Haplology effects

Plag (1998) proposes a theory of morphological haplology according to which morphological haplology results from a family of constraints named OCP, which prohibit adjacent identical elements in the output.¹ The Obligatory Contour Principle (OCP) was originally suggested to ban adjacent identical tones from the lexical representation of a morpheme in tone languages (e.g. Leben 1973), but it seems that the concept can be fruitfully extended to non-tonal phenomena. In the case of -ize derivatives at least one OCP constraint plays a significant role, the one against identical onsets in adjacent syllables (OCP-ONSET):

(1) 

Note that OCP-ONSET is only terminologically different from Raffelsiefen's *O_¡RO_¡. The reason for renaming this constraint is that my terminology expresses the fact that seemingly different phenomena can be subsumed under one family of related constraints on the repetition of identical phonological structures (see Plag 1998 for elaboration and discussion).

As already mentioned above, disyllabic base words are never affected by haplological truncation, whereas dactylic bases always are. This patterning can be explained if we introduce OCP-ONSET into the hierarchy we have established so far. The data show that OCP-ONSET must be ranked between •CLASH-HEAD and R-ALIGN-HEAD. AS we will shortly see, OCP-ONSET must be crucially non-ranked with respect to MAX-C. Crucial non-ranking is expressed by a dotted line in the tableaux below and by '=' in the constraint hierarchies. Pertinent derivatives are evaluated in the tableaus in (2) and (3):

Although it may seem at first sight that MAX-C is violated in order to satisfy OCP-ONSET, we see that it is really the alignment constraint which triggers deletion with féminìze. Recall that C-final dactylic bases not leading to identical onsets (hóspitalìze) cannot be truncated due to MAX-C. If either MAX-C or OCP-ONSET must be violated, the candidate with the least violations of R-ALIGN-HEAD is optimal.

* CLASH-HEAD becomes operative only with trochaic bases, as we can see in (3):

The minimally truncated form (3c), *strýchniìze, emerges falsely as optimal in this tableau, but this is only because the effects of DEP and ONSET are not presented. Thus candidate (3c) either violates DEP (if a glide is inserted in onset position) or ONSET (if no glide is inserted). The attested candidate (3b) violates neither of these constraints and is therefore selected as optimal.

¹ See also Yip (1996:7) who assumes the following OCP constraints: OCP (feature), OCP (segment), OCP (affix), OCP (stem). The latter two, however, are not phonological in nature. In Plag (1998) the extension of OCP constraints to morphological elements is rejected. See that paper for a more detailed discussion.