Italian s-Voicing and the Structure of the Phonological Word

Marc van Oostendorp

Appeared in S.J. Hannahs and Mike Davenport, Issues in Phonological Structure. Benjamins, 1999, 197-214;

1. Introduction

Certain phonological processes may be limited to a domain which on the one hand is bigger than the Foot, while on the other hand it is not precisely congruent with any morphological or syntactic category (e.g. Nespor and Vogel 1986, Selkirk 1986, Inkelas 1989). This category is usually called the Phonological or Prosodic Word; I will adopt the former term here. In earlier stages of generative phonology, this Phonological Word was derived by rule from morphological structure. More recently, some authors have suggested that prosodic structure and morphological structure are checked in parallel (Sadock 1990, Mohanan 1993, McCarthy and Prince 1993, 1995). In both approaches, it is an important question the question exactly which factors determine the relation between morphological and prosodic structure.

One example of this is the asymmetry between prefixes and suffixes with respect to resyllabification which is attested in many languages, including Italian, German and Indonesian. In all of these languages, prefixes behave more as independent prosodic units than (most) suffixes. When we are confronted with this observation, we can do one of two things.

The first option is to stipulate this asymmetry for every single language. This is the option chosen in many studies hitherto. We might assume for instance that prefixes automatically get a separate prosodic word while suffixes are incorporated into the prosodic word formed by the stem. Generalised Alignment theory (McCarthy and Prince 1993) provides an example of such a theory, in which it is stipulated that the left boundaries of words are stronger than the right boundaries. I will briefly discuss this point in section 2.

In this paper, I present part of the results of a project that is aimed at giving an alternative to this stipulation (cf. Van Oostendorp 1994, 1997 on Dutch and German). My goal is to derive the precise structure of the Phonological Word from independently motivated principles. One such principle is that in the unmarked case stem and affixes are not in the same Phonological Word. In other words, every morpheme forms its own prosodic domain:

PrWd PrWd PrWd PrWd PrWd
| | | | |
prefix prefix stem suffix suffix

Different structure can only be assigned to a combination of a stem plus an affix if there are sufficiently strong reasons to do so. One such reason may be a minimality requirement to the effect that Phonological Words have to contain at least two syllables or moras. Since affixes often are monosyllabic, assigning a separate Phonological Word to them would yield a violation of this minimality requirement.

In this article, I will concentrate on another prosodic well-formedness condition that forces affixes to be parsed into a Phonological Word together with the stem. This force is syllabification. I assume that syllables never cross the boundaries of Phonological Words, at least in the languages under discussion. In order to get a good syllabification, we therefore sometimes put affixes and stems together in a Phonological Word. I argue that this assumption explains certain asymmetries between prefixes and affixes, and among certain classes of prefixes, in Italian as well as in the Bantu language Kíhehe. Both languages display similar prefix-suffix asymmetries although the phenomena involved in the two languages are quite different.

This article is structured in the following way. In section 2 I discuss the phenomenon of s-voicing and I show how it interacts with prosodic word formation. In section 3 I show that Kíhehe prosodic word formation interacts in exactly the same way with reduplication. Section 4, finally, is devoted to a conclusion.

2. S-Voicing

2.1 Nespor and Vogel's (1986) definition of the word

Certain northern dialects of Italian have a process of intervocalic s-voicing. It applies in the contexts of (2a) but not in those of (2b) (These examples have been copied from Vogel 1993):

  1. a.
    i[z]ola 'island' (morpheme internal)
    cau[z]ava 'he/she was causing' (before inflection)
    famo[z]öissimo 'very famous' (before superlative)
    di[z]öuguale 'unequal' (end of prefix)


    riö[s]alare 'resalt' (after prefix)
    portaö[s]igaretta 'cigarette case' (in compound)
    lavandoö[s]i 'washing oneself' (before enclitic)
    la [s]era 'the evening' (after proclitic)
    bella [s]era 'beautiful evening' (in phrase)

In what way do the examples in (2a) form a natural class vis à vis the examples in (2b)? Some authors have proposed that the context for s-voicing is best analysed in prosodic terms. The relevant domain for sövoicing is then the Phonological Word. The rule could then be formulated in the following way:

(3)s -> z / Vi ___ Vj
condition: Vi and Vj are in the same Phonological Word

Intervocalic voicing of fricatives is not uncommon in languages of the world (Laver 1994) and probably a more principled explanation of the phenomenon could be given. Yet (3) is the most explicit statement of what happens in s-voicing at the segmental level that I need for my purposes. The problem I am interested here is to define the notion of a Phonological Word for Italian so that we can account for the environments in which (3) does and does not apply.

