or a
participle, preceded or not by la, noted 1".
(2) a noun in the construct state, i.e., the bound allomorph of a noun, symbolized
as N', preceded or not by la, or a sequence of two such nouns, symbolized as
(la) N'(N')+gen.
(3) a personal suffix.
(4) a substantive or an adjective.
(5) the nominative ending.

Examples for the subject string :

Formula : (2) (3) (4) (5)
sa (la) N' (N')+gen. poss. substl nom.
sa:t adj. }
§u:t
rab(i)
(la)
subst
part

Example : sa la-amat +gu

la ma:r be:l parç +i(m)
sa tabt +i +su
ma:r Sar: +i(m) rabi + u
su:t re:s +i +3u
sa sibit: +i(m) nasr + u
be:l la il +i(m)
rab(i) sikkat +i(m) (+u)
ma:rat Enlil 0 qarit: + u
be:l awat +i + SU
be:l +u
be:l +§u
132 APPENDIX

(i) with pl. ma:r-f i: sar: +i

a:sib-f-u:t kultar +i
be:l ma:t+a:t + i
(ii) with inversion Anun:aki rabi+u:t + u = rabu:tu Anun:aki
andul: +ka raps + u = rapsu andul:aka
The subject may be formalized in the following rules :
ProWOMHnomlnatlve + Num
Í1' + 2 + Nutria + (3 + Num)
Subject -»· Í2 + Num + (3 + Num))] + (4 + Numa
1" + Numa + 3 + Num J
5 + Num
(All the occurrences of the subscripted Num have to be sub-
stituted for by the same selected item, either Sg or PI, whereas
5 + JVwma) any occurrence of a non-subscripted Num may be substituted
for by different, mutually independent, selections from the
items Sg or PI).
Sg
Numa
PI
Pra«oM«nominative + Sg {ana:ku, at:a, ...} (i.e., the subject case ("nominative") of
the personal pronoun, free form, see [Link])
Pra«oM«nominative + PI -* {ni:nu, atrunu, ...} (i.e., the subject case of the personal
pronoun, free form, see [Link])
1' -> {§a, sa:t, §u:t, rab(i)}
1" + Sg -»· {bi:tu, be:lu, ...} (i.e., the nouns - substantives and participles - in the
lexicon)
1" + PI {bi:ta:tu, be:lu:, ...} (i.e., a noun plus the corresponding plural ending, as
listed in the lexicon)
Members (1) through (5) of the subject string, as exemplified on p. 131 f., and as re-
duced to the above formula, do not all have equal status. Members appearing without
parentheses are obligatory, others, which are enclosed in parentheses, are optional.
Among the optional members, the selection of some requires the selection of some
others; conversely, the omission of some is contingent upon the omission of some
others.
a) The obligatory member is (1) which may be represented by either 1' or 1".
1' must be followed by (2); 1" must be followed by either (2), (3), or (5). If 1" is
followed by (2) or (3) or both, it cannot be followed by (5) unless member (4) is also
selected. 1' and 1" are mutually exclusive.
b) Member (4) is optional, but if it occurs, it must be followed by (5).
c) Member (5) is obligatory after member (4), and after member 1" when the latter
is not followed by (2) or (3) or both; it does not occur after 1', nor after 1" + 2 ( + 3).
APPENDIX 133

d) The only restriction governing the occurrence of the optional member (3) is that
it does not follow immediately after 1".
e) Member (2) is obligatory after l'. 2
Two optional modifications of the subject are possible : (i) inversion ; (ii) pluraliza-
tion.
(i) Opt 1" + Numa + 2 + Num + (3 + Num) + 4 + Numa + 5 + Numa
4 + Numa + 1" + Numa + 2 + Num + (3 + Num) + 5 + Numa
i.e., the sequence {1" + 2 + (3)} is reversible with 4.
(ii) Pluralization, as indicated in the formula and in the expansions Pronoun-
nominative + PI and 1" + PI, affects only Pronounnomm&tive + Num and 1" + Num.
The selection of {PI} after (2) and (3) is not equivalent to the pluralization of the
subject, e.g., it does not govern plural concord either between the subject and the
predicate, or with any of the members of the subject string. Note also that subgroup
1' {§a, ..., rabi} cannot be pluralized, as its expansion was so specified as to include
the formally plural sa:t and su:t.

A peculiar feature of the subgroup 1' {sa ... rabi} is the fact that in some cases, when a
member of this subgroup is followed by a member of column (2), the member of
column (2) is not followed by /i/ but by the nominative ending of column (5). In other
words, a member of column (2) behaves after such words of column (1) as if it itself
belonged in or moved into column (1). For this reason, the words sa ... rabi may be
considered in some cases not as independent, free morphs, but as proclitics that form
compounds. Such examples are cited in the illustration, where the more common
rab(i) sik:ati(m) is paralleled by a rab(i) sik:atu(m), and the string $a+la+(a)mat+(i)
+su occurs as Sa-la-mat-su, a writing which indicates the morpheme-boundary
assimilation of /t/ and /§/ which would not take place if /amat/ were followed by /i/.
Besides members of the subgroup sa ... rabi, a few other words which appear as
members of column (1), such as be:l, ma:r, bi:t, and perhaps still others, exhibit this
feature.
The same shift in membership is observable for some words with first members as
ust cited when pluralization is applied; since we said above that member (2) is not
aifected by the pluralization of member (1), in those cases when it is affected by it, or
is pluralized instead of member (1), it must be considered as having shifted into column
(1). Such examples are:

sg. sa biltim pl. sa bila:tim

bi:t sibit:i bi:t sibta:te
rab bu:li rab burla: nu
rab a:li rab a:la:ni
be:l piqit:i be:l piqine:ti (see [Link])
2
This last statement is trivial insofar as it comments on the above formula. However, an omission of
this statement is the only modification needed to write the formula for the subject of a relative clause;
for this formula, the expansion of 1 ' will have to be slightly revised.
134 APPENDIX

The optional modifier la which may appear before a member of column (1) or (2) may
also be considered a compounding proclitic, but this view does not change the gram-
matical behavior of the members of the column.
Note that the inversion rule, by placing the adjective in column (1) and the sub-
stantive in the optional column (4), permits subjects consisting of an adjective or an
adjective preceded by la alone, which are followed only by members of column (5),
such as mahru:, arku:, la hassu, etc.; however, by permitting the inversion for these
members but specifying that member (2) can occur only if member (1) is present, we
have excluded a subject string consisting of an adjective followed by a construct (N').
This inversion rule also expresses the fact, usually called "substantivization of the
adjective", or "adjective in substantival use", that an adjective may appear in the syn-
tactical positions of the substantive.
The only adjective that may be followed by a construct (N') is rab(i), and this is the
reason why it was included among the members of subgroup 1' of column (1).
An adjective which does not occur in its base-form but is followed by an ending
-a(m) or the feminine gender ending -at may occur followed by a construct, i.e., the
affixation of -a(m) or -at permits the adjective base to appear in column (1) (see 7.3.1).
A further use for the subject string above specified is that it simplifies the statement
about the subject of a relative clause. For the subject of the relative clause, the same
string can be set up, with the modification that the members sa, sa:t and Su:t (plus
some other words: mala, etc.) may by themselves form the subject.
The illustrations of the string on p. 131 f. above indicate the grammatical morphemes
of the case inflection that were not specified under the grammatical object. We see
that the nominative ending is -u(m), and the genitive ending is -i(m), and that, of
course, the two are mutually exclusive. This, together with the accusative ending -a(m)
discovered above, p. 130, establishes the case inflection system for the singular noun
which is not followed by another noun.
The plural case inflection is established as nominative (V = u) and non-subject
case (V = i/e); the latter appears both with the object and in column (3), i.e., as geni-
tive, that is, the plural has only two cases.
The category of the personal suffix, represented in the above list by the ending -su,
has as members the suffixes -i, -ka, -ki, -Su, etc. (see [Link]) and their distribution,
according to the person and gender of the noun in column (2) or of the pronoun which
may be substituted for it, is thus not dependent on any member of the string.
It may be seen also that this personal suffix appears in the form -su, etc., in the syn-
tactic constructions encountered until now, i.e., with the object and the subject; in
other syntactic constructions, namely in the one exemplified in the string by column
(1) + (2) + (3), i.e., one consisting of a bound form (N') preceded by either a substan-
tive or participle or by ία ... rab(i), and in the syntactic position of a modifier, pro-
visionally specified as preceded by ina, ana, ... (all prepositions), it appears preceded
by -ζ-. We can thus reinterpret the possessive suffix as
APPENDIX 135

a) -su b) ( l ) + ( l a ) Ν' (N') + Ì + SU

-sa ( i n a , . . . ) + Jsubstl
sa
[part J
etc.
etc.

2. A GENERATIVE STATEMENT OF THE INFLECTED VERB

The following rules apply to the verb inflected for person and for the prefixed moods
precative and vetitive.

