LinguisticChange and

GenerativeTheory
INDIANA UNIVERSITY STUDIES IN THE Essays from the
HISTORY AND THEORY OF LINGUISTICS UCLA Conference on Historical Linguistics
EDITORIAL COMMITTEE in the Perspective of Transformational
THOMAS A. SEBEOK, Chairman Theory February 1969
Indiana University
HANS AARSLEFF
Edited by
Princeton University
ERIC P. HAMP Robert P. Stockwell
University of Chicago University of California, Los Angeles
DELL H. HYMES and
University of Pennsylvania
WILLIAM LABOV
.. Ronald K. S. Macaulay
University of Pennsylvania
' Pitzer College, Claremont
ROBERT STOCKWELL
University of California
Los Angeles

Indiana University Press

Bloomington and London
274 Vennemann

--. 1968b. Some informal remarks on phonological change. Department of

Germanic Languages, University of California, Los Angeles, dittoed.
9.
---. 1969. H.istorical German phonology and the theory of marking: Grimm's
Law_ Department of German and Russian, University of California, Irvine,
mimeographed. Prepared for the conference documented in this volume,
but not presented. Note on a PhonologicalHierarchy
in English
Arnold M. Zwicky,
The Ohio State University

1
1. Introduction
A major point of contact between theoretical work in generative
grammar and more traditional activities in historical linguistics is
the search for conditions on the form and content of grammars.
Such conditions function indirectly as predictions of the possibility
of certain kinds of linguistic change; as a result, known changes
can be used as a source of fruitful hypotheses about conditions in
grammatical theory, and such changes can be inspected as sources
of evidence for, or counterevidence to, particular systems of hy-
potheses. Much is concealed in my facile use of the phrases "con-
ditions on the form and content of grammar" and "known changes,"
the latter in particular, for the pursuit of specific hypotheses
normally entails a careful examination of accepted presentations
of linguistic changes. But I shall not explore these issues here.
Rather, I shall provide a few preliminary examples and then move
to a consideration of some aspects of English phonology which
supply evidence about the content of grammatical theory and thus,
derivatively, about linguistic change.

1. 1. PHONOLOGICAL HIERARCHIES

The following are some of the hypotheses which have been put
forth about phonological theory:
275
 276 Zwicky Note on a Phonological Hierarchy in English 277

(1) If a language has a rule dropping [b]2 between vowels, that rule also ticular, establish hierarchies with respect to individual rules (inter-
drops [d] and [g] ; and if a language has a rule dropping [d] between vocalic deletion of voiced stops and lowering of vowels, respective-
vowels, that rule also drops [g] . But the converseis not true in either ly). It happens that both these hierarchies apply to the class of
case (see Foley's- contribution to this volume). Put another way: a rule
segments affected by the rules in question. There is, however, no
deletingintervocalic[b], or intervocalic[d], or both, is not a possible
reason to suppose that such orderings are limited in this way; simi-
phonological rule.
(2) If a language has a rule lowering [i] to [a,], that rule must also lower lar hierarchies might appear in the environments of rules. Sections
[e] to [~] (David Stampe, personal communication).Put another way: 2 through 7 of this paper will, in fact, describe hierarchies applica-
a rule lowering [e 1·to [~] is not a possible phonological rule. ble to rule environments.
A discussion of hierarchies is significant only insofar as the same
The form of such hypotheses is familiar from Jakobson's investi- or similar orderings (or their inverses) reappear at many points in
gations of implicational universals: the grammars of many languages. That is, an argument that a par-
(3) "If, in the languages of the world, or in child language, the fricative ticular hierarchy requires representation in phonological theory
consonants are limited to a single phoneme, this phoneme is as a rule will resemble arguments supporting particular choices of feature
represented bys" (Jakobson 1968: 55). systems or other notational conventions (see Chapters 7 and 8 of
(4) "The existence of back consonants in the languages of the world pre- Chomsky and Halle 1968, also Bach 1968). A ramified theory
supposes accordingly the existence of front consonants" (Jakobson would require an account of the relationships and interactions
1968: 53). among the various hierarchies, and to be completely satisfying,
All four hypotheses lead to specific predictions about possible the tenets of such a theory should have some external justification,
historical changes, given the assumption (as in. Halle 1962) th•t for example, in terms of phonetics. In the following sections, which
r~: among the mechanisms of linguistic change are (i) moderate change~) largely constitute a report on work still in progress, I pursue the
in existing rules and (b) the addition of new rul~We-ca:n··corlcllicle hypothesis that the hierarchy
from hypothesis (3), for example, that an unconditioned replace-
ment of [sl by [fl is an impossible phonological change, because (5) Vowels Glides [rl [ll [nl [ml [~l Fricatives Stops
it would yield a system in violation of (3). From hypothesis (1)
we can conclude that a language with a rule deleting intervocalic must be represented in phonological theory. I assume the rather
[g] cannot change so as to generalize that rule to drop both [gl modest burden of illustrating the reappearan_ce of this sequence
and [bl (but not [d] ), and that the deletion of intervocalic [dl in English, without attempting to locate it within a ramified theory,
(without [gl also) or [bl (without [dl and [gl also) is an impos- without attempting to produce external explanations for the mem-
sible phonological change, because all these changes would resnlt bership and arrangement of this class, and without attempting
in a system in violation of (1 ). Similar principles can be adduced seriously to relate my remarks to current discussions of feature rep-
in morphology-there is some predictability in case syncretisms, resentations and markedness.
for example-and in syntax-some principles of word order imply I use the features and notational conventions of Chomsky and
others, for instance (cf. Greenberg 1963 and Ross 1967). Still more Halle (1968), except that to refer to subsets of (5) I occasionally
predictions can be made on the basis of additional specific assump- employ the ad hoc notations in (6).
tions about linguistic change itself, such as Kiparsky's (1968) hy-
potheses concerning the reordering of rules. (6) [ir ➔ p 1 to refer to the set of all elements in (5) beginning with ,r and
Hypotheses (1) through ( 4) all refer to hierarchical arrangements ending with p (regardlessof whether 1T precedes or follows pin (5));
among the sounds of a language. Hypotheses (1) and (2 ), in par- [ir] + to refer to [ir ➔ Stops]; [ir i- to refer to [ir ➔ Vowels]
278 Zwicky Note on a Phonological Hierarchy in English 279