Nespor and Vogel (1986) argue that the Phonological Word in Italian has the following definition:

  1. Italian Phonological Word (Nespor and Vogel 1986)
    The Phonological Word consists of
    a.a stem and all suffixes,
    b.a prefix ending in a vowel,
    c.a prefix ending in a consonant and the following stem or prefix.

This definition works as a structure-building rule, deriving prosodic structure from morphological structure. It has as an advantage that it makes exactly the correct predictions when used in combination with the rule in (3). It assigns the proper prosodic structure to each of the forms in (2), so that application of (3) will give the correct results for each of these forms. I take the following forms to be representative:

/ | \
s s s
| | |
i so la

/ | \
s s s
| | |
cau sa va

/ | \ \
s s s s
| | | |
di su gua le

| / | \
s s s s
| | | |
ri sa la re

Application of the rule in (3) in these contexts assigns the appropriate structure to each of these forms.

While empirically (4) thus assigns the correct structure, conceptually it does not seem the most attractive solution. It has to stipulate no less than two asymmetries: an asymmetry between prefixes and suffixes on the one hand and within the class of prefixes an asymmetry between vowel-final and consonantöfinal morphemes on the other. It does not seem very plausible that (4) would have a primitive status in the theory. It would be desirable to derive the effects of this definition from more basic principles. This is what I will try to do in this paper.

At the outset, I want to remark on the representation I gave in (5d). Recently some authors have argued that the soöcalled Strict Layer Hypothesis, stating that prosodic categories of level n should have daughters of level n-1 and mothers of level n+1 which was adhered to by Nespor and Vogel should be weakened. Affixes (or clitics) which are not incorporated into the Phonological Words nevertheless do not form Phonological Words of their own, but are 'subminimal' sisters to other Phonological Words, and for instance direct daughters of prosodic constituents such as the Clitic Group or the Phonological Phrase (Vogel 1993). One important argument comes from the prosodic minimality requirement mentioned above: clitics are often monosyllabic, even in thos languages in which lexical words are not allowed to be smaller than bisyllabic. My own proposal is strictly speaking neutral concerning the question whether prefixes form their own Phonological Word or not. For the sake of concreteness I adopt (5d) rather than the structure in (6):

/ \
| PWd
| / | \
s s s s
| | | |
ri sa la re

With a few minor changes, the theory presented here could also be made to work for the structure in (6) but I will not try to do that here. I now return to the two asymmetries in the definition in (4).

2.2 Asymmetries among prefixes

The most curious one is that in the class of prefixes. Why would a prefix ending in a consonant like dis- be incorporated into the prosodic word of the base, while the same is not true for a prefix ending in a vowel? How does the phonological (segmental) shape of the affix affect its prosodic behaviour?

The only really strong evidence for this division is exactly sövoicing. This means we only have good evidence that dis- is incorporated into the stem PWd when the final -s is actually voiced. But this in turn can only happen if the stem starts with a vowel, because sövoicing applies only intervocalically. In a word like discontento we simply do not have a test to decide what the prosodic structure is.

It seems that we get the same results for sövoicing if we assume that not the phonological shape of the affix but rather the shape of the stem is the deciding factor in building the Phonological Word. If a stem starts with a vowel, it tends to incorporate a preceding prefix. If it starts with a consonant, it does not incorporate that prefix. This assumption would give us the following structures:

(7)segmentsprosodic structure


b.V][Vstra+ordinario (??)

c.C][C(dis) (contento)

d.V][C(ri) (salare)

The structures in (7a) and (7d) are the same as those assigned by Nespor and Vogel (1986). (7b) and (7c) are different. For (7c) I already argued it is hard to find evidence which proposal is right since the only strong test for prosodic wordhood of affixes in Italian is s-voicing and this rule by definition does not apply (but see 2.5 for some discussion). I will return to (7b) in section 2.4 below.