1. Vb -> M AM S
Sg
1
PI
Sg Gl
2. AM 2
PI 1
Sg
Í3
3
PI G]

3. G

~2Pl ' ~2 Sg m' a

3 PI m 3 Sg u
4. + s +
3 Pif 3 Sg a
_2 S g f _ 2 Sg m i

M vet
5. M Mprec
0
6 . Mprec 2 Sg m + S S Imp

8 [il] + s + S

ΓίM:1Ί
9. S ιJ + R χ
η P"
0 Vs

In some dialects, AM 3 G, and 3 Sg f -*• 2 Sg m

136 APPENDIX

10
{\
f
i
u env. \I\
w
11. R CCC
P'- "ui "
12. P" Imp ai
ys Vs

13. P"

ia 6 |
14. P'
"Ί
Mprec rrLl V
e n v
15. —>
y ·- C
.Mvet . aj
16. La - lu
Ώ. L -y I
18. aj -+ e env.

* The vocalic patterns of the imperative.

5
This rule states that Stem IV (N) tenses are formed either as those of Stem I, with stem vowel
retained, or as those of Stems II and III, with a vocalic pattern, see [Link].
' The vocalic pattern of the present.
' The vocalic pattern of the preterit.
GLOSSARY OF AKKADIAN FORMS

The English glosses are meant only as identifications of the Akkadian words and are often approxi-
mations of their meaning; they may, however, be of use to those for whom the grammar serves as
an introduction to the study of Akkadian.
The phonemic notations V: and C: are equivalent to the traditional notations V or V (long vowel)
and CC (double consonant).
Abbreviations specifying grammatical categories and functions of glossary entries are patterned
as follows (the items in parentheses are omitted as a rule, the +-marked items only if not applicable):

after nouns: (s) or adj or part, and (m) or /, and (sg) or pi, and (nom) or acc or gen or loc or term
or bf or t f , and + suf or + gen, and (SB) or OAkk or OB or Ass or Ν A ;

after verbs either: ν or ν stem or inf SM,* and + bf;

or : imp and (m) or /, and (sg) or pi, and (/) or SM,* and + vent, and + suf, and + cl;
or: 1 or 2 or 3, and sg or pi, and près or prêt or pf or prec or stat, and (I) or SM*
and + vent, and + suf, and + cl.

The formal equivalent of the above verbalization is {(s), adj, part} + {(m),f} + {(sg), pi} + {(noni),
acc, gen, loc, term, bf, t f } + {(+ suf), (+ gen)} + {(SB), OAkk, OB, Ass, NA}; etc.

acc accusative pers pron personal pronoun

adj adjective Pf perfect
adv adverb pl plural
Ass Assyrian dialect prec precative
bf bound form prep preposition
cl clitic près present
conj conjunction prêt preterit
cp compound pron pronoun
f feminine s substantive
gen genitive SB Standard Babylonian dialect
imp imperative sg singular
inf infinitive stat stative
loc locative suf personal suffix
m masculine term terminative
NA Neo-Assyrian dialect tf terminal form
neg negation V verb (infinitive simple stem)
nom nominative vent ventive
OAkk Old Akkadian dialect 1 first person
OB Old Babylonian dialect 2 second person
part participle 3 third person

* Where SM ( = Stem Marker) stands for II or III or IV or 1/2 or III2 . . . (see pp. 73f.).
1 38 GLOSSARY OF AKKADIAN FORMS

(a)maluktu sf (a priestess) 38
amäriä loc see ama:ru 68
amartu sf 'brick pile' 115
ama:ru V 'to see' 68, 81, 108, 109
amaitu see amartu 115
amatu sf 'word' 37, 113
ammiu pron 'that' 99
amnu 1 sg prêt see manû 107
amSi 1 sg prêt see maàû 110, 114
ana prep 'to, for' 69, 100, 103
an:a: acc see annû 113
a-nam-çar 1 sg près see naçâru 114
an:e:m acc see annû 114
an:i:am acc see annû 113, 114
an-mu-us 1 sg prêt see namaáu 107
annanna pron 'so-and-so' 99
annannïtu f see annanna 99
annû pron 'that' 113, 114
an:u see arnu 114
anza:nis 'like the ...bird' 69
apkakatu f see apkal:u 120
apkaku s 'wise' 120
aqartu f see aqru 118
aqru adj 'precious' 118
ara^'.i 1 sg près see rebû 115
ara:ni§ 'like an eagle' 69
ardu s 'slave' 113
arki prep 'after' 103
arku adj 'long' 64
arku: adj 'later' 134
arnu s 'sin' 114
arqu adj 'green' 118
aruqtu f see arqu 118
asa:tu f see asu: 123
asirtu f see asi:ru 121
asi:ru s 'prisoner' 121
asi:tu sf 'pillar' 116
askup:u s 'slab' 116
asmara:niJ 'like a lance' 69
asmaru: s 'lance' 61
asu: s 'physician' 123
aça:ah 1 sg près Ί laugh' 48
asar conj 'where' 27
aä-ku-un-ma 1 sg prêt + cl see äakänu 107
aSru s 'place' 107, 110
aäruä:u loc + suf see aSru 110
aä-ru-un-ni loc + suf see a§ru 107
aääu conj 'because' 103
aääumi prep 'on account of' 103
aStakan 1 sg prêt see äakänu 113
atalkam imp 1/2 + vent see aläku 53
atlakam imp 1/2 + vent see aläku 53
attamannu pron 'whoever you are' 99
attimannu f see attamannu 99
atu:du s 'wild sheep' 38
at:umuS 1 sgpf see namäSu 108
140 GLOSSARY OF AKKADIAN FORMS

at:ardak:um 1 sg pf + vent + suf see tarad u 111

atruda§:u I sg prêt + vent + suf see (arädu 110
awat bfOB see amatu 113
a-wa-tam acc OB see amatu 37
awi:lu(m) s OB 'man' 37, 113
awi:lu:tim s fgen OB 'mankind' 37,69
awutum nom Ass see amatu 113
ba:alu(m) V 'to be bright* 48,49
baba:lu V 'to carry' 93
[Link]ni s pi 'gates' 63
bäPertu sf 'catcher' 122
baril 3 sg stat see ba:alu(m) 48
ba:it sfbf 'desired' 48
bajjartu sf 'catcher' 122
bakru s 'camel foal' 64,65
*bâlu V see ba:alu(m) 49
ba:ma:nis adv 'in halves' 69
banû V 'to build' 86, 88
baqa:mu V 'to pluck' 116
baqa:nu V see baqa:mu 116
bartu Sf 'revolt' 44
barû V 'to see' 88
ba:u V 'to walk' 48
ba:ula:ti sfpl 'mankind' 48
be:l imp 'rule!' 84
be:l piqine:ti s pi see be:l piqit:i 108, 133
be:l piqitri s 'officeholder' 133
be:latu sf see beltu 121
beltu sf 'lady' 121
be:lu s 'lord' 121
bibla:ni s pi 'gifts' 63
bi:ilti see bi:iltu 48
bi:iltu adjf 'stinking' 44
•bfltu see bi:iltu 49
bina:tu sfpl 'limbs' 113
bina:tu$:u s f p l loc + suf see bina:tu 69
bi-ni imp see banû 88
bini:a:tum see bina:tu 113
bi-ri imp see barû 88
birka :5a s pi + suf '(on) her knees' 69
bi:t sibit:i cp 'prison' 133
bi:ti s gen see bltu 113
bi:tim s gen see bltu 113
bi:tiS emu:tim term + gen 'to the wedding house' 69
bltu s 'house' 68, 69, 113
bi:tuSka toc + suf see bltu 68
burtu sf 'cow' 121
buru: s 'reed mat' 61
bu:ru s 'calf* 121
bur:umtu adjf 'multicolored' 47
bu:ura s acc 'hunt' 48
<da-an> see da:n 124
dabbibu adj 'gossipy' 46
däbibu adj 'gossipy' 46
dabu:e£ 'like a bear' 48
da:iku part 'killer' 48
GLOSSARY OF AKKADIAN FORMS 141

da:iq tf Ass see damqu 113

da:iS part 'threshing' 48
damraqu pltf see damqu 52
damiq tf see damqu 113
damiqtu f see damqu 38, 53, 120
damqatu f see damqu 53, 120
damqu adj 'good' 64, 113
da:n 3 sg stat see dan:u 43, 52, 96
dan:at 3 sgf stat see dan:u 52
dan:atu f see dan:u 119
dan:u adj 'strong' 64, 119
dap:a:nu s pi 'boards' 63
daq:aqu pltf see daqqu 52
daq:atu f see daqqu 119
daqqu adj 'small' 64, 119
da:ri:a:tiS u:mi: f pi term + gen 'forever* 69
da:riS u:mi: term + gen 'forever' 69
da:ri:tu f see da:ru: 119
da:ru: adj 'eternal' 119
-dbub ν stem 'speak' 43, 96, 111
de:iq tf Ass see damqu 49, 114
di:i s gen 'headache' 48
di:na:ni s pi 'laws' 63
dital:u s 'ashes' 112
dita:nu s 'aurochs' 112
-dnin ν stem 'become strong' 96
du:tu s 'virility' 112
du-ub-ba imp pi see -dbub 43
du:um adj tf 'dark' 48
e:a(:u) s 'leather bearing of door pivot' 49
eb:etu f see eb:u 120
<e-bi> tf see eb:u 124
ebru s 'friend' 64
eb:u adj 'clean' 64, 120
eb:u:bu s 'flute' 114
ebu:ra:nu s pi 'crops' 63
e:d tf see ed:u 87
ed:etu f see ed:u 120
ediStu f see eääu 51
ed:u adj 'pointed' 120
<e-eb> tf see eb:u 124
[Link]lu V 'to bind' 115
<e-ez> tf see ez:u 124
ek:al 3 sg près Ass see akälu 114
ek:ulu 3 pi près Ass see akälu 90
eia prep 'apart from' 103
elami:tu f see elamu: 119
elamu : adj 'Elamite' 119
el:etu f see el:u 120
eli prep 'over' 103
el:u adj 'pure' 120
em(a) prep 'where' 103
embu:bu see eb:u:bu 114
em:etu f see em:u 120
emSu s 'hypogastric region' 110
em:u adj 'hot' 120
142 GLOSSARY OF AKKADIAN FORMS