Thus, [ 1 ➔ Fricatives] denotes the class comprising [ 1], the nasals, phonology. Consider the rule Gliding, which shifts [i] to [y] oblig·
and the fricatives, while [ I ➔ Glides] denotes the class of liquids atorily in such words as pavilion, Pennsylvania, and invasion. Toe
and glides; [ 1] + refers to [I] together with all elements of (5) to process is extended in fast speech to Lithuania, colonial, lithium,
the right of [I] (that is, to [ l] and all true consonants), and [ ! ]- accordion, criterion (to which Gliding does not apply in slow speech
refers to [ 1] together with all elements of (5) to the left of [ 1] because no + boundary precedes the [i] ), spontaneous, permeate
(that is, to liquids, glides, and vowels). (to which Gliding does not apply in slow speech because the vowel
in question is [e], n.ot [i] ), marsupial, oblivion, Kentuc/d(JJ'l (to
1.2. ALLEGRO VARIANTS which Gliding does not apply in slow speech because the conso-
The rules under discussion are for the most part processes nant preceding the [i] is noncoronal), and familiarity and pecu-
triggered in some fashion by speed of speech, style, fatigue, and liarity (to which Gliding does not apply in slow speech because the
the like. I have been obliged to make rather fine judgments about vowel following the [i] has primary stress). The slow-speech re-
what is acceptable in "normal," "fast," and "very fast" speech. strictions on Gliding are discussed in Chomsky and Halle (1968:
The difficulties of such an undertaldng are obvious; in addition to 225-27).
having to judge the acceptability of many different versions of the More than simple generalization is taking place here, however.
same utterance at varying speeds and in varying styles, the investi- Chomsky and Halle (1968: 228) argue that the failure of Gliding
gator is required to judge whether unrelated processes are operative to apply in the word emaciate (similarly, propitiate, appreciate,
at the same speed and in the same style. Nevertheless, seeing no initiate, and many others) should be explained by the assumption
alternative to sensitive introspection and listening, I have attempted that Gliding precedes the rule Alternating Stress, at least in those
to malce such judgments. I do not suppose that my decisions are en- dialects (among them mine) in which Gliding is obligatory in words
tirely consistent. Certainly they will differ in many details from the like beneficiary and auxiliary; if emaciate has final stress at the
decisions of other speakers. stage in its derivation when Gliding might apply, then Gliding will
The phenomena of fast speech merit much closer examination be blocked (as in familiarity), and a following application of Al-
than they have received in the literature. The very richness of the ternating Stress will shift the stress from the ultima to the ante-
data seems to have convinced many investigators that allegro vari- penult. But in fast speech the extended Gliding rule must apply to
ants are merely automatic consequences of faster speech, except emaciate after Alternating Stress applies, for otherwise the allegro
for certain words that have idiosyncratic variants. However, the form *[omoseyt] (instead of the correct [omeysyeyt] or [omey-
view that fast-speech forms have a direct explanation in phonetic seyt]) would be predicted.
terms presumes a coherent and detailed theory of linguistic pho- It is tempting to suppose,in light of this example and the numer-
netics, which cannot be said to be available yet, despite the work ous situations in which allegro rules must be assumed to apply quite
of Ladefoged 1967 and of Jakobson and Halle (notably Jakobson, late in the sequence of rules, that all specifically fast-speech rules
Fant, and Halle 1961 and Chomsky and Halle 1968, Chapter 7). The are ordered after all obligatory rules. S. Jay Keyser has suggested
view is nevertheless encouraged by a number of fast-speech pro- the following counterexample to me. There are English dialects in
cesses, for example, the rule Pre-Stress Contraction discussed which the Flapping rule is obligatory in words like writer and rider.
briefly in Section 2, and the transitional stop variation that is the In these dialects, as in mine, words like winter do not have a flap
subject of Section 5. in slow speech. But in faster speech the (specifically fast-speech)
rule of Nasal Dropping (ordered after Nasalization of vowels) works
1.3. RULE EXTENSION to place the [t] of winter in the proper environment for Flapping,
One notabk characteristic of many allegro variants is that they which then applies to yield [wiDr]. Flapping is also called into play
involve extensions of (usually obligatory) rules of slow-speech by the operation of the fast-speech rule Glide Deletion, which drops
280 Zwicky Note on a Phonological Hierarchy in English 281
initial [hl (generally) and initial [wl (in a few words only) be- sis, the reordering of Alternating Stress before Gliding is to be ex-
fore unstressed vowels; phrases like might have [mayDo(v)l and pected.
what would [waDadl thus have flaps in fast speech. Another Bailey's hypothesis should be extended in some fashion to in-
example is the rule Auxiliary Contraction, which (in conjunction corporate the observation that when specifically fast-speech rules
with Glide Deletion) is responsible for the reduction of is and has are added, markedness in rule ordering tends to be minimized.
to [zl, would and had to [dl, have to [vl, will to [ll, am to [ml, This imprecise formulation of the principle is intended to cover
and are to [rl .3 Is and has reduce to [zl after consonants as well as such cases as the ordering of Nasalization before Nasal Dropping,
vowels, and if the preceding consonant is voiceless, the obligatory Glide Deletion before Auxiliary Contraction, Nasal Dropping and
rule of Progressive Voicing Assimilation, familiar from its applica- Glide Deletion before Flapping, and Auxiliary Contraction before
tion to various formatives realized as [s, z,-izl or [t, d,-idl, yields Progressive Voicing Assimilation.
[sl, as in Chuck's not here and Pat's been sick. Before turning to other matters, here is another example of rule
C.-J. Bailey has pointed out to me that a number of fast-speech extension in fast speech. Further instances are noted in the follow-
phenomena can be explained as the result of a rule deleting bound- ing sections.
aries in a variety of contexts. Although few of the phenomena dis- The rule Nasal Assimilation, which assimilates the final [nl of
cussed in this paper can be so explained (the applicability of Glide the prefixes con-, in-, syn-, and en- to the position of any follow-
Deletion, Auxiliary Contraction, and Flapping in phrases contain- ing obstruent except [fl (conflation), [vl (convection), [kl (con-
ing unstressed auxiliaries or pronouns being the principal class of cussion), or [gl (congratulation), is extended in fast speech to apply
exceptions), it should be noted that this deletion must apply be- (a) before [f v k g], (b) to final [n] of the prefixes un- and non- and
fore a considerable number of obligato:ry slow-speech rules-among of the first elements of compounds, which do not assimilate at all
them, the Flapping rule already mentioned and the rules of Palatal- in slow speech, and (c) progressively, to nasals following obstru-
ization and Y Dropping, which apply in slow speech in the deriva- ents, as in eleven, bacon, and hypnotize. I have specified that Nasal
tion of actual, gradual, sensual, and visual (Chomsky and Halle Assimilation applies generally before obstruents, and not merely
1968: 230-32), and in fast speech to yield [diJol for did you, before [p bl, because I take this to be the same rule as the one
[w6wcal for won't you, and the like. specifying the nasal before [kg] within morphemes, as [~l (rink.
Also due to Bailey is the hypothesis that if the order of two ring), and the nasal before [p b fl within morphemes, as [ml
rules with respect to each other changes in fast speech, it changes (camp, amber, camphor). That is, the obligatory (slow-speech)
from the marked to the unmarked order (in the sense of Kiparsky rule applies generally within morphemes and across the + bound-
1968), never in the opposite direction. Let us examine the order- ary (the boundary in comfort and congress, which show assimila-
ing of Gliding and Alternating Stress within this framework. In tion in show speech-cf. convent, concourse, infant), but only be-
slow speech the order Gliding-Alternating Stress is marked, because fore [p b] across the= boundary.•
if Alternating Stress were to apply first the domain of applicability It is natural to suppose that rather than the fast-speech rules being
of G,liding would be increased by the emaciate class, whereas the extensions or generalizations of normal (slow-speech) rules, the
slow-speech form of Gliding neither decreases nor increases the normal rules are restricted versions of the more general processes. I
domain of Alternating Stress. In fast speech this order remains am indebted to David Stampe for the idea. The associated view of
marked, because the extended version of Gliding removes cases, language acquisition is that the child embarks upon the task oflan-
the emaciate class again, from the domain of Alternating Stress, guage learning with innate knowledge of a set of quite general rules,
whereas Alternating Stress neither decreases nor increases the do- and that what he must learn is (a) the way in which these rules are
main of the fast-speech version of Gliding. On Bailey's hypothe- restricted in his language, plus (b) their order, insofar as it must be
282 Zwicky Note on a Phonological Hierarchy in English 283