If we look at the stems rather than the prefixes, it is easier to understand why the division is made the way it is in the Italian dialects under discussion. Normally, we want to have a prosodic word boundary between the prefix and the stem. If there are no other forces at work, we ideally get representations like (7c), that is to say representations in which each morpheme has its own prosodic node (cf. (1)).

Yet vowelöinitial stems start with an onsetless syllable, a type of syllable that is undesirable. Suppose that it is so undesirable that it can force the prefixes to be incorporated into the prosodic word of the stem. This would give us the representation of (7a) for consonant-final prefixes and vowel-initial stems.

We can easily convert these informal statements into an analysis in terms of Optimality Theory. Suppose then that we have the constraints listed in (8), ordered in the way stated in (9) (if we would use the representations of Vogel 1993 rather than those of Nespor and Vogel (1986), the definitions of Pr=Lx and Lex=Pr would have to be changed accordingly). This would give us the tableaux in (10):

(8)Onset:Every syllable should have an onset.
Pr=Lx:Every Phonological Word boundary should correspond to a morpheme boundary.
Lx=Pr:Every morpheme boundary should correspond to a Phonological Word boundary.

(9)Onset >> Pr=Lx >> Lx=Pr


(ris)( alare)

There is one constraint on syllable structure, the well-known Onset constraint and there are two constraints on the interface between phonology and morphology. The most important aspect of this representation is that in Italian requirements on syllable structure have more weight than requirements on respecting morphological and prosodic boundaries. The Italian prosody tries to build the best syllables we can, even at the cost of disrespect for syntactic or morphological boundaries.

Constraints are freely rankable in Optimality Theory. We therefore do not expect to find the same ranking in all natural languages. In German for instance a word like ent+ehren, Îto disgraceâ, with a consonant-final prefix and a vowel-initial stem gets syllabified as rather than as, which would be the ÎItalianâ syllabification (for the sake of simplicity I have put the constraints Pr=Lex and Lex=Pr together under the heading 'Interface' in this tableau):



In German, the interface conditions play an important role in the determination of syllabification. In Italian, it is necessary to take into consideration the prosody-morphology interface only if syllable structure does not play a role, as in the case of risalare, which consists of four perfect core syllables under any (reasonable) division into Phonological Words.

It therefore does not seem necessary to stipulate the difference between two types of prefixes as in the definition of Nespor and Vogel (1986). The relevant facts of s voicing follow from the way interface constraints and syllable structure interact.

2.3 Asymmetries between prefixes and suffixes

I now turn to the second asymmetry stipulated by Nespor and Vogel (1986), the distinction between prefixes and suffixes. Many languages have this type of asymmetry and if they have it, they seem to have it in the same way as Italian: prefixes are more independent from the stem than suffixes (cf. Cohn 1989, McCarthy and Cohn 1994 on Indonesian, Booij 1977, 1981 on Dutch, Szpyra 1992 on Polish, Levergood 1984 on Maasai, etc.). This stipulation for Italian might therefore look somewhat more acceptable than the one within the class of prefixes. Yet it remains a stipulation and a theory that could derive the asymmetry from some other well-established fact seems certainly preferable.

For instance, McCarthy and Prince (1993), developing a theory of ÎGeneralised Alignmentâ, have proposed that we should split the constraint called Lx=Pr in this paper into two separate constraints, one dealing with the left edge and one with the right edge of prosodic words. These constraints are formulated in (12). Their relative ranking is given in (13):

(12)Align-Left: Every left edge of a stem should correspond to a prosodic category (a prosodic word).
Align-Right: Every right edge of a stem should correspond to a prosodiccategory (a syllable).

(13)Align-Left >> Align-Right

The ranking established in (13) seems to have a universal status. McCarthy and Prince (1993) mention only a few possible examples of languages in which the constraint ordering may be reversed, and these can perhaps be analysed in a different way.

It is also not clear whether we need to distinguish between left and right edges of words in this way. This is particularly true for Italian, where nearly all suffixes start with a vowel. Let us take causava as an example. It consists of two morphemes, the root and the suffix (I will return to the status of the stem-final vowel below):

(14)caus<a> +ava

The suffix starts with a vowel. This vowel needs to be the head of a syllable and this syllable in turn needs to have an onset. It can get this onset by incorporating the final consonant of the stem, which means that suffix and stem should be one prosodic domain:



Since almost all Italian suffixes start with a vowel, i.e. with a potentially onsetless syllable, we can explain why they behave this way: the Onset requirement is relevant for all of them. The tableau for the stem+suffix combination in (15) is not essentially different from the tableau for the prefix+stem combination in (10).

