The architecture of the english lexicon
Prefixes in unsuffixed words
Yüklə 2,22 Mb.
|
- Bu səhifədəki naviqasiya:
- 5.5 Summary of lengthening effects
- Stem-Coherence and SuffixMoras
- Stem-Ft-R
- NonFin( s ²)
- Moraic Prominence
- 6. Stress retraction and constituency 6.0 Retraction and OT
- 6.1 Words in /-ate/ and /-ize/
- Edgemost
- Ft’ to MWd
- NonFin(Ft’)
- Edgemost and Stem-Ft-Rt
5.4.3 Prefixes in unsuffixed words Prefixes are also involved in effects involving the right edge of the word, motivated by NonFinality constraints. As was discussed in ¤ 3.3.4, there are some classes of nouns and adjectives which correspond to prefixed verbal stems of the type con-dœct. Some show the same stress pattern as the verb, e.g., resœlt, effŽct, while others show stress on the prefix, e.g., ’ncrease, rŽbound, rŽcess, dŽfect, prŽfix. The former group clearly shares the same structure as the corresponding verb forms, the suffix /-æ/ being retained and accounting for final stress. The latter group differs in that the suffix /-æ/ is apparently not included in the nominal form, and such words show initial stress on the prefix due to NonFin(Ft’). However, unlike the case in bisyllables formed from unitary stems, mora augmentation is possible because the prefix is not part of the morphological stem. Thus, lengthened forms are expected in both monomoraic prefixes and stems in such cases, without incurring violations of Stem-Coh: (5.61)
(5.62)
(5.63)
The optimal candidates in these cases would have been eliminated by Stem-Coh if they had formed in themselves complete stems, e.g., br’gand. For longer stems, e.g., dŽ-ficit, rŽ-primand, and for multiply prefixed forms such as dŽ-re-lict, the larger foot domains preclude lengthening: (5.64)
(5.65)
In both cases, the adoption of the prefix into the head foot, combining with the initial light syllable of the stem, is the most optimal choice available. There is another type of prefixed word which does not show these effects, but rather shows a short vowel in the prefix, e.g., rŽcord, pr—duct, pr—ject. It will be proposed that this type, which is much more restricted in distribution than the type of word described above, is treated as if both prefix and root formed a single stem, i.e., such forms are treated like br’gand. How this different treatment of subcategorization and constituency in these forms can be lexically handled using OT will be discussed in ¤ 7.1.1. Finally, there are word-initial sequences which are apparently monomoraic prefixes, but which never show lengthened or stressed alternants, e.g., appr‡isal, av—wal, asŽssment. These stand in contrast to forms like appr—ve Ü ˆpproba³tion, attŽst Ü ˆttesta³tion, where the prefix may appear with vowel quality when stressed and can be given an underlying form /ad-/, assimilation taking place in the adjacent obstruents. While forms like appra’sal could be given the same interpretation, it is simpler for the grammar to rather regard such forms as prefixed in /æ-/. The segment /æ/ is never lengthened regardless of moraicism, and also cannot be stressed.93 Since such syllables never appear with any vowel quality, it is difficult to say how the grammar would be able to assign any other segment to their nuclei. This is further evidence for an underlying segment /æ/ within the grammar, beyond its role in the verbal and adjectival suffix / æ/.
Most of the lengthening effects described in this chapter are the direct result of morphological constituency. Feet which are restricted, for a variety of reasons, to align with open, light syllables, are forced to become bimoraic due to the Min-2 constraint. There are a variety of strategies for achieving this bimoraicity, but all produce the similar effect of a lengthened vowel in an otherwise monomoraic morpheme. The overgeneration of forms with lengthened vowels is prevented by Stem-Coherence and SuffixMoras, which ensure that such lengthening is licit only within monomoraic stems and prefixes (¤ 5.2). This results in the attested distribution of arguably lengthened vowels, which are never seen in multimoraic stems or in unstemmed suffixes. Vowel lengthening as seen in forms affixed with the / al/ group of suffixes (e.g., to³nal, repr´sal) is enforced by both Stem-Ft-R, which aligns the foot to the right edge of the stem, and NoOrphans, which favors the association of the mora projected by the suffix to the stem, where it can be footed. Among the / ent/ group, which subcategorize similarly to the / al/ forms but are intrinsically bimoraic, lengthening occurs only when insufficient moras are present in the stem, e.g., so³nant. When more material, such as a prefix, is present in the word, it can be incorporated into the stem foot, e.g., rŽsident. The contrast between the segmentally similar forms repr´sal and rŽsident is due to the effect of NoOrphans, which does not apply to the latter as its suffix is footable (¤ 5.1.2). Monomoraic prefixes such as /re /, /pro / can also be lengthened when they occur alone in a foot. In some cases, constraints enforcing extrametricality effects such as NonFin(Ft’) prevent stems from appearing within the primary stressed foot, and the prefix alone appears as the main stressed foot, e.g., the noun de³fect. In other cases, the prefix stands in a stem and morphological word constituent of its own due to its subcategorization, e.g., reø-cr«a³te ‘create again’, which can be