Abstract
A common theoretical framework for visual short-term memory (VSTM) is one in which capacity is limited by the fixed number of items that can be stored in a high-level store (usually estimated to be approximately 4). However, this type of high threshold model (HTM) has been shown to contain problematic assumptions, and previously we have shown VSTM for color can be better described using a signal detection (SDT) approach. Here we describe a series of six experiments designed to test the generality of this approach using two unidimensional features: spatial frequency (SF) and orientation. Two arrays were sequentially presented composed of multiple gabors, arranged on an imaginary circle. Each array as presented for 100 ms, with an ISI of 1500 ms. The phase of the gabors was randomized between arrays. Each feature involved two separate experiments, one in which set size was changed and target number kept constant, and another in which set size was constant and target number changed. In the orientation experiments, orientations in the first display were chosen randomly; a change could be either pi/2 or pi/4 rotation. In the SF experiments, the frequencies of gabors were randomly selected from three fixed values. Observers were asked to report in a yes/no manner whether there was a change in the appropriate dimension between displays, and rate their confidence in this response on a 1–4 scale. Only a max model was shown to provide a good fit for the ROCs for the 4 experiments. Overall our results suggest that features are encoded in parallel within VSTM, with no obvious limit to the number of items that can be encoded.
WJM is supported by the Swartz Foundation.