The principal objective of this article is to provide a characterization of the Standard Non-linear conception of extraprosodicity. This concept involves the claim that in some given string, certain constituents (autosegments, syllables, feet, etc.) may not count for the purpose of assigning prosodic structure. In this sense, they are extraprosodic.
To achieve this goal, the present paper is divided into three sections. First, we proceed by defining this concept, delimitating its types and assessing each one against the other to work out their similarities and differences. The second one deals with the phonetics and phonology of extraprosodicity. The goal of this section is to show that extraprosodicity is phonetically well-motivated and phonologically well-founded. The third section deals with the representation of extraprosodic material. Different proposals are sketched and assessed against each other. In doing so, we are seeking to acquaint the unfamiliar reader with the issues extraprosodicity has raised.
Table of Contents
Introduction
1- Definition
2- Types of Extraprosodicity
2.1- Extrametricality
2.2- Floatingness and Extratonality
2.3- Templatic Inertness
2.5 - Extrasyllabicity
3- Phonology and Phonetics of Extraprosodicity
4- The Representation of Extraprosodic Segments
References
Introduction
The principal objective of this article is to provide a characterization of the Standard Non-linear conception of extraprosodicity. This concept involves the claim that in some given string, certain constituents (autosegments, syllables, feet, etc.) may not count for the purpose of assigning prosodic structure. On this sense, they are extraprosodic.
To achieve this goal, the present paper divides to three sections. First, we proceed by defining this concept, delimitating its types and assessing each one against the other to work out their similarities and differences. The second one deals with the phonetics and phonology of extraprosodicity. The goal of this section is to show that extraprosodicity is phonetically well-motivated and phonologically well-founded. The third section deals with the representation of extraprosodic material. Different proposals are sketched and assessed against each other. In doing so, we are seeking to acquaint the unfamiliar reader with the issues extraprosodicity has raised.
1- Definition
As far as the etymology of the term extraprosodicity is concerned, we notice that it is composed of three morphemes. The first one, 'extra-', is a prefix with the meaning “additional” or “outside” (cf. Oxford Dictionary). The second morpheme is 'prosodic', which derives from prosody. Following Crystal (1985, pp249-50), prosody is a term used “in suprasegmental phonetics and phonology to refer collectively to a variation in pitch, loudness, tempo and rhythm. Sometimes it is used loosely as a synonym of suprasegmental”. In this study, we take prosody to be a combination of the two senses above (see Hockett, 1947). In other words, prosody simply means non-linear hierarchical units above the segment (tonal tiers, syllables, feet, and word, etc.). As for '-ity', it is a noun-forming suffix. So prosodicity can be taken as a synonym of suprasegmental and of hierarchical structure above the segment. Therefore by combining “extra-“and prosodicity we derive extraprosodicity with the meaning outside prosodic/ hierarchical structure.
With hindsight, in the non-linear paradigm, the concept of extraprosodicity involves the claim that in some given string, certain constituents (autosegments, syllables, feet, etc.) may not count for the purpose of assigning prosodic structure. On this sense, they are extraprosodic.
Within this context, linguists working in different areas of morphophonology have encoded this notion into their analyses –namely Metrical Stress Theory, Syllable Theories, Autosgmental phonology, Templatic Morphology and Prosodic Circumscripional Morphology. In the Metrical theory of stress, the term extrametricality is often used. In the Autosgmental theory, the term Extraprosodicity is rendered by the term floatingness and extra-tonality. In Templatic Morphology, the term templatic inertness is found. In Circumscriptional Morphology, the label negative circumscription is come across. Finally, in Syllable Structure theories, the term extrasyllabicity is often utilized.
It should be pointed out that the term “extrametricality “is more popular than “extraprosodicity“. This is so because extrametricality claims a certain historical precedence over extraprosodicity and equally over the other terms (extrasyllabicity, negative circumscription, extratonality, etc.). For example, linguists working in the area of syllable structure often use the term extrametricality more than extrasyllabicity (cf. Durand, 1986; Prince & Smolensky, 1993; Steriade, 1982 and Levin, 1985, etc.). On the other hand, the term extraprosodicity has been coined, we believe, to discard any confusion the term extrametricality may inspire. On this sense, extraprosodicity is a neutral term while extrametricality is not (cf. Goldsmith, 1990; Hdouch, 1994).