contrasted with rŽcr«a³te ‘refresh’, where the prefix is part of a single morphological word constituent and fails to undergo lengthening since it forms a bimoraic foot with the stem syllable /cre/ (¤ 5.4.2). The prefix can also lengthen where it forms a secondarily stressed foot, appearing to the left of the main stress, e.g., r«v—lt, d«ba³te, pr¯v´de. However, such feet are understood as stressless, and may appear optionally with reduced vowels, e.g., ræv—lt, dæba³te, præv´de (¤ 5.4). This is due to the proximity of the main stress and the restriction of adjacent stresses enforced by NonFin(s²) (¤ 5.3). These lengthenings can be contrasted with another, well-known type of lengthening, so-called "CiV lengthening". This takes place within the stem, and is the result of a mis-correspondence between moras projected by segments and moras which associate to segments. An unassociated mora is projected (for an /i/) by Moraic Prominence, and associates to the vowel of the preceding syllable in the optimal candidates, because the /i/ links to the onset of the following syllable due to the Onset constraint. This shift does not violate Stem-Coherence, as the mora count within the stem remains constant. This lengthening is different from the lengthenings described above as it is the result of interactions between prosodic, rather than morphological, constituents. Another purely prosodic lengthening effect is seen when two vowels appear in hiatus, and is also enforced by the Onset constraint (¤ 5.2.2). The morphologically conditioned lengthenings identified herein went previously undetected due to theoretical assumptions about the underlying lexical forms of these words, which were thought to contain long vowels. This assumption has led to the various proposals of shortening discussed in chapter 2, and the subsequent difficulties this approach posed for accounts of English stress. The distribution of long and short vowels in the English data, as presented in chapter 3, demands a new understanding of the underlying forms seen in "level I" words, and the OT accounted presented in chapters 4 and 5 provide an analysis which successfully treats both stress and vowel alternation. The next chapter applies the OT hierarchy arrived at here, as well as the conclusions drawn about underlying forms, to further cases of "level I" affixation and "stress retraction" in English.
6. Stress retraction and constituency 6.0 Retraction and OT In the preceding chapter, most of the forms discussed took their main stress on one of the final three syllables, in accordance with Hayes’ principle of extrametricality, now captured in an OT context by using various constraints. However, there are various suffixed types which can show their main stress further back in the word, or further back than expected purely on segmental grounds. These are the forms which have been described as showing "stress retraction" by Kager (1989) and others, as described in chapter three. As with previous cases of suffixation, these will be accounted for using subcategorization constraints and the constraint hierarchy arrived at in chapter five. In this chapter, a number of the retraction effects noted by Myers (1987), Kager (1989) and others are discussed and accounted for structurally, most prominently the retraction seen in the /-ate/ and /-ize/ series of suffixes (¤ 6.1). The constraint hierarchy under discussion will be augmented to account for new phenomena, and further insight will be given into the role of morphological subcategorization in the prosodification of English words, as a number of affix types are treated in detail (¤¤ 6.2-6.6). Some contrasts between British and American pronunciations are accounted for by proposing slightly different constraint hierarchies for these two major dialects. Following the treatment of various affixes, some of the notable retraction effects discussed in the literature are addressed (¤ 6.7), and finally the role of the non-Latinate vocabulary will be touched upon (¤ 6.8). 6.1 Words in /-ate/ and /-ize/ Two sets of very common verbal affixes which show a consistently "retracted" stress pattern are /-ate/ and /-ize/. About 95% of such forms trisyllabic and longer show antepenultimate stress, regardless of the weight of the penult: (6.1) (hŽsi)tate (c‡ndi)date (hy³phe)nate (dŽsig)nate (Žx)(pur)gate (po³)(ten)tate (dŽle)gate (c—n)(cen)trate (dŽmon)strate
Since the Edgemost constraint as formulated above in (5.41) aligns the head foot to the right edge of the morphological word (rather than the prosodic word), placing the head away from its edge (which in this case corresponds to the stem edge, the right edge of /-sign-/) will incur violations; however, the more highly ranked constraint NonFin(s²) prevents the penult (which is final in the morphological word) from ever taking main stress.95 Thus, the unusual presence of (LH) stems within the word, and the general antepenultimate stress seen in these forms, is explicable via suffix subcategorization, the morphological word boundary appearing before, rather than after, the level I suffixes in question. For monosyllabic stems suffixed in /-ate/, some British pronunciations fail to show stress on the final stem syllable, e.g., rot‡te, castr‡te, vac‡te, migr‡te, as predicted by the current constraint hierarchy:
