Haplology and stress

We now move on to the discussion of *O_iRO_¡, which is argued by Raffelsiefen to account not only for the truncation of base forms suffixed by -ize (e.g. feminize), but also to be operative with other suffixes like -ity or -ify. If this were feasible, this constraint could be an important step towards the long-sought solution to the notorious problem of morphological haplology.¹ However, the constraint as it stands seems to be too powerful because it rules out a whole range of attested complex and simplex forms. Raffelsiefen herself concedes that this constraint "is often violated in English words" (1996: 199), but explains this away by restricting the constraint to potential words. Actual words may violate this constraint without any consequences. But even if *O_iRO_¡ is restricted to possible words, too many counterexamples come to mind. Raffelsiefen only mentions the lexicalized words entity, identity, quantity, sanctity, but the following counterexamples of the putatively illegal *nRn, *rRr, *sRs, and *tRt combinations with -ize can be added, namely strychninize, mirrorize, terrorize,² classicize, potentize, dilletantize.

Thus it seems that other constraints must be ranked higher than *O_¡RO_¡. However, ignoring the effect of M-PARSE, the constraint ranking given by Raffelsiefen (see (44) above) would predict that truncated forms like *classize or *strychnize would be optimal, since they would only violate lower-ranked constraints (in this case * CLASH and IDENT). But if M-PARSE is taken into account, violators of *O_¡RO_¡ will never surface at all, because M-PARSE is ranked lower. A similar argument applies to the other forms mentioned. In any case, we are again faced with wrong predictions concerning the absence or presence of certain forms.

It is obvious that the truncation of the base in some of the derivatives mentioned would lead to a derived form whose base is no longer recognizable, which may be an important general factor constraining the deletion of base-final segments in complex words. Raffelsiefen mentions this problem in a footnote and proposes an additional constraint to restrict possible truncations. She claims that minimally the first two syllables in the base must correspond to identical syllables in the derived form (p. 202). This putative constraint, like some others Raffelsiefen mentions in passing, is not incorporated into her tableaux, nor into the constraint ranking she advocates, which makes its status somewhat unclear. In any case, the constraint is designed to ensure the presence of enough phonological material so that the base can still be related to its non-affixed forms. Thus a form like *crisize (instead of the *O_¡RO_¡ violator *crisisize) would be ruled out because it is no longer recognizable as derived from crisis. However, Raffelsiefen's putative constraint on truncation is inadequate since it rules out all licit formations with consonant-final disyllabic bases like randomize. With such derivatives, the last segment of the stem is syllabified as the onset of the new final syllable (ran.do.mize). This means that in such cases there are no two identical syllables in the base and the derivative (cf. ran.dom). Thus cri.si.size would violate Raffelsiefen's base-recognition constraint just as cri.size would. Going strictly by the constraint ranking in (44), truncated forms like *potenize, *potize, or *classize would be better candidates than the attested ones. It seems that a more adequate solution is needed to account for the truncation restrictions.

In sum, the empirical evidence for the operation of *O_¡RO_¡ with -ize is equivocal³⁴, and the analysis of some counterexamples makes contradictory constraint rankings necessary. Thus with her examples, *O_¡RO_¡ is ranked higher than IDENT, but with the counterexamples IDENT must be ranked higher than *O_¡RO_¡.

A reconsideration of the contradictory data reveals, however, a striking regularity, namely that all base words where one of the identical onsets is deleted are polysyllabic (emphasis, metathesis, feminine, maximum, etc.), whereas all base words with surviving identical onsets (except one, dilletante, to be discussed shortly) are disyllabic. At first sight the constraint appears to be sensitive to the number of syllables. However, the stress pattern is also different between the violators of *O_¡RO_¡ on the one hand, and the conformers on the other. Thus, all the conformers have base words with antepenultimate stress followed by two unstressed syllables i.e. they are dactyls, while none of the violators has a base word that exhibits a stress lapse. That this stress-based description is superior to the simple counting of syllables is corroborated by the behavior of dilletántize, whose base word has a main stress on the final syllable (which all the other polysyllabic forms cited lack). In sum, the operation of *O_¡RO_¡ needs to be restricted to those cases in which two unstressed syllables precede -ize. This is similar to the situation concerning Raffelsiefen's *VV constraint discussed above, where stress lapses are in general only tolerable if the final segment of the base is a consonant. Therefore, the haplology facts should be related to the stress lapse generalization formulated in (Stress lapses are only allowed if a consonant precedes -ize.), as it is proposed in (1):

(1) 

By way of illustration, a form like *fémininìze exhibits a stress lapse, which would only be tolerable if the onset and the coda of the were not identical (as in forms such as féderalìze, sénsitivìze). Since they are identical, *fémininìze is an illicit formation. Derivatives like térrorìze do not exhibit a stress lapse, hence they behave according to the generalization. With vowel-final dactyls, stress lapses are avoided by vowel truncation (mémory - mémorìzé)

I will propose a more sophisticated constraint model to account for the variable toleration of identical onsets in -ize formations.

¹ See e.g. Stemberger (1981), Menn and McWhinney (1984) for discussion. More recent approaches to haplology are Yip (1996), and Plag (1998), both within the framework of OT.

² Raffelsiefen considers terrorize a French borrowing. According to the OED this is incorrect. Semantically, this word is also completely regular.

³ Some of the examples Raffelsiefen mentions in favor of her identical onset constraint could be ruled out on independent grounds, like for example *Hittitize (stress clash), and *horrorize (blocked by horrify). The existence of Leninize is a problem for Raffelsiefen's model, so is the form Leninite, which equally violates * O_¡RO_¡.

مواضيع ذات صلة

Non-uniform preservation of secondary stress

No vowel deletion: trochaic vowel-final bases

Vowel deletion: dactylic bases

Preservation of main stress

Stress assignment in English nouns: Pater (1995) OT model

Phonological constraints on -ize derivatives: A new account

Summary: Hiatus, truncation and stress

Truncation and stress

Towards well-formedness conditions on -ize derivatives Hiatus and stress

Previous analyses

Optimality Theory: An introduction

The phonology of -ize derivatives Introduction