Individual visual short-term memory capacity predicts the number of conjunction errors in Treisman's illusory conjunction task

Abstract

Listen

Translate article

Treisman & Schmidt (1982) used a divided attention to ask to investigate whether attention is required to bind the features of visual objects (letter identity and letter color). The term “illusory conjunctions” refers to observers' responses where two features belonging to different objects in the scene are reported as belonging to the same object. Several experiments in the literature have studied whether the phenomenon of illusory conjunctions in Treisman & Schmidt's task is purely a visual phenomenon or whether it is influenced by the memory demands of the task. Here, we use a different approach to understand illusory conjunctions. We measured each observer's visual short term memory (VSTM) capacity (K) using a simply color change detection task, and we then look for correlations between K and the different types of errors participants produced in the illusory conjunction task. The illusory conjunction task replicated Treisman & Schmidt's original pattern of results, with higher illusory conjunction errors than feature intrusion errors (e.g., reporting correctly a letter identity conjoined with a color that was not presented in the display). We expected to find a negative correlation between K and intrusion errors (with lower capacity leading to more frequent guessing), yet, somewhat counter-intuitively, there was no correlation between these measures. However, K did significantly positively correlate with illusory conjunctions, r = 0.6, suggesting that VSTM had access to features present in the display, yet not to the precise form/color conjunctions they were originally presented in. Our results support the view that VSTM can, under conditions of divided attention, act as a non-differentiated feature storage (remembering visual features without knowing what objects they belonged to) and that illusory conjunctions, at least in the Treisman & Schmidt methodology, are strongly dependent on this form of VSTM storage.