e:netu / see e:nu 121

entu / see e:nu 121
e:nu Í 'high priest' 121
epe:äu V 'to do' 87, 109
e:piStu:a part f + suf see e:piäu:a 47
e:piäu:a part m pi + suf 'sorcerer' 47
êpltu f see ëpû 119, 122
êpû s 'baker' 119, 122
eqla:tu Pi see eqlu 121
eqlu s 'field' 26, 121
era:sis term OAkk 'for seeding' 68
ere:bu V 'to enter' 84, 87, 90f., 107
erçetu sf 'earth' 27
erub imp see ere:bu 84
eääetu f see eääu 51, 120
eSSu adj 'new' 51
ez:etu f see ez:u 119
<e-zi> tf see ez:u 124
ezib prep 'apart from' 103
ez:u adj 'furious' 119
gabaru: s 'copy' 61
gadu prep 'together with' 103
gala:du see gala:tu 115
gala:tu V 'to tremble' 115
garu: Ass see geru: 114
gaz:atu / see gaz:u 119
gaz:u adj 'shorn' 119
geru: V 'to become hostile' 114
guq:u: s 'offering' 61
habâtu V 'to make a razzia' 81
t)abul:u Ass see bubul:u 75
badû V 'to rejoice' 88
balqu adj 'run-away' 64
bamsu adj 'fifth' 108
banbu adj 'luxuriant' 64
bapa:ru see bepe:ru 85
fearu: s 'vat' 61
baru: see beru: 115
baä:u see bamSu 108
beperru V 'to dig' 85
beru: V 'to dig' 88, 115
bi-di imp see badû 88
¡¡ji-mi-iç-çe-tim cp see bimit and ?e:tim 111
himit sbf 'fever' 111
hi-pa-a imp pi 'break!' 88
¿i-pi imp 'break!' 88
hi-ri imp see beru: 88
hisib sbf 'yield' 107
bi-?i-im-ma-tim see biçib and ma:tu 107
bubul:u s 'debt' 75
bu-di-i impf see badû 88
buptu s 'hole' 116
butpu s 'hole' 116
bu-ú-du imp see badû 88
i:ad:aru 3 pi près IV 'become eclipsed' 48
i:ar:ur 3 sg près 'trembles' 86
GLOSSARY OF AKKADIAN FORMS 143

iba:a 3 sg près see ba:u 48

ib:al 3 sg près see aba:lu 52
ib:aru see imbaru 114
ib:at:aq 3 sg près IV 'is cut off* 81
ib:atiq 3 sg prêt IV 'was cut off' 81
ibi:ad 3 sg près 'spends the night' 52
idrabub 3 sg pf see -dbub 111
ida:§a spi + suf '(at) her side' 69
idbub 3 sg prêt see -dbub 96
Idiglat-eä term 'to the Tigris' 69
id:ik:um 3 sg prêt + suf see nada:nu 112
iduS ti:amat loc + suf 'at the side of Tiamat' 68
igan:ib 3 sg près 'coughs' 115
igdamar 3 sg ρf 'finished' 114
igen:ib see igan:ib 115
i[Link]t 3 sg près IV see ljabätu 81
it>:ar:a$ 3 sg près IV 'is deducted' 81
ifrari? 3 sg prêt IV 'was deducted' 81
it:as:as 3 sg près IV 'is remembered' 81
itcasis 3 sg prêt IV 'was remembered' 81
i-ib-bu-ú-m s OB 'loss' 87
i-iç-tum see içtum 111
i-iä-tum see i$tum 111
i-ka-al 3 sg près see akälu 87
ik:al 3 sg près see akälu 114
ik:an:ak 3 sg près IV 'is sealed' 81
ik:anik 3 sg prêt IV 'was sealed' 81
ik:is 3 sg prêt see naka:su 109
ik-ki-im-ma 3 sg prêt + cl see nakäpu 107
ik-ki-ip-ma 3 sg prêt + cl see nakäpu 107
iktanar:u 3 sg près 1/3 'becomes short' 115
iktener:u see iktanar:u 115
iku:an 3 sg près Ass see ku:n 52
ik:ul:atu ila:tim loc + gen 'among the totality of goddesses' 68
iku:n 3 sg prêt see ku:n 52
iku:nu 3 pi prêt see ku:n 52
i-ku-ul 3 sg prêt see akälu 87
ila:ni Pi see ilu 63
ila:tim fpi gen see ilu 68
ildu see iSdu 113
ilid bf 'offspring* 125
<i-lid/t> bf see ilid 125
<i-li-id/t> bf see ilid 125
<i-li-it-ti> bf see ilitru 125
il:ikunik:um 3 pi prêt + vent + suf see aläku 112
iiiS term see ilu 68
<i-lit-ti> bf see ilit:u 125
ilit:u sf 'offspring' 113
iltakan 3 sgpf see Sakänu 113
iltur 3 sg prêt see Satäru 113
ilu s 'god' 27
im:angar 3 sg près IV see maga:ru 114
imaä:a:u 3 pi près see maáa?u 86
ima&u 3 pi près see maSû 86
imbaru s 'fog' 114
imbi 3 sg prêt see nabû 114
144 GLOSSARY OF AKKADIAN FORMS

imdud 3 sg prêt 'measured' 114

im:ertu f see im:eru 117
im:eru s 'sheep' 117
imna adv 'at the right' 107
imti 3 sg prêt 'diminished' 114
ina prep 'in' 103
in:ad:ar 3 sg près IV 'becomes eclipsed' 81
in:adir 3 sg prêt IV 'became eclipsed' 81
in:ah:az 3 sg près IV 'gets married* 81
in:ahiz 3 sg prêt IV 'got married' 81
in:ak:al 3 sg près IV see akälu 81
in:am:ar 3 sg près IV see ama:ru 81
inambu 3 sg près 'moans' 114
in:amir 3 sg prêt IV see ama:ru 81
inandi 3 sg près 'throws' 114
inanziq 3 sg près 'is worried' 114
in:ap:al 3 sg près IV 'is paid' 81
in:apil 3 sg prêt IV 'was paid' 81
inbu s 'fruit' 108
indud see imdud 114
in(i see imti 114
ipar:as 3 sg près see -prus 52
ip:aris 3 sg prêt IV see -prus 108
iplah 3 sg prêt 'feared' 77
i-qa-am-ma see i-qa-ap-ma 107
i-qa-ap-ma 3 sg près + cl 'buckles' 107
iqar:ib 3 sg près see qara:bu 114
iqbe:m 3 sg prêt + vent see qabû 114
iqbiram 3 sg prêt + vent see qabû 114
iqer:ib 3 sg près see qara:bu 114
iqtibi 3 sg pf Ass see qabû 51, 114
ira?ub 3 sg près 'rages' 92
ira:ubu 3 pi près 'rage' 86
ired:iam 3 sg près + vent 'follows' 89
ireb:i 3 sg près see rejjû 115
iri:ab 3 sg près see ra:bu 52
irnit:aka bf + suf 'your victory' 125
irnit:i bf(+ suf) '(my) victory' 125
irtuS loc + suf 'on his chest' 68
irub:u 3 pi près 'quake' 86
ir:ubu 3 pi près see ere:bu 90 f.
ir:ur 3 sg près 'trembles' 86
i-ru-ub-ma 3 sg prêt + cl see ere:bu 107
i-ru-um-ma 3 sg prêt + cl see ere:bu 107
is:ah:ar 3 sg près IV 'is turned' 81
is:akan 3 sg pf Ass see äakänu 114
is:anah:ur 3 sg près I¡3 'turns' 110
isi:tu see asi:tu 116
is-ki-im-ma see is-kip-ma 107
is-ki-in-ma see is-kip-ma 107
is-kip-ma 3 sg prêt + cl 'threw down' 107
i?:abat 3 sg Ρ f 'seized' 110
i§:ab:at 3 sg près IV 'is seized' 81
is:abit 3 sg prêt IV 'was seized' 81
içtum adjf 'few' 111
iäa:am 3 sg près 'buys' 92
GLOSSARY OF AKKADIAN FORMS 145

