Quantitative models of backward masking

Click on one of the buttons below to work with a model simulation.

This page provides simulations of several quantitative models of backward masking. Seven models are available:
  1. Dual-channel model: This model was described by Weisstein (1972). It hypothesizes two neural pathways, one for the target and one for the mask. Cross channel inhibition produces the masking effect.
  2. Recurrent inhibition model: This model was described by Bridgeman (1978). It uses a set of neural cells that recurrently sends signals back and forth amongst neighbors. Masking is measured as the average (across time) correlation of activity among cell activities when only a target is present and when both a target and mask are presented.
  3. Decaying trace model: This model was described by Anbar & Anbar (1982). They extended models of brightness perception and proved that u-shaped masking could occur in such models when the mask interacted with the decaying trace generated by the target.
  4. Efficient masking model: This model was described by Francis (2000, 2003). It is not a model as much as a method that can be used in a variety of models. The simulation here uses the system described by Francis (2003) to demonstrate the method.
  5. Reentrant processing model: This model was described by Di Lollo, Enns, & Rensink (2000). It was created to account for properties of object substitution.
  6. Simple mask-blocking: This model was described by Francis & Cho (to appear) as a more simple way of introducing the properties of mask-blocking. As described in Francis (2000), mask-blocking is a computational approach that is used by many models to account for the appearance of u-shaped masking functions. [Added 10 June 2004.]
  7. Perceptual retouch: This model was described in the appendix of Bachmann (1994). It is based upon non-specific signals from the thalamus influencing specific cortical signals in cortex. [Added 16 June 2004.]
To fully understand how to use and interpret the simulations you will need to understand the properties of each model. You are advised to read the original descriptions of the models. Also see Francis (2000), who discussed and analyzed the dual-channel model, the recurrent inhibition model, the decaying trace model, and the efficient masking model.

In each simulation you can select an independent variable and its range of values. The independent variable will then define the x-axis, while the y-axis will always correspond to the model's value of the target percept. For every model, larger values on the y-axis indicate stronger target percepts (e.g., brighter percepts, better detection, better discrimination,...).

Each simulation loads with a default configuration of parameters and stimulus properties that corresponds to what was a standard design for each model. Most models were interested in accounting for the u-shaped masking function that appears when the timing between the target and mask is varied. For most models, the parameters and stimulus properties are set to produce a typical u-shaped masking function. The one exception is the reentrant processing model. This model was designed to investigate properties of mask duration and attention. The default configuration sets the independent variable as mask duration.

It should be noted that this is not an exhaustive list of models of backward masking. Francis (1997) and Purushothaman, Ogmen, and Bedell (2000) describe two neural network models with interesting properties. However, the computation involved in those models is much more substantial than the models presented here. For these more complicated models, a run of the simulation may take hours or days. Researchers interested in those models are advised to create their own simulation, or to contact the authors.

Additional details about the simulations can be found in Francis (2003b). Source code for the programs can be downloaded as a compressed file (OnLineMaskingSims.tgz, 43KB).


This material is based upon work supported by the National Science Foundation under Grant No. 0108905. Development of the perceptual retouch simulation was supported by a NATO Collaborative Linkage Grant (Michael Herzog, PI).


Last updated 16 June 2004