marked. The associated view of linguistic change is that the pri- (8)(a) Before [r] : hindering, 6 blunderer, puckering, elaborate
mary mechanism of change, aside from reordering, is the removal (adj.), mackerel, amorous, doggerel, cadaverous, camera,
of restrictions on rules. What corresponds to "addition of a rule" every, laboratory, ephemeral, operative, separate, treasurer,
within this framework is the removal of an absolute prohibition reference, federal, impoverish, inaugural, lateral, directorate,
against the rule, so that it is to be expected that the earliest evi- dangerous, et cetera, temperature, Everest, Barbara
dences of a rule will appear in considerably restricted environments (b) Before [!]: pedaling, twinkling, suppler, awfully, re-
(restricted in the class of segments affected, in the contexts in spectfully, erratically, reciprocally.javelin, Emily, especially,
which the rule applies, and in the lexical items to which the rule excellent, equivalent, acidophilus, chocolate, easily, ambu-
applies). Some discussion of rule spread, in time and space, within latory, benevolent, desolate (adj.), leveling, channeling,
similar conceptual frameworks, can be found in Bach (1968: 135- Lancelot, Evelyn, chancellor
37) and in Labov's paper in the present volume. Foley's contribu- (c) Before [n]: happening, stiffening, opener, pardoning.
tion to this volume puts forth a theory of linguistic change along seasonal, reasoning, reckoning, resonant, coordinate, per-
similarlines. sonal, definite, infinite, arsenal, reasonably, gelatinous, ef-
Although this promising direction of investigation will not be feminate, rational, larceny, traditional
pursued further here, it should be noted that not all phonological Slur applies obligatorily in business, and in my speech to a number
processes "extend" in the same fashion as Gliding and Nasal As- of other words, e.g., camera, every, celery, general, mystery, choco-
similation. Thus, for example, the rule N+Resonant, which assimi- late, and family; most of these are instances of syncope before [r J,
lates the final [n] of the prefixes con-, in-, syn-, and en- complete- although there are a few cases before [I].
ly' to a following [r Im n], does not apply to a number of forms I wish to emphasize that this rule is a general process, in contrast
(e.g., enrage, enlist, enmesh, inlay). In fast speech the unassimilated to many other English contraction rules, for example Auxiliary
clusters remain, and moreover the rule does not extend to [n] in Contraction (see section 1.3), which is restricted to a small set of
un- ancl non-, nor to [n] at the end of first elements of compounds words and is restricted in a rather complex way by the phono-
(tin wrap, as opposed to the Nasal Assimilation example pinb·all), logical and syntactic environment of those words (see Zwicky
nor to [n] before resonants in such names as Conroy and Conlon. 1969).
Instead, in definitely fast speech Nasalization and Nasal Dropping On the other hand, the deletion in Slur is not governed by any
apply to yield [areyJ] for enrage, [n~Iiygl] for nonlegal, [kilroy] simple or obvious conditions on the "pronounceability" of the
for Conroy, etc. result, in contrast to the contraction rule Pre-stress Contraction

a ➔ !/)/#C-C [~~~~s]
2. The Rule Slur 9
( )
Let us begin with the following syncope rule: + stress
which seems to have no relationship to the hierarchy in (5). Pre-
O ➔ !/J/C-{~}
(7) - cons ]
+voe stress Contraction manifests itself in quite fast speech in the drop-
[ - stress ping of the unstressed vowel in the first syllable of such words as

This rule applies in moderately fast or casual speech to delete the (10) derivative, united, 7 development, galoshes, demonstra-
reduced unstressed vowels in examples like the following: tive, subordinate, senility, vicinity, ferocious, coordinate
284 Zwicky Note on a Phonological Hierarchy in English 285

