Processing of Arabic diacritical marks: phonological-syntactic disambiguation of homographic verbs and visual crowding effects.

Abstract

Diacritics convey vowel sounds in Arabic, allowing accurate word pronunciation. Mostly, modern Arabic is printed nondiacritized. Otherwise, diacritics appear either only on homographic words when not disambiguated by surrounding text or on all words as in religious or educational texts. In an eye-tracking experiment, we examined sentence processing in the absence of diacritics and when diacritics were presented in either modes. Heterophonic homographic target verbs that have different pronunciations in active and passive (e.g., [see text]/d(a)r(a)b(a)/, hit; [see text]/d(o)r(i)b(a)/, was hit) were embedded in temporarily ambiguous sentences in which in the absence of diacritics, readers cannot be certain whether the verb was active or passive. Passive sentences were disambiguated by an extra word (e.g., [see text]/b(i)j(a)d/, by the hand of). Our results show that readers benefitted from the disambiguating diacritics when present only on the homographic verb. When disambiguating diacritics were absent, Arabic readers followed their parsing preference for active verb analysis, and garden path effects were observed. When reading fully diacritized sentences, readers incurred only a small cost, likely due to increased visual crowding, but did not extensively process the (mostly superfluous) diacritics, thus resulting in a lack of benefit from the disambiguating diacritics on the passive verb.