0% found this document useful (0 votes)

42 views34 pages

Prosodic Phonology With A New Foreword - (PG 59 - 92)

Uploaded by

claudianunes00.cn

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)

42 views34 pages

Prosodic Phonology With A New Foreword - (PG 59 - 92)

Uploaded by

claudianunes00.cn

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

Chapter 2

Motivation for Prosodic Constituents

2.0. Introduction

Among the processes that involve a modification of the sound pattern

of a language, there is a qualitative difference between those processes
that must be formulated with rules that make direct reference to spe-
cific morphological or syntactic properties of the elements involved
and those that can be formulated without making direct reference to
such information. It is the latter type of processes, those in which there
is no systematic correspondence between the domains that must be re-
ferred to in the formulation of the process and the constituents of the
morpho-syntactic hierarchy, that constitute the strictly phonological rules
of a language, henceforth referred to as purely phonological or simply
phonological rules, a subset of which consists of the prosodic rules that
are the focus of this book.
In the first part of this chapter, we will briefly discuss the types of rules
that are not subsumed under our definition of purely phonological rules.
In the second and third sections, a number of arguments will be presented
that demonstrate why morphological and syntactic constituents cannot
constitute the domains of application of certain phonological rules. This
failure of morpho-syntactic constituents is taken as evidence of the need
for some other type of constituents, which we propose are the constitu-
ents of the prosodic hierarchy, discussed in Chapters 3 through 8. The
specific types of criteria used in order to motivate the postulation of a
prosodic constituent will be discussed in the last section of this chapter.

2.1. Phonological processes in nonphonological contexts

Since we are concerned here with purely phonological rules, that is,
those rules that make reference only to phonological elements in their
formulation, we must be able to distinguish these rules from other pro-
cesses in which phonological information alone is not adequate. In order
to make clearer the type of phenomena we will consider in this book, we
will briefly examine below the properties of the types of rules we do not
intend to discuss further. Strictly phonological rules, and in particular
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

Nespor, Marina, and Irene Vogel. Prosodic Phonology : With a New Foreword, De Gruyter, Inc., 2007. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/lancaster/detail.action?docID=3049945.
Created from lancaster on 2024-09-16 09:46:24.
28 Prosodic Phonology

prosodic rules, will be discussed in subsequent chapters in relation to the

prosodic constituents that serve as their domains of application.

2.1.1. Morphological contexts

There are many phonological processes that apply only under specific
morphological conditions. These may be divided into two general groups
on the basis of the type of morphological information required. That is,
there are some rules that only need to 'see' morphological structure, and
others that also need to 'see' specific morphemes or types of morphemes.
The rules that need to have access to morphological structure are those
that have typically been accounted for by different types of boundaries in
their formulations within the SPE framework. For example, the fact that
Nasal Assimilation in English applies to the prefix in- but not to the pre-
fix un- was captured in the SPE framework by allowing the rule to apply
across '+' boundaries but not across ' # boundaries. The rule, therefore,
applies in the case of illegal and irresponsible, but not in the case of
unlawful and unreliable, because of the different boundaries in their un-
derlying representations:

(1) a. in + legal illegal

in + responsible -*• irresponsible
b. un# lawful *ullawful
un# reliable -»· *urreliable

In more recent proposals, such as the Level Ordering Hypothesis (see

Siegel, 1974, 1977; Allen, 1978) and Lexical Phonology (see Kiparsky,
1982; Mohanan, 1981), the boundaries have been replaced by other
mechanisms that also take morphological structure into account, and
thus allow assimilation to apply in cases like those in (la), but not those
in (lb). 1 Regardless of the mechanism chosen to represent the facts of
Nasal Assimilation, it is clear that whether or not the rule applies is not a
purely phonological issue. The words below in (2) show that assimilation
does not apply in the absence of a morphological juncture, and, as the
examples in (lb) above and the additional examples in (3) show, it does
not apply at all morphological junctures either.

(2) only -• olly

Stanley *Stalley
Henry -> *Herry

(3) sudden- ly -· suddelly

man - like -> *mallike
sun - less -*• *sulless
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

Thus, the formulation of the rule of Nasal Assimilation must take into
account information about the morphological structure of the words in
question. This type of rule is, therefore, different in nature from the
strictly phonological rules we will consider in subsequent chapters of this
book.
Another example of a rule that must take morphological structure into
account is the Vowel Deletion rule of Italian that deletes an unstressed
vowel before another vowel only when the vowels are separated by a
morpheme juncture. Thus, the rule applies in the words in (4a), but not in
those in (4b). 2

fama - oso famoso 'famous'

giallo - astro -ν giallastro 'yellowish'
fiore - aio fioraio 'florist'
castoro - i castori 'beaver + pi.'
paura *pura 'fear'
caotico *cotico 'chaotic'
meandri *mandri 'meanders'
In both the English and Italian examples, in order for the relevant pho-
nological processes to apply, the formulation of the rule must contain
not only phonological information, but also information about the mor-
phological structure of the items involved. It is not necessary, however,
to provide any morphological information other than that relating to the
structure. Structural information is not adequate, however, for all
morpho-phonological rules. For example, there are rules that apply only to
specific lexical categories or in the presence of specific morphemes. The
first type can be illustrated on the basis of the Main Stress Rule in English
(see SPE). This rule applies differentially to different lexical categories,
such that primary stress is assigned to the first syllable of the noun and to
the last syllable of the verb in the following set of minimal pairs:

(5) a. [record vs. [record ]y

b. [pervert]^ vs. [pervert]y
c. [siibject]^ vs. [subjectjy
d. [extract]^ vs. [extractjy

The second type of rule can be exemplified by the rule of ζ-Devoicing

in English. As was mentioned in Chapter 1, this rule applies only in the
presence of a particular morpheme, -ive (see SPE), as illustrated by its
application in (6a), but its lack of application in (6b).

(6) a. abu[z]e + ive abu[s]ive

b. abu[z]e + ing->· abu[z]ing (*abu[s]ing)
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

Another rule that applies only in the environment of a specific morpheme

is the rule of Affrication in Italian (see Scalise, 1983). This rule changes
[t(:)] to [t(:)s] before the suffix -ione (after the application of Vowel
Deletion), as illustrated in (7).

(7) corre[t:]o - ione -+ corre[t:s]ione 'correction'

distin[t]o - ione distin[ts]ione 'distinction'
danna[t]o - ione danna[ts]ione 3 'damnation'

Affrication does not apply in the same segmental context (i.e. before
[j]V...) present in other suffixes, as seen in (8a), nor does it apply in
the same segmental context where no morpheme juncture intervenes,
as seen in (8b).

carre[t:]o - iere *carre[t:s]iere 'wagoner'

(cf. carre[t:]iere)
den[t]e - iera -» *den[ts]iera 'denture'
(cf. den[t]iera)
insala[t]a - iera *insala[ts]iera 'salad bowl'
(cf. insala[t]iera)
ottiene *o[t:s]iene '(he) obtains'
(cf. o[t:]iene)
sen[t]iero *sen[ts]iero 'path'
(cf. sen[t]iero)
e[t]iopico -> *e[ts]iopico 'Ethiopian'
(cf. e[t]iopico)

What all the rules discussed in this section have in common is the fact
that they carry out phonological modifications in a context that must
contain morphological information as well as phonological information.
They cannot, therefore, be considered purely phonological rules, and their
domains of application cannot be expressed in terms of prosodic pho-
nological constituents. This is not to say that all rules that apply at or
below the word level are necessarily morpho-phonological rules. It will
be seen in Chapters 3 and 4, in fact, that there are also purely phonolog-
ical rules that apply at and below the word level, their domains of applica-
tion being formulated in terms of prosodic phonological constituents.
We will assume that the former type of rule is to be handled by a
mechanism such as lexical phonology, while the latter will be handled by
the prosodic subsystem of phonology. Consequently, only the latter type
of rules will be discussed further in this book.
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

2.1.2. Syntactic contexts

Given that there are (at least) two types of rules that change the sound
pattern of a language at or below the word level - those that belong to
the prosodic subsystem of phonology and those that belong to another
subsystem - an empirical question arises as to whether a parallelism exists
between the phonology at and below the word level, on the one hand, and
that above the word level, on the other hand. That is, one can ask whether
above the word level, too, there are two types of phonological rules,
those that make reference to syntax and those that belong to the prosodic
subsystem of phonology. In the discussion of phonological rules operating
above the word level, it will be shown in Chapters 5 through 8 that a
number of rules that were thought to have a syntactic domain of applica-
tion belong, instead, to prosodic phonology, in that direct reference to
syntactic structures does not yield a proper definition of their domains
while reference to prosodic constituents can handle such cases. The ques-
tion we will now address is whether there is also a group of phonological
rules above the word level that belongs to another, not purely phonolog-
ical, subsystem of the grammar, as is the case for word-internal rules.
In early generative grammar, no attempt was made to specify the
domains of application of sandhi rules applying across words. The pioneer-
ing work in the field is that of Selkirk (1972), who proposed that by
translating the syntactic constituent analysis in such a way that infor-
mation about the bracketings is made accessible to phonological rules,
it is possible to account for the domain of application of phonological
rules applying across words, for example Liaison in French. The specific
way in which this translation is carried out is by inserting word boundaries
into the phrase marker, following the conventions proposed in SPE and
Selkirk (1972). While phonological rules could not look into the syntactic
structure of a sentence, they could see such boundaries. For example,
as was mentioned in Chapter 1, it was claimed that Liaison could apply
across one boundary, but was blocked if more boundaries intervened
between two words. That syntactic constituents could not be exactly
identified with the domains of application of phonological rules, however,
was already clear in Selkirk's work itself, where special conventions were
needed both to delete a word boundary in certain syntactic structures,
thus allowing Liaison in contexts in which it would otherwise be blocked,
and to insert a word boundary when the opposite effect was needed
(see also Rotenberg, 1975, 1978).
Once it became clear that syntactic constituents were not the answer
to the problem of defining the domain of application of at least some
phonological rules, several linguists attempted to define the domains of
a number of rules in terms of other types of syntactic notions. A syntactic
distance analysis was proposed by Rotenberg (1975, 1978) to account for
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