The rule does not apply in such words as Decameron, revised,

(I 8) [+ cons] [- cons ]
pedestrian, and deflation, presumably for phonetic reasons.' Slur, a ➔ 0/C- - obs! + voe
in contrast, fails to apply in a great many cases in which the result + cor - stress
would be easily pronounceable:
However, I shall argue that the exclusion of [ml from the en-
(11) Before syllables bearing stress: degenerate (vb.), compen- vironment of Slur is a relative rather tha'.nan absolute exclusion
satory, imaginary, polarize, scandalize, hyphenate, Eucha- and that the acceptability of the output of Slur varies, being high-
rist, vocalism, intellect, parallel est in position before [r l, lowest before [nl. In fact, I shall argue
(12) Before clusters: development, honestly, graciously, ear- that the hierarchy in (5) corresponds to a differential in the accept-
nestly ability of the outputs of Slur, ranging from normally entirely ac-
(13) Before obstruents: historical, relevant (cf. referent), di- ceptable before [r l, to less acceptable before [nl, to normally un-
visible, voracity, Arabic, imperative, analogy acceptable before [ml and [~ l, to entirely unacceptable before ob-
(14) Before [m] :9 element, minimum, monogamy, astronomer, struents (as in (13)).
consummate (adj.), ultimate First, the contraction before [ml and [~l- The Slurred versions
(I 5) Across strong (i.e., compound or word) boundaries: of the examples in (14) strike me as unacceptable at the same rate
counter-attraction, counterintelligence, inter-American, of speech at which the examples in (8) are normal. But in faster
underexposed, castle adventure, button across speech they are not quite so bad, certainly much better than the
Note also that the correct operation of Slur depends on the prior Slurred versions of the obstruent examples in (13). A few words-
application of a number of other rules-vowel-reduction rules, be- Quadrigesima, handsomer, perhaps unanimous-are good even in
cause the rule applies only to unstressed vowels that have been re- only moderately fast speech. Examples of [al in the Slur environ-
duced to [al ,and also an extended (fast-speech)version ofY Drop- ment before [~ l are not easy to find, but one such occurs in the
ping (see sections 1.3 and 7), which eliminates the glide of [ya l proper name Durringer (when pronounced with simple [~l rather
(from careful speech [yuwl) in examples like ambulatory and in- than with [~gl or [J1JlJ; Slur is inapplicable in this case.
augural and of [ya l in examples like auxiliary, because there is no Next, the exceptions to Slur. They are somewhat more numer-
trace of the [yl in the contracted forms of these words. Presum- ous before [ll than before [r l, and somewhat more numerous be-
ably, for speakers who do not obligatorily palatalize [ ty dy l to fore [nl than before [ll. The exceptions before [r l are, so far as
[c J] before unstressed vowels, an extension of Palatalization (and I have been able to determine, all of a single type: They are all cases
Y Dropping) precedes Slur in the derivation of such words as" of a failure of deletion before [riyl. My judgments on some words
ending.in Loriyl with respect to Slur in moderately fast speech are
(16) lecturer, cultural, congratulate, spatula, credulous, modu- given in (19):
lar, fraudulent, fortunate
(19) (a) Slur applies: discovery, menagerie, factory, summary,
all of which show palatals-e.g. [f6rcnitl for fortunate. hosiery, misery, machinery, mastery, scenery, illusory,
In the notation of ( 6), Slur has the formulation dispensary, shrubbery, bindery, every,. nursery, cursory,
robbery, slippery, hickory, memory, Bowery
(17) [- cons ] (b)? Slur applies: haberdashery, quackery, battery, chic-
a ➔ 0/C-[r ➔ nl +voe
- stress ory, napery, creamery, usury, surgery, perjury, feathery,
periphery, sorcery, mockery, refinery, thievery, Calvary
which appears to be merely an unmotivated variant of (c) Slur does not apply: infirmary, nunnery, summery,
286 Zwicky Note on a Phonological Hierarchy in English 287

lechery, perfumery, buttery, buggery, granary, cannery, solute, because there are some forms (e.g., effeminate, scrutiny,
gunnery, crockery, Hungary, plenary, greenery, notary, gelatinous, fattening) that reduce comfortably in fast speech, and
papery, snobbery, rotary, primary, rosary, rookery, dean- because the domain of the rule is extended as the rate of speech in-
ery creases. The following are some examples:
Note especially the robbery-snobbery and summary-summery con- (22) (a) [mn] ;feminine, geminate_(adj.), Gethsemane, stamina,
trasts. The classification of words in [ ariy] according to the appli- phenomenal, nominal, abdominal, dominant, hominy, alu-
cability of Slur seems to be largely arbitrary, although some general- minum, voluminous, Germany, prominent
izations can be made (e.g., adjectives in -y, "like," do not contract (b) [tn]: intestinal, Latinist, monotony, botany
in moderately fast speech: silvery, blustery, blistery, splintery, The reduction "infunctional, inflectional, and similar forms is rela-
cindery, thundery, powdery, gingery, leathery, gossamery, sum- tively unacceptable, probably because of the complex [ksn] clus-
mery, watery, papery, peppery, coppery, butteryu)_ It is clear that ter created. I find the reduction of traditional, emotional, and the
Slur is being extended to new vocabulary items, 12 just as it is ex- like quite natural, as opposed to the reduction of confessional, pro-
tended in very fast speech. cessional, and the like. And I find the contraction of progeny, nitrog-
The exceptions to Slur before [I] are of two types. First, [a] be- enous, misogyny, etc. (to [Jn]) less acceptable than the contrac-
fore the suffix -ly does not drop after a stressed syllable, with few tion of fortunate (to [en]).
exceptions (easily, especially, finally): A final source of evidence for the ranking of the resonants in the
(20) (a) -al-ly: mentally, eternally, morally, vocally, orally, environment of Slur consists of cases in which the rule could apply
totally, monumentally to either of two different resonants within the same word:
(b)-i-ly: readily, happily, prettily, jerkily, wheezily, chub- (23) (a) [oral]: federally, generally, literally, laterally, curso-
bily, cloudily, stuffily, sunnily, drowsily rily, naturally, minerally, electorally
(Cf. the words in-ically, which is always reducible to [ikliy] ). The (b) [anal]: personally, subliminally, seasonally, conven-
second class of exceptions contains most occurrences of [ ~] de- tionally, terminally, marginally, criminally
rived from [yuw], except in ambulatory, perhaps in particular, and In (23a) the contraction before [r] is clearly preferable to the con-
in the words in which the [y] of [yuw] has been absorbed by traction before [I]; [Jenraliy], for example, is much better than
Palatalization and Y Dropping (e.g., spatula): [Jenrliy l, although both forms can occur in fast or casual speech
(for the latter form, see the next section). In (23b) the matter is
(21) binocular, jocular, populous, popular, amulet, fistula, cu-
not so clear; for some words (e.g., personally) I prefer the reduc-
mulus, truculence, circular, muscular, tabular tion before [I], in others (e.g., criminally) the reduction before
Y Dropping, but not Slur, applies to these words in fast speech, so [n]. The former cases provide some evidence for the dominance
that popular, for example, has the variants [papyuwlr], [papyalr], of [r] over [!], while the latter cases are neutral with respect to
[papalr] in increasingly rapid speech. · · the question of the ordering of [l] and [n] -but this ordering is
The exceptions to Slur before [n] are diverse. The principal re- the one that is supported most strongly by the argument from ex-
strictions are against [mn] and [tn] clusters that would result from ceptionality.
the operation of Slur. These restrictions must refer specifically to
the rule Slur, for the clusters are not, in general, unacceptable in 3. The Rules Ruh-reduction and Ruh-lessness
English (cf. the proper names Hamnet, Simnall, Putney, Courtney, A number of words (e.g., separable, preferable, cooperative, cor-
and also amnesia, insomnia); moreover, the restrictions are not ab- roborative, tolerably) have, in addition to the variants generated by
288 Zwicky Note on a Phonological Hierarchy in English 289

