Contrasting Predictions of Low- and High-Threshold Models for the Detection of Changing Visual Features

Abstract

Change blindness is the failure of observers to notice otherwise obvious changes to a visual scene when those changes are masked in some way (eg by blotches or a blanking ofthe screen). Typically, change blindness is taken as evidence that our representation of the visual world is capacity limited. The locus of this capacity limit is thought to be visual short-term memory (vSTM). The capacity of vSTM is usually estimated with a high-threshold model which assumes that each element in the stimulus array is either fully encoded or not encoded at all, and, furthermore, that false alarms can arise only by guessing, not by noise. Low-threshold models, by contrast, suggest that false alarms can arise by noise at the level of detection/discrimination and/or decision. In this study, we use a well-controlled stimulus display in which a single element changes over a blanking of the screen and contrast predictions from a popular high-threshold model of vSTM with the predictions of a low-threshold model (specifically, the sample-size model) of visual search and vSTM. The data were better predicted by the low-threshold model.

Detecting Unidentified Changes

“…Because we were able to show that this model did not provide an adequate account of our observers' data we were able to conclude that our observers sometimes had only partial information about the changes that occurred; that is, enough information to determine that a change had occurred but not enough to determine what has changed. Our analysis is therefore consistent with a previous study that shows that high threshold models do not provide an accurate account of change detection [28]. Our analysis goes beyond this previous study by showing that observers can sometimes detected changes without being able to identify or localize them.…”

“…Our model of observer behaviour could be described as a high threshold model in that it assumes that a change will never be detected if it does not occur [28]. False alarms are assumed to be due to guessing, not false detections.…”

Attribute Amnesia in the Auditory Domain

“…Hauser and Schawartz (2016) found that MTurk participants actually outperformed laboratory participants in their comprehension of instructions. Findings from laboratory studies tend to be replicated in MTurk samples (Horton et al., 2011; Huh et al., 2014; Paolacci et al., 2010) and the test–retest reliability of MTurk data is generally quite high (Burmester & Wallis, 2012). Kittur et al.…”

Characterizing the time course of decision-making in change detection.

“…Using a visual search task where participants searched for either one or two targets among three or two distractors, they found that this capacity limitation was well described by the sample-size relationship 1= ffiffi ffi 2 p for the double-target deficit. Burmester and Wallis (2012), like Wilken and Ma (2004), compared a low-threshold model with a high-threshold (i.e., "slots-based") account. Specifically, they were interested in testing a sample-size account (Palmer, 1990;Sewell et al, 2014;Shaw, 1980), which assumes, like signal-detection theories, that stimuli are represented by a normally distributed noisy strength signal, that observers set a threshold for responding that can be exceeded based on noise alone, and that noise increases with the number of items which must be encoded in order to make a decision.…”