the domain of application of French Liaison; the relevance of left branches

in a syntactic tree was proposed by Napoli and Nespor (1979) in the
form of the Left Branch Condition, to account for the domains in which
the Italian rule of Raddoppiamento Sintattico applies. Similarly, Clements
(1978) proposed that certain tonal rules of Ewe are sensitive to the distinc-
tion between left and right branching structures. Most of these rules have
been subsequently reanalyzed in terms of phonological, or prosodic,
constituents (see Selkirk, 1978b, for Liaison; Nespor and Vogel, 1982,
for Raddoppiamento Sintattico and Liaison; Hayes, to appear, for the
tonal rules of Ewe; and Chapters 5 through 8 below for a variety of
other rules). Since the specification of the domain of application of a rule
in terms of a constituent is theoretically simpler - that is, more restric-
tive - than a specification of such a domain that refers to subsets of specific
types of constituents, such a solution is to be preferred to one that refers
to syntactic structures.
There are, however, rules that apply across words that need to refer
to information expressed by the (syntactic) labeled bracketings. That is,
certain rules apply across two words only if they are contained within a
specific syntactic constituent; other rules apply only if the word that
undergoes the modification has a specific syntactic label.
The first type of rule is exemplified by two vowel deletion rules of
Greek discussed in Kaisse (1977). The first rule, Unrounded First Vowel
Deletion, deletes the final vowel of word t in a sequence word } word 2
if the vowel is unrounded and word 2 also begins with a vowel, but only
under the condition that the two words are within an NP. The second rule,
Less-Sonorant Vowel Deletion, deletes one of the two vowels in a sequence
word j word 2 , where word t ends with a vowel and word 2 starts with a
vowel. The specification of which vowel is deleted is complicated and not
relevant to the present discussion. The relevant point here is that the two
words must belong to the same VP.
The second type of rule, that which must see the syntactic label of the
word which undergoes the rule, can be exemplified by a vowel deletion
rule of Italian (see van Hoorn, 1983; Vogel et al., 1983). Thus rule, Verb
Final Vowel Deletion, optionally deletes the final vowel of a word if a) the
+corona
vowel is immediately preceded by a ^ consonant, preceded,
[_+sonorant_
lin a specific domain,
in turn, by a vowel, b) another word follows with
and c) the word where the deletion takes place is a verb. Thus, while in
(9a) the final vowel of the verb may be deleted, deletion of the final vowel
of a noun in a similar segmental context yields the ungrammatical sentence
given in (9b).4
(9) a. So che vuol[0] nuotare. (<vuole)
Ί know he wants to swim.'
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

b. *Ho le suol[0] nuove. (< suole)

Ί have new soles.'

It should be noted that the different syntactic relation between the

words directly involved in the rule is not the crucial factor. Consider
the items in (10), where in both cases the word that presents the seg-
mental environment for deletion is the head of a phrase and the following
word is its complement. Deletion, nevertheless, applies to the verbal
but not to the nominal head (with the exception of a small group of
nouns ending in -re (see van Hoorn, 1983; Vogel et al., 1983)).

(10) a. parlan[0] bene (<parlano)

'they speak well'
b. *un gabbian[0] bianchissimo (<gabbiano)
'a very white seagull'

Similarly, syntactic categories play a crucial role in the Down Step rule of
Igbo where, in the sequence . . . V # VCV (where " ' indicates high tone),
a rule applies to lower the initial tone of the second word in certain
domains, if this word is a noun (see Weimers and Weimers, 1969; Wei-
mers, 1973; Kenstowicz and Kisseberth, 1977). The application of this
rule is illustrated below, where the initial tone of άηύ, a noun, is lowered
in example (11a), whereas the initial tone of ocd, an adjective, is not
modified in example (1 lb).

(11) a. wetää'nü (V' = vowel with a down-stepped tone)

'bring meat'
b. üwe oca
'a white garment'

From these and similar examples described in various works (Kenstowicz

and Kisseberth, 1977; Kaisse, 1977; among others), the conclusion must
be drawn that a separate subsystem of rules exists in which reference is
made to the information borne by labeled bracketings. Thus, the answer
to the question raised at the beginning of this section is that a parallelism
does exist between the types of rules operating at and below the word
level and the types of rules operating above the word level. That is, in both
areas, a distinction must be drawn between those phonological processes
that are to be handled by purely phonological rules and those that are to
be handled by rules that also take other than phonological information
into account.
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

2.2. Failure of morphologically specified contexts of prosodic phenomena

As has been seen in the preceding sections, there are certain phonological
rules that apply in contexts that are specified directly in terms of morpho-
syntactic constituents. This is not the case, however, for all phonological
rules. It will be shown in this section that there are rules whose domain of
application cannot be formulated in terms of morpho-syntactic constitu-
ents. The first language we will consider here as an illustration of this
point is Yidiji, a language spoken in Northern Queensland Both the data and
analyses reported here are based on Dixon (1977a, b). In Yidip, there is
a phonological rule that lengthens the penultimate vowel of any underived
word with an odd number of syllables, as can be seen below, where the
examples in (12a) are words with an odd number of syllables, and thus
show Penultimate Lengthening (PL), while the examples in (12b) are
words with an even number of syllables, and therefore do not undergo the
rule.

(12) a. guda:ga 'dog'

mad^ndar) 'walk up'
b. müd,am 'mother'
gumbiraqa 'pick up'

In derived words, whether or not the sequence root + suffix forms a

domain of application of Penultimate Lengthening depends on the num-
ber of syllables contained in the suffix. If the suffix is monosyllabic
and is affixed to a root with an even number of syllables, thus yielding
the desired environment for Penultimate Lengthening - that is, a derived
word with an odd number of syllables - then the rule applies, as shown in
(13).

(13) a. gali:-na
go (purp.)
b. qunar)gara:-nda
whale (dat.)

If, however, a bisyllabic suffix is affixed to a root with an odd number of

syllables, thus forming a derived word with an odd number of syllables,
PL does not apply, as seen in (14), although the word contains the desired
environment consisting of an odd number of syllables. Instead, the penul-
timate syllable of the stem is lengthened.

(14) mad|,i:nda-galig
walk up (pres.)
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

It should be noted that the two types of suffixes, monosyllabic and bi-
syllabic, cannot be assigned to different morphological categories. That
is, there are no morphological regularities, such as linear ordering with
respect to each other, that might suggest the existence of two morpholog-
ical classes of suffixes. They behave in different ways phonologically
because of different phonological characteristics (i.e. number of syllables)
that cannot be captured, in a non ad hoc way, in morphological structure.
Further evidence that morphological constituency cannot define the
context of application of Penultimate Lengthening is offered by derived
words that contain the sequence root + bisyllabic suffix + monosyllabic
suffix. The sequence formed by the two suffixes behaves like an underived
word as far as PL is concerned, as shown in (15).

(15) gumari-daga-jiu -> gumä:ridagä:jiu 'to have become red'

red (inch.)(past)

While the lengthening in the root can easily be expressed in terms of

morphological boundaries, there is no way to formulate the lengthen-
ing in -daga:-flu in terms of morphological boundaries, since there is no
constituent of morphological structure that exhaustively includes the two
suffixes. Another rule deletes the final -u in the surface form.
Additional support for the claim that the domains of application
of phonological rules are not necessarily isomorphic to morphological
constituents can be found in another phonological rule of Yidiji, Stress
Assignment, which assigns stress to alternating syllables. In the morpholog-
ical word in (15), the stressed syllables are the second and the fifth ones.
That is, there are two adjacent unstressed syllables, the third and the
fourth ones. Since this is a forbidden sequence in Yidiji, we are led to the
conclusion that the domain of Stress Assignment is not the morphological
word, and that, more generally, the morphological word is not a constitu-
ent as far as phonological rules are concerned.
Further evidence that constituents of the morpho-syntactic hierarchy
cannot be considered the domains of application of at least some word-
level phonological rules comes from two assimilation rules of Greek (see
Householder, 1964: Nespor, 1986). One rule, Nasal Assimilation, as-
similates a nasal consonant in point of articulation to a following non-
continuant consonant; the other rule, Stop Voicing, voices a stop when
it is preceded by a voiced consonant. The application of the two rules
is exemplified in (16), where it is shown that the rules apply across mor-
phemes within a word.

(16) συν+πλέκω '(I) knit'

sin+pleko si[mb]leko
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

Nasal Assimilation and Stop Voicing, however, are not strictly word in-
ternal but may apply in certain cases across word boundaries as well.
For example, they may apply across two words when wordj is an article
and word 2 a noun, or when wordj is the negative element bev and word 2
a verb, as shown in (17). They may not apply, however, when wordj
is an auxiliary and word 2 a verb, as shown in (18).

(17) a. τον#πατέρα 'the father (acc.)'

ton#patera [tombatera]
b. δβν#πβφάξί '(it) doesn't matter'
öen#piräzi [öembiräzi]

(18) έχουν#πλέξει '(they) have knitted'

exun#pleksi -*• *[exumbleksi]

Since in each of these examples the two words in question are separated
by a single word boundary, but the phonological rules apply in the first
two cases but not in the third, a theory of phonology that makes use of
boundaries to define the domains of application of phonological rules is
inadequate. In order to account for the facts just illustrated within such
a theory, it would be necessary to posit a morpheme boundary rather
than a word boundary between articles and nouns and between negative
elements and verbs, so that the rules would be word internal. This so-
lution, however, aside from being ad hoc, is untenable for syntactic
reasons. That is, since boundaries reflect morpho-syntactic structure,
positing a morpheme boundary between articles and nouns and between
negative elements and verbs would amount to claiming that these strings
are dominated by a single terminal node in syntactic structure. Such a
claim is false, however, since other words may intervene to separate the
two elements of the string.
To summarize, what the examples from Yidiji and Greek show is that
a phonological theory in which the domains of application of phonolog-
ical rules must be expressed by making reference to morphological
boundaries and word boundaries is inadequate, given that the constituents
of morphological structure are not necessarily isomorphic to the domains
of application of a set of phonological rules. An account of these and
other phenomena in a variety of languages, whose domains of application
must be expressed in terms of another type of constituent, the phonolog-
ical word, forms the foundation of Chapter 4.