is + al 3 sg prêt 'asked' 35
isa:m 3 sg prêt 'bought' 92
isdu s 'foundation' 113
iSparu s 'weaver' 116
iS-pu-ra-an-ni 3 sg prêt + suf 'sent me' 107
iStakan 3 sgpf see Sakänu 114
iätaäni see uätaäni 116
iäti prep 'with' 103
iätu prep 'from' 103
iätum see içtum 111
iätur 3 sg prêt see äatäru 113
it:akis 3 sgpf see naka:su 108
ital imp 'lie down!' 77
it:andin 3 sg prêt IV¡3 see nada:nu 109
it:anmar 3 sg prêt IV/3 see ama:ru 109
it:apras 3 sg prêt ¡VI3 see -prus 108
itapras imp IVI3 see -prus 108
itaprusu infIV/3 see -prus 108
itat:uku infIV/3 'to drip' 108
it:enpu$ 3 sg prêt IV¡3 see epe:§u 109
it-tan-am-pa-hu 3 pi près 1/3 'become distended' 114
itti prep 'with' 27, 103
itu: s 'border' 61
ituar 3 sg près Ass see târu 36
ituwar see ituar 36
iz:anak:ar 3 sg près I¡3 'speaks' 110
jamattu proti 'each' 99
jamuttu pron 'each' 99
kaba:ru ν 'to be thick' 85
kabtu adj 'heavy' 64
kada:ru ν 'to fence in' 85
kadur:u /lis see kudunu 75
kaju s 'goad' 38
kak:aba:ni s pi 'stars' 63
kakiabis Sama: mi term + gen 'like the stars of the heavens' 69
kalbatu f see kalbu 120
kalbu s 'dog' 120
kalbu: s pi see kalbu 66
kan§u adj 'humble' 108
kapa:ru ν 'to wipe' 85
kara:bu ν 'to bless' 85
karaSki bf + suf 'your belly' 53
karSaki see karaski 53
kaça:ru ν 'to tie' 85
kaä:aptu f see kaä:a:pu 121
kaá:a:pu s 'sorcerer' 121
kaziztu f see kazzu 120
kazzu adj 'cut' 120
keze:ru V 'to trim' 85
kî prep 'like' 103
kil- ν stem see -kla 77
ki:nu adj 'true' 121
kiä[Link]nus:u loc + suf 'on his neck' 69
kitpad imp J/2 'mind!' 98
kit:u f see ki:nu 121
-kla ν stem 'hold back' 77
146 GLOSSARY OF AKKADIAN FORMS

kuatim pers prort 2 sg 'you' 36

kudur:u s 'boundary' 75
kul:atu sf 'totality' 68
kümu prep 'instead' 103
kumu:a prep + suf see kümu 48
ku:n ν stem 'be firm' 52
kunuk bf 'seal' 52
kuru: adj 'short' 64
kuwatim see kuatim 36
la neg 'not' 25
la bassu neg + adj 'fool' 134
laba:nu V 'to make bricks' 85
laba:ru V 'to grow old' 85
lab:atu / see lab:u 120
lab:u s 'lion' 120
lam(a) prep 'before' 103
la-wi 3 sg stat OB 'surrounded' 37
laziz tf see lazzu 43
laziztu f see lazzu 120
lazzu adj 'persistent' 120
-IbiS ν stem 'dress' 110
lebe:nu see laba:nu 85
lemnu adj 'evil' 118
lemut:u f see lemnu 114, 118
le:u adj 'able' 48
li:ab:it 3 sg prec IV 'let be destroyed' 48
lil:ikak:im 3 sg prec + vent + suf see aläku 112
limd- ν stem see -lmad 77
liqia imp pi Ass 'take!' 89
litabSu see litbuSu 53
litamdu see litmudu 53
litbuSu in/I/2 see -lbiä 53
litmudu in/1/2 see -lmad 38, 53
<lit-$u-ru> 3 pi prec see naçâru 111
-lmad ν stem 'learn' 77
lu conj 'or' 25
lumnu s 'evil' 107
luSab:i:u 3 pi prec II Ass 'may they satisfy' 48
lu-äal-bu-ku-nu 3 pi prec III + suf Ass 'may they surround you' 37
lu-uS-ku-um-ma 1 sg prec + cl see äakänu 107
m[Link]u V 'to be numerous' 35, 48
magarru ν 'to be favorable' 85, 114
mahar prep 'before' 103
maba:ru V 'to accept' 85
ma^ru: adj 'former' 134
mafru: s 'ecstatic' 119
mat:u:tu f see mafru: 119
mala pron 'as much as' 134
malkatu f see malku 120
malku s 'ruler* 120
mamman pron 'whoever' 99
mannu pron 'who' 98
manû V 'to count* 88, 107
manzaztu sf 'position' 111
mar:atu f see mar:u 119
ma:ra:tu fpl see ma:ru 121
GLOSSARY OF AKKADIAN FORMS 147

marfraçu s 'lotion' 30
marçu adj 'sick' 118
ma:rtu f see ma:ru 121
martu f see ma:ru 121
ma:ru s 'son' 121
mar:u adj 'bitter' 119
maruStu f see marçu 114, 118
mas:u bf + suf see ma:tu 109
ma$a?u V 'to plunder' 86, 89
maáú V 'to forget' 86, 89, 110, 114
ma:tu s 'land' 56
maz:aätu see manzaztu 111, 114
meke:ru V 'to irrigate' 85
mimma pron 'what' 99
mimmü pron 'whatever' 98
minduda 3 f p l stat I¡2 'measured' 111
minu pron 'what' 98
miSilSa bf+ suf 'her half' 68
miälu$:a loc + suf 'her half' 68
mit^a? imp I¡2 'fight!' 98
miti:tu sf 'diminution' 112
mu:a:tiá term Ass 'to death' 68
mubal:it:u sf 'fishpond' 111
mu:ir:u s 'leader' 48
rau-na-a imp pi see manû 88
mupanistu part f II see -prus 117
mupar:isu part II see -prus 117
mu$:abirtu f see muç:abru 118
mu?:abru part 112 'squinting' 118
mu:çu: s 'exit' 61
naba:lju V 'to bark' 94
na:bal(u) i 'dry land' 109
nabû ν 'to name' 114
nada:nu ν 'to give' 109,112
nafrsat see nashat 116
nakäpu ν 'to gore' 107
naka:su ν 'to cut' 94, 108, 109
nakru ί 'enemy' 64
na:mar(u) i 'tower' 109
namäSu ν 'to move' 107, 108
namzi:tu if 'mixing vat' 112
nanduru inflV 'to become enraged' 108
nanduru inflV 'to become dark' 84
nanmuru inflV see ama:ru 108
napla:su§:a loc + suf 'at her glance' 68
naqa:ru V 'to tear down' 94
na:rim s gen 'river' 114
nasljat ftf 'excerpted' 116
nasi:ka:ni s pi 'sheikhs' 63
naçâru V 'to guard' 114
nenduru inflV 'to embrace' 84, 108
ne:piS(u) s 'ritual' 109
nidintaka bf + suf 'your gift' 125
nidinti bf 'gift' 125
nidnatka bf + suf 'your gift' 125
nitbi 1 pi prêt 'we got up' 108
148 GLOSSARY OF AKKADIAN FORMS