Slur, fast-speech forms that appear to show Slur in operation be- Ruh-reduction is applicable in definitely fast or casual speech (al-
fore obstruents: [seprb!l, [prefrb!l, etc. 13 But in fact, this reduc- so in the speech of many children). Hence, forms like [prefrb!l are
tion is dependent not on a resonant following the [a], but rather less acceptable than forms like [prefrn bl] . In still faster speech
on a resonant (more precisely, an [r]) preceding the [a], as can some speakers eliminate the retroflex coloring of [rl, when un-
be seen from the examples in (24), which exhibit a variation be- stressed and flanked by cons on an ts, in many words. This rule, Ruh-
tween normal speech [ra I and fast speech [fl in diverse positions.
14 lessness, is common in my speech in only a few forms (surprise,
governor, paraphernalia, thermometer, and particular), for which
(24) reciprocal, segregate, instrument, intricacy, introduction, dissimilatory influences are usually cited as the cause, as in Kenyon
hydroplane, profusion, pretend, professor, demonstrative, and Knott 1953: xlvi. Although the best examples of Ruh-lessness
corporal, natural, pirate, hieroglyph, chiropractor, moral are in words with two [r] 's, the explanation in terms of dissimila-
Clearly a rule distinct from Slur, moreover one that can apply to tion is not entirely convincing, because (a) some of the best ex-
the output of Slur, is required. This rule, Ruh-reduction, is not con- amples would involve anticipatory dissimilation over long distances,
ditioned by the stress on (or even the existence of) a neighboring e.g., three syllables in particular and thermometer, and (b) many
syllable: speakers of otherwise [r]-ful dialects have [a] for unstressed [r]
in allegro pronunciations of words lacking a second [r] (instrume~ t,
(25) ra ➔ r / [- word boundary] - [- word boundary] profession, introduction, permission). Nevertheless, Ruh-lessness
The context in (25) prohibits the rule from applying in initial posi- is favored in dissimilatory environments, so that the reduction in
tion (Ramona, reduction, risotto, romantic; cf. brassiere and pro- governor is more acceptable than the reduction in governess. 16
fessor, which reduce) or in final position (hydra, pellagra, Capra, Ruh-reduction does not simplify [ra] only. It applies also, in a
Barbara; cf. apron and corporal, which reduce). The formulation very restricted fashion, to [la I, [na], and [ma]. These reductions
of the change itself requires some discussion. It does not seem pos- are characteristic of quite fast or careless (for example, alcoholic)
sible to frame the rule as a straightforward [a ]-deletion rule (con- speech. I find the reduced [na] and [ma] forms somewhat less ac-
tracting the sequence CraC to CrC, upon which the independently ceptable than the reduced [la] forms.
required rule Syllabication would operate to yield CrC), because The generalized version of (26a) will assimilate some unstressed
the reduction takes place after vowels (even lax vowels, as in syllabics to some preceding resonants in the features coronal
caraway) as well as after nonsyllabics. Some speakers, as reported lateral, and nasal. The details of the assimilations are not entire!;
to me by David Stampe, have a clear intermediate stage [rr] be- clear to me. The generalized version of (26b) will then drop [r J be-
tween [ra] and [r] , so that two rules are involved, the first an as- fore coronal vowels, [!] before lateral syllabics, and nasals before
similation, the second a simplification of [q] that is required in nasal syllabics, Without giving a concrete form to these revisions I
any event to account for the common pronunciation of mirror as list here a few examples (many of which show, in addition to Ruh-
[mlrl (similarly, horror, purer, pairer): reduction, assimilation rules not discussed here):

(26) (a) a ➔ [ +cor] /[-word boundary] r - [-word boundary] (27) (a) [la] from Slur: chocolate [cakJ?], ambulatory [ii:bl-
toriy], equivalent [akwivl'] . · ·
(b)r ➔ 0/-r
15

