Cortical dynamics of lateral inhibition: Metacontrast masking.

Abstract

The dynamic properties of a neural network model of visual perception, called the boundary contour system, explain characteristics of metacontrast visual masking. Computer simulations of the model, with a single set of parameters, demonstrate that it accounts for 9 key properties of metacontrast masking: Metacontrast masking is strongest at positive stimulus onset asynchronies (SOAs); decreasing target luminance changes the shape of the masking curve; increasing target duration weakens masking; masking effects weaken with spatial separation; increasing mask duration leads to stronger masking at shorter SOAs; masking strength depends on the amount and distribution of contour in the mask; a second mask can disinhibit the masking of the target; such disinhibition depends on the SOA of the 2 masks; and such disinhibition depends on the spatial separation of the 2 masks. No other theory provides a unified explanation of these data sets. Additionally, the model suggests a new analysis of data related to the SOA law and makes several testable predictions.

Modeling spatial and temporal aspects of visual backward masking.

“…In accordance with previous experimental findings and modeling results (Bridgeman, 2001;Francis, 1997), the SOA yielding strongest masking is predicted to shift to lower values when the mask intensity is increased (Figure 8). For high intensity masks, strongest masking is found for an SOA of 0 ms, which means that the curve has become monotonic; that is, A-type masking occurs.…”

“…Common onset masking can be modeled by many of the existing feed-forward and lateral inhibition models of masking (Bischof & Di Lollo, 1995;Francis, 1997;Francis & Hermens, 2002) and by Growney (1978) and Growney et al (1977) experiments. ISI ϭ interstimulus interval.…”

“…Although there are several possible ways to solve the energy summation problem, such as the use of reset signals (Francis, 1997;Ö gmen, 1993) or gain control (Grossberg, 1982), they all Figure 12. For four pairs of masks, the difference in Fourier power spectra (⌬ ) was computed and compared with the psychophysical thresholds ( ).…”

Unconscious task set priming with phonological and semantic tasks

“…First of all, we were intent on making sure the participants really were unaware of the prime. Even if it is difficult to demonstrate statistically that visibility is exactly zero, it is known that metacontrast masking rarely produces complete masking at any level of SOA (Francis, 1997).…”

On-line simulations of models for backward masking

“…Anbar and Anbar (1982) elaborated a model of brightness perception to show that it accounted for some properties of backward masking. Francis (1997) showed that a neural model of cortical visual processing accounted for many properties of backward masking. Purushothaman, Ogmen, and Bedell (2000) showed that a recurrent neural network could produce oscillations that accounted for new data in backward masking.…”