Vowel deletion: dactylic bases

Although the ranking in (Preservation of main stress (2)) and (3) rules out all kinds of candidates which do not preserve the prosodic head of the base, possible candidates involving the truncation of segments under preservation of main stress would emerge as optimal because of their better satisfaction of R-ALIGN-HEAD. The puzzling fact now is that consonant-final dactyls are always left intact (on the cost of R-ALIGN-RIGHT), whereas vowel-final dactyls always lose their final vowel (on the cost of some MAX constraint). The pair ho̒spital - ho̒spitalı̒ze as against mémory - mémorìze are canonical examples of this generalization. In terms of OT, this means that a constraint against the deletion of consonants (MAX-C) must be ranked above R ALIGN-HEAD and that a constraint against the deletion of vowels (MAX-V) must be ranked below R-ALIGN-HEAD:

(1) 

This splitting up of MAX into one vocalic and one consonantal constraint has been proposed before by Kager (1997) to account of rhythmic vowel deletion in Macushi Carib and South-eastern Tepehuan. What English -ize derivatives and these two languages have in common is that the deletion of vowels is conditioned by constraints on rhythmic structure (such as RALIGN-HEAD).

Let us compare the selection of optimal candidates with consonant-final and vowel-final dactylic bases, using ho̒spitalìze and mémorìze as canonical examples:

So far this accounts only for the non-uniformity of truncation with V-final and C-final dactylic bases. Disyllabic vowel-final bases will be dealt with below.

There are a number of derivatives in the corpus that go against the pre dictions made by our model as it stands. As we will shortly see, the only unsystematic counterexample is lìbráryìze. With regard to this form, the constraint model would predict the truncation of the base-final vowel instead of stress shift, contrary to the fact as given in the OED.¹ Since lìbràryìze is the only counterexample of this kind, it should be treated as exceptional.

Apart from lìbràryìze, putative counterexamples involve Greek bases ending in schwa. It was already argued above that the epenthesis of [t] is conditioned by the selection of the pertinent bound stem allomorph. The alternative to stem allomorph selection would be the assumption that the insertion of [t] is forced by prosodic constraints. This assumption faces two major problems. First, it is empirically inadequate, since, according to the proposed ranking of the constraints, truncated forms like *cinemize should be more optimal than the attested cinematize because vowel truncation is more easily tolerated than stress lapses (i.e. three violations of R-ALIGN HEAD). Of course one could argue that possibly additional constraints are needed and/or different constraint rankings. However, this is unlikely in view of the firm generalizations that fall out from the constraint ranking as proposed above, and, more importantly, in view of the fact that intermediate [t] almost exclusively occurs in a very special, well-defined set of words, those of Greek origin. Thus it seems that we are dealing here with borrowed morphology.

This becomes more evident if we consider the second major problem, which is that [t] as an epenthetic consonant cannot be independently justified. The default epenthetic consonant in English (as in many other languages) is the glottal stop, and not [t]. This can be observed, for instance, in colloquial speech, when speakers do not use the expected allomorph of the article before a following vowel [ði] or [ən]), but instead separate schwa and the following vowel by a glottal stop. Hence the selection of [t] is phonologically highly unnatural. As already mentioned, putative epenthetic [t] surfaces only in words of Greek origin or occasionally also in words that are coined in analogy to the Greek examples. In essence, we are faced with two different stem allomorphs (both borrowed from Greek), one occurring in free forms and compounds, the other in suffixed forms. The important difference between the stem allomorphy of the Greek bases and the stem allomorphies of the other bases discussed in this study is that the latter are truly phonologically conditioned, whereas the Greek allomorphs are lexically conditioned, i.e. derivatives in -ize on schwa-final Greek bases must obligatorily be based on the bound stem allomorph, the one ending in [t]. Perhaps one reason why this kind of exceptional behavior survives in English is that the derivatives coined in that way fit into the overall prosodic structure of -ize derivatives: those with penultimate stress (e.g. rhématìze, aro̒matìze) are totally unproblematic and those involving a stress lapse have a base-final consonant (cf. cìnematìze). In both cases the generalizations in (Towards well-formedness conditions on -ize derivatives Hiatus and stress((2) and (Haplology and stress (1)) are unviolated. It has to be noted, however, that in the case of schwa-final Greek bases morphology (i.e. lexically conditioned allomorphy) overrules the phonological constraints (cf. *cinemize vs. cinematize).

This account leaves us, however, with the problem of schwa-final bases that are not of Greek etymology. Unfortunately, there is only one form of this type among the neologisms, pátinìze. Another pertinent derivative mentioned in the OED is ìotìze (rare, based on iota), which is however not a 20th century innovation (first attested in 1880). It seems that no additional machinery is necessary to select pátinìze as optimal, since it is evaluated in the same fashion as derivatives like mémorìze in (2) above, the final schwa is truncated in order to satisfy more highly ranked constraints:

¹ Note, however, that the penultimate stress of libràryìze relates the verb to librárian. Thus it could even be argued that there is no stress-shift. The semantics and the fact that the verb is not !librarianize, however, do not suggest that librarian is the base word.