Conceptualized herein as englobing many phenomena, Extraprosodicity may be phenomenologically characterized as a theory of Exceptionality or Invisibility (cf. Hung, 1993-1994)). It can also be seen as an attempt to form a synthesis between different theories of the generative paradigm, hence the label cover term above.
After having defined the term “extraprosodicity”, we now turn to providing a review of its types.
2- Types of Extraprosodicity
The purpose of this section is to provide a synopsis of the specificity of each type of extraprosodicity. As stressed above, terminological variation follows from the particular subject under study. Phonologists distinguish between five instances of extraprosodicity: extrametricality; floatingness and extratonality; templatic inertness; negative circumscription and finally extrasyllabicity.
2.1- Extrametricality
The concept of extrametricality has been introduced in the metrical analysis by Liberman and Prince (1977) and Nanni (1977). The term has been developed by Hayes (1979) and much subsequent work including Hayes (1981, 1984, 1993), Archangeli (1984), Pulleybank (1986), Sauzet (1989), Inkelas (1989), Barker (1989), Buckey (1991) (cf. Hayes, 1993 for a complete list).
This notion involves the claim that in some given string certain constituents do not count for the purpose of assigning metrical structure. In other words, extrametricality means that some constituents (segments, consonants, vowels, syllables, feet, etc.) are systematically ignored in the computation of stress patterns (cf. Selkirk, 1984). Hayes (1993) proposes the following rule for extrametricality:
(1) x < x > / -----] D where < > = [+extrametrical]
Where x is some phonological constituent and where [-----] D is the edge domain in which the stress rules of the language apply. This domain is usually the word (cf. Hayes, 1982; Selkirk, 1989).
Hayes (1993) argues that extrametricality is a diacritic feature attached to constituents which cause stress rules to treat constituents as invisible entities. Hayes (1980, 1993) and Harris (1983) argue that extrametricality, a powerful device, should “be severely” constrained so as to characterize only constituents found at the extremes of particular domains”. In this respect, Hayes (1993) proposes the following restrictions on extrametricality:
(2)
a) Constituency: Only constituents (e.g. segment, syllable, foot, phonological word…) may be marked as extrametrical.
b) Peripherality: A constituent may be extrametrical only if it is at a designated edge (left or right) of its domain.
c) Edge Markedness: The unmarked edge for extrametricality is the right.
d) Non-Exhaustivity: An extrametricality rule is blocked if it would render the entire domain of stress rules extrametrical.
Hayes (1993) argues that the Peripherality Constraint was proposed by Harris (1983) as a well-formedness condition, rather than as a constraint on extrametricality rules. The fact that the right edge of a domain is the unmarked one is supported by a large body of cases. Some cases where the left edge is marked for extrametricality have been attested as well.
As for the provision Non-Exhaustivity, it is necessary to allow for the stressing of monosyllabic words in languages with syllable extrametricality (e.g. Latin, Hayes, 1993).
Hayes (1980, 1982, and 1993) argues that extrametricality may sound as a counter-intuitive device. For this reason, he supplies a body of arguments on its behalf. First, it severely restricts the inventory of possible foot types. It is a device that eliminates ternary feet. Indeed, a template like the one in (3a) or (3b) is eliminated in languages with syllable extrametricality. The favored template is (3c).
Abbildung in dieser Leseprobe nicht enthalten
Where x a head of foot, (.) a light syllable and σ, a syllable.
To ensure the full integration of extrametrical material into prosodic structure, Hayes (ibid.) invokes the convention of Stray Adjunction. This convention simply states that once a given layer of metrical structure has been created, any element that is stray on that layer is adjoined as a metrically weak member to the adjacent constituent. This Stray Adjunction Convention (SAC) is well-motivated because prosodic constituents always exhaust the string. An adjoined constituent will be realized phonetically. In case it is not adjoined, it will be erased by the universal convention of Stray Erasure (See Steriade, 1982; Itô, 1986, 1989).
