Purchase this article with an account.
Qingleng Tan, Kazuhisa Shibata, Yuka Sasaki, Takeo Watanabe; Explaining anterograde and retrograde interference in visual perceptual learning by a limited plasticity resource model. Journal of Vision 2015;15(12):31. doi: 10.1167/15.12.31.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Visual Perceptual Learning (VPL) refers to a long-term enhancement in visual task performance as a result of visual experience (Sasaki et al., 2009). To understand the mechanism of VPL, it is crucial to examine how VPL is temporally developed. One key phenomenon is interference, which occurs when training on a first task is followed by training on a second and competing task within a one-hour interval. Disruption of VPL of the first task by training on the second task is called retrograde interference (Seitz et al., 2005, PNAS), whereas disruption of VPL of the second task by training on the first task is called anterograde interference (Yotsumoto et al., 2009, Vis Res). Interference suggests that the state of VPL is fragile immediately after training. However, there is no model that successfully explains both retrograde and anterograde interference. Here, we built a Hebbian learning model (Δωij=α*xi*yj, where the changes of weight equal to a learning rate α times the input xi times response yj and α=exp(-time2/σ2), where the learning rate decays with time) in which there is a limited plasticity resource for VPL within a certain time window (Σi,jωij≤c). The model indicates that successive learning competes for the limited plasticity resource. The result of simulation indicates that the model can explain recent psychophysical and brain imaging results. First, short-period training leads to retrograde interference, whereas long-period training leads to anterograde interference. Second, the concentration of glutamate divided by the concentration of GABA in the primary visual cortex, termed the E(excitatory signal)/I(inhibitory signal) ratio increased after the short-period training, whereas it decreased after the longer-period training. These results indicate the current model can explain how the two types of interference occurs and would be useful for a better understanding of time-course development of VPL.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only