(b) Other [la I : ablative [;;,bJdiv], complicate [kiipJkey,J,

Because there is some independent evidence for (26b) and because Raglan [rregJn], restless [resjs]
the rules in (26) have a sort of internal motivation lacking in (25), (c) [na] from Slur: coordinate [kw6rd,1,], definite
I shall suppose that (26) is closer to the correct formulation of Ruh- [defi,,]
reduction than (25), even though the intermediate stage [rr] is not (d) Other [na] : abnegation [reb9geys,1], ignominious
prominent in my speech. [lgIJ1mfnyas], sadness [sredi;is]
290 Zwicky Note on a Phonological Hierarchy in English 291
(e) (26b) applying to Jn]: kennel [k~\l, denim [d~1]1], (highest for the examples in (29), lower for vetoing and tangoing,
linen [lfn], canner [kref] . and very low for hulaing and subpoenaing) and to some extent with
(f) [ma]: admonition [aodnmisn], admiration [a,d- the quality of the preceding vowel (highest for front vowels, es-
1]1~Slll • _ pecially [iy] , as in seeing and carrying).
(g).(26b) applying.to [ml: maximum [mreksal]1]. camel In summary, my judgment is that the contraction is most likely
[kre)J, summon [s.Z1]1] to take place before [r] and [I], then in the cases before [n] and
Note that my judgment is that [raj is most likely to contract, then [ml given in (29), then before [~] and in the cases before [n] and
[Ia], then [no] and [ma]. [ml given in (30).
Ruh-lessness applies generally only to [r L In extremely fast or
careless speech it may affect a few occurrences of [I] ([caka ?] for 5. Transitional Stops
chocolate, [kipakey?] for complicate), but not the syllabic nasals. A variety of CCC clusters have CC variants in fast speech, with the
middle C dropped. Conversely, a variety of CC clusters have vari-
4. The Rule VVR ants with a transitional stop intervening between the two conso-
Still another deletion of [a] occurs, as in the case of Slur, before nants. Thus, although some speakers of English have a distinction
resonants, but otherwise in an environment quite different from between [ns] in prince and [nts] in prints (in slow speech), in
the Slur contexts: faster speech the two forms vary freely with each other. Some
speal<ers do not appear to have the distinction at all. In general,
(28) a ➔ ,-_ cons] [- cons7 _ [+ cons] the greater the number of feature distinctions between the first and
0 / l+voc - voe J - obst last consonants, the more likely the variation. Without exploring
VVR applies to the words in (29) to reduce a sequence VGaR to in detail the principles of this variation, I provide first some ex-
17 amples with nasals (in conjunction with fricatives in (31) and (32),
VGR:
with stops in (33) and (34):
(29) (a) [r] : dire, flower, diary. prior, theory, payer, fiery
(b) [I]: jewel(er), dial, duel(is/), fuel(ing), royal(ly), con-
(31) tenth, thousandth; prince, prints, answer. pencil, land-
scape; friendship, French, trench; warmth; Damson,
strual, denial, real
(c) [n]: Brian, Rowan, lion(ize), Zion(ist), Cohen
damps; Canfield, panful, handful, grandfather; length;
rinks; compunction, juncture; drumful, lumpful; Hamp-
(d) [ml: Noam, jeroboam, Hyams
shire; ringful, tankful; gowns, grounds, lens, lends; strange,
(e) [~]: doing, sowing, saying, pawing, seeing
fringe; dams, damsel; rings, stirrings; Convair
The reduction before liquids is usual in fairly fast speech for me. (32) fishnet, rashness, rationing; casement, Westminster; Kast-
The reduction before the nasals is less normal; 18 some reductions are ner, Flessner; Fishman, trashman; Hefner, roughness;
impossible for me except in very fast speech: business, Gesner
(33) dreamt, jumped; ranked; dreamed; ringed
(30) (a) [n] : Malayan, Siouan, Korean, Samoan
(34) happening, raptness; weakening, weakness; Hickman"
(b) [m] : museum, atheneum
Now observe that [l] is involved "in-asimilar variation. However,
Also I find the reductions before [~] definitely less natural than in contrast to the variation for the nasals, which is normal in fast
the dther reductions, with the acceptability of the results correlated
speech for all spealcers, and regular at all speeds for some, the vari-
to some extent with the degree of stress on the preceding syllable
ation for [I] is com.man only in definitely fast speech, and it is not
294 Zwicky Note on a Phonological Hierarchy in English 295

must appear in the environment of these rules, the most notable tentative way) the question of the extent to which it can be pre-
of which is the part of the Main Stress rule that incorporates es- dicted when the hierarchy is relevant to a given rule.
sentially the Classical Latin stress rule, by which vertebra is stressed The hierarchy in (5)-vowels, then glides, then liquids, then
on the antepenult but umbrella on the penult. In the nonce nota- nasals, then fricatives, then stops-corresponds to a gradation of
tion of (6) a weak cluster is represented as sonority, from greatest to least. The sequence of the major classes
(47) [r]+ ([r]-) is governed by the principles in (48), with (48a) applied before
( 48b):
Note that in this instance, as in the case of Dentdel and transi-
tional stops, the problem to be solved is the exclusion of [r] (to- (48) (a) (i) [- cons] precedes[+ cons]
gether with the glides) from a larger class to which [r] naturally (ii) [- obs!] precedes [+obs!]
belongs. Another example of this type may be the following re- (b) (i) [+ voe] precedes [- voe]
striction on the occurrence of [s] and [s] in initial position: Be- (ii)[+ cont] precedes [- cont]
fore [r], [s] but not [s] occurs; while before [l m n], [s] but not By the principles in ( 48a) all vowels and glides precede all liquids,
[s] occurs. Dialect variation in the treatment of [yuw] (when it nasals, fricatives, and stops (i), and all vowels, glides, liquids, and
bears some stress and follows a coronal nonsyllabic) might pro- nasals precede all fricatives and stops (ii). The result is a three-way
vide still another example. There are two extreme dialects with division, with all vowels and glides preceding all liquids and nasals,
respect to Y Dropping (see section 1.3), one preserving the [y] in which in turn precede all fricatives and stops. Principle (48bi) orders
words like tune, duty, assume, nuisance, lute, and ruin, the other vowels before glides within their class, and liquids before nasals
(mine) without a [y] in any of these words. I am aware of two within their class (but does not affect fricatives and stops), while
intermediate dialects of interest in an examination of the hier- principle (48bii) orders liquids before nasals within their class, and
archy in (5). The first preserves the [y] in all cases except after fricatives before stops within their class (but does not affect vowels
[r l], the second in all cases except after [r] (lute, Lucifer, illu- and glides).
minate, and lubricate maintaining the [y], ruin, ruby, and -rule Within the portion of the hierarchy of immediate interest to us,
lacking it). 24 namely the liquids and nasals, the sequence is governed by the
The processes in which [r] is excluded contrast with Slur, Ruh- principles in (49):
reduction, Ruh-lessness, and VVR (perhaps also Syllabic Resonants
-see note 9), in which the problem (aside from gradation in ac- (49) (a) (i) [- back] precedes [+back]
ceptability) is the exclusion of one or more of the nasals from a (ii) [+ car] precedes [- car]
larger class to which they naturally belong. (b) [- ant] precedes[+ ant]
Principle (49a) orders [nm] before [~] (i) and [n] before [m~]
8. Prospectus (ii), without affecting the liquids. Principle (49b) orders [r] be-
fore [I] (and must be ordered after (49a) so as not to affect the
Let us suppose, on the basis of the English rules discussed in sec- nasals).
tions 2 through 7, that the hierarchy in (5) requires representation
Note that the principles in (49) are applicable to oral consonants
in phonological theory. I shall not consider the question of how
as well as to resonants; Jakobson's observations quoted earlier in
this representation is to be accomplished, but rather shall offer a
(3) and (4) follow from (49a). In other words, the ordering of the
few remarks on the relationship of the hierarchy to feature compo-
nasals in (5) corresponds to the ordering with respect to marked-
sition and to markedness, and then I shall approach (in a most
ness, from least to greatest, and not to the ordering with respect
296 Zwicky Note on a Phonological Hierarchy in English 297