2.3. Failure of syntactically specified contexts of prosodic phenomena

In considering phonological rules operating across words, a problem arises

that is analogous to the one seen in relation to rules that apply within
the word. That is, it is necessary to define the domains of application of
such rules. While in traditional generative theory it was supposed that
these domains directly correspond to syntactic constituents (see Chomsky
and Halle, 1968; Selkirk, 1972), it has been demonstrated in more recent
work that such a claim cannot be maintained. In this light, several alter-
native proposals have been made in which different syntactic notions have
been used to characterize the domains of phonological rules operating
between words (Bierwisch, 1966; Rotenberg, 1978; Clements, 1978;
Napoli and Nespor, 1979). While all of these proposals have contributed
to our understanding of the phenomena in question, they nevertheless
have certain shortcomings in that they fail to correctly predict all cases
of application of a given rule, or are not sufficiently general to be appli-
cable to a large number of other phenomena.
In the following sections, we will examine in some detail several of the
problems associated with the claim that syntactic constituents are the do-
mains of application of phonological rules. Specifically, we will show that
syntactic structures fail in three respects. First of all, it will be demonstrat-
ed that direct reference to syntactic constituents either results in the loss
of generalizations or leads to incorrect predictions about the application
of a number of phonological rules, evidence that these constituents cannot
be the appropriate domains for the rules in question (section 2.3.1).
Secondly, while it follows from a strictly syntactic approach in which
phonology directly refers to s-structures that empty elements are visible
to phonological rules, we will demonstrate that such a position is unten-
able (section 2.3.2). Finally, we will examine in some depth the inade-
quacy of syntactic constituents as the domains of intonation contours, a
point that has already been touched on in the traditional generative
literature (section 2.3.3).

2.3.1. Noncorrespondence between syntactic constituents and domains of

phonological rules
The inappropriateness of syntactic constituents as the domains of appli-
cation of phonological rules will be demonstrated on the basis of three
different types of problems. First of all, it will be shown that direct
reference to syntactic constituents does not make the correct predictions
about the domains of phonological rules. Secondly, whereas syntactic
constituency is determined uniquely in terms of structural factors, it will
be shown that a nonstructural factor, the length of a given string, is rel-
evant to the phonology in that constituents of the same syntactic nature
but different lengths exhibit different behaviors as far as the application
of phonological rules is concerned. Finally, in contrast with the implicit
prediction made by a syntactic constituent approach to phonology that
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

the largest possible domain of application of a phonological rule is the

sentence, it will be demonstrated that there exist phonological rules that
apply in larger domains.

Bracketings
If we try to express the domains of certain phonological rules in terms
of syntactic constituents, we find that the bracketings of these constit-
uents do not delimit units that account for the application of the rules,
as can be illustrated by Raddoppiamento Sintattico in Italian.
Raddoppiamento Sintattico (RS) is a rule of central and southern
varieties of Italian that lengthens the initial consonant of word 2 in a se-
quence word j word 2 - RS does not apply, however, in just any sequence of
two words, but requires that two phonological conditions be met. These
conditions vary a great deal according to the regional variety of Italian.
We will consider here only the variety spoken in the Tuscan region, the
one most commonly described in traditional works on Italian (Fiorelli,
1958; Camilli, 1965; Pratelli, 1970; Lepschy and Lepschy, 1977). The
phonological condition on word ( is that it must end in a stressed vowel.
Thus, RS applies in (19a), but not in (19b). 5

(19) a. La scimmia aveva appena mangiato metä [b: ]anana.

'The monkey had just eaten half a banana.'
b. II gorilla aveva appena mangiato quättro [b]anane.
'The gorilla had just eaten four bananas.'

The phonological condition on word 2 requires that the onset of the first
syllable be either a single consonant or a cluster other than s followed by
another consonant. RS applies, therefore, in (20a) and (20b), but not in
(20c).

(20) a. II ragno aveva mangiato metä [f:]arfalla.

'The spider had eaten half a butterfly.'
b. II ragno aveva mangiato metä [g:]rillo.
'The spider had eaten half a cricket.'
c. II ragno aveva mangiato metä [s]corpione.
'The spider had eaten half a scorpion.'

These conditions on the two individual words in a given sequence, how-

ever, are not sufficient to determine the occurrence of RS. Thus, although
the phonological conditions on word } and word 2 are met in (21), RS does
not apply.

(21) a. La volpe ne aveva mangiato metä [p]rima di addormentarsi.

'The fox had eaten half of it before falling asleep.'
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

b. II gatto aveva catturato un colibri [m]olto pregiato.

'The cat had caught a highly valued hummingbird.'

The examples in (21) illustrate that in order for RS to apply, wordj and
word 2 , in addition to fulfilling the phonological requirements just men-
tioned, must be in a particular relation to each other. It is the specifi-
cation of this relation that is problematic if it is to be expressed in terms
of syntactic constituents. The cases that are most problematic are those
in which the same types of syntactic constituents are treated different-
ly by a given phonological rule, a situation that a purely syntactic analysis,
by definition, cannot handle.
Consider first the following examples which illustrate that different
sister nodes behave differently with respect to RS.
W1 W
2 W3
(22) a. Ha appena comprato un colibri [b:]lü [k]on le ali sottilissime.
'He just bought a blue hummingbird with very thin wings.'
w w w
1 2 3
b. Caccerä [k:]aribu
'He will hunt [k]olwith
caribous fucile e cervi
a rifle andcon l'arco
deer withe alebow
frecce.
and ar-
row.
Wj w2 w3
c. Un levriero costerä [s:]uppergiü [m]ezzo milione.
Ά greyhound will cost about half a million (lire).'
w1 2w w3 w 4
d. Disegno [b:]alene blu [k]on inchiostro di lapislazzuli.
'He draw blue whales with lapislazuli ink.'

Although the syntactic structures of (22a-c) differ somewhat, in each

sentence the three words relevant to the present discussion are in the same
abstract relation to each other, as represented schematically in (23a),
where "H" stands for the head of the phrase and "C" for a complement.
The abstract relation among the four crucial words in (22d), on the other
hand, is represented in (23b).
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

While in each of these four sentences the segmental environment for RS

is met in two positions, that is, between Wj and w 2 and between w 2 and
w 3 in (22a-c), and between Wj and w 2 and between w 3 and w 4 in (22d),
the rule applies only in the first position. There is no way, however, in
(22a-c) for the head of the phrase to be grouped into a syntactic con-
stituent together with its first complement (w 2 ), but not with the second
complement, whose initial word is w 3 - Similarly, in (22d), syntactic con-
stituency does not allow us to group the head of the phrase ( w t ) with
its first complement (w 2 and w 3 ) to the exclusion of its second comple-
ment, whose initial word is w 4 . Since the head and complements are sister
nodes, there is no way to form a unit that consists only of the head and
its first complement. Thus, the domain of application of RS cannot be
identified with any of the syntactic constituents available in this case.
That syntactic constituents cannot account for the domain of applica-
tion of RS can be further illustrated with the following example, where
a series of right branching elements does not uniformly exhibit RS even
though the proper phonological conditions are present.

(24) a. Hovistotre [k:]olibri [b:]rutti.

Ί saw three ugly hummingbirds.'
b. Hovistotre [k:]olibri [k]osi [b:]rutti.
Ί saw three such ugly hummingbirds.'

In (24a), RS applies in both of the environments in which the phonolog-

ical conditions on the individual words are satisfied, that is, between tre
'three' and colibri 'hummingbirds' and between colibri and brutti 'ugly'.
In (24b), however, there are three environments in which the phonological
conditions for RS are met, that is, between tre and colibri, between colibri
and cosi 'such' and between cosi and brutti, but the rule applies only in
the first and third cases. Since all of the words in question are grouped
in a parallel way in syntactic structure, as can be seen in (25), there is no
reason to expect RS not to apply between w 2 and w 3 in (25b), since it
applies in all the other positions.

(25) w, w2 w3
a. ...[tre [colibri [brutti]]]
Wj w>2 w3 w4
b. ...[tre [colibri [cosi [brutti]]]]

While in (25a) the repeated application of RS seems to indicate that the

rule applies throughout an entire constituent, it can be seen that this is not
the case in (25b), where the rule does not apply throughout a similar
type of constituent. That is, if it is claimed that the domain of application
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

of RS is a syntactic constituent, it would be necessary to claim, further-

more, that in the example in (25b), and w 2 form an exhaustive consti-
tuent. There is, however, no syntactic constituent that groups these two
words together to the exclusion of w 3 and w 4 -
Another rule that shows the inappropriateness of syntactic constituents
as the domains of phonological rules in still another respect is the much
discussed phenomenon of Liaison, as illustrated in (26). While Liaison
applies in all styles of French, we are concerned here only with its applica-
tion in informal speech.

(26) a. Les giraffes et les elephants sont ses meilleurs^amis.

'Giraffes and elephants are his best friends.'
b. Claude a des perrochets / intolerables.
'Claude has some intolerable parrots.'

In syntactic terms, meilleurs 'best' is the complement of amis 'friends',

just as intolerables 'intolerable' is the complement of perrochets 'parrots'.
That is, meilleurs and amis in (26a), just as perrochets and intolerables
in (26b), are sister nodes directly dominated by the same type of node.
It is, therefore, impossible to predict in terms of syntactic constituents
that Liaison applies in the first case but not in the second.
The three problems just discussed show that we cannot identify
the domains of application of phonological rules with syntactic constit-
uents, since it was shown that neither of the two rules under consideration
applies uniformly throughout a constituent of a given type. That is, there
is no way to capture in terms of syntactic constituent structure the fact
that similar constituents behave differently with respect to the same
phonological rule.

Length of constituents
Another problem that arises if syntactic constituents are posited as the
domains of application of phonological rules is related to the length of the
constituents involved. Since constituents in syntax are defined in terms of
certain structural relations among the words of a given string, the number
of words involved is irrelevant. That is, as far as the syntax is concerned,
a constituent of a certain type composed of one word is structurally
equivalent to another constituent of the same type composed of five,
ten, or any number of words. If syntactic constituents are taken as the
domains of application of phonological rules operating above the word
level, this means that, under the appropriate segmental conditions, a given
phonological rule should operate uniformly within all syntactic constit-
uents of a certain type. That is, a constituent composed of two words,
the minimum necessary for the application of a phonological rule operat-
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

ing above the word level, should exhibit the same phonological behavior
as the same type of constituent of any other length. It will be demonstrat-
ed here that certain phonological rules, in fact, do not apply through-
out all consituents of a given type, but rather are sensitive to the length
of the constituents.
Let us consider the rule of Gorgia Toscana (GT) in Italian. This rule
is traditionally described as a phonological phenomenon of Tuscan Italian
that results in varying degrees of so-called aspiration of the voiceless stops
[p], [t], [k] in intervocalic position, or more precisely, between two
[-consonantal] segments (see, among others, Lepschy and Lepschy, 1977;
Giannelli and Savoia, 1979). Since GT varies somewhat in different areas
of Tuscany, we will limit our discussion to the variety spoken in the
province of Florence. The most common form of GT changes [p], [t],
and [k] into [φ], [0], and [h], respectively, both within and across words,
as the examples in (27) and (28) illustrate.

