Abstract
Visual working memory (VWM) has been extensively investigated using a change-detection paradigm and a High Threshold (HT) model in which stimuli are encoded and remembered perfectly up to a capacity limitation, and observers guess otherwise. However, the visual system does not function flawlessly and is limited by various sources of internal and external noise. Wilken and Ma (2004) used a signal detection theory (SDT) approach to model VWM performance at high contrast, and suggested that performance was limited by internal noise that increases with display size. Their experiment provided only modest constraints on the estimation of internal noises. The external noise approach, i.e., joint manipulations of the amount of external noise (white Gaussian random noise) and contrast of signal stimuli (oriented Gabor patches), provides a more stringent test of Wilken and Ma's claims and sufficient constraints on observer models. Observers detected a 45° change in the orientation between sample and test displays with 1, 2 or 4 randomly oriented Gabor objects in either clear or noisy displays. Contrast psychometric functions were measured and modeled with the perceptual template model (PTM) of the observer (Lu & Dosher, 2008). Noise terms associated with stimulus enhancement and external noise exclusion (often associated with attention) varied with display size in the VWM task. Our results extend and provide explicit measurement for Wilken and Ma's (2004) claim that VWM performance is limited by internal noises that increase with display size; the finding that the impact external noise increases with display size suggests that stimulus encoding also deteriorates with memory load. We also consider a HT model. The PTM observer model provides a natural context within which to understand the functioning of VWM in a broad range of stimulus conditions.