The Psychological Reality of OCP-Place in Arabic

The psychological reality of an abstract consonant dissimilation constraint is demonstrated in an experiment with native speakers of Jordanian Arabic. In this experiment, novel verbs containing constraint violations and those without violations were presented orthographically for judgments of well-formedness. Native speaker well-formedness judgments reflected knowledge of the phonotactic constraint. Systematic gaps were rated much less wordlike than accidental gaps that were equivalent in their lexical characteristics. Judgments for novel verbs containing constraint violations were also gradiently influenced by consonant pair similarity. The experimental study supports previous dictionary-based phonotactic analyses that propose that the native speaker’s knowledge of consonant cooccurrence constraints in Arabic is based on emergent generalizations over the lexical items in an abstract root lexicon.*

The phonotactic constraints of a language define the set of possible words in that language. Typically, phonotactic constraints are restrictions on the possible consonant clusters in a language or conditions on the minimal prosodic unit that can be considered a word. Such constraints undoubtedly have an impact on productive word formation and foreign word accommodation, and thus these constraints are assumed to be active in the synchronic grammar. In the case of the Arabic dialects, there have been a number of studies (based on the analysis of a dictionary of Standard Arabic as an approximation to the mental lexicon) that claim that there is a phonotactic restriction on possible consonant sequences within the abstract verbal root (Greenberg 1950, McCarthy 1986, 1988, 1994, Pierrehumbert 1993, Frisch et al. 1997). The Arabic case is of particular interest as the root consonants do not usually appear adjacent to one another in the words of the language. Arabic has a root-and-pattern morphology, where consonants from the verbal root are interleaved with vowels in a manner prescribed by a lexical template. Due to the abstractness of lexical roots, an investigation of the psychological reality of consonant cooccurrence constraints defined over root consonants provides a strong test of abstract generative theories of competence.

In generative phonological theory (e.g. Kenstowicz 1994, Prince & Smolensky 1993) phonotactic constraints are typically categorical statements of well-formedness. Recent analyses of Arabic phonotactics, however, have found systematic quantitative regularities that cannot be explained by categorical constraints (McCarthy 1994, Pierrehumbert 1993, Frisch et al. 1997). But these analyses have been based on statistical studies using a dictionary as an approximation of the knowledge of lexical items possessed by native speakers. Thus, it is additionally unclear whether the detailed quantitative patterns in the lexicon are a part of the grammars of native speakers. It may be that quantitative lexical patterns are regularized and encoded as categorical in the grammar. Alternately, it could be the case that quantitative phonotactic constraints have no effect on the grammar at all. The statistical lexical patterns may nonetheless affect productive word formation or metalinguistic judgments of well-formedness via analogical processes. In this case, we might say that these patterns are part of a native speaker’s synchronic knowledge, but they are external to the grammar.

The potential for lexical statistics to influence well-formedness judgments was taken into account in the design of stimuli for an experiment on native speaker knowledge of Arabic phonotactic patterns. Lexical statistics and the presence of phonotactic constraint violations were carefully manipulated in an attempt to determine the relative importance of these factors in speakers’ well-formedness judgments for novel roots. By controlling lexical factors, we obtain unambiguous evidence for the psychological reality of abstract, gradient phonotactic constraints in Arabic. Our results support recent studies of Arabic phonotactic patterns based on statistical analyses of dictionary corpora. Further, we demonstrate that these constraints must be abstract, and not merely a reflection of superficial lexical statistics or analogical comparison of relatively unanalyzed surface forms (e.g. as in the analogy-based grammar of Skousen 1992, see Eddington 2000). Due to the quantitative nature of this abstract constraint, the phonotactic knowledge employed by native speakers of Arabic cannot be described in current generative theory. Thus, we find phonotactic knowledge is more fine grained than generative theory allows, and more abstract than analogical word comparison would...