Are letter features weighted asymmetrically during the early moments of lexical access? Evidence from masked priming in Serbian Cyrillic.

Models of visual word recognition assume that abstract letter identities are encoded from visual features; but the principles governing this mapping remain underspecified. In Latin-script languages, masked primes that omit a diacritic facilitate recognition almost as much as identity primes (e.g., facil → FÁCIL [easy]), whereas primes that add a diacritic produce much weaker facilitation (e.g., féliz → FELIZ [happy]). This asymmetry has been explained by the perceptual salience of diacritics, or by assuming that feature presence carries more weight than feature absence. We tested these alternatives in Serbian Cyrillic, which contains letter pairs that differ by a single integrated letter feature (e.g., Л - Љ; H - Њ). We conducted a masked priming lexical decision experiment in which a target stimulus could be preceded by an identity prime, a visually similar prime created by feature omission or feature addition, or a visually dissimilar control prime. For targets containing complex letters (Љ/Њ), omission primes were as effective as identity primes (e.g., кoницa → КOЊИЦA). For targets containing simple letters (Л/H), addition primes produced a substantial cost relative to identity primes (e.g., кoњaц → КOHAЦ). Thus, the addition-omission asymmetry reported for diacritics generalizes to letters that differ by an integrated visual feature. These findings favor a general account in which the presence of a feature is weighted more heavily than its absence during the earliest stages of orthographic processing.