A few suffixes do start with a consonant. Examples of these are:

(16)-mente, -mento , -bile.

Since they do not have an onsetless syllable, we expect these suffixes to form a Phonological Word of their own. As far as I know, there are no suffixes starting with /s/. We therefore cannot use s-voicing as a test for these forms.

In the case of these other suffixes, there is one other test for Phonological Wordhood however. This is stress-assignment, at least if we make the assumption that the domain of stress is the Phonological Word. Italian has a familiar type of three-syllable window at the right edges of words for stress.

Yet the results of the stress test are not very clear. If we take -mente as an example, on the one hand -mente and the adjectival stem to which it is attached seem to behave as two independent domains for (secondary) stress assignment (Peperkamp 1994). On the other hand, Nespor and Vogel (1986) have already pointed out that there is only one primary stress in adverbs that are derived by -mente. This suggests that these adverbs have a prosodic structure that is different from compounds that do behave as if there is more than one 'primary' stress.

Maybe these two results taken together merely indicates that an approach along the lines of Vogel (1993) is correct: the suffix does not form a Phonological Word of its own, but it forms a Foot which is incorporated into the base.

Interestingly there is a clitic that starts with /s/, si, used among others as a reflexive clitic (Monachesi 1996). There is no voicing of the clitic-initial /s/ in the Italian dialects under discussion here.

(17)lavando-[s]i (*[z]i)
Îwashing himselfâ

Nespor and Vogel (1986) took this as evidence for the existence of a separate prosodic category, the Clitic Group. The clitic behaved not as a suffix, but in a somewhat different way. It therefore was taken to be also in a different domain than a suffix. A somewhat diverging analysis is possible under the assumptions laid out above. Suppose that clitics have the same prosodic and phonological status as affixes, phrasal affixes if one likes. Because the clitic starts with a consonant, syllable structure does not force it to incorporate into the base. Therefore, the constraints on the morphology-prosody interface will come into play. They will assign independent status to both the base and the clitic.

The constraint tableau will look as follows:



The advantage of the proposal made here is thus that it can derive all aspects of Nespor and Vogelâs (1986) definition --- the asymmetry between prefixes and suffixes, the asymmetry between different classes of prefixes and the special behaviour of clitics --from one single stipulation: the stipulation that the Onset constraint in Italian outweighs considerations of respecting morphological boundaries.

2.4 Vowel deletion and prefixes ending in a vowel

Before we can conclude on this proposal, however, we have to study the configuration in which a vowel-final prefix is followed by a vowel-initial stem. I have listed some examples of this in (19):


What we expect is that in these cases, prefix and stem form independent domains, since the prefix cannot help to provide an onset to the stem. The Onset constraint is therefore irrelevant and we expect these forms to get the structures in (20):

(20)(stra) (ordinario)

(pre) (avviso)

There is no way to resolve the hiatus, and therefore we separate stem and affix. However, some scholars might argue that in the case of nouns and adjectives, stem+suffix combinations underlyingly have exactly the same type of hiatus. But in this case the hiatus is resolved by deletion of the first of the two vowels:

(21)fama+oso > famoso

This leaving unparsed of the /a/ might be seen as another result of the wish for syllables to be provided with an onset. The first syllable of the suffix is so badly in need of an onset that it is even allowed to skip the final vowel of the stem. If this would be true, we would have a real asymmetry between prefixes and suffixes, because the same reasoning did not apply in (19).

As a first observation, I might point out that the final vowel is also deleted in cases such as (22):

(22)eroe+ina > eroina

However in this case it is not very clear how deletion of the vowel could help us make better the syllable structure of the suffix. It is therefore not sure that the constraint Onset is involved in the analysis of these facts at all.

The analysis is further complicated by the fact that there are some words that end in a vowel that cannot be deleted. This vowel is invariably stressed, as pointed out by Scalise (1983):

(23)blu + astro > bluastro

caffé + uccio > caffeuccio

Scalise (1983, 1994) proposed to analyse this by the following rule:

  1. V[-stress] -> 0 / __ +V

It is necessary to mention the boundary symbol '+' in this rule, because otherwise the rule would also be incorrectly applied in cases such as creatura 'creature' [kreatura], *[kratura] and beato 'benified' [beato], *[beato]. Furthermore, Scalise has pointed out that the rule also has to be restricted in some other way in order to avoid overapplication in cases such as (22).