to sonority, which is[~ n ml .25 On the other hand, the ordering of elimination of secondary articµlations, as in Ruh-lessness, or by de-
[r] before []] in (5) accords with the ordering by decreasing so- letion, as in Slur and VVR) and when assimilations characteristic
nority, not with the ordering by increasing markedness, [r] being of vowels are operative (as in the first part of Ruh-reduction and
marked in contrast to [l]. That is, the hierarchy in (5) is complex the )j9 nasalization rule).' In contrast, the nasals are favored when
with respect to the natural dimensions of sonority and marked- the "strength" of a consonant cluster is involved. Put very suc-
ness. cinctly, the distinction is one between "vowel" rules and "conso-
English contraction rules supply some evidence about other sub- nant" rules. These excessively vague notions await much further
hierarchies of (5). For example, the rule Glide Deletion (recall sec- investigation.
tion 1.3) drops initial (h] quite extensively, initial [w] in very re- Finally, notice that other hierarchies interact with the one in
stricted environments (only in the words will, would), and initial (5). Thus, Dentdel is most likely when a consonant (especially an
[y] not at all. The rule is sensitive to factors of stress, speed, and obstruent) follows the [t] or [d] to be deleted. Slur is normally
syntax in a complex way, but the sequence is not difficult to dis- inapplicable before stressed syllables, but in quite fast speech (and
cern. 26 in some dialects) it extends to this position as well. The appli-
Although it is impossible at present to predict when a hierarchy cability of VVR is in part conditioned by the stress on and quality
will apply to a rule, or to predict which hierarchy (or hierarchies) of the preceding syllable. The result is in each instance a complex
will apply and in which order (recall the contrast between transi- pattern of acceptability, made still more elaborate when idiosyn-
tional stop variation, Dentdel, and the weak cluster condition, on cratic lexical markings, rather than statable phonological or syn-
the one hand, and Slur, Ruh-reduction, Ruh-lessness, and VVR, on tactic environments, contribute to the determination of accept-
the other), some suggestive observations can be made. ability, as in the case of Slur applying to [a] before [riy].
Note first that all the rules discussed so far in which the hier- A considerable amount of work will be required to sort out these
archy in (5) is applicable are rules in which the resonants appear processes and the restrictions on them-among other things, a sur-
in the environment. In contrast, rules that change the major class vey of synchronic processes involving liquids and nasals, of histori-
features of resonants are governed by principles cutting a"Cross cal changes involving these sounds, of principles of borrowing in-
considerations of sonority and markedness. The relationships be- volving them, and of their treatment in language acquisition, all of
tween [r] and [s], [z], [n], []], [d], and [ w], and similar relation- these in a wide variety of language. 27 To this survey I offer the
ships between [l] and various stops and glides, and between nasals above observations as prolegomenon.
and sounds of other classes all appear to be quite independent of
the hierarchy in (5). Other processes that affect the resonants
directly, in particular the superimposition of secondary features NOTES
on the resonants, invoke the hierarchical principles. For example, I am indebted to Ann D. Zwicky for her many suggestions about the
in Kolokuma jj9 (Williamson 1965: 16) vowels and [w yr], but form and content of this paper. I have incorporated into the text a number of
not [l], are nasalized in the vicinity of nasals. observations made by other participants in this conference. This is the ver-
Although the characterization of the circumstances in which [r] sion of February 16, 1969.
2. Throughout this paper I use italics to cite forms in conventional orthog-
is favored, as opposed to those in which the nasals are favored, is
raphies, surrounding brackets to cite transcriptions, which are to be taken as
impossible on the basis of the tiny sample of rules considered phonetic except where the context makes it clear that a more abstract repre-
here, a rough conceptualization of the distinction is not out of the sentation is intended. My phonetic transcriptions are not uniform, being rather
question: [r] is favored when the strength of a syllabic nucleus is narrow for features under discussion, broad in other respects. In particular,
reduced (by elimination of an initial resonant, as in (26b), or by I do not mark many distinctions of quality and quantity in vowels.
298 Zwicky Note on a Phonological Hierarchy in English 299
3. Auxiliary Contraction is related to the rule VVR discussed in Section 4
delphia, skeleton, Europe (the relevant examples differ from dialect to dialect),
below, although the nature of the association is still unclear to me. none of which are reducible in my speech except at high speeds.
4_ For discussion of the= boundary see Chomsky and Halle 1968. Na~es 14. The word iron has no form with nonsyllabic Ir] in my speech. Thus, in
like Canby (with assimilation only in fast speech) must be treated as contam.M
slow speech I have a contrast between irony, "like iron," {a.Y:rniy] and the
ing a # boundary associated with the formative vby (appearing also in Whitby,
figure of speech irony [iyraniy] (although the contrast is eliminated in fast
with failure of voicing assimilation, and in Rugby, with its unusual cluster of
speech by the application of Ruh-reduction to the latter form). Iron then re-
voiced obstruents. quires either the obligatory application of Ruh~reduction or else the unusual
5 _I propose to demonstrate, on another occasion, that both this rule and
underlying representation /'fVrn/.
the "assimilation" rule operative in words like affect, support, and attest are
15. For the appearance of the feature coronal in (26a), see Chomsky and
not rules of complete assimilation, but rather deletion rules, at least in my
Halle 1968: 304. The reduction of mirror and similar forms indicates tlrnt
dialect. (26b) is not restricted by occurrences of #in the way that (26a) is. Note that
6. In hindering and many similar examples the [a] is first inserted by n:le,
the rule Syllabic Resonants mentioned in note 9 might be treated essentially
then deleted optionally by another rule, in this case Slur. The underlymg
as the lllrrror image of (26).
form of hinder does not have a vowel preceding the Ir] because hindrance has
16. In a few cases the second fr] is deleted instead of the first, as in inter~
two, not three, syllables in all dialects (compare dangerous). Schwa I~serti_on
pret [intfpit], where the first [r] is stressed and the second unstressed.
is required because in many dialects (mine among them) hinder and hznderzng
17. In examples like burial, Albanian, criterion, geranium, and linoleum,
have [ a r], rather than [~], in slow speech. VVR does not apply, because the extension of Gliding applies first, changing
7. The resultant [yn] cluster in united is pronounced []1]. . . [iyo J to [yo J .
8. But note that there is no obvious phonetic explanation for the co~dition
18. Except in a few words like diamond and diaper, where VVR is obligatory
that the rule does not apply to initial [a] , as in alembic, aroma, anemic, and for me.
amusing, or to [a] in syllables after t.he first, as in Canaletto, paronymic, homi- 19. Words like tentful and Pitman (also saltpeter, with [1] , and cartful, with
letic, and analytic. However, some of the examples cited in note 13 bel_ow can [r]) do not have acceptable variants without a medial stop, in my speech.
be interpreted as undergoing medial applications of Pre-stress Contract10~. What happens is that a Glottalization rule replaces [tJ by[?] in these words,
9. The exclusion of the environment before Im] (and [l]J) makes 1t un- yielding [ti?f]J and [pf?mlJl]. Glottalization (which also applies to [t] in
likely that Slur is the same rule as Syllabic Resonants, which applies to the
mountain and mutton and to [p] in popbottle) must therefore precede the
sequences [ar al an ;;im al)] to yield [~ ! 1:1!1 !J]in such words as pal'.or, ~edal, processes of variation with transitional stops, if the latter are to be blocked.
bacon, ransom, and baking. There does appear to be some gradation m t~e Glottalization is much more extensive in some dialects, Scots English for one,
operation of Syllabic Resonants, at least to the extent that [!Jl appears m than it is in mine.
faster speech than the other syllabic resonants, and perhaps that [~] appears 20. An extension of Glottalization applies to damp and junk in vecy fast
in faster speech than [:r ~µ].I have not yet made a systematic investigation of speech, even before vowels and in phrase-final position, to yield [dEe?] and
these matters. [J.4.?J,which cannot be further reduced. Glottalization applies to [tJ as well
10. Kenyon and Knott (1953) give palatals for all of these words, but many as [p k]: note [k!"e?] for can't in fast speech. The reduced versions of words
speakers have clusters in fairly slow speech. like went [wen] cannot be considered as resulting from Glottalization followed
11. I know of but one exception, wintry. by a [?]-deletion rule, for two reasons: (a) [dJ drops in the same environments
12. In many British dialects ...Slur applies to the words in (19b) and (19c) as as [t], but has no intermediate form[?], and (b) after the operation of Glot~
well as to those in (19a). These dialects also reduce (and subsequently delete) talization, original {p t k] have fallen together as [?] and cannot be distin~
the penultimate vowel in words like legendary, imaginary, and secondar]!, guished in the operation of a putative [?] -deletion rule, which nevertheless
which cannot have reduced vowels in my dialect, and in many of the words m would have to drop[?] from original [t] but not from [k p].
(20) and (22) below. 21. William Labov has pointed out that in some Black English dialects the
13. The words probably, mathematics, and vaudeville are subject to an ex- rule is extended to the past tense ending as well.
ceptional deletion of [a], in the case of probably and vaudeville before an 22. Dentdel, or a rule having a similar effect, is obligatory in words like
obstruent and in the case of mathematics before [m] followed by a stressed fasten, soften, hasten, and moisten.
vowel. Note that the corresponding reduction of probable is unacceptable. 23. Final [d] remains, and final It] is replaced by [?J in very fast speech
c.-J.Bailey has pointed out to me that various dialects syncopate [a] follo--:v- (see notes 19 and 20 above).
ing [r 1], in words like America, orange, Carolina, guarantee, caramel, Phrla~ 24. In my wife's dialect the [y] is preserved except after [s r 1]. This
treatment of the [y] involves a principle separate from the hierarchy in (5),
300