nudun:u: s 'dowry' 61
nutar:ak:um 1 pi près II + vent + suf 'we give back to you' 108
pagra:nu s pi 'carcasses' 63
patitila term + suf 'to her chest' 69
panu:a spi + suf 'my face' 47
pa:nuk:a loc + suf 'before you' 68
pa:nuSka loc + suf see pa:nuk:a 68
papa:ha:ni s pi 'chapels' 63
pa:risa:tu part f pi see -prus 121
pa:ristu partf see -prus 117, 121
paristu f see parsu 118
pa:risu part see -prus 117
parsat ftf see parsu 52
parsa:tu fpl see parsu 121
parsu adj 'separated' 118
pasa:mu V 'to veil' 116
pasa:nu see pasa:mu 116
pa£a:lu V 'to crawl' 115
pa&Su adj 'anointed' 109
pata:ru V 'to loosen' 85
peSe:lu see pa$a:hi 115
pilal) imp •fear!' 77, 98
piqit:u *f 'commission' 108
piriätu sf 'secret' 114
pirku s 'wrong' 115
pi§:aä imp 1/2 'anoint yourself!' 111
piSb- see -pSab 77
piSku s see pirku 115
pitarías imp 1/3 see -prus 98
-prus ν stem 'separate' 52, 76, 98, 108, 117, 121
-pSab ν stem 'quiet' 77
puhriä:un term + suf 'to their assembly' 69
pursi:tu sf 'jar' 112
puräumtu f see puräumu 117
pursumu s 'old man' 117
qabaltu $e:ri loc + gen 'in the middle of the plain' 68
qabli§:a term + suf 'to her middle' 69
qabû V 'to speak' 51, 114
qadiStu / see qaSdu 51, 118, 120
•qadäu see qaSdu 109
qal:atu / see qal:u 119
qal:u adj 'light' 119
qaq:arsum loc + suf 'to her earth' 69
qara:bu V 'to approach' 85, 114
qardu adj 'valiant' 111
qarit:u f see qardu 111
qa£:atu f see qaSdu 51, 120
qaSdu adj 'holy' 50, 109, 118
qata:ru V 'to smoke' 85
qa:ta:tu sfpl 'shares' 121
qa:ti£:u term + suf see qa:tu 69
qa:tu sf 'hand' 121
qebe:ru V 'to bury' 85
qere:bu V see qara:bu 85
qibi:tuä:a loc + suf 'at her command' 68
qi:S ν stem 'give as present' 110
GLOSSARY OF AKKADIAN FORMS 149
rab a:li cp 'mayor' 133
rab bu:li cp 'overseer of cattle' 133
raba:tu fpl see rabu: 121
rab:atu f see rab:u 119
rab-bu:la:nu Pi see rab bu:li 63, 133
rabirat ftfOB see rabu: 49, 119
rabi:tu f see rabu: 119, 121
rabiu OB see rabu: 119
rabu: adj 'great' 64, 119
ra:bu V 'to accrue' 52, 113
rab:u adj 'soft' 119
rapaätu f see rapsu 118
rapäu adj 'broad' 118
raq:atu f see raq:u 119
raq:u adj 'thin' 119
re^û V 'to impregnate' 115
ri:a:bum OB see ra:bu 113
rigim bf 'noise' 113
rigmiSka term + suf 'at your noise' 69
rikab imp see -rkab 98
rikb- ν stem see -rkab 77
rikis:u bf+ suf 'his bond' 109
ritap:ud imp 1/3 'roam!' 98
-rkab ν stem 'ride- 77, 98
ruba:tu f see rubu: 44, 119
rubu: s 'prince' 119
ru:u:a s pi + suf 'my friends' 48
sekerru V 'to dam up' 85
suqtu s 'chin' 115
ç[Link]i:tu f see $[Link]u 121
ç[Link]u adj 'roaming' 121
sabru see ?el)ru 114
safrurtu Ass see çujjurtu 75
$alma:ni s pi 'statues' 63
salu:lu Ass see $uhi:lu 75
?ara:m inf + bf 'wish' 113
sebertu f see çeftru 118, 120
çehretu f see çehru 120
sehru adj 'small' 64, 114, 118
serxtu f see $er:u 120
ser:u s 'rival' 120
?e:ruä:u loe + suf 'toward him' 68
se:tim gen 'heat' 111
*$il:i§ Dagan term 'into the shade of Dagan' 69
sil:um Suen loc 'in the shade of Suen' 68
çimdatu sf 'decree' 122
çimittu sf 'team' 122
çiru: s 'chieftain' 61
$i:taä adv 'at sunrise' 112
si:tu sf 'sunrise' 112
çu^urtu sf 'adolescence' 75
sulu:lu s 'protection' 75
?upra:nus:u loc + suf 'in his claws' 69
Sa pron 'that, which' 99
sa biltim cp 'carrier' 133
sa libbakni pron + bf + suf + cl 'what your heart (prompts)' 53
150 GLOSSARY OF AKKADIAN FORMS

Sada:ni í pi 'mountains' 63
sakänu V 'to place' 107, 112ff.
Sa:l imp 'ask!' 84
SamSu s 'sun' 108
äamü s 'heaven' 27
Sanatka bf+ suf 'your year' 125
Sapa:ru V 'to send' 84, 107
Sa-pi-in-na-ri-im see Sa:pir + na:rim 114
Sa:pir + na:rim cp 'overseer of canals' 114
Saplirtu f see Saplu: 119
saplu: adj 'lower' 119
Sar:a:qi:tu f see sar:a:qu 121
sar:a:qu s 'thief' 121
Sar:a:tu fpl see Sar:u 121
sar:atu f see sar:u 120, 121
äarruttu sf 'kingship' 45
sar:u s 'king' 120
SaS:u see samSu 108
Sat:i bf + suf 'my year' 125
satäru V 'to write' 113
Sebe:ru V 'to break' 84
Seltu / see se:lu 121
Se:lu adj 'sharp' 121
Se:mu:m adj 'hearer' 64
Se:puä:u loc + suf 'at his feet' 68
Sibtu f see Si:bu 121
si:bu s 'old man' 121
äim- ν stem see -Sma 77
Si:ma:tu sf p l 'fate' 69
Sipir bf 'message' 125
SipirSu bf + suf see Sipir 125
Sipri bf+ suf see Sipir 125
Sitamme:a imp pi 1/3 see -Sma 112
Sitannu see Sitnunu 53
äi-ti imp 'drink!' 88
Si-ti-í impf 'drink!' 88
Sitnunu inflß 'to rival' 53
-Sma ν stem 'hear' 77, 112
sulma:niS term 'in prosperity' 69
áulum bf 'well-being' 113
sumí bf + suf 'my name' 125
Sumka bf + suf 'your name' 125
Sum + ud infill + bf see m[Link]u 35
ta-an-zi-im-ti sfbf 'complaint' 109
ta-aä-pur-ra-am-ni-si 2 sg prêt + vent + suf see Sapa:ru 107
taba:a 2 sg près see ba:u 48
tabarlu V 'to carry off' 93
taljsistu sf 'memorandum' 111
tahsiätu see tajjsistu 111
taklu s 'trusted' 64
tak$i:r bf 'bandage' 30
tak?i:ra:ni s pi see tak$i:r 63
talaprap 2 sg près 'you wrap' 30
tal-bi-tu Ass see tal-wi-tu 37
*tal-wi-tu sfOB 'surroundings' 37
tarak:as 2 sg près 'you tie' 30
GLOSSARY OF AKKADIAN FORMS 151

târu ν 'to return' 36

ta$a:ab 2 sg près 'you laugh' 48
te:a(ti) s 49
ti-ba-a imp pi 'get up!' 88
ti-bi imp 'get up!' 88
tidkuSat 3 sgf stai 1/2 'is punctured' 112
tidu:ku 3 pi stat 1/2 'were fighting' 112
tikl- ν stem see -tkal 77
tiçbut 3 sg stat 1/2 'is joined' 112
ttëab imp 'sit!' 77
tiSamme:a see äitammera 112
ti:u:tu sf 'food' 48
tizkar imp 1/2 see zaka:ru 77
tizqar adj tf 'sublime' 112
-tkal ν stem 'trust' 77
tuSeäkun infIII/2 + bf see sakänu 112
tuzaq:uru infll/3 'to be sublime' 112
ta:bat ftf see ta:bu 52
tabtu f see ta:bu 52, 121
ta:bu adj 'sweet' 121
tabu: see tebû 115
tarädu V 'to dispatch' 110, 111
tefcu: V 'to approach' 115
te:m bf 'counsel' 113
titabbû infl/3 'to become submerged' 112
tup bf 'tablet' 125
tup:aka bf + suf 'your tablet' 125
tup:a:ni s pi 'tablets' 63
tup:i bf 'tablet* 125
u COtlj 'and' 25
ubar:ak:u£:u 1 sg près II + vent + suf + suf Ί convict him to you' 110
uba:u 3 pi près II 'they search' 48
ubluniS:um 3 pi prêt + vent + suf 'they brought to him' 110
ubap:i see ul)ep:i 115
u¿ep:i 3 sg prêt II 'he broke' 115
uftultu sf 'lye' 75
ukultu sf 'food' 75
ullû pron 'that' 99
urdu Ass see ardu 113
usbat Ass see waSbat 114
uSa:bal 3 sg près III 'he sends' 93
uSab:al see u§a:bal 93
uSa:bil 3 sg prêt III 'he sent' 93
<uSamsak> 3 sg près III 'he has (it) thrown' 114
uSapras 3 sg près III see -prus 52
u£:is:ina:ti 3 sg prêt II + suf 'caused them troubles' 109
uSparu see iäparu 116
uS-Sa-ab-ma 3 sg près + cl 'he sits' 107
uä-äa-am-ma see u$-Sa-ab-ma 107
uStaSni 3 sg pf III 'he gave a second time' 116
utaz:umu infII/3 'worry' 108
wardum OB see ardu 113
waSbat 3 sg f stat OB 'is sitting' 114
wilitrum OB see ilit:u 113
zab:atu f see zab:u 120
zabbilu s 'carrier' 46
152 GLOSSARY OF AKKADIAN FORMS

zäbilu s 'carrier' 46
zab:u s 'ecstatic' 120
zakarru ν 'to speak' 77, 85
zaki:tu / see zaku: 119
zaku: adj 'clean' 119
zakuat f t f Ass see zaku: 119
zap:u s 'bristle' 38
zaqi:pa:ni Pi see zaqi:pu 63
zaqi:pu s 'stake' 116
zikaru s 'male' 64
ziqirpu see zaqi:pu 116
*zlru s 'hatred' 49
zitlunu: adj 112
zit:u sf 'share' 112
zuqtu see suqtu 115
zu:uzti bf 'division' 48
GLOSSARY OF SELECTED LINGUISTIC TERMS

Terms used in reference to the writing system :

GRAPHEME: A grapheme is a bundle of concurrent graphic distinctive features, representing one or
more VALUES.
GRAPH : A graph is a realisation of a GRAPHEME.
ALLOGRAPH: An allograph (or diagraph) is a variant of a GRAPHEME.
SIGN: A sign is a GRAPHEME in the cuneiform writing system.
VALUE : A value of a GRAPHEME is a PHONEME o r a sequence of phonemes represented by the grapheme.
TRANSLITERATION: A transliteration of a sequence of GRAPHEMES is a corresponding sequence of
their VALUES.
SYLLABLE: A syllable (in writing, see pp. 27f.) is a GRAPHEME referring to a PHONEME or a sequence
of phonemes, one of which is syllabic (here, a vowel).
SYLLABARY: A syllabary is an inventory of (written) SYLLABLES; e.g., the cuneiform syllabary is the
inventory of cuneiform SIGNS, i.e., the written syllables, and their corresponding VALUES.

