Word Length and Visual-Noise Texture in Backward Masking

Visual-noise masks composed of randomly scattered letter fragments disrupt the recognition of briefly viewed letter arrays when the masks are presented immediately after these arrays. According to the feature-similarity assumption, maximal disruption should occur when the target letters and the letter fragments comprising the mask share a common stroke width. This assumption was disconfirmed in the present experiment. Two-, four-, and six-letter arrays, whose stroke width subtended a visual angle of 3·8′, were viewed tachistoscopically in conjunction with visual-noise masks whose letter fragments subtended visual angles of 1·2′, 1·9′, 3·8′, 6·6′, or 9·7′. The most effective mask for all three letter-array lengths was the one composed of letter fragments subtending 6·6′. The results, together with those of previous experiments, are accounted for in terms of a neural line-detecting mechanism.