(27) a. lu[0]o ( < lu[p]o) 'wolf

b. crice[0]o ( < crice[t]o) 'hamster'
c. fo[h]a (<fo[k]a) 'seal'
W W
1 2
(28) a. Lo zoo ha appena comprato una nuova [0]antera.
( < [p]antera)
'The zoo has just bought a new panther.'
Wj w2
Ci sono tantissimi [Θ ]arli nella mia scrivania.
«Marli)
'There are lots of wood-worms in my desk.'
wj w2
Sta arrivando uno stormo di [h]orvi neri.
(< [k]orvi)
Ά flock of black crows is arriving.'

Gorgia Toscana, however, does not apply across just any two words.
In the examples in (29), cases are seen in which the rule operates within
a VP, while in the examples in (30), cases are seen in which the rule does
not normally apply between an NP and a VP, even though the appropriate
phonological conditions are present (i.e., the consonant in question is
in intervocalic position).

(29) a. aveva [hjonosciuto Arcibaldo (<[k]onosciuto)

'(he) had met Arcibaldo'
b. viaggia [h]ol cammello (<[k]ol)
'(he) travels by camel'
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

c. si lamenta [hjostantemente ( < [kjostantemente)

'(he) complains constantly'

(30) a. Le zanne dell'elefante bianco dell'Africa Orientale [k]ostano

sempre di ρίύ in Europa.
(?*[h]ostano)
'The tusks of the white elephant of eastern Africa cost more and
more in Europe.'
b. Quella banda segreta di ragazzi temuta da tutti [kjacciaorsi
ferocissimi solo per divertirsi.
(?*[h]accia)
'That secret band of boys feared by all hunts very ferocious
bears just for fun.'
c. Certi tipi di uccelli trovati solo in Australia [k]ostruiscono nidi
complicatissimi a due piani.
(?*[h]ostruiscono)
'Certain types of birds found only in Australia construct very
complicated two-story nests.'

If the domain of application of GT is to be expressed in terms of syntactic

constituents, for example the verb phrase, the prediction is made that the
rule will not only apply in the VPs listed in (29), but in all VPs, including
those in (31) below. By the same token, just as GT does not apply be-
tween the NPs and VPs in (30), it is expected that the rule will not apply
between the NPs and VPs of the sentences in (32). As it turns out, how-
ever, contrary to the predictions, GT does not always apply in all the
positions in which its segmental context is present within a VP, as indic-
ated in (31). It normally does apply, however, across the NPs and VPs
in (32). "_" and in (31) indicate those segments to which GT normal-
ly does and does not apply across words, respectively.

(31) a. Osservano il rarissimo colibri peruviano con le penne azzure

con un cannocchiale particolarmente adatto alia situazione.

'(They) observe the very rare Peruvian hummingbird with blue

feathers with binoculars that are particularly suitable for the
situation.'
b. Hanno comprato un sacco di caramelle liquerizia e lattine di

coca cola con i soldi che hanno guadagnato vendendo giornalini.

'(They) bought a lot of candies, licorice, and cans of coca cola

with the money they earned selling comic books.'
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

(32) a. Gli struzzi [hjorrono velocemente. ( < [kjorrono)

Ostriches run quickly.'
b. Un levriero [h]osta moltissimo. ( < [k]osta)
Ά greyhound is very expensive.'

It is obvious from these examples that the length of a given string is

crucial in determining whether or not a phonological rule, in this case GT,
applies. Thus, while it seemed from the examples in (29) that the domain
of application might be, at least in part, identified with a phrasal node,
specifically VP, the items in (31) show that when a VP is particularly
long, GT tends not to apply in all positions. Similarly, while the sentences
in (30) might lead one to the conclusion that GT is blocked across the
boundary of two phrasal nodes, specifically NP and VP, those in (32)
demonstrate that when the NP and VP in question are relatively short,
GT in fact normally does apply.
This type of situation is by no means limited to the application of GT,
or to Italian, for that matter. An analogous problem arises, for example, in
specifying the domain of application of Nasal Assimilation (NA) in
Spanish.
It has frequently been observed that in Spanish nasals are homorganic
to a following consonant both within a word and across words (see among
others, Navarro Tomas, 1957; Harris, 1969; Hooper, 1976). Examples of
homorganic nasal and following consonant within and across words are
given below:

(33) a. ga[m]ba 'shrimp'

b. co[m] piedad 'with pity'
c. elefa[n]te 'elephant'
d. si[n] tardar 'without delaying'
e. laßjgosta 'lobster'
f. come [η] carne '(they) eat meat'

NA does not, however, apply between just any two words. As was seen
above in relation to GT, an examination of the domains in which NA
does and does not apply reveals that they do not necessarily coincide
with any syntactic constituent. While NA applies in the VPs in (34), it
tends not to apply in those in (35).

(34) a. tiene[rj] cuatro gatos ( < tiene[n])

'(they) have four cats'
b. canta[m] bien ( < canta[n])
'(they) sing well'
c. no se vea aquella especia de escorpiofn] ( < escorpiofn])
tan frecuentemente en Brasil
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

'(one) doesn't see that species of scorpion very frequently in

Brazil.'

(35) a. Colecciona en todo el mundo plumas de tucä[n] para su som-

brero preferido.
(?*tuca[m])
'(She) collects all over the world toucan feathers for her
favorite hat.'
b. Estudia el sistema comunicativo de algunos tipos de delfifn]
con aparatos muy sofisticados.
(?*delfi[rj])
'(She) studies the communicative system of several types of
dolphin with very sophisticated apparatus.'
Furthermore, while Ν A appears to be blocked between NPs and VPs
on the basis of sentences such as those in (36), the sentences in (37)
show that it is not blocked between all instances of NP and VP.

(36) a. El nuevo canario de mi amiga Carmefn] canta solo cuando estä

solo.
(?*Carme[r)])
'My friend Carmen's new canary sings only when it is alone.'
b. Su nuevo sombrero con tres plumas de tucä[n] cuesta sin duda
mas del sombrero de su hermano.
(?*tuca[r)])
'Her new hat with three toucan feathers undoubtedly costs
more than her brother's hat.'

(37) a. Mi faisä[rj] corre siempre. (<faisa[n])

'My pheasant always runs.'
b. Eso tucäfm] parece enfermo. (< tucafn])
'That toucan seems sick.'

There are still other phonological rules, for example, Intervocalic Spi-
rantization in Italian and s-Assimilation in Greek, that exhibit the same
general pattern just illustrated with GT and NA, in that they apply in a
given type of constituent if it is short, but not if it is long, or are blocked
across a particular type of boundary if the constituents on either side are
relatively long, but not if they are short. We will not illustrate these rules
here, but refer the reader to a more detailed discussion of this problem
in Chapter 7. The relevant point here is that such rules provide further
evidence that the domains of application of phonological rules operating
across words cannot be coextensive with the constituents provided by
syntax. That is, while it has just been demonstrated that the application
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

of certain phonological rules is sensitive to the length of the strings in-

volved, the principles that define syntactic constituents do not take into
account nonstructural factors such as length. There is, thus, a fundamental
difference between the nature of the domains relevant for phonological
rules and the constituents of syntax.

Beyond the sentence

We have just seen cases in which the domains of application of various
phonological rules could not be expressed in terms of syntactic consti-
tuency either because the syntax provides inappropriate bracketings of
the elements of a given string or because the phonology may be sensitive
to length, a factor the syntax does not take into account. In both cases,
the problem was essentially that the phonological rules in question did
not apply uniformly throughout the constituents provided by the syntax.
The last problem we will consider in this section is somewhat different
from the first two in that it does not so much represent a mismatch be-
tween syntactic constituents and domains of application of phonological
rules, but rather demonstrates that syntactic constituents, by definition,
cannot delimit domains large enough to account for the application
of certain phonological rules.
Let us first consider the well-known rule of Flapping in American
English. The examples in (38) show that this rule applies both within
words and across two words in a sentence.

(38) a. water wa[r]er

b. capital capi[r]al
c. Wait a minute. Wai[.f]a . . .
d. The white rabbit escaped from its cage. -*• . . .rabbi[r]escaped...

Flapping, however, is not restricted to applying within a sentence; it

may also apply across two words in different sentences, as the examples
in (39) illustrate.

(39) a. Please wait. I'll be right back. ...wai[>]Fll...

b. It's hot. Open the window. -> . . . h o f r J O p e n . . .
c. Don't shout. It's rude. -*• ...shou[>]It's...

It is not the case, however, that Flapping can apply across just any pair
of sentences. The examples in (40) show that in some cases (i.e. where
the two sentences are unrelated) the rule is blocked in exactly the same
segmental contexts in which it was seen to apply in (39).

(40) a. They didn't wait. I'll be right back. ->-*.. .wai|>]I'll...

b. Where's Scott? Open the window. -»· *.. .Sco[.r JOpen...

c. Don't shout. Is Ed here? *.. .shou[.r]Is...

Since the largest constituent in syntax is the sentence, there is no way

in which the domain of application of Flapping can be identified with a
syntactic constituent. That is, beyond the sentence there is no unit that
allows us to group together, for example, the pairs of sentences in (39),
and thus characterize the domain of Flapping in terms of this unit. The
fact that Flapping does not apply across just any two sentences compli-
cates the problem even further, since it automatically rules out the pos-
sibility of defining the rule's environment as being any place within the
largest syntactic constituent, or across two such constituents.
A similar problem arises with two other phenomena in another variety
of English as well, Linking-r and Intrusive-r in Received Pronunciation.
Both rules account for the pronunciation, in certain contexts, of an
r that is not pronounced when the word in question is uttered in
isolation. In the first case, the r is present orthographically, and in the
second case, it is not present (see among others, Gimson, 1970). These
rules apply, as does Flapping, within words, across words within a sen-
tence, and across words in different sentences, as illustrated in (41), where
"t" indicates an orthographic r that is not pronounced when the word is
uttered in isolation.