Setting the main stress on the suffix violates only the Edgemost constraint, while stressing the stem would violate the more highly ranked NonFin(s²). Some dialects, however, including most American dialects, do not permit the main stress to fall outside the Morphological Word, and the stress is retained on the stem (i.e., mgrte). This suggests a constraint restricting the main stress to fall within the morphological word, Ft’ to MWd, which outranks NonFin(s²) in those dialects. This constraint can be formally defined as follows: (6.5) Ft’ to MWd: NI(PrWd: PrWd, MWd, Ft’) This states that for each prosodic word, no head foot intervenes between the prosodic word and the morphological word. This will be satisfied only if the head foot is inside of some morphological word. In a position outside the morphological word, it will intervene between a prosodic word edge and a morphological word edge, incurring a violation. Adding this constraint to the hierarchy does not affect previous results, as all optimal candidates among word types dealt with above would satisfy it. Monomoraic stems of the type seen in mögra³te would be typically lengthened as in the previous cases discussed, explaining the dearth of light stems of this kind (e.g., n‡rrate, a geminate stem). Note that as with the s—nant type, the lengthening is no longer seen when the stem is extended by a prefix, e.g., “mmigr‡te. The subcategorization of /-ate/ and /-ize/ for the morphological word rather than for the stem, as with the /-al/ and /-ent/ suffixes, suggests that the stress pattern seen in words with these suffixes should resemble that seen in their unsuffixed forms, and that suffixed words that are further extended with /-ate/ or / ize/ would retain their stress patterns. For example, we find pairs such as —rient Ü —rientate, n‡tional Ü n‡tionalize, dit—rial Ü dit—rialize, “ndiv’dual Ü “ndiv’dualize. This is expected, because the role of the stem remains the same within both forms, the main-stressed foot unable to encompass the suffix /-al/, which must follow the stem due to subcategorization. However, to yield the correct forms, it must be recognized that the definition given to Edgemost above (5.41), which was based on previous derivational conceptions of extrametricality, is in fact insufficient to account for this (although this has no effect on the validity of previous tableaus). Previously, Edgemost stated that the head foot of a given morphological word preferred to align to the right edge of that word, and was only thwarted in classic extrametricality cases like (—ri)gin, re(vŽr)sal, by the higher-ranking constraint NonFin(Ft’). However, when the morphological word is not congruent with the prosodic word, as in the / ate/ and /-ize/ forms, NonFin(Ft’) will not apply to morphological word-final syllables which are not word-final in the prosodic word. Thus, nothing would prevent the incorrect form *[vi(sœ-al)]ize. Instead, a new definition for the Edgemost constraint should be introduced, formulated as follows:96 (6.6) Edgemost: NI-R(MWd: Ft’, Stem, s) This replaces the morphological word with the morphological stem as the second argument of the constraint. Thus, the head foot attempts to align to the right edge of the stem (rather than the morphological word). This correctly penalizes the misaligned stem in the morphological word constituent *[vi(sœ-al)], and enforces the correct [(v’su)-al]. It is the consistent form of the stem throughout such complex words in English that produces, for most stems, the apparent uniformity of stress under affixation which inspired derivational theory (¤ 1.4). Only when the stem itself is altered, for example by the addition of an /-ity/ suffix, which subcategorizes within the stem, is the apparently "basic" stress pattern of a given morpheme altered. Changing the formulation of Edgemost, which had referred to the morphological word as both the domain and the second argument of the constraint, does not affect the results of the preceding tableaux. In most cases, the words being treated were unsuffixed and the right edges of the morphological word and stem were identical. For the suffixed cases discussed previously, e.g., rŽsident, repr’sal, the suffixes were both monosyllabic and standing at the right edge of the word; thus, any candidate *re(pr’sal) which satisfied the old Edgemost constraint by aligning to the right edge of the morphological word would have been eliminated by the higher-ranking NonFin(Ft’). Even in the / ize/ and / ate/ cases discussed above, e.g., dŽsignate, the right edge of the stem was congruent to that of the morphological word. It is only in cases where more than one syllable intervenes between the stem and the prosodic word edge that the new formulation becomes necessary. Applying this new definition of Edgemost, the complex multiply suffixed forms in / ate/ and / ize/ can be accounted for: (6.7)
Both Edgemost and Stem-Ft-Rt, which restrict the head foot to the right stem edge, enforce its position within the morphological word. Thus, words like v’sualize will have morphological words stressed identically to those seen in unaffixed forms such as v’sual. The fact that most of the given constraints, such as NoOrph, govern the morphological word means that the constituent relationships seen within morphological words will remain constant despite suffixation, e.g., revtalize, mo³torize, as long as that suffixation subcategorizes outside the stem. Yüklə 2,22 Mb. Dostları ilə paylaş:
|