Parafoveal processing of Arabic diacritical marks.

Abstract

Diacritics are glyph-like marks on letters that convey vowel information in Arabic, thus allowing for accurate pronunciation and disambiguation of homographs. For skilled readers, diacritics are usually omitted except when their omission causes ambiguity. Undiacritized homographs are very common in Arabic and are predominantly heterophones (where each meaning sounds different), with 1 version more common (dominant) than the others (subordinate). In this study the authors investigated parafoveal processing of diacritics during reading. They presented native readers with heterophonic homographs embedded in sentences with diacritization that instantiated either dominant or subordinate pronunciations of the homographs. Using the boundary paradigm, they presented previews of these words carrying either: identical diacritization to the target; inaccurate diacritization, such that if the target had dominant diacritization, the preview contained subordinate diacritization, and vice versa; or no diacritics. The results showed that readers processed the identity of diacritics parafoveally, such that inaccurate previews of the diacritics resulted in inflated fixation durations, particularly for fixations originating at close launch sites. Moreover, our results clearly indicate that readers’ expectation for dominant or subordinate diacritization patterns influences their parafoveal and foveal processing of diacritics. Specifically, a perceived absence of diacritics (either in no-diacritics previews, or because the eyes were too far away to process the presence of diacritics) induced an expectation for the dominant pronunciation, whereas the perceived presence of diacritics induced an expectation for the subordinate meaning.

Insights from the study of Arabic reading

“…The first study suggested that skilled readers only use diacritics if they are necessary to disambiguate a word; otherwise, the diacritics are treated as redundant information, with readers spending no extra processing time to analyse the diacritics (Hermena et al, 2015). The second study (Hermena, Liversedge, & Drieghe, 2016) examined the processing of diacritics on upcoming, parafoveal words 2 and showed that skilled readers actually develop expectations that, if the upcoming word lacks diacritics, its pronunciation will correspond to the most frequent one in print. This expectation was evident if the parafoveal word did not carry diacritics or if the fixation prior to the word was too distant to afford perception of the diacritics.…”

“…With regards to resolving the ubiquitous ambiguity in Arabic print, eye‐tracking research suggests that the different pronunciations of Arabic words are somehow represented and accessed from memory in terms of their relative frequency of occurrence (Hermena et al, 2016). Readers also rely upon the sentence context that surrounds the ambiguous word to compensate for the under‐specified nature of the lexical information that is available in print; this claim is supported by experiments using both eye tracking (Hermena et al, 2015) and other measures (Abu‐Rabia, 1997, 1998, 2001; Taouk & Coltheart, 2004).…”

“…Given the findings reviewed so far, it is perhaps not surprising that the increased visual and informational density of Arabic jointly attenuate the parafoveal processing of Arabic words. This is best evidenced by the fact that, whereas readers of English and other European languages skip approximately 30% of the words in a given text (e.g., Rayner, 2009; Rayner & Pollatsek, 1981), readers of Arabic exhibit only modest word‐skipping rates (on average 8%; Hermena et al, 2015, 2016, 2017; Hermana, et al, in press). Similarly, the magnitude and reliability of the effects that reflect parafoveal processing are more limited to spatially proximal pre‐target fixation positions (Hermena et al, 2016), presumably because these positions afford the high visual acuity required to identify the small features of Arabic words.…”

A dynamic adjustment account of word skipping in reading: Evidence from simulations and invisible boundary experiments

“…The manipulation was also associated with inflated foveal fixation durations, suggesting that the degradation also interfered with higher-level orthographic and lexical processes. Hermena et al (2016) reported tentative evidence that skipping is insensitive to subtle diacritical marks of the upcoming word in reading Arabic, while the same diacritics resulted in clear preview benefits when the words were later fixated. This finding suggests that the visual information that is available for saccadic modulation is functionally poorer than that later extracted via presaccadic attention.…”

Procura-PALavras (P-PAL): A Web-based interface for a new European Portuguese lexical database

“…For instance, both caso ['kazu] and cash ['kɛ ] are orthographic neighbors of casa, but only caso is a phonographic neighbor, since it takes more than one operation to transform casa ['kaz ] into cash ['kɛ ]. Finally, it is worth noting that since EP spelling makes use of different diacritics (e.g., ç á, à, â, ã, é, í, ó, õ, ú, ê, ô) that change both the visual form of the word and its pronunciation (e.g., the cedilla indicates that <ç> is pronounced [s] and not [k], as in <c>), affecting word processing (see Hermena, Liversedge, & Drieghe, 2016, for a recent eyetracking study showing diacritic effects on reading), the orthographic statistics provided in P-PAL take diacritics into account. Therefore, words such as avô [grandfather] and avó [grandmother] are considered orthographic neighbors.…”