(41) a. clea^ + est clea[r]est

b. gnaw + ing gnaw[r]ing
c. That type of spidej1 is dangerous. ...spide[r]is...
d. The giant panda eats pounds of bamboo a day.
...panda[r]eats...
e. There's my brothe^. I have to go. -> ...brothe[r]I.. t
f. Try that sofa. It's softer. . . . sofa[r]It's...

As was the case with Flapping, not all combinations of sentences per-
mit the application of Linking-r and Intrusive-/·. For example, the two
rules in question would most probably be blocked in the pairs of sentences
in (42), even though it was just seen that they apply in identical segmental
contexts in (41e,f).

(42) a. There's my brothe^. I have a cold. ->·*.. .brothe[r]I...

b. Try that sofa. It's after midnight. -»· *...sofa[r]It's...

Once again, we are faced with the problem of how to represent the do-
main of application of phonological rules that can apply across two
sentences, since the largest constituent provided by the syntax is the sen-
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

tence. Furthermore, the fact that the rules apply across some sentences
but not others provides additional evidence that we are dealing with rules
that operate in specific contexts larger than the sentence, not simply in
any sequence of sentences.
Finally, the Voicing Assimilation rule of Mexican Spanish discussed
by Harris (1969) appears to be another example of a phonological rule
that operates beyond the sentence. This rule, that voices s before a voiced
consonant, may, under certain circumstances, apply across two sentences,
as illustrated by the pair of sentences below (from Harris, p.60).

(43) Los dos. Dämelos. ,.,do[s z ]Dämelos.

'Both of them. Give them to me.'

Harris points out further that the rule does not always apply across sen-
tences; rather, it is blocked by an interruption in 'phonational activity'.
Thus, although Harris does not provide any other examples, it seems
quite likely that Voicing Assimilation represents, in fact, the same type
of phenomenon we have just examined in American and RP varieties of
English.
What all of these rules demonstrate is that syntactic constituency
cannot provide the appropriate domains for the application of a specific
type of phonological rule, that is, any rule that can apply across certain
sentences but not others. Since the largest syntactic constituent is the
sentence, there is by definition no way in which the domain of applica-
tion of such rules can be identified with a constituent of syntactic struc-
ture. It seems, instead, that a more appropriate way to account for the ap-
plication of the phonological rules in question would be in terms of a dif-
ferent type of unit, one that is not limited by the constituent structure
of the syntactic component.
The combination of the problems examined above in which it was seen
that a number of phonological rules do not apply uniformly throughout
the strings delimited by syntactic constituents, and the problem of phono-
logical rules operating beyond the sentence that has just been addressed
here, quite clearly reveals the inadequacy of syntactic constituents as the
domains of application of (at least some) phonological rules. That is, there
is no simple one-to-one relation between the constituents of syntax and
the strings within which phonological rules operate. Instead, it seems that
units other than those provided by the syntactic component are necessary
in order to account for the application of phonological rules operating
above the word level.

2.3.2. Phonetically null syntactic constituents and phonological rules

It has been proposed by a number of linguists that phonological rules
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

that apply between two words are sensitive to the presence of syntactic
elements that do not have a phonetic matrix, such as traces and PRO
(see, among others, Selkirk, 1972; Chomsky and Lasnik, 1977; Rizzi, 1979;
Vanelli, 1979; Jaeggli, 1980). This position follows directly from the
theoretical claim that the labeled bracketings of s-structure are carried
directly into the phonological representation in the form of boundaries.
That is, since s-structure consists of bracketed constituents that may
contain phonetically null elements, it is claimed that such elements will
affect the application of phonological rules applying across words. It is
argued, specifically, that the effect of an empty element will show up
where adjacency is a necessary condition for the application of a pho-
nological rule. If a syntactic constituent is present between two words,
these words obviously can not be considered adjacent and thus are not
subject to the application of the phonological rule in question. There is no
unanimous agreement, however, on which phonetically empty syntactic
constituents are relevant for the definition of adjacency in phonology.
Research in this area has taken essentially two directions, which we will
examine below as the strong hypothesis and the weak hypothesis. Accord-
ing to the strong hypothesis, all empty elements have the same status that
lexical items have in determining adjacency. That is, the occurrence of any
such element between two words will prevent the application of a pho-
nological rule whose environment is otherwise present (see, among others,
Rizzi, 1979; Vanelli, 1979). According to the weak hypothesis, only cer-
tain types of empty elements, that is, those marked for case, are capable
of blocking the application of phonological rules that apply across words
(see Jaeggli, 1980; Chomsky, 1981). We will demonstrate below that syn-
tactic constituents not represented phonetically are not capable, under
any circumstances, of blocking the application of phonological rules and
thus that both of the hypotheses are untenable.
The empty elements to which recent proposals attributed the power
of blocking phonological rules are: traces of clitics, PROs, and traces of
wh. Of these, the last type is clearly marked for case, the second type is
clearly not marked for case, and the first type is somewhat controversial,
in that it is considered by some linguists to be marked for case (Longobar-
di, 1980) and by others not to be so marked (Jaeggli, 1980; Chomsky,
1981). We will consider each of these categories separately.

Traces of clitics
If traces of clitics have an influence on the application of phonological
rules, as has been proposed, for example, by Rizzi (1979), we would
expect both a) that in a given language these traces would block the
application of all phonological rules whose environment is defined on
adjacent words, and b) that either all or none of the traces of the same
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

type would block the application of a particular rule. Consider, however,

the sentences in (44) and (45), which include the segmental environment
for the phonological rule of Italian, Raddoppiamento Sintattico (see sec-
tion 2.3.1 above). In the examples in (44) the rule is blocked, while in
the examples in (45) the rule applies, despite the presence of traces of
clitics in both sentences.

(44) a. Ne voglio comprare rnetd ? CL [s]ubito. (no change)

Ί want to buy half of it immediately.'
b. Ne incontrero tre f C L [vjenerdi. (no change)
Ί will meet three of them Friday.'

(45) a. Ne comprero i C L [s:]ubito. (< [sjubito)

'I'll buy some immediately.'
b. Lo incontrero t C L [v:]enerdi. (< [v]enerdi)
'I'll meet him Friday.'

The inconsistent behavior of RS6 with respect to traces of clitics in these

examples indicates that the traces cannot be the factor that determines
in which contexts the rule is blocked, at least if a unified analysis of the
phenomenon is to be given.
A possible solution in which traces of clitics are present at the moment
of application of some phonological rules, for example the vowel deletion
rule mentioned by Rizzi, and absent at the moment of application of
others, for example RS, must be excluded since this would require the
introduction of a new type of rule within the phonological component
which would substantially deviate from the nature of the other rules of
the component. That is, it would be necessary to introduce a type of
rule that has as its only effect the deletion of syntactic material in order to
create adjacency between two words that were at some other point consi-
dered structurally nonadjacent. It should be noted that the ungrammatical
sentences given by Rizzi to show that Specifier Vowel Deletion (SVD)
is blocked by phonetically null syntactic constituents can be excluded
for independent reasons, as shown in Nespor and Scorretti (1985).

Pro
We will now consider the blocking effect of PRO, which was proposed
by Rizzi (1979) and Vanelli (1979) to account for the context of applica-
tion of the same rule used by Rizzi to argue in favor of the blocking effect
of traces of clitics: Specifier Vowel Deletion. An argument similar to the one
just given for traces of clitics can also be made for PRO on the basis of
two other phonological rules of Italian: Gorgia Toscana (GT) and Inter-
vocalic Spirantization (IS).
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

GT, as shown in section 2.3.1 above, applies both within words and at the
juncture between two adjacent words. The rule applies, however, in the sen-
tences in (46), in which, according to the analysis proposed by Rizzi
(1979) and Vanelli (1979), w 1 and w 2 are not structurally adjacent since
the empty element PRO intervenes between them.
w
i
(46) a. Tu dai da mangiare al puma australiano e io a quello PRO
W
2
[</>]eruviano.
(< [pjeruviano)
'You feed the Australian puma and I the Peruvian one.'
w, w2
b. Ci sono due leoni obesi e quattro PRO [0]utti magri.
(< [tjutti)
'There are two obese lions and four very thin ones.'
Wj w2
c. Ho visto un passero pennuto e uno PRO [h]alvo.
(< [kjalvo)
Ί saw a feathered sparrow and a bald one.'
Similarly, in the sentences in (47), GT applies between w l and w 2 in
additional types of constructions in which it is commonly accepted
that PRO separates the two words in question, although Rizzi and Vanelli
do not explicitly mention these cases.

(47) a. II mio cavallo e stato tutto un mese senza PRO [</>]ascolare.

(< [pjascolare)
'My horse has gone a whole month without grazing.'
Wj w2
b. II pappagallo ha trovato come PRO [0]orturarci con le nostre
parole.
(< [t]orturarci)
'The parrot has learned how to torture us with our own words.'
w2
c. II mio barbagianni crede di PRO [hjantare meglio di un
usignolo.
(< [k]antare)
'My barn owl thinks he sings better than a nightingale.'
Another segmental rule of Tuscan Italian is Intervocalic Spirantization,
the rule that changes the affricates [t/] and [d3] into the corresponding
fricatives [/] and [3] between two vowels (see Lepschy and Lepschy,
1977). As is the case with GT, IS applies both within words and across
words:
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

(48) a. cri[/]eto ( < cri[t/]eto) 'hamster'

b. fa[3]ano (<fa[d3]ano) 'pheasant'

(49) a. Questo dev'essere un uovo di [/]efalo.

( < [t/]efalo)
'This must be a mullet egg.'
b. Le [giraffe si abbeverano al tramonto.
(<[d3]iraffe)
'Giraffes drink at sunset.'

The sentences in (50) illustrate that IS applies in the same environments

as GT across intervening PROs in the type of analysis proposed by Rizzi
and Vanelli.

(50) a. Hanno trascurato l'armadillo zoppo per concentrarsi tutti su

W W
1 2
quello PRO [/]eco.
( < [t/]eco)
'(They) neglected the limping armadillo so all could concentrate
on the blind one.'

b. Parlava di un pinguino nano e di uno PRO [3]igante.

( < [d 3 ]igante)
'(He) was talking about a dwarf penguin and a giant one.'