This somewhat suspicious restriction has led Peperkamp (1994) to assume the final vowel in the case of most adjectives and nouns is an inflectional element. Some nouns and adjectives exceptionally have no (overt) inflectional element. These are exemplified by the words blu and caffé. It is important to note that these forms have no overt inflectional element. The fact that these forms have final stress might indicate that these forms end in some kind of covert empty segment, a catalectic element as proposed by Kiparsky (1991) and Kager (1993).

Affixation might then be analysed as affixation to a root, without the inflectional element. So in (21), -oso might be affixed to fam- in stead of to fama. No vowel deletion rule is needed at all. We could now assume that the final vowel of monosyllabic prefixes is not an inflectional element. Since there is no vowel deletion, also this vowel will not be deleted. In bisyllabic prefixes, the issue is notoriously more complicated. Scalise (1994) gives the following table:

  1. Example Vowel deletion
    extra+acido no
    ultra+ubbidiente No
    anti+estetico No
    semi+analfabeta No
    arci+energico No
    vice+usciere No
    sovra+affolato Yes
    oltre+alpe Yes
    sotto+esposto Facultative
    contro+offensiva Facultative

It might be that the final vowels of the prefixes sovra and oltre, and optionally those of sotto and contro can be analysed as theme vowels. The vowels in stra or extra on the other hand are 'fixed'.

2.5 Nasal assimilation as evidence for the Phonological Word

Apart from intervocalic s-voicing, Nespor and Vogel (1986:130-134) provide some extra evidence for their definition of the Phonological Word. Most of these facts are irrelevant to the present discussion, because the analysis presented here provides us with the same prosodic structure as Nespor and Vogel's account, so that the predictions of the two analyses are not different.

One case is interesting however, because it involves the underlying forms such as in+regolare. For Nespor and Vogel, these words form one Phonological Word. As evidence for this, they cite a rule of Total Nasal Assimilation, which supposedly also has this Word as its domain:

  1. in+raggiungibile -> irragiungibile 'unreachable'
    in+legale -> illegale 'illegal'
    in+morale -> immorale 'immoral'
    con+rispondere -> corrispondere 'to correspond'
    con+legare -> collegare 'to put together'

The constraint ranking considered until now would put in (or con) and the following stem in two separate words:

  1. /in+legale /

However, it will be clear that we will need a new constraint in order to get the assimilation pattern in the first place. For the sake of simplicity might assume for instance that this constraint has the following form:

  1. Assimilate: /n/ should be assimilated to a following consonant in the surface

Assuming that this constraint gets the same ranking as Onset, we again obtain the correct result:

  1. /in+legale /

One of the advantages of Optimality Theory is that we can form rather subtle prosodic structures on the basis of simple constraints. We can assume that dis+contento gets the structure (dis)(contento) and in+legale the structure (, because in the latter nasal assimilation requires the two morphemes to be in the same Phonological Word, whereas no such requirement is operative in the former word.

3. Kíhehe reduplication

3.1 Basic facts

One important difference between Nespor and Vogelâs (1986) definition and the one presented here is that in the former theory it was the major class of the final segment of the prefix that was relevant while in the approach I proposed it was the first segment of the stem. It would be interesting to find a language in which it is undeniably either the stem or the prefix that determines the prosodic shape of the derived form. In this section I provide an example of such a language. This language is Kíhehe, a language from Tanzania, which is discussed in Odden and Odden (1985). This language has some syllable-related phonological processes that make it remarkably similar to Northern Italian.

I have to remark at the outset that I can only discuss facts having to do with syllabification. Odden and Odden (1985) point to another interesting factor that can draw prefixes into the prosodic word of the stem. There is a prefix which consists just of the consonant /ñ/. This consonant fuses with the first consonant of the stem (if any). The resulting structure as a whole counts as the initial segment of the Phonological Word. This shows that in Kíhehe not only syllable factors, as in Italian, but also segmental factors can overrule considerations of alignment. However, as I said I will not go into this small complication (cf. Rowicka 1994 for discussion of a very similar process in Polish).

