Abstract
The fidelity of visual short-term memory (VSTM) representations become compromised with time. A novel technique of adaptive noise estimation is shown to be a useful methodological tool for measuring the time course of memory decay. Differences in noise equivalence estimates between retrieval type suggest task-dependent effects of time on memoranda. Method: Sixteen observers memorized the orientation of a sample Gabor. After a short (1 sec) or long (4 sec) delay, a second test Gabor was presented. For recall trials, observers reproduced the sample orientation by rotating the test Gabor. For discrimination trials, observers indicated a binary same/different response. Adaptive methods estimated discrimination and recall error psychometric functions. Filtered noise was adaptively added to the sample Gabor on short-delay trials to produce equivalent performance to that measured on low-noise, long-duration trials. Results: Increased storage duration impaired discrimination (-20% accuracy) and recall error (+8% magnitude). Performance under adaptive noise successfully tracked that of low-noise, long-delay conditions. A mixed-effects regression analysis provided a common currency for translating between retention delay and stimulus noise. Discrimination noise equivalence, the amount of stimulus noise required to produce impairments similar to long-delays, was estimated to have a mean of 0.39, 90% CI [.18, .68]. For recall, mean noise equivalence was 0.16, 90% CI [.03, .37]. Conclusion: The equivalent noise method for estimating VSTM decay provides a novel technique for measuring the time course of memory fidelity. Trending differences in equivalence across tasks suggests that more noise is necessary to offset the effect of retention duration for discrimination compared to recall. A likely interpretation is that decay is non-linear, such that delay-induced effects are larger for the brief response time (RT) profile of discrimination (mean RT = 924 msec) versus recall (mean RT = 2396 msec).
Meeting abstract presented at VSS 2015