Terms used in reference to phonology and morphology:

PHONEME: A phoneme is a bundle of concurrent (acoustic) DISTINCTIVE FEATURES serving primarily
to differentiate MORPHEMES and whole words. One or more phonemes may constitute one
morpheme; for zero phoneme, see pp. 35 and 37.
PHONE: A phone is a realisation of a PHONEME.
DISTINCTIVE FEATURES: Distinctive features are the ultimate components of the speech-stream
capable of differentiating MORPHEMES from each other; their combinations into bundles of
concurrent features are termed PHONEMES.
PROSODY : A prosody (or prosodie feature) may be either distinctive or non-distinctive. Prosodies
are of three types : tone, stress, and vowel and consonant duration (length).
PHONEMICS: Phonemics deals with the distinctive function of speech sounds.
MORPHEME: A morpheme is the smallest meaningful unit in the language (such as a stem, a root, an
affix, a PATTERN). A morpheme represents a class of MORPHS and is here sometimes referred to
by naming one of its morphs.
M O R P H : A morph is a realisation of a MORPHEME.
ALLOMORPH : An allomorph is a (positional or free) variant of a MORPHEME.
MORPHOPHONEMIC ALTERNANT: A morphophonemic alternant is a phonemic alternant from which
other alternants can be predicted in terms of morphological environment.
DISCONTINUOUS MORPHEMES: Discontinuous morphemes are MORPHEMES which are interdigitated
into words or stems (see p. 54) ; in the Semitic languages, a consonantal discontinuous morpheme
(the root) and a vocalic discontinuous morpheme (the PATTERN) are interdigitated.
PATTERN: A vocalic DISCONTINUOUS MORPHEME is called pattern.
BIBLIOGRAPHY

Aro, 1953 = Aro, J.( Abnormal Piene Writings in Akkadian Texts ( = Studia Orientalia,
XIX/11) (Helsinki, 1953).
Aro, 1955 = Aro, J., Studien zur mittelbabylonischen Grammatik (= Studia Orientalia, XX)
(Helsinki, 1955).
Birkeland, 1954 = Birkeland, H., Stress Patterns in Arabic (Oslo, 1954).
Cantineau, 1946 = Cantineau, J., "Esquisse d'une phonologie de l'arabe classique", BSL, 43 (1946),
93-140.
Cantineau, 1950 = Cantineau, J., "La notion de schème et son altération dans diverses langues
sémitiques", Semitica, 3 (1950), 78-83.
Chao, 1961 = Chao, Y. R., "Graphie and Phonetic Aspects of linguistic and mathematical
symbols", Proceedings of the Symposium on the Structure of Language and its
Mathematical Aspects, American Mathematical Society, vol. 12 (1961), 69-82.
Chomsky, 1964 = Chomsky, N., "The logical basis of Linguistic Theory", Proceedings of the
Ninth International Congress of Linguists, August 27-31,1962, Cambridge, Mass.,
(The Hague, 1964), pp. 914-1008.
Christian, 1913 = Christian, V., Die Namen der assyrisch-babylonischen Keilschriftzeichen
(= MVAG, 18/1) (1913).
Dávid, 1945 = Dávid, Α., "Le terme [Link].-siga", Acta Societatis Hungaricae Orientalis, 5-12
(= Orlens Antiquus) (Budapest, 1945), 5-19.
Falkenstein, 1960 = Falkenstein, Α., "Kontakte zwischen Sumerern und Akkadern auf sprachlichem
Gebiet", Genava, VIII (1960), 301-314.
Fleisch, 1956 = Fleisch, Η., VArabe Classique. Esquisse d'une Structure Linguistique (= Re-
cherches publiées sous la direction de l'Institut de Lettres Orientales de Beyrouth,
Tome V) (Beyrouth, 1956). (Also = Mélanges de V Université Saint Joseph, Tome
XXXIII, Fase. 1).
Gelb, 1948 = Gelb, I. J., Memorandum on Transliteration and Transcription of Cuneiform sub-
mitted to the 21st International Congress of Orientalists, Paris (Chicago, 1948).
(Mimeographed.)
Gelb, 1952 = Gelb, I. J., A Study of Writing (Chicago, 1952), 2d Rev. Ed. (1963).
Gelb, 1955 = Gelb, I. J., "Notes on von Soden's Grammar of Akkadian", BiOr, 12 (1955),
93-111.
Gelb, 1961a = Gelb, I. J., Old Akkadian Writing and Grammar (= Materials for the Assyrian
Dictionary, No. 2). Second edition, revised and enlarged (Chicago, 1961).
Gelb, 1961b = Gelb, I. J., "WA = aw, iw, uw in Cuneiform Writing", JNES, 20 (1961), 194-196.
Gleason, 1961 = Gleason, Η. Α., An Introduction to Descriptive Linguistics. Revised Edition
(New York, 1961).
Goetze, 1946 = Goetze, Α., "The Akkadian Masculine Plural in änü/i and its Semitic Back-
ground", Language, 22 (1946), 121-130.
Goetze, 1958 = Goetze, Α., "The Sibilants of Old Babylonian", RA, 52 (1958), 137-149.
Greenberg, 1950 = Greenberg, J. H., "The patterning of root morphemes in Semitic", Word, 6
(1950), 162-181.
Greenberg, 1952 = Greenberg, J. H., "The Afro-Asiatic (Hamito-Semitic) Present", J AOS, 72
(1952), 1-9.
BIBLIOGRAPHY 155
Greenberg, 1960 = Greenberg, J. H., "An Afro-Asiatic Pattern of Gender and Number Agree-
ment", JAOS, 80 (1960), 317-321.
Harris, 1941 = Harris, Z., "Linguistic Structure of Hebrew", J AOS, 61 (1941), 143-167.
Heidel, 1940 - Heidel, Α., The System of the Quadriliteral Verb in Akkadian (= AS, 13) (Chi-
cago, 1940).
Kienast, 1961 = Kienast, B., "Weiteres zum R-Stamm des Akkadischen", JCS, 15 (1961), 59-61.
Kienast, 1963 = Kienast, B., "Das System der zweiradikaligen Verben im Akkadischen", ZA,
N.F. 21 (55), 138-155 (1963).
Landsberger, 1935 =Landsberger, Β., "Die Gestalt der semitischen Wurzel", Atti XIX Congr. degli
Orientalisti (1935), 450-452.
Mcintosh, 1956 = Mcintosh, Α., "The Analysis of Written Middle English", Transactions of the
Philological Society, 62 (1956), 26-55.
Moscati, 1960 = Moscati, S., Lezioni di linguistica semitica (Roma, 1960).
Poebel, 1939 = Poebel, Α., Studies in Akkadian Grammar (= AS, 9) (Chicago, 1939).
Reiner, 1964 = Reiner, E., "The Phonological Interpretation of a Subsystem in the Akkadian
Syllabary", Studies Presented to A. Leo Oppenheim (Chicago, 1964).
von Soden, 1960 = von Soden, W., "Status Rectus-Formen vor dem Genitiv im Akkadischen und
die sogenannte uneigentliche Annexion im Arabischen", JNES, 19 (1960),
163-171.
von Soden, 1961 = von Soden, W., "Akkadisch", in Levi della Vida, ed., Linguistica Semitica:
Presente e Futuro ( = Università di Roma - Centro di Studi Semitici, Studi Semi-
tici, 4) (Roma, 1961).
JANUA LINGUARUM

STUDIA MEMORIAE NICOLAI VAN WIJK DEDICATA

Edited by Cornells H. van Schooneveld

SERIES PRACTICA

1. MARILYN CONWELL and ALPHONSE JUILLAND : Louisiana French Grammar. I : Phonology,

Morphology, and Syntax. 1963. 207 pp., 2 maps. Cloth. Gld. 36 —
4. MORRIS F. GOODMAN : A Comparative Study of Creole French Dialects. 1964.143 pp., map.
Gld. 22.—
5. ROLAND HARWEG: Kompositum und Katalysationstext, vornehmlich im späten Sanskrit.
1964.164 pp. Gld. 25.—
6. GUSTAV HERDAN: The Structuralistic Approach to Chinese Grammar and Vocabulary.
Two Essays. 1964. 56 pp., 4 figs. Gld. 18 —
7. ALPHONSE JUILLAND: Dictionnaire Inverse de la Langue Française. 1965. 564 pp., 9 figs.
Cloth. Gld. 80.—
8. A. HOOD ROBERTS: A Statistical Linguistic Analysis of American English. 1965. 437 pp.,
11 figs., 6 tables. Cloth. Gld. 54.—
9. VALDIS LEJNIEKS: Morphosyntax of the Homeric Greek Verb. 1964. 92 pp. Gld. 15.—
10. ROBERT E. DIAMOND: The Diction of the Anglo-Saxon Metrical Psalms. 1963. 59 pp.
Gld. 10.—
11. JOSEPH E. GRIMES: Huichol Syntax. 1964. 105 pp. Gld. 18.—
12. CLARA Ν. BUSH:Phonetic Variation and Acoustic Distinctive Features. A Study of Four
General American Fricatives. 1964. 161 pp., 64 figs., 84 tables. Gld. 30.50
13. WILLIAM E. CASTLE:The Effect of Selective Narrow-Band Filtering on the Perception of
Certain English Vowels. 1964. 209 pp., 53 figs., 84 tables. Gld. 38 —
14. ANN SHANNON: A Descriptive Syntax of the Parker Manuscript of the Anglo-Saxon
Chronicle from 734-891. 1964. 68 pp. Gld. 13.—
15. EICHI KOBAYASHI: The Verb Forms of the South English Legendary. 1964. 87 pp.
Gld. 17.—
17. WOLF LESLAU: Ethiopian Argots. 1964. 65 pp. Gld. 16.—