Second, in syllable weight measuring, domain final-position is special. Most often a CVC syllable is heavy non-finally but light (=CV) finally. This asymmetry is attributed to final consonant extrametricality. Of relevance here is the fact that the final consonant in a CVC syllable is syllabified in the coda but in another context it is marked as extrametrical. The last argument proposed by Hayes (1993) is that in some languages different lexical categories may have “distinct but related stress patterns.” English is a case in point. In fact, Hayes (1981, 1982) shows that extrametricality offers a unified account of stress in nouns and suffixed adjectives on the one hand and stress in verbs and unsuffixed adjectives on the other.
Harris (1983) goes a step further in equating markedness with extrametricality. He argues that a word like [sabana] has a marked stress pattern which must be recorded lexically by a representation like [sa’ban]>a] N. (underlining designates the constituent in which markedness lodges). If the final rime in the derivational stem is extrametrical, then extrametricality becomes the correspondent of marked stress in the lexicon.
To sum up, it should be noted that extrametricality and the constraints imposed on its use have been extended to other theories of phonology and morphology. We now turn to see how Autosgmental Phonology has encoded the assumptions laid down above.
2.2- Floatingness and Extratonality
Autosegmental representations consist of two or more tiers of segments. Segments on each tier differ from the other with respect to which features are specified in them (cf. Goldsmith 1976, 1979, 1985 and 1990). Tones and phonemic melodies are related by means of association lines. That is, to relate tones and vowels, appeal is made to what is known in autosegmental parlance as ‘Association Convention’ which states that (cf. Goldsmith 1990, p19):
(4)
When unassociated vowels and tones appear on the same side of an association line, they will be automatically associated in a one-to-one fashion, radiating outward from the association line.
The following schema demonstrates this state of affairs:
Abbildung in dieser Leseprobe nicht enthalten
However, it is the case that the Association Convention is not exhaustive. In other words, tones may be left unassociated to vowels and vowels may surface associated to no tones. See (6) below:
Abbildung in dieser Leseprobe nicht enthalten
Where H and L = High and Low
To paraphrase, given the nature of autosegmental representations and the character of the Association Contention (AC), the theory predicts that tones and vowels have an independent life of their own. On this view, the theory predicts two situations. The first one involves cases where tones are just inert or floating; that is, they are not docked onto a vowel on the corresponding tier. In this respect, Goldsmith (1990) warns us that the term floating is used in two ways. First, it refers to a morpheme that is lexically only tonal. Second, if a certain vowel is deleted in a derivation, the tone associated to it may be left floating. For the sake of illustration, we present the following example from Mixtecan (Goldsmith, 1990- p20), which has an underlyingly unassociated suffixal high tone, and which associates rightward to the following word. See (7) below:
Abbildung in dieser Leseprobe nicht enthalten
(Where M=Mid) ‘the child will eat’
The stray high tone gets associated with the following word to be prosodically licensed.
Related to the issue of floating tones is the fact that in some languages, like Supyire (Goldsmith, 1990; Carlson, 1985) a particular pattern may show up whereby tones may outnumber vowels. The final floating tone is then associated by rule to the final vowel. See (8) below:
Abbildung in dieser Leseprobe nicht enthalten
Supyire contributes another important point. In this language nouns have a lexically specified floating low tone on the left as in (9):
Abbildung in dieser Leseprobe nicht enthalten
Goldsmith argues that instead of modifying the AC as stated in (4) above “it appears preferable to allow some means for certain lexically-specified elements to be marked as inert, that is as not participating in the autosegmental rules at a given stage in the derivation.” For him, the notion “inert” is a generalization of the notion extrametricality. On this sense, only peripheral lexically-specified tones may be ignored by the AC the same way metrical structure rules ignore certain peripheral constituents. This lexical inertness may be schematized as follows:
(10) T < T > / [-----] D where< > =[+inert] or [+ extrametrical]
and D = Domain of tone association.
The edge may be left or right.
From the above schema, we conclude that 'inertness', 'floatingness' or 'tone extrametricality' is a diacritic feature imposed on lexical tones.