Odden and Odden (1985) present some interesting facts of reduplication in Kíhehe. Reduplicational morphology only affects a stem, i.e. a root plus its suffixes whilst prefixes remain outside of the stem:

  1. kú-haáta
    Îto fermentâ
    Îto start fermentingâ
    Îto be happyâ
    Îhe's sort of happyâ
    Îto buyâ
    Îwe shopped a bitâ
    Îhe's cooking it (7) casuallyâ
      kú-fi-gulagúla (*kú-fi-gula-fi-gúla)
    Îto buy bit of them (8)â
      kú-tovatóva (*kú-tova-ku-tóva)
    Îto beat a bitâ

If we adapt Odden and Oddenâs terminology to the one used here, the base of reduplication seems to be the Phonological Word and prefixes are outside of the Phonological Word while suffixes are inside:

  1. PWd

    tu(gul it e)

All stems and suffixes (except for the very last one, an obligatory suffix which is called final vowel and which seems irrelevant for our purposes) end in a consonant while all suffixes start with a vowel. In other words, the constraint Onset could again be held responsible for the fact that the suffixes are forced into the Phonological Word, just as in Italian. Also in Kíhehe, we might conjecture that syllable structure constraints outrank the constraints on the prosody-morphology interface:

  1. /ku+tes+a/

3.2 Vowel-initial stems

What happens with vowel-initial stems? In Italian, they were the only ones that allowed (consonant-final) prefixes to be incorporated into the stem. Kíhehe does not seem to have consonant-final prefixes. Yet some (many) prefixes end in a high vowel. This high vowel can turn into a glide before another vowel, in order to satisfy Onset (see Rosenthall 1994 for discussion of Bantu high vowel gliding in Optimality Theory):

  1. kú-teléka

    Îto cookâ


    Îto spillâ


    Îheavy (1,3)â


    Îfast (1,3)â


    Îheavy (8)â


    Îfast (8)â


    Îto buy it (11)â


    Îto pour it (11)â

If my analysis of Italian is correct, it should apply to Kíhehe as well. Because in (28) the constraint Onset is clearly at work, we expect prefixes to form a prosodic unit with their base in this case as well as suffixes normally do. kwíita should get the follwoing prosodic structure:

  1. PWd

    / \

    s s

    kwíi ta

The reason for this is that Onset has more weight than any of the interface constraints has:

  1. /ku+ íit+a/

Reduplication applies to Phonological Words, so we predict that prefixes are in the reduplicant if and only if the stem starts with a vowel. This prediction appears to be correct:íita+kw+iíta

Îto pour a bitâ


Îto pour it (11) a bitâímba+kw+iímba

Îto sing a bitâ


Îto sing a bit to himâ

The examples in (31b) and (31d) show that the incorporation of one prefix into the prosodic word does not imply incorporation of all prefixes. This seems to me to be a straightforward case of minimal violation. Incorporation of affixes, violating the interface constraints, is only allowed if it helps to satisfy Onset. This is tested for every prefix separately.

Before we draw a conclusion from these observations, I have to discuss one complication for the sake of completeness. This complication arises because there is one prefix which is both vowel-initial and vowel-final, /i/. Kíhehe can thus form sequences like /kú-i-eénda/. At the surface, this would yield two violations of Onset. The simplest way to avoid these violations probably would be to glide the i, so that we would have *[kú.(jeénda)]. This however is not the actual output, which is [kwíijeénda]. Odden and Odden (1985) claim that high vowels are lengthened when appearing after another high vowel and furthermore that glide formation does not apply to long vowels. The /j/ is the result of a different insertion rule, which belongs only to the stem, not to the prefix. Evidence for this is that only the glide is copied in the reduplication process: kwíijeendajeénda.

Quite important for our purposes is that it in the Kíhehe example it cannot possibly be the segmental make-up of the prefix that is the determining factor in the interface. The prefix is the same in the forms to cook and to spill, yet the prosodic structures assigned to the prefix+stem combination are different. The only important factor can be the make-up of the stem.

4. Conclusion

In this article I have argued that affixes and stems normally do not form one Phonological Word together; rather, it is preferred to assign separate prosodic domains to every morpheme. Yet there might be other factors at work which force affixes to enter the prosodic domain of the base. Here we have seen one such Îforceâ: the constraint Onset. When a suffix or a stem starts with a vowel, it can be convenient to parse the consonant (or high vowel) at the end of the preceding stem or prefix into the onset, thus creating a CV(C) syllable.