M O U T O N & CO · P U B L I S H E R S · THE H A G U E

(Mouton Grammar Library 37) Knut J. Olawsky - A Grammar of Urarina-Mouton de Gruyter (2006)
100% (1)
(Mouton Grammar Library 37) Knut J. Olawsky - A Grammar of Urarina-Mouton de Gruyter (2006)
964 pages
Hoffner-Melchert Hittite PDF
100% (1)
Hoffner-Melchert Hittite PDF
574 pages
Alan S. Kaye - Phonologies of Asia and Africa (Including The Caucasus) Volume 1&2
No ratings yet
Alan S. Kaye - Phonologies of Asia and Africa (Including The Caucasus) Volume 1&2
1,074 pages
Adiego 2007 (The Carian Language)
100% (1)
Adiego 2007 (The Carian Language)
541 pages
Amurru Akkadian - A Linguistic Study - Shlomo Izre'el - 1991
100% (1)
Amurru Akkadian - A Linguistic Study - Shlomo Izre'el - 1991
257 pages
Militarev Semitic2011
No ratings yet
Militarev Semitic2011
27 pages
An Akkadian Handbook Helps, Paradigms, Glossary, Logograms, and Sign List (Douglas B. Miller R. Mark Shipp) (Z-Library)
100% (4)
An Akkadian Handbook Helps, Paradigms, Glossary, Logograms, and Sign List (Douglas B. Miller R. Mark Shipp) (Z-Library)
330 pages
Khan The Neo Aramaic Dialect of The Assyrian Christians of Urmi I Phonology and Morphology 2016
No ratings yet
Khan The Neo Aramaic Dialect of The Assyrian Christians of Urmi I Phonology and Morphology 2016
635 pages
Azevedo J - A Simplified Coptic Dictionary Sahidic Dialect
No ratings yet
Azevedo J - A Simplified Coptic Dictionary Sahidic Dialect
235 pages
Rewriting Dialectal Arabic Prehistory
100% (1)
Rewriting Dialectal Arabic Prehistory
393 pages
Axenov S., The Balochi Language of Turkmenistan. A Corpus-Based Grammatical Description
No ratings yet
Axenov S., The Balochi Language of Turkmenistan. A Corpus-Based Grammatical Description
336 pages
The Hittite Name For Garlic
No ratings yet
The Hittite Name For Garlic
5 pages
English - Hittite Dictionary
No ratings yet
English - Hittite Dictionary
73 pages
Proto-Semitic Deverbal Derivation Insights
No ratings yet
Proto-Semitic Deverbal Derivation Insights
25 pages
Dictionary Khotan Saka
No ratings yet
Dictionary Khotan Saka
579 pages
Oraham Dictionary
No ratings yet
Oraham Dictionary
574 pages
Harris Linear Elamite Texts 31.10.11
100% (1)
Harris Linear Elamite Texts 31.10.11
25 pages
Beginning Cairo Arabic PDF
No ratings yet
Beginning Cairo Arabic PDF
309 pages
Proto SemiticSoundsInDaughterLanguages PDF
No ratings yet
Proto SemiticSoundsInDaughterLanguages PDF
6 pages
(CIIL Phonetic Reader Series - 17) K. Rangan - Balti Phonetic Reader (1975, Central Institute of Indian Languages)
No ratings yet
(CIIL Phonetic Reader Series - 17) K. Rangan - Balti Phonetic Reader (1975, Central Institute of Indian Languages)
131 pages
Introduction To Akkadian PDF
No ratings yet
Introduction To Akkadian PDF
116 pages
Das Akkadische Syllabar by Wolfram Von Soden
No ratings yet
Das Akkadische Syllabar by Wolfram Von Soden
71 pages
Beginner S Assyrian
100% (3)
Beginner S Assyrian
183 pages
The Khoesan Languages (Routledge)
No ratings yet
The Khoesan Languages (Routledge)
1,073 pages
(Richard Caplice) Introduction To Akkadian PDF
No ratings yet
(Richard Caplice) Introduction To Akkadian PDF
116 pages
English Traced To Arabic
No ratings yet
English Traced To Arabic
181 pages
Gelb, Amorite AS21, 1980
No ratings yet
Gelb, Amorite AS21, 1980
673 pages
SCHMITZ2008
100% (1)
SCHMITZ2008
6 pages
Bomhard - Afrasian Comparative Phonology and Vocabulary
100% (1)
Bomhard - Afrasian Comparative Phonology and Vocabulary
393 pages
Breton Grammar for English Speakers
No ratings yet
Breton Grammar for English Speakers
104 pages
Materials For A Comparative Grammar of The Dene-Caucasian (Sino-Caucasian) Languages
No ratings yet
Materials For A Comparative Grammar of The Dene-Caucasian (Sino-Caucasian) Languages
74 pages
Hoffner Ed - 1997 - The Hittite Dictionary of The U of Chicago, Vol P - Eng
100% (1)
Hoffner Ed - 1997 - The Hittite Dictionary of The U of Chicago, Vol P - Eng
440 pages
The Semitic Languages International Handbook
100% (1)
The Semitic Languages International Handbook
130 pages
Aramaic Language Marshall
No ratings yet
Aramaic Language Marshall
261 pages
L N
No ratings yet
L N
506 pages
The Tasaday Language: A Key To Tasaday Prehistory
100% (1)
The Tasaday Language: A Key To Tasaday Prehistory
21 pages
Semitic Language Classification
100% (1)
Semitic Language Classification
2 pages
Semitic Root Incompatibilities
100% (1)
Semitic Root Incompatibilities
18 pages
Proto Uralic
No ratings yet
Proto Uralic
25 pages
Nicolls-A Grammar of The Samaritan Language-1838 PDF
No ratings yet
Nicolls-A Grammar of The Samaritan Language-1838 PDF
152 pages
Bernd Heine, Tania Kuteva - The Changing Languages of Europe
100% (2)
Bernd Heine, Tania Kuteva - The Changing Languages of Europe
375 pages
Saxena-The Linguistic Landscape of The Indian Himalayas Languages in Kinnaur 2022
100% (2)
Saxena-The Linguistic Landscape of The Indian Himalayas Languages in Kinnaur 2022
494 pages
Assyrian Dict
No ratings yet
Assyrian Dict
472 pages
Chaldean Course
No ratings yet
Chaldean Course
16 pages
Mushin Ilana A Grammar of Western Garrwa
100% (1)
Mushin Ilana A Grammar of Western Garrwa
495 pages
A5 Parpola 2007 Assyrian-English-Assyrian Dictionary
100% (7)
A5 Parpola 2007 Assyrian-English-Assyrian Dictionary
320 pages
The Origin of Arabic Words in The Ancient Egyptian Language
No ratings yet
The Origin of Arabic Words in The Ancient Egyptian Language
12 pages
Hebrew Phonology PDF
100% (12)
Hebrew Phonology PDF
386 pages
1575061295
100% (6)
1575061295
386 pages
Moscati Comparative Grammar of Semitic Languages
83% (6)
Moscati Comparative Grammar of Semitic Languages
193 pages
Greek Grammar: Accidence Overview
No ratings yet
Greek Grammar: Accidence Overview
376 pages
Von Siebenthal, Heinrich - Ancient Greek Grammar For The Study of The New Testament
100% (3)
Von Siebenthal, Heinrich - Ancient Greek Grammar For The Study of The New Testament
764 pages
(UC Publications in Linguistics) Wallace Chafe-A Grammar of The Seneca Language-University of California Press (2014)
100% (1)
(UC Publications in Linguistics) Wallace Chafe-A Grammar of The Seneca Language-University of California Press (2014)
251 pages
T. Muraoka & B. Porten - A Grammar of Egyptian Aramaic 1998 PDF
100% (2)
T. Muraoka & B. Porten - A Grammar of Egyptian Aramaic 1998 PDF
439 pages
A Grammar of Egyptian Aramaic PDF
100% (1)
A Grammar of Egyptian Aramaic PDF
439 pages
Rosenthal A Grammar of Biblical Aramaic B
No ratings yet
Rosenthal A Grammar of Biblical Aramaic B
108 pages
Degif Petros Banksira - So
No ratings yet
Degif Petros Banksira - So
365 pages
Biblical Hebrew Grammar Overview
100% (1)
Biblical Hebrew Grammar Overview
16 pages
Smyth, Herbert Weir: Greek Grammar
88% (8)
Smyth, Herbert Weir: Greek Grammar
798 pages
Acta Classica 1 (1958) 31-44 (Knight, Vergil's Latin)
No ratings yet
Acta Classica 1 (1958) 31-44 (Knight, Vergil's Latin)
15 pages
Acta Classica 2 (1959) 70-71 (Baumbach, Petroniana)
No ratings yet
Acta Classica 2 (1959) 70-71 (Baumbach, Petroniana)
3 pages
Acta Classica 3 (1960) 90-92 (Hewitt, Petroniana)
No ratings yet
Acta Classica 3 (1960) 90-92 (Hewitt, Petroniana)
4 pages
Acta Classica 4 (1961) 46-52 (Conomis, Varia Graeca)
No ratings yet
Acta Classica 4 (1961) 46-52 (Conomis, Varia Graeca)
8 pages
Bell, The Duration of Life and Conditions Associated With Longevity
No ratings yet
Bell, The Duration of Life and Conditions Associated With Longevity
78 pages
Acta Classica 2 (1959 43-57 (Louw, On Greek Prohibitions)
No ratings yet
Acta Classica 2 (1959 43-57 (Louw, On Greek Prohibitions)
16 pages
Babcock, The Syntax of Spanish Reflexive Verbs
No ratings yet
Babcock, The Syntax of Spanish Reflexive Verbs
97 pages
Aòhmadåi, The Daeva Cult in The Gathas
No ratings yet
Aòhmadåi, The Daeva Cult in The Gathas
353 pages
Ivić, The System of Serbocroatian Cases Denoting Spatial Relations
No ratings yet
Ivić, The System of Serbocroatian Cases Denoting Spatial Relations
7 pages
Samuels, Linguistic Evolution
100% (1)
Samuels, Linguistic Evolution
212 pages
Elementary Modal Verbs & Signs Guide
No ratings yet
Elementary Modal Verbs & Signs Guide
3 pages
Unit One PDF
No ratings yet
Unit One PDF
6 pages
Grammar Quize
No ratings yet
Grammar Quize
33 pages
Article 2-1
No ratings yet
Article 2-1
18 pages
Word Formation Processes Explained
No ratings yet
Word Formation Processes Explained
84 pages
NLP Unit1
No ratings yet
NLP Unit1
28 pages
7-Ci Sinif Diaqnostik 2 Variant and English Books
No ratings yet
7-Ci Sinif Diaqnostik 2 Variant and English Books
2 pages
Programa Inglés Nivel 4 - 2019
No ratings yet
Programa Inglés Nivel 4 - 2019
5 pages
TENCE
No ratings yet
TENCE
1 page
Pronoun&Antecedent
No ratings yet
Pronoun&Antecedent
9 pages
Use of English
No ratings yet
Use of English
7 pages
Consolidation 2 of 9th Grade
No ratings yet
Consolidation 2 of 9th Grade
6 pages
IB ACIO English Course Schedule
No ratings yet
IB ACIO English Course Schedule
2 pages
Grammar
No ratings yet
Grammar
3 pages
Grammar and Vocabulary Exercises
No ratings yet
Grammar and Vocabulary Exercises
4 pages
Comparatives and Superlatives and The and The
No ratings yet
Comparatives and Superlatives and The and The
14 pages
Irregular Plural Nouns: Latinate Plurals Changing Vowel Sounds
No ratings yet
Irregular Plural Nouns: Latinate Plurals Changing Vowel Sounds
2 pages
Verbs: State-Of-Being Verbs/ Verbs of Being
No ratings yet
Verbs: State-Of-Being Verbs/ Verbs of Being
4 pages
English Grammar & Vocabulary Syllabus
100% (1)
English Grammar & Vocabulary Syllabus
5 pages
Grammar Guide: Subject-Verb Agreement
No ratings yet
Grammar Guide: Subject-Verb Agreement
2 pages
Noun Pronoun Quantifiers
No ratings yet
Noun Pronoun Quantifiers
3 pages
Present Simple Grammar Test Worksheet
No ratings yet
Present Simple Grammar Test Worksheet
2 pages
Opposites Adjectives Song For Kids: and Learn The Chant
No ratings yet
Opposites Adjectives Song For Kids: and Learn The Chant
8 pages
Present Continuous Tense
0% (1)
Present Continuous Tense
18 pages
Narmada River Assignment Detailed
No ratings yet
Narmada River Assignment Detailed
14 pages
Tuwali Dictionary Grammar Sketch
No ratings yet
Tuwali Dictionary Grammar Sketch
112 pages
Sentence Types & Tenses Guide
No ratings yet
Sentence Types & Tenses Guide
6 pages
GRE Vocabulary 2000 Words
No ratings yet
GRE Vocabulary 2000 Words
279 pages
Vocabulary 8 2
No ratings yet
Vocabulary 8 2
2 pages
ESL Writing: Proper Nouns
No ratings yet
ESL Writing: Proper Nouns
4 pages