namely the functioning of [s1 with [r] as continuants,

stops, both oral and nasal.
Zwicky

in opposition to the
' Note on a Phonological Hierarchy in English 301

Ladefoge_d, Peter. 1967. Linguistic phonetics. Working papers in phonetics, 6.

Phonetics Laboratory, University of California, Los Angeles.
Ross, John Robert, 1967. Gapping and the order of constituents. Paper pre-
25. Also [g d b] . Foley's hierarchy in (I) is ordered by sonority, presumably
sented at the 10th International Congress of Linguists, Bucharest.
on the basis that the most sonorous stops, the voiced velars, are most likely to
Williamson, Kay. 1965. A Grammar of the Kolokuma Dialect of Jjo. West
drop when surrounded by the maximally sonorous segments, vowels (especial-
~frican language monographs, 2. Cambridge: Cambridge University· Press.
ly stressed vowels). Note Fletcher's values for "phonetic power" cited by
Zw1cky, Arnold. 1969. Englisp. contractions: interactions between syntax and
Jakobson, Fant, and Halle 1961: 28).
phonology. The Ohio State University. Dittoed.
26. A complicating factor is the treatment of initial [O], which is affected
by Glide Deletion in approximately the same way as [ w].
27. Hans Hock has called my attention to a number of difficult cases in the
development of various Inda-European languages. The common feature of
Hock's examples is the exclusion of one nasal from a larger class containing
the other nasal and the liquids. Presumably, principles other than the one dis-
cussed in this paper are to be invoked, although these matters are by no
means clear. An apparent, rather than real, difficulty is the Latin rule assimi-
lating the [n] of con- to a following resonant, a rule sometimes said (as in
Hale and Buck 1966: 25) to apply before nasals and [r] but not before [1];
as Eric Hamp and C.-J. Bailey have pointed out to me, what appears to be
the case is that new formations with con-, with no assimilation only at first
(regardless of whether [r] or [1] followed), were made at different periods
in the history of Latin.

REFERENCES
Bach, Emmon. 1968. Two proposals concerning the simplicity metric in
phonology. Glossa 2: 128-49.
Chomsky, Noam, and Morris Halle. 1968. The Sound Pattern of English. New
York: Harper and Row.
Greenberg, Joseph H. 1963. Some universals of grammar with particular refer-
ence to the order of meaningful elements. Universals of Language, Joseph H.
Greenberg, ed., pp. 58-90. Cambridge: M.I.T. Press.
Hale, William G., and Carl D. Buck. 1966. A Latin Grammar. University, Ala.:
University of Alabama Press. Reprint of 1903 edition.
Halle, Morris. 1962. Phonology in generative grammar. Word 18: 54-72.
Jakobson, Roman. 1968. Child Language, Aphasia and Phonological Universals.
The Hague: Mouton. Translation by Allan R. Keller of Kindersprache,
Aphasie und allgemeine Lautgesetze, 1941.
Jakobson, Roman, C. Gunnar M. Fant, and Morris Halle. 1961.Preliminaries to
Speech Analysis, 4th printing. Cambridge: MJ.T. Press.
Kenyon, John S., and Thomas A. Knott. 1953. A Pronouncing Dictionary of
American English, 4th ed. Springfield, Mass.: G. & C. Merriam.
Kiparsky, Paul. 1968. Linguistic universals and linguistic change. Universals in
Linguistic Theory, Emmon Bach and Robert T. Harms, eds., pp. 170-202.
New York: Holt, Rinehart and Winston.