IS, like GT, applies across other instances of PRO as well, as seen in (51).
Wj w2
(51) a. Non hanno ancora capito come PRO [/]ercare l'upupa.
( < [t/]ercare)
'They haven't yet figured out how to locate the hoopoe.'
w
i
b. La mia foca non puo resistere un solo giorno senza PRO
w2
[3]ocare a palla con Pierino.
( < [d 3 ]ocare)
'My seal can't go even one day without playing ball with
Pierino.'

These sentences lend further support to the point made above in relation
to traces of clitics, that is, that a phonetically empty element does not
have any effect on the application of phonological processes. That is,
GT and IS apply between two words that contain the relevant phonolog-
ical characteristics, the adjacency of the two words being totally unaf-
fected by the presence or absence of PRO. We must conclude, then, that
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

PRO does not have the same status as phonologically non-null consti-
tuents, in that it does not count in determining whether two words are
adjacent for the purposes of phonology.
At this point we must reject the strong hypothesis according to which
all phonetically empty elements, including those not marked for case,
may intervene between two words in the way a lexical item may, to
interrupt the adjacency of the words.

Traces of wh
We will now examine data relevant to the weak hypothesis, that is,
sentences in which it is case-marked empty elements, specifically traces
of wh, that intervene between two words that otherwise satisfy the re-
quirements for the application of phonological rules. Raddoppiamento
Sintattico and Gorgia Toscana are both examples of rules that apply freely
across case-marked traces, as shown in (52) and (53), respectively.

(52) Raddoppiamento Sintattico

a. Cosa filmerä i w h [d:]omani?
(< [djomani)
'What will he film tomorrow?'
b. Filippo e il cavallo che montero t ^ [d:]omani.
(< [djomani)
'Filippo is the horse that I will ride tomorrow.'

(53) Gorgia Toscana

a. Chi hai fotografato [hjol pappagallo sulla spalla?
«[k]ol)
'Who did you take a picture of with the parrot on his shoulder?'
b. Questi sono i picchi che abbiamo comprato f ^ [hjol sussidio
statale.
«[k]ol)
'These are the woodpeckers that we bought with the national grant.'

Phonological rules that apply across intervening case-marked traces are

found not only in Italian, but in all the languages we have investigated.
Consider, for instance, Nasal Assimilation in Spanish. While the sentences
in (54) show that NA applies between two words that are structurally
adjacent, the sentences in (55) show that the rule also applies freely across
traces of wh.

(54) a. El faisafm] peruano tiene la cola mas larga.

b. Quisiera ver un airo[n] tambien.

( < airofn])
Ί would like to see a heron too.'
c. Ha escrito un articulo sobre el sistema comunicativo del
delfi[r)] canadiense.
( < delfifn])
'He has written an article about the communicative system
of Canadian dolphins.'

(55) a. Que comefm] ^ para navidad las tortugas?

( < come[n])
'What do turtles eat for Christmas?'
b. Este es el coyote que fotografia[m] f ^ para la rivista.
( < fotografia[n])
'This is the coyote that they are taking a picture of for the
magazine.'

The last phonological rule we will discuss to illustrate the irrelevance of

the feature "case" for phonological rules is Linking-r in Received Pro-
nunciation. While a word-final r is usually deleted, it is retained in con-
nected speech if the following word begins with a vowel, as illustrated in
(56). (See Chapter 8 for a detailed analysis of this rule.)

(56) a. I'd prefer a monkey.

b. A caterpillar is a wormlike larva of a butterfly or moth.

Linking-/· applies, as do the other rules seen in this section, also when there
is an intervening case-marked trace, as is shown in (57).

(57) a. What would you order t ^ instead?

b. I can't think of what I could wear- twn. otherwise.

These facts, along with those of Italian and Spanish, lead us to the conclu-
sion that phonological rules are not affected by the presence of phoneti-
cally empty syntactic constituents marked for case.
Let us now consider a stress rule in Italian and one in English. The
Italian rule exhibits a mixed type of behavior in that it is never blocked
when the trace of wh intervening between the two words involved in the
rule is left either by relative clause formation or by indirect questions,
while it may be blocked when the trace is left by the movement involved
in the formation of direct questions. The English stress rule, however,
shows a more consistent pattern, in that it typically applies across all
traces of wh, regardless of their origin.
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

Consider first the rule of Stress Retraction (SR) found in standard

northern Italian. In a sequence of wordx word 2 , where w l has primary
stress on the last syllable and w 2 is stressed on the first syllable, SR
retracts the primary stress of (see Nespor and Vogel, 1979). This rule
applies freely across f ^ in (58a) and (58b), examples of relative clauses
and indirect questions, respectively, but it does not apply in (58c), a direct
question.

(58) a. Ho gia capito quello che färo i w h dopo.

(< faro) 7
'I've already understood what I will do afterwards.'
b. Ci si domanda cosa fära i w h dopo.
(< fara)
One wonders what he'll do afterwards.'
c. Cosa dirä ? wh dopo?
(?*dira)
'What will he say afterwards?'

Iambic Reversal (IR), a stress retraction rule of American English (see

Liberman and Prince, 1977), however, does not usually exhibit the mixed
behavior just seen in Italian with respect to traces of wh. That is, IR ap-
plies in (59a), (59b), and (59c), examples of relative clauses, indirect
questions, and direct questions, respectively.

(59) a. The picture that I'm going to reproduce t ^ later is the one
Emily took.
(< reprodüce)
b. They asked me which company Martha represents f ^ now.
(< represents)
c. What are they going to export t ^ next?
(< export)

As Nespor and Scorretti (1985) observe, attributing the blocking of SR

in Italian in (58c) to the presence of a trace would pose serious problems
to the commonly accepted unified derivation of the three phenomena
under consideration: relative clause formation, indirect questions, and
direct questions. It also seems to be in conflict with the uniform behavior
of IR in the same three types of constructions in English. Following Nes-
por and Scorretti (1985), we believe that the low acceptability of the
direct questions in Italian in which SR has applied cannot be attributed to
the presence of a trace but rather must be attributed to the intonation pat-
tern of these questions. Specifically, the unacceptability of SR in (58c)
is due to the fact that the word in question, dirä, contains the peak of an
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

intonation contour, and in general material in this position is resistant to

any type of reduction. This analysis is further confirmed by the facts of
English. In direct questions, such as the one in (59c), the peak of the in-
tonation contour is not typically on the word which is subject to IR, and
thus the primary stress may be retracted. Under certain circumstances,
however, it is possible for the word in question to be the peak of the con-
tour, and in such cases IR is in fact blocked, as SR is in Italian. This hap-
pens, for example, when there is contrastive stress or emphasis on the
word that would otherwise undergo Iambic Reversal, as illustrated by the
following sentences:

(60) a. Who are they going to introduce f ^ next?

(< introdiice)
b. I said, who are they going to introdiice ?wh next, not who
are they going to invite next.
(* introduce)

We can thus conclude that it is not the syntactic properties of the con-
structions under consideration, but rather their phonological properties
that are responsible for the application or nonapplication of Stress Re-
traction and Iambic Reversal.
Given that traces of wh are invisible to phonological rules that depend
on the adjacency of two words, as are the other types of phonetically
empty elements seen above, we can conclude that the weak hypothesis
about the influence of empty nodes on phonology must be rejected also,
at least if it is formulated in the general fashion found in Chomsky (1981).
While the hypotheses discussed above clearly have no general validity,
it could still be claimed that they are valid for a subset of the rules of the
phonological component. That is, it might be possible that a subcom-
ponent of the phonology is sensitive to the presence of empty elements,
while another subcomponent, presumably ordered after the first, is blind
to nonphonetic material. To show that this is not the case, it would be
necessary to demonstrate that all the rules that have been claimed to be
sensitive to empty elements are in fact only sensitive to phonological
material. These rules include to-Contraction and Auxiliary Reduction in
American English (see King, 1970; Lakoff, 1970; Zwicky, 1970; Selkirk,
1972; Kaisse, 1983, among others) and Specifier Vowel Deletion in
Italian (see Rizzi, 1979; Vanelli, 1979). We will not provide a reanalysis
of these rules here, since it would involve a syntactic discussion that would
lead us too far from the topic of the present chapter. An analysis of these
rules that does not rely on the presence of traces, however, can be found
in Nespor and Scorretti (1985), where it is shown that these rules, like all
phonological rules, are not sensitive to nonphonological material. We may
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

thus conclude that the hypotheses made about the influence of empty
categories on phonological rules must be rejected.

2.3.3. Noncorrespondence between syntactic constituents and domains

of intonation contours
The noncorrespondence between the limits of intonation contours
and major syntactic breaks is a phenomenon often mentioned in the lin-
guistic literature. The type of sentence most often cited as an illustration
of this phenomenon is one containing a restrictive relative clause. It
has been noted that complex sentences containing restrictive relative
clauses present a problem for intonation in that 'the intonation breaks are
ordinarily inserted in the wrong place' (Chomsky, 1965:13). Underlying
Chomsky's statement is the assumption that if intonation breaks were
in the 'correct' place, they would directly reflect the surface syntactic
structure of the sentence, in particular, the direction of embedding.
Compare the bracketings of the sentences in (61) and (62), where those
in (61) illustrate the syntactic constituent analysis and those in (62)
reflect the intonational structure.

(61) This is [the cat that caught [the rat that stole [the cheese]]]

(62) [This is the cat] [that caught the rat] [that stole the cheese]

According to Chomsky and Halle (1968:372), this discrepancy between

how things are and how things should be is 'obviously a matter of per-
formance rather than of grammatical structure'. In this regard the notion
of 'phonological phrase' is introduced, and rules that insert the boundaries
of such phrases 'will have to take account of syntactic structure, but
they will also involve certain parameters that relate to performance,
e.g. speed of utterance'.
While it seems clear that the presence and position of intonation
breaks are to some extent connected to such factors as the rate of speech
and the length of an utterance, this does not necessarily mean that the
occurrence of intonation breaks is not rule governed like other per-
formance phenomena such as hesitations. Instead, the specification of the
positions in which intonation breaks may occur follows from the consti-
tuent structure of prosodic trees, and is thus a matter of competence, as
will be shown in Chapter 7. 8
The flexibility of the domains over which intonation contours are
spread is another area in which it can be seen that intonation domains
cannot be directly determined by syntactic constituent structure. In
section 2.3.1, we have illustrated the flexible nature of the domains of
application of sandhi rules and have argued, on the basis of the inflex-
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

ibility of syntactic structure, that syntactic constituents cannot form the

domains of application of phonological rules applying across words. The
same flexibility can also be seen in relation to the domains over which
intonation contours extend, and thus provides further evidence that syn-
tactic constituents are not the appropriate structures for defining the
domains of intonation contours. For example, while it is possible to read
the sentence below with three separate intonation contours, as indicated
by the bracketings in (63a), a pronunciation with the intonation contours
indicated by the bracketings in (63b) is also possible. Even a pronunciation
without internal intonation breaks, that is, with one single intonation
contour as in (63c), is possible.

