Effect of surface characteristics and style of production on naming and verification of pictorial stimuli.

Theories of object recognition that are based purely on part decomposition do not take into account the role of textural, shading, and color information, nor do they differentiate between stylistic factors in the preparation of line-drawn pictorial stimuli. To investigate these factors, naming and verification experiments were performed using line drawings, monochrome photographs, and color photographs of common objects. For line drawings, it was shown that line width, exposure, and contrast affected naming latency, which increased for lines of narrow width and extremes of exposure. Naming latencies were compared for objects drawn by a professional artist, with varying degrees of surface detail, and objects produced by a computer-aided design (CAD) system, with no surface detail. The mean naming latencies for the artist set were faster than for the CAD set, though not significantly, with a significant degree of object correlation being observed. However, in certain cases there were significant differences between objects. These were investigated in a further experiment in which subsets with common properties of present or absent surface detail were selected from the artist-drawn stimuli. It was found that the presence of surface features resulted in lower response latencies even for those objects that intuitively could be recognized by parts alone. The time to name photographic and line-drawn stimuli was compared, and a progressive decrease in naming latency from line to monochrome to color stimuli was observed. In a verification task, no significant advantage for color or monochrome photographs over line drawings was found, either when comparing stimuli of equivalent or of different mode. However, there was a tendency for the comparison of different modes to take longer than the comparison of same modes. The results are discussed in terms of theories of human visual processing and cognitive and computational models of object recognition.