The role of morphology in lexical processing and representation is a critical issue in psycholinguistics. Most research into this has been in languages with a concatenative morphology, with the exception of recent work in Hebrew. Another Semitic language, Modern Standard Arabic (MSA), which also has a non-concatenative morphology, offers additional opportunities for probing not only the use of discontinuous morphemes that never surface on their own, but also discontinuous morphemes that do not carry semantic information.
In MSA surface word forms fall into three categories: primitive nouns, deverbal nouns and verbs. Each of these comprises three interwoven morphemes: a consonantal root morpheme that carries semantic information, a vocalic morpheme conveying syntactic information and a skeletal morpheme providing a canonical shape associated with a particular meaning or grammatical function. Thus, the word [katam] ("conceal") is composed of the root morpheme {ktm} with the semantic load "concealment", the vocalic morpheme {a-a} with the syntactic information "active" and the skeletal morpheme {CVCVC} with the information "past tense". The combination of vocalic and skeletal morphemes gives rise to an abstract complex morphological unit called the Word-Pattern.
In a series of cross-modal priming experiments we investigated the processing and representation of these root and word pattern morphemes. When prime and target are primitive nouns sharing the two components of the word-pattern (e.g., [qamarun]-[maraqun], "moon-sauce") no priming is obtained. When deverbal nouns sharing a word-pattern (e.g., [xuDuuAun]-[HuduuTun], "submission-happening") are used as primes and targets, cross-modal priming occurs only if the overlap is both at the level of the vocalic and the skeletal morphemes. Pairs like [sujuunun]-[ HuduuTun] ("prisons-happening") which share the form of the word-pattern but not its meaning fail to yield any priming effects. Word pattern priming is further obtained with verbs both when prime and target share the form and meaning of the word pattern and when they share solely its form. This suggests that the use of the word pattern morpheme as a "phonological structure" or as a "meaning-carrying unit" depends on the transparency and the productivity of the surface form at hand.
Turning to the root morpheme in MSA, we focussed on verbs and deverbal nouns, using primes and targets sharing the root morpheme while co-varying the semantic relationship between them. Significant cross-modal priming was obtained even in the absence of a transparent semantic relationship between prime and target.
Taken together the results point clearly to word-patterns and roots being lexical units in MSA. They indicate that the language processor picks up on highly abstract morphological units, which never surface on their own. The morphological priming effects observed both with opaque root morphemes and with word patterns are hard to account for as simply an interaction between form and meaning.