Purchase this article with an account.
Hongsup Shin, Ronald Van den Berg, Wei Ji Ma; Visual short-term memory resource is not shared among features. Journal of Vision 2013;13(9):458. doi: 10.1167/13.9.458.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
There is an ongoing debate on whether the basic units of visual short-term memory (VSTM) are objects or features. The traditional way of framing this debate is: if capacity is four and each item has two features, can observers remember four or only two items? The problem here is that discrete-capacity models of VSTM have been discredited. According to the modern view of VSTM, memory resource is continuous, as well as variable across items and trials. The more resource an item receives, the higher is its mnemonic precision. In this new view, the question of how resource is distributed over features remains unaddressed. To examine whether VSTM resource is shared among features in the context of the variable-precision model, we used a change localization paradigm. On each trial, human subjects (n=8) briefly viewed four ellipses with random orientations and colors. After a 1-second delay, a second display containing four ellipses appeared, exactly one of which had changed. In the one-feature conditions, the change occurred always in the same feature – either orientation or color, depending on the experimental session. In the two-feature condition, on every trial, the change occurred randomly in either feature. Observers reported the location of the change. We tested two optimal-observer models, which differed in their resource allocation. In the independent-resource model, mean mnemonic precision for a given feature was identical between the one-feature and two-feature conditions. In the shared-resource model, the mean precision for a given feature was a proportion of the mean precision in the corresponding one-feature condition. We found that the independent-resource model better explained subject behavior; the log likelihood difference was –75.1±11.9 (mean±SEM). This suggests that the mnemonic precision of a feature is not affected by the number of features, and thus that VSTM resource is not shared among features.
Meeting abstract presented at VSS 2013
This PDF is available to Subscribers Only