On the other hand, the second situation predicted by the theory is the one in which certain vowels may be lexically extratonal. In other words, vowels may be invisible to the autosegmental rules associating tones to vowels. For example, in Kurundy (Goldsmith, 1990), words that are vowel-initial have no tone assigned to them by the word-level phonology. In this sense, all word-initial vowels are inert or extratonal. A similar case of vowel extratonality is reported by Akinbiyi (1995)5. In Bini’s polysyllabic forms, all low tones of the head noun become high except the first. The associative high tone whose domain is the head noun fails to associate to the first vowel, less it would create a rising tone, a tonal structure not allowed in this language. According to AkinBiyi (p.272) “the standard rule-based treatment considers the peripheral segment [vowel] as extratonal”.
The rule assigning extratonality to vowels may be formulated as follows:
(11) V < V > / [-----] D (Edge left or right)
Floatingness and extratonality present similarities and differences with the notion of extrametricality sketched above. As far as similarities are concerned, we notice that in the case of “Tone Floatingness” certain tones may be diacritically marked as floating in the underlying representation the way underlying material is marked for extrametricality. In addition to that, a floating tone may be the target of a later association rule the way Stray Adjunction adjoins the extrametrical material to the left (or right) syllable, foot or prosodic word. Vowels may also be marked for extratonality, a diacritic imposed on certain vowels. Finally, it appears that in both cases a unit is marked as 'inert' to secure the operation of a certain rule.
The implementation of the concept extrametricality for tonal phenomena shows differences with the standard metrical view. Autosegmental phonology makes use of two terms to express what in metrical stress theory is one entity. Indeed, we need tone floatingness to describe the failure of tones to associate to a licenser (vowels) and extratonality to strictly express vowels’ inertness. The reason behind this terminological bifurcation lies in the fact that tones and phonemic melodies are placed on independent tiers. As a matter of fact, we cannot attribute extratonality to tones because a tone cannot be extratonal in itself.
This comes to the conclusion that unlike metrical stress theory where lower units are extrametrical only in relation to their licenser (a syllable is extrametrical vis-à-vis a foot and not vice versa) autosegmental phonology cherishes the claim that both lower units (tones) and their licensers (vowels) can be marked as extraprosodic vis-à-vis each other.
Another difference between the two models lies in the implementation of the Edge Markedness constraint. Metrical stress theory shows a bias towards the right edge. In contrast, in Autosegmental phonology, floating tones and extratonal vowels may appear on either the left or the right edge. On this account no edge can be said to be marked than the other.
As for the Non-Exhaustivity constraint, Hewitt (1989), reported in Prince and Smolensky (1993), argues that extrametricality with respect to tonal association may indeed exclude entire monosyllabic domains. This case points towards a difference in the implementation of the Non-Exhaustivity constraint. In fact, in metrical stress theories, extrametricality would be revoked if it would lead to the exclusion of an entire domain (e.g. Latin - Hayes (1993)). Nevertheless, in autosegmental analyses, it is just the case that a whole monosyllabic domain may be discarded by “Tonal Extrametricality“, to use P and S term.
To sum up, the differences pointed out above in the implementation of extrametricality (i.e. extraprosodicity) in Metrical and Autosegmental theories respectively show that it [i.e. Extraprosodicity] is not a unified notion (cf. P and S (1993)). In the next section, we deal with extraprosodicity in Templatic Morphology.
2.3- Templatic Inertness
Incorporating the autosegmental formalism (Goldsmith 1976, 1980, 1985), the Theory of Templatic Morphology (McCarthy 1979, 1980, 1981, 1982..) conceives of phonological representations as being multi-tiered. Of relevance here is the fact that extraprosodicity has played a major role in the description of word-formation in languages with non-concatenative morphology. The arguments presented below concern the Par Excellence studied example of Classical Arabic verbal stem morphology (McCarthy (1979, 1980, 1981, 1982); Goldsmith, 1990) in which consonants and vowels are placed on different tiers.
Within the classical account of the language, "there are fifteen conjugations, formal categories with strict phonological definitions and rough semantic meaning," argues Goldsmith (1990). The choice of Cs and Vs does not depend on consonantal morphemes. This change of the syllable structure (defined on the skeleton) is an integral part of the derivational component of the language. The second conjugation of the verb “ktb” (write) is defined by the skelata in (12) below:
[...]