Reiner, A Linguistic Analysis of Akkadian

Uploaded by

Reiner, A Linguistic Analysis of Akkadian

Uploaded by

A LINGUISTIC ANALYSIS OF A K K A D I A N

LONDON · THE H A G U E · PARIS

LIBRARY OF CONGRESS CATALOG CARD NUMBER : 65-24780

Printed in The Netherlands by Mouton & Co., Printers, The Hague.

This grammar, my first attempt toward a formulation of the structure of Akkadian

2. THE WRITING SYSTEM 23

3.0 Basic assumptions 33

4.3 Consonant clusters 49

[Link] Declension (2) 67

[Link].2 "Primae η" 93

6. MORPHOPHONEMIC ALTERNATIONS 105

6.2 Non-morphophonemic changes 112

7. PHONOTACTIC ALTERNATIONS 117

GLOSSARY OF AKKADIAN FORMS 137

GLOSSARY OF SELECTED LINGUISTIC TERMS 153

ABL R. F. Harper, Assyrian and Babylonian Letters.

0.1. PURPOSE AND SCOPE

The chapter headed Phonology is my attempt to classify the information that is

1.1. THE LANGUAGE

2.0. PRELIMINARY STATEMENT

2.1. THE SCRIPT

2.1.1. Graph and referent

A graph in the cuneiform writing system is a cluster of wedges imprinted in clay, or

ous with other definitions of transliteration used by Assyriologists1 - is of importance

2.2. ELEMENTS OF THE TRANSLITERATION SYSTEM

The system of transliteration as defined above is a composite system in two respects:

2.2.2. Word signs

2.2.3. Phonetic complements

2.2.4. Syllabic signs

etymology. Since Akkadian was recognized to be a Semitic language, phonemes

[Link]. Types of syllabic signs.

[Link]. Sequences of signs.

2.3. CHANGES IN WRITING HABITS

2.4. PROBLEMS CAUSED BY THE WRITING SYSTEM

For the phonemes here cited in phonemic notation, see 3.1.

3.0. BASIC ASSUMPTIONS

3.1.2. Configurative feature

3.1.3. Phones without phonemic status

[Link]. Distribution of glides.

Stress is not phonemic; this is indicated by the following considerations: While

3.3. DISTINCTIVE FEATURE ANALYSIS

The inventory of phonemes can be most economically divided as follows :

4.0. QUESTIONS TO BE TREATED

4.1. VOCALIC AND CONSONANTAL LENGTH

Since it is preferable, wherever possible, to assume a symmetrical phonological

4.1.1. Cluster of identical consonants

[Link]. Evidence of the writing.

[Link]. Evidence of morphophonemic alternations.

4.1.2. Clusters with length

[Link]. Occurring clusters.

[Link]. Members of clusters.

[Link]. Inventory of clusters.

short vowel : VCV long vowel: V:CV

(ii) pattern with consonant cluster : VC

[Link]. Free variation of /:C/ and /C:/.

4.2. SEQUENCE OF TWO VOWELS

Akkadian has no diphthongs; moreover, as is the case in most Semitic languages,

4.2.1. Boundary between two vowels

[Link]. Indication of boundary by writing conventions.

[Link]. Indication of boundary by a syllabic sign.

ku-mu-u-a ~ ku-mu-ia /kumu:a/

4.3. CONSONANT CLUSTERS

4.3.1. Non-occurring clusters

[Link]. Notes on non-occurrence of certain clusters.

[Link]. Morphological significance.

4.3.2. Distinctive feature analysis

4.4. RESOLVING OF CLUSTERS

4.4.1. Two consonants

[Link]. Consonant and length.

4.4.2. Three consonants

[Link]. Two consonants and length.

Final C : + C # > C : V C # (dan:+t > dan:at); final C + : C # does not occur.

[Link]. Consonant and two lengths.

[Link]. Positions of the epenthetic vowel.

5.1. MORPHOPHONEMIC ALTERNATIONS

Whenever two morphs come into contact, in inflection, derivation, or in compounds,