Purchase this article with an account.
Daryl Fougnie, Christopher L. Asplund, Tristan J. Watkins, René Marois; Object features limit the precision of working memory. Journal of Vision 2010;10(7):741. doi: 10.1167/10.7.741.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
An influential theory (Luck & Vogel, 1997) suggests that objects, rather than individual object features, are the fundamental units that limit our capacity to temporarily store visual information. This conclusion was drawn from paradigms in which the observer must detect whether a change occurred between a sample and a probe array when the arrays are separated by a short retention interval. Such ‘change detection’ paradigms reveal that increasing the number of objects, but not the number of distinct features, affects working memory performance (Luck & Vogel, 1997; Olson & Jiang, 2002). Using instead a paradigm that independently estimates the number and precision of items stored in working memory (Zhang & Luck, 2008), here we show that the storage of object features is indeed costly. We collected estimates of the precision and guess rate of working memory responses as participants had to remember the color, orientation, or both the color and orientation of isosceles triangles. We found that while the quantity of stored objects is largely unaffected by increasing the number of features per object (no change in guess rate), the fidelity of these representations dramatically decreased. Moreover, selective costs in precision depended on multiple features being contained within the same objects, as effects on both guess rate and fidelity were obtained when the orientation and color features were presented in distinct objects. Thus, in addition to providing evidence against cost-free conjunctions, our results demonstrate that storage of objects and features both limit visual working memory capacity. We argue that previous reports of cost-free conjunctions were due to the insensitivity of the tasks to changes in representational precision. Consistent with this interpretation, we found, using a change detection task, that manipulations of feature load do affect performance when the task places demands on the precision of the stored visual representations.
This PDF is available to Subscribers Only