SPATIAL FREQUENCY TUNED COVARIANCE CHANNELS UNDERLYING SCOTOPIC CONTRAST SENSITIVITY

摘要

The processes underlying photopic contrast sensitivity functions (CSFs) have been modeled in terms of multiple channels selective for spatial frequency (De Valois & De Valois, 1988). The lowest frequency channel obtained foveally using stational sinusoidal gratings typically has its peak sensitivity near 1 c/deg (Greenlee et al. 1988; Peterzell & Teller, 1996, 2000; Tolhurst, 1973). Less is known about the processes underlying scotopic contrast sensitivity. The channels underlying scotopic and photopic vision may differ considerably. Hess and Howell (1988) demonstrated that contrast sensitivity peaks near 0.2 c/deg when stimuli are presented at scotopic luminances. This low-frequency peak cannot be modeled using only a bandpass channel that peaks near 1 c/deg. Hence, the authors concluded that several spatial frequency channels exist at very low spatial frequencies but may operate at scotopic luminances only (or, similarly, the peak of a channel might shift to lower spatial frequencies at low light levels, due, perhaps, to a reduction of the influence of the surrounds of receptive fields). Greenlee et al. determined that the lowest adaptable frequency channel obtained using scotopic stationary gratings occurred well below 1 c/deg (as measured in rod monochromats). They concluded that rod monochromats differed from normals. Equally likely from their results, however, is the possibility that scotopic vision, unlike photopic vision, contains multiple spatial channels below 1 c/deg. Over the last 15 years, about 20 psychophysical and electrophysiological studies with adults and infants have examined normal individual differences, using statistical covariance analyses in order to quantify the number of spatiotemporal channels, and to measure channels' spatial and temporal frequency tuning (reviews: Peterzell & Teller, 1996, 2000; Peterzell et al. 2000). The paradigm uses simple detection data to assess the unadapted, unmasked visual system. It requires relatively few complex theoretical assumptions or theoretical structure to estimate the number of channels, and provides a direct estimate of channel tuning. It has revealed mechanisms that map well onto channels derived using masking and adaptation.