(63) a. [The frog] [ate a fly] [for lunch]

b. [The frog] [ate a fly for lunch]
c. [The frog ate a fly for lunch]

While it would be possible to describe the data in (63) in terms of syntac-

tic constituents by saying that an intonation break may optionally occur
after each syntactic phrase, i.e. at the end of the maximal projection of
any category X, this is not the correct generalization. For example, it
can be seen that such a principle would yield the unacceptable division
into intonation domains in (64), where each break indicated corresponds
to the end of a syntactic phrase.

(64) *[Bruce] [never] [understood], [I believe], [why Paul]

[can't see the Southern Cross] [from his home in Brittany]

In conclusion, we have provided some evidence that the analysis into

constituents offered by the syntax does not correspond to the division
into possible domains of intonation contours. This issue will be taken up
again in Chapter 7.

2.4. On motivating a phonological constituent

Up to this point, we have shown that morpho-syntactic constituents

cannot account for the domain of application of all types of phonolog-
ical rules. We will now motivate the postulation of phonological consti-
tuents that do account for such domains. The specific criteria that will
be used to motivate phonological constituents throughout this book
derive, in part, directly from the notion of constituent in general; they
are the same criteria that are used to motivate a constituent analysis
in syntax. There are other criteria, however, that are specific to phono-
logy. The first two criteria mentioned below are the more general ones
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

related to the notion of constituent. The last two, however, are specific
to phonology since they are related to the fact that phonological consti-
tuents represent an analysis of strings of sounds.
In generative grammar, the formal configuration adopted to represent
the internal structure of a sentence is the constituent structure tree.
In both phonology and syntax this tree is a representation of the hierar-
chical grouping of the elements of a string into constituents and the left-
to-right order of these constituents. A string is considered a constituent
in phonology, as in syntax, if a) there are rules of the grammar that
need to refer to it in their formulation, or b) there are rules that have
precisely that string as their domain of application. In phonological the-
ory, however, these are not the only motivations for positing a consti-
tuent (see also Nespor, 1983). It is also possible for a string to be a consti-
tuent even if the first two criteria are not met, if the string is the domain
of phonotactic restrictions. Thus, a constituent X would be motivated,
for example, in a language in which a sequence of a nasal consonant
immediately followed by a liquid is ill formed within a given domain X,
but is allowed when the two segments are in different Xs.
Finally, even in the absence of phonological phenomena that need
to refer to the domain X for one of the reasons mentioned above, X
may be posited as a constituent of the phonological hierarchy for still
another reason. That is, a phonological constituent, unlike a syntactic
constituent, can also be motivated on the basis of a fourth criterion, the
relative prominence relations among the elements of a string. Consider
the case in which, for example, a constituent X" cannot be motivated
in a given language on the basis of any of the first three criteria. It might
nevertheless be the case that in this language, stress peaks within the con-
stituent X n + 1 are neither distributed in a regularly alternating pattern
nor located in positions that are specifiable in terms of syntactic con-
stituents. The lack of periodicity in the recurrence of these stress peaks,
together with the lack of connection between the location of the peaks
and a given position within syntactic constituents, indicates that there
must be some principle responsible for the distribution of stress other than
an abstract alternation rule or a rule that makes direct reference to syn-
tactic structure. In such a case, another type of constituent that accounts
for the stress pattern would be motivated. We propose that it is precise-
ly in such situations that a phonological constituent X" can be posited
on the basis of the fourth criterion mentioned above: relative prominence
relations within a given string. It should be noted that such a constituent
has the function of delimiting the domains of stress patterns, independent-
ly of the mechanism chosen to represent the actual prominence relations
and the rules that modify them.
While prosodic and morpho-syntactic constituents are in principle
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

nonisomorphic, it should be noted that an occasional case of isomorphism

between the two does not constitute grounds for rejecting a specific pro-
sodic constituent in a given language. Replacement of a prosodic consti-
tuent by a syntactic constituent in the case of isomorphism would, in fact,
have undesirable results. First of all, since prosodic rules may refer to more
than one constituent, in the case that one of these constituents is morpho-
syntactic, two different hierarchies would be referred to by a single phono-
logical rule. Secondly, even in the case of prosodic rules that need to
mention only one category, allowing the rules of a given level to refer to
a morpho-syntactic constituent would create a heterogeneous situation.
That is, the majority of the prosodic rules would refer to one hierarchy
while the rules of a single level would refer to the other hierarchy.
To summarize, we have claimed that there are four possible types of
motivation for positing a constituent in the phonological hierarchy. It
is on the basis of such motivations that each of the constituents of the
prosodic hierarchy will be established in Chapters 3 through 8.

NOTES

1. It should be noted that Siegel actually proposes that in- and un- are both Class I
affixes. She thus fails to account for the difference in assimilation patterns, a diffi-
culty that is resolved within the same theoretical framework by Allen, however, who
assigns in- to Level I and un- to Level II.
2. See Scalise (1983) for a more detailed discussion of this rule.
3. A subsequent rule that lengthens /ts/ in intervocalic position will give the pro-
nunciation danna[t:s\ione.
4. It should be noted that we are concerned here only with a colloquial style of
speech, not with poetic language, in which deletion phenomena are more widespread.
5. Here and elsewhere in the book, primary stress is indicated with an acute accent,
even when this is in conflict with the orthographic conventions of Italian.
6. The same argument can be made on the basis of other phonological rules of
Italian, such as Gorgia Toscana, Stress Retraction, and Intervocalic Spirantization.
7. The acute accent mark on a nonfinal syllable of a word that undergoes Stress
Retraction should not be interpreted as being equivalent to a primary stress.
Thus, for example, faro (< faro) in (58a) is not necessarily phonetically identical
to the word faro 'lighthouse'. Stress Retraction changes the prominence relations
within a word by destressing the last syllable and stressing a previously unstressed
syllable. The fact that it does not necessarily create a stress as strong as word primary
stress on the newly stressed syllable is not relevant here.
8. It has also been suggested by Langendoen (1975), though for different reasons,
that the position of intonation breaks is dependent on competence rather than
performance.
Copyright © 2007. De Gruyter, Inc.. All rights reserved.

