Figure 2.

The Binary Sampling Model. According to McGonigle and Chalmers (1977), an animal attempts to solve the BD pair as if it were either pair BC, CD, or BD. B+C− is a training pair and so it will select item B. C+D− is also a training pair where the animal is taught to select C. Unfortunately, because the animal has been presented with pair BD, there is no item C being displayed and hence no item C to select. But if the animal were trying to solve the BD pair as if it was item C+D− being displayed, it also learned to avoid D, which is what the animal does, and again selects item B. Finally, if the animal attempts to solve the BD pair as if it were BD, it has received no training with these two stimuli presented together, and randomly chooses between them. The net effect is the animal will select B 83% of the time and D 17% of the time, which happens to be very close the performance levels many animals achieved with the BD test.