The Arabic definite article does not assimilate

In this paper we argue that the popular account of the phonology of the Arabic definite article in terms of assimilation to a following coronal consonant is not justified. The accepted account holds that the definite article has an underlying phonological form /ʔal/ (or /l/ in some versions) which surfaces as [ʔal] when the following word begins with a non-coronal consonant, but when the following word begins with a coronal consonant the /l/ completely assimilates to the coronal resulting in a geminate coronal consonant: compare for example al-bint [ʔalbint] ‘the girl’ and al-zaffa [azːafːa] ‘the wedding procession’.

We present theoretical and empirical grounds for rejecting the assertion that the /l/ of the definite article assimilates to a following coronal consonant in any synchronically meaningful sense of ‘assimilation’. We argue that for something to count as synchronic assimilation it must be optional, meaning that an unassimilated pronunciation must also be allowed by the grammar. In the absence of counter-evidence, we also assume that optionality implies gradience. Historical assimilation, by contrast, admits neither optionality nor gradience, only unsystematic token-to-token variation in the realisation of the product of a historical process of change. Using illustrative acoustic and electropalatographic data, the situation with the /l/ of the definite article when followed by a coronal consonant is compared to within-word sequences of /l/+coronal consonant in alzam /alzam/ ‘most necessary’, alṭaf ‘most kind’, the form I doubled verbal coronal geminate in hazza /hazza/ ‘to shake’, and the optional assimilation of word-final /l/ to word-initial /r/ in ḥabil rafī‘ ‘a thin rope’. It is also compared with the optional assimilation of the definite article /l/ to a following dorsal stop in Cairo Arabic. Electropalatographic and acoustic data are presented to support the argument that forms such as [azːafːa] should be regarded synchronically not as assimilation but as a type of ‘true’ or ‘lexical’ geminate resulting from phonologically-determined allomorphy.