Week 6
No ratings yet
Week 6
4 pages
Phonetic Uas
No ratings yet
Phonetic Uas
2 pages
Analysis8 PDF
No ratings yet
Analysis8 PDF
14 pages
Chapter 3 Phonological Rules and Phonological Processes
100% (1)
Chapter 3 Phonological Rules and Phonological Processes
4 pages
Unit 2
No ratings yet
Unit 2
19 pages
Intro To Phonological Analysis
No ratings yet
Intro To Phonological Analysis
270 pages
Auxiliary Reduction in English
No ratings yet
Auxiliary Reduction in English
15 pages
E004 Phonall
No ratings yet
E004 Phonall
294 pages
Phonological Process
100% (1)
Phonological Process
11 pages
Linguistics
No ratings yet
Linguistics
10 pages
Phonology-Morphology Interface
No ratings yet
Phonology-Morphology Interface
60 pages
Apa Phonetics and Phonology
No ratings yet
Apa Phonetics and Phonology
7 pages
Phonology Reviewer
No ratings yet
Phonology Reviewer
3 pages
English Phonology Phological Rules Week 3
No ratings yet
English Phonology Phological Rules Week 3
3 pages
Studies in Lexical Phonology (Phonetics and Phonology, Volume 4)
No ratings yet
Studies in Lexical Phonology (Phonetics and Phonology, Volume 4)
440 pages
What Is Phonology Report
No ratings yet
What Is Phonology Report
4 pages
The Handbook of Phonological Theory
100% (1)
The Handbook of Phonological Theory
25 pages
Class 21 Allomorphy
No ratings yet
Class 21 Allomorphy
13 pages
Phonological Rules With PRACTICE
100% (2)
Phonological Rules With PRACTICE
55 pages
Understanding Phonological Rules
No ratings yet
Understanding Phonological Rules
21 pages
Generative Phonology Overview
No ratings yet
Generative Phonology Overview
28 pages
Phonological Rules
No ratings yet
Phonological Rules
41 pages
Modern English Segmental Phonology
No ratings yet
Modern English Segmental Phonology
61 pages
Roger Lass - Phonology - An Introduction To Basic Concepts (Cambridge Textbooks in Linguistics) - Cambridge University Press (1984)
No ratings yet
Roger Lass - Phonology - An Introduction To Basic Concepts (Cambridge Textbooks in Linguistics) - Cambridge University Press (1984)
306 pages
Phonological Processes in English
No ratings yet
Phonological Processes in English
240 pages
English Phonology Rules Guide
No ratings yet
English Phonology Rules Guide
3 pages
Phonological Rules and Notation
No ratings yet
Phonological Rules and Notation
10 pages
Chp.15 Modes of Application
No ratings yet
Chp.15 Modes of Application
26 pages
Phonology: Speech Sound Systems Explained
No ratings yet
Phonology: Speech Sound Systems Explained
2 pages
The Rules of Phonology
No ratings yet
The Rules of Phonology
6 pages
DocScan 03 09 2024
No ratings yet
DocScan 03 09 2024
4 pages
Phonology Phonemes
No ratings yet
Phonology Phonemes
52 pages
Introduction to Phonology Concepts
No ratings yet
Introduction to Phonology Concepts
65 pages
Table 3
No ratings yet
Table 3
86 pages
Connected Speech and Assimilation in English
No ratings yet
Connected Speech and Assimilation in English
4 pages
Processes PDF
No ratings yet
Processes PDF
24 pages
Phonetics Nesselhauf Part IV
No ratings yet
Phonetics Nesselhauf Part IV
15 pages
Prosodic Phonology (Nespor, Marina Vogel, Irene) (Z-Library) PDF
No ratings yet
Prosodic Phonology (Nespor, Marina Vogel, Irene) (Z-Library) PDF
359 pages
Unread - On Derived Domains in Sentence Phonology
No ratings yet
Unread - On Derived Domains in Sentence Phonology
18 pages
Syllables
No ratings yet
Syllables
35 pages
Hbet1203 830319125458001
No ratings yet
Hbet1203 830319125458001
13 pages
2015 Revising Phonology
No ratings yet
2015 Revising Phonology
5 pages
English Phonetics 2
No ratings yet
English Phonetics 2
8 pages
Corrected Lxcs Exam S4 SN
No ratings yet
Corrected Lxcs Exam S4 SN
5 pages
Prinsip 1&2 Summary
No ratings yet
Prinsip 1&2 Summary
3 pages
Meeting 2 Ling in ELT (2024)
No ratings yet
Meeting 2 Ling in ELT (2024)
36 pages
Roca, I. & Johnson W. (1999) - A Course in Phonology
No ratings yet
Roca, I. & Johnson W. (1999) - A Course in Phonology
380 pages
679699980-7ab - Phonology
No ratings yet
679699980-7ab - Phonology
3 pages
Phonology - The Context of Phonology - Mariana PDF
No ratings yet
Phonology - The Context of Phonology - Mariana PDF
11 pages
Phonological Processes in English and BCS
No ratings yet
Phonological Processes in English and BCS
14 pages
Nonlinear Phonology: Further
No ratings yet
Nonlinear Phonology: Further
26 pages
Phonology Simplified Notes
No ratings yet
Phonology Simplified Notes
4 pages
3) Phonological Awareness
No ratings yet
3) Phonological Awareness
7 pages
General Linguistics Group 5 - Phonology
No ratings yet
General Linguistics Group 5 - Phonology
7 pages
Chapter 2 Producing Speech Sounds: 2.1 How Humans Produce Speech
No ratings yet
Chapter 2 Producing Speech Sounds: 2.1 How Humans Produce Speech
20 pages
Understanding Phonological Rules
No ratings yet
Understanding Phonological Rules
6 pages
English Phonetics & Phonology Guide
No ratings yet
English Phonetics & Phonology Guide
20 pages
Summary of English Phonolog1
No ratings yet
Summary of English Phonolog1
2 pages
Test Unit 789 TA5
No ratings yet
Test Unit 789 TA5
3 pages
GRE Vocabulary 2000 Words
No ratings yet
GRE Vocabulary 2000 Words
279 pages
A Guide To Using Quantifiers in English
No ratings yet
A Guide To Using Quantifiers in English
11 pages
U1 - U7 Final Exam
No ratings yet
U1 - U7 Final Exam
9 pages
English Grammar Test With MCQ (1) Beginner
No ratings yet
English Grammar Test With MCQ (1) Beginner
5 pages
Contoh dan Latihan Adverb of Manner
No ratings yet
Contoh dan Latihan Adverb of Manner
11 pages
E0 LSK Reading 3 PDF
No ratings yet
E0 LSK Reading 3 PDF
17 pages
Contoh Silabus English For SMA
No ratings yet
Contoh Silabus English For SMA
7 pages
Englishhawaiian Dictionary
100% (1)
Englishhawaiian Dictionary
268 pages
Vocabulario Básico de Ingles 3
No ratings yet
Vocabulario Básico de Ingles 3
1 page
Infinitive Past Simple Past Participle Traducción
No ratings yet
Infinitive Past Simple Past Participle Traducción
6 pages
ThinkAhead2 StandardsAssessmentCharts-Murcia 22599
No ratings yet
ThinkAhead2 StandardsAssessmentCharts-Murcia 22599
7 pages
Understanding Question Words in English
100% (1)
Understanding Question Words in English
6 pages
Talk To Me in Korean - Level 4
No ratings yet
Talk To Me in Korean - Level 4
5 pages
Old French Lessons
No ratings yet
Old French Lessons
210 pages
Class VI Syllabus Overview
No ratings yet
Class VI Syllabus Overview
3 pages
Lexical Semantics CH 4 - GrB4
No ratings yet
Lexical Semantics CH 4 - GrB4
20 pages
(Ebook PDF) Business English 12th Edition by Mary Ellen Guffey PDF Download
100% (2)
(Ebook PDF) Business English 12th Edition by Mary Ellen Guffey PDF Download
49 pages
Pre Posicion Esy Prepositions
No ratings yet
Pre Posicion Esy Prepositions
97 pages
Word Formation in FCE and CAE
No ratings yet
Word Formation in FCE and CAE
4 pages
Turkish Grammar
No ratings yet
Turkish Grammar
45 pages
Start Where You Are. Use What You Have. Do What You Can.: Phrase
No ratings yet
Start Where You Are. Use What You Have. Do What You Can.: Phrase
34 pages
English Grammar Practice Guide
No ratings yet
English Grammar Practice Guide
1 page
9th Grade English Test Paper
No ratings yet
9th Grade English Test Paper
1 page
Focus4 2E Grammar Quiz Unit7.2 GroupA 1kol
No ratings yet
Focus4 2E Grammar Quiz Unit7.2 GroupA 1kol
2 pages
Understanding Personal Pronouns
No ratings yet
Understanding Personal Pronouns
19 pages
Barbara K. Gaines - Idiomatic American English, A Step-By-Step Workbook For Learning Everyday American Expressions (1986)
80% (5)
Barbara K. Gaines - Idiomatic American English, A Step-By-Step Workbook For Learning Everyday American Expressions (1986)
123 pages
German Language Learning This Book Includes Learn German For Beginner PDF
100% (2)
German Language Learning This Book Includes Learn German For Beginner PDF
534 pages
Use of Adverb
No ratings yet
Use of Adverb
6 pages
Prepare 5 Tests Book-61
No ratings yet
Prepare 5 Tests Book-61
1 page

Prosodic Phonology With A New Foreword - (PG 59 - 92)

Uploaded by

Prosodic Phonology With A New Foreword - (PG 59 - 92)

Uploaded by

Chapter 2

Motivation for Prosodic Constituents

Among the processes that involve a modification of the sound pattern

2.1. Phonological processes in nonphonological contexts

prosodic rules, will be discussed in subsequent chapters in relation to the

2.1.1. Morphological contexts

(1) a. in + legal illegal

In more recent proposals, such as the Level Ordering Hypothesis (see

(2) only -*• *olly

(3) sudden- ly -*· *suddelly

fama - oso famoso 'famous'

(5) a. [record vs. [record ]y

The second type of rule can be exemplified by the rule of ζ-Devoicing

(6) a. abu[z]e + ive abu[s]ive

Another rule that applies only in the environment of a specific morpheme

(7) corre[t:]o - ione -+ corre[t:s]ione 'correction'

carre[t:]o - iere *carre[t:s]iere 'wagoner'

2.1.2. Syntactic contexts

the domain of application of French Liaison; the relevance of left branches

b. *Ho le suol[0] nuove. (< suole)

It should be noted that the different syntactic relation between the

(10) a. parlan[0] bene (<parlano)

(11) a. wetää'nü (V' = vowel with a down-stepped tone)

From these and similar examples described in various works (Kenstowicz

2.2. Failure of morphologically specified contexts of prosodic phenomena

(12) a. guda:ga 'dog'

In derived words, whether or not the sequence root + suffix forms a

If, however, a bisyllabic suffix is affixed to a root with an odd number of

(15) gumari-daga-jiu -> gumä:ridagä:jiu 'to have become red'

While the lengthening in the root can easily be expressed in terms of

(16) συν+πλέκω '(I) knit'

(17) a. τον#πατέρα 'the father (acc.)'

(18) έχουν#πλέξει '(they) have knitted'

2.3. Failure of syntactically specified contexts of prosodic phenomena

In considering phonological rules operating across words, a problem arises

2.3.1. Noncorrespondence between syntactic constituents and domains of

the largest possible domain of application of a phonological rule is the

(19) a. La scimmia aveva appena mangiato metä [b: ]anana.

(20) a. II ragno aveva mangiato metä [f:]arfalla.

These conditions on the two individual words in a given sequence, how-

(21) a. La volpe ne aveva mangiato metä [p]rima di addormentarsi.

b. II gatto aveva catturato un colibri [m]olto pregiato.

Although the syntactic structures of (22a-c) differ somewhat, in each

While in each of these four sentences the segmental environment for RS

(24) a. Hovistotre [k:]olibri [b:]rutti.

In (24a), RS applies in both of the environments in which the phonolog-

While in (25a) the repeated application of RS seems to indicate that the

of RS is a syntactic constituent, it would be necessary to claim, further-

(26) a. Les giraffes et les elephants sont ses meilleurs^amis.

In syntactic terms, meilleurs 'best' is the complement of amis 'friends',

(27) a. lu[0]o ( < lu[p]o) 'wolf

(29) a. aveva [hjonosciuto Arcibaldo (<[k]onosciuto)

c. si lamenta [hjostantemente ( < [kjostantemente)

(30) a. Le zanne dell'elefante bianco dell'Africa Orientale [k]ostano

If the domain of application of GT is to be expressed in terms of syntactic

(31) a. Osservano il rarissimo colibri peruviano con le penne azzure

con un cannocchiale particolarmente adatto alia situazione.

'(They) observe the very rare Peruvian hummingbird with blue

coca cola con i soldi che hanno guadagnato vendendo giornalini.

'(They) bought a lot of candies, licorice, and cans of coca cola

(32) a. Gli struzzi [hjorrono velocemente. ( < [kjorrono)

It is obvious from these examples that the length of a given string is

(33) a. ga[m]ba 'shrimp'

(34) a. tiene[rj] cuatro gatos ( < tiene[n])

'(one) doesn't see that species of scorpion very frequently in

(35) a. Colecciona en todo el mundo plumas de tucä[n] para su som-

(36) a. El nuevo canario de mi amiga Carmefn] canta solo cuando estä

(37) a. Mi faisä[rj] corre siempre. (<faisa[n])

of certain phonological rules is sensitive to the length of the strings in-

Beyond the sentence

(38) a. water wa[r]er

Flapping, however, is not restricted to applying within a sentence; it

(39) a. Please wait. I'll be right back. ...wai[>]Fll...

(40) a. They didn't wait. I'll be right back. ->-*.. .wai|>]I'll...

b. Where's Scott? Open the window. -»· *.. .Sco[.r JOpen...

Since the largest constituent in syntax is the sentence, there is no way

(2) only -• olly

(3) sudden- ly -· suddelly