This type of constraint interaction can be nicely described in a theory such as Optimality Theory, which is specifically designed to describe the way conflicts between demands on output structure are solved. The complicated definitions that were used in earlier approaches can be replaced by a limited set of independently motivated interacting constraints.


Aronoff, M. and S.N. Sridhar (1983) 'Morphological levels in English and Kannada or Atarizing Reagan,' Papers from the parasession on the interplay of phonology, morphology and syntax. Chicago Linguistic Society 19: 3-16.

Cohn, A. (1989) ÎStress in Indonesian and Bracketing Paradoxes', Natural language and linguistic theory 7, 167-216.

Cohn, A. and J. McCarthy (1994) 'Foot Alignment and Apparent Cyclicity in Indonesian', ms. Cornell Universty and University of Massachussetts.

Hannahs, S.J. and E. Stotko (forthcoming) 'The Non-Isomorphism of Phonological and Morphological Structure: Evidence from Acquisition', in: M. Young-Scholten and S.J. Hannahs (eds.) Current Issues in the Acquisition of Phonology, Amsterdam: John Benjamins.

Inkelas, S. (1989) Phonological Constituency in the Lexicon,â PhD Dissertation, Stanford University.

Kager, R. (1995) 'Consequences of Catalexis',

Kiparsky, P. (1991) Catalexis, manuscript, Stanford University.

Laver, J. (1994) Principles of Phonetics, Cambridge, Cambridge University Press.

Levergood, B. (1984)\Rule-Governed Vowel Harmony and the Strict Cycle,\ in Proceedings of NELS 14,275-293.

McCarthy, J. and A. Prince (1993) ÎGeneralized Alignment,â in G. Booij and J. van Marle, eds. Morphological Yearbook, Kluwer.

McCarthy, J. and A. Prince (1995) 'Faithfulness and Reduplicative Identity', in J. Beckman, L. Dickey and S. Urbanczyk (eds.) Papers in Optimality Theory (University of Massachusetts Occasional Papers 18). Amherst: GLSA. Pp. 249-384.

Mohanan, K.P. (1993) \Fields of Attraction in Phonology,\ in J. Goldsmith, ed. The Last Phonological Rule, Chicago, Chicago University Press.

Monachesi, P., (1996) A Grammar of Italian Clitics, Doctoral Dissertation, Tilburg University.

Nespor, M. and I. Vogel (1986) Phonological Phonology, Foris, Dordrecht.

Odden, D. and M. Odden (1985) ÎOrdered reduplication in Kíhehe'. Linguistic Inquiry, 497-503.

Oostendorp, M. van (1994) ÎAffixation and syllable integrity in Dutch', in R. Bok-Bennema and C. Cremers, eds, Linguistics in the Netherlands 1994, John Benjamins, Amsterdam.

Oostendorp, M. van (1997) 'On an asymmetry between prefixes and suffixes', manuscript University of Leiden.

Peperkamp, S. (1994)\Derivational Affixation in Italian,\ ms., Universiteit van Amsterdam.

Rosenthall, S. (1994) Vowel/Glide Interaction in a Theory of Constraint Interaction, PhD Dissertation, University of Massachussetts at Amherst.

Rowicka, G. (1994) \Polish Palatal Assimilation in Prefixed Words,â in R. Bok-Bennema and C. Cremers, eds. Linguistics in the Netherlands 1994, Amsterdam, John Benjamins.

Sadock, J. (1990) Autolexical Syntax, Chicago, University of Chicago Press.

Scalise, S. (1983) Morfologia Lessicale, Padova, Clesp.

Scalise, S. (1994) Morfologia, [****]

Selkirk, L. (1986) 'On Derived Domains in Sentence Phonology', Phonology Yearbook 3,


Szpyra, J. (1992)`The Phonology of Polish Prefixation', In: J. Fisiak and S. Puppel (eds.) Phonological investigations, John Benjamins, Amsterdam.

Vogel, I. (1993) \Phonological interfaces in Italian', in M. Mazzola, ed., Proceedings of Linguistic Studies on Romance Languages XXIII.