Models of visual word recognition in Latin script assume rapid abstraction of letter identities from perceptual features, reflected in the absence of visual similarity effects for pseudowords (e.g., viotin and viocin yield similar response times in lexical decision). Whether this principle extends to Arabic-derived scripts remains unclear because visual similarity in Arabic can be confounded with preservation of consonantal root structure in a Semitic language. We examined this issue in Persian, an Indo-European language written in an Arabic-derived script that preserves dot-based letter contrasts but lacks Semitic root-based morphology. In two lexical decision experiments, pseudowords were formed by replacing one internal letter in words with (1) a visually similar letter (e.g., موسیفی /musifi/ [base word: موسیقی /musiqi/, music]), (2) a visually dissimilar letter preserving ligation (موسیلی /musili/) or (3) a visually dissimilar letter altering ligation (موسیزی /musizi/). Experiment 1 presented the stimulus until response, whereas Experiment 2 limited presentation to 200 ms. Response times were slower for visually similar pseudowords than for visually dissimilar pseudowords preserving ligation, and the two visually dissimilar conditions did not differ reliably. These findings support the view that lexical access in Arabic-derived scripts operates over abstract letter identities, while dot-based visual similarity can produce a processing cost.