Academic journal article Canadian Journal of Experimental Psychology

Parameterizing the Attentional Blink Effect

Academic journal article Canadian Journal of Experimental Psychology

Parameterizing the Attentional Blink Effect

Article excerpt


The attentional blink effect (AB) is used to examine the limits of attention in dual-task paradigms. However, since the effect is nonlinear, it is sometimes difficult to characterize the results. Furthermore, it is difficult to assess the significance of the effect between groups because the results are highly variable both within and across subjects. In this paper, we propose a method to quantify four characteristics of the AB curve: the minimum performance, the amplitude between the minimum and the asymptotic performance, the amount of Lag-1 sparing, and the width of the effect. The method, based on curve fitting, allows easier comparisons of the results across experiments, can test only one characteristic at a time, and yields more powerful statistical tests.

(ProQuest Information and Learning: ... denotes formulae omitted.)

The attentional blink effect is a decrease in report accuracy for the second of two targets (T^sub 2^) presented in rapid succession (Raymond, Shapiro, & Arnell, 1992; Shiffrin & Schneider, 1977). The effect is usually maximal for T^sub 2^ presented about 200-300 ms after the first target (T^sub 1^) and often lasts up to 300 ms. Typical results are illustrated in Figure 1. The effect has been shown to depend on a number of factors, such as the duration of central processing of the first target (Jolicoeur, 1999), the similarity relationship between targets and distractors (Chun & Potter, 1995) or the presence/absence of a mask following the first target (Raymond et al., 1992; Seiffert & DiLollo, 1997) or the second target (Giesbrecht & DiLollo, 1998).

In many investigations, the influence of a variable of theoretical interest (e.g., duration of central processing in Task 1) is represented by manipulating a factor (call it Cond) crossed with the factor Lag (the temporal separation between the two targets) in a factorial experimental design. Imagine a simple example in which the factor Cond has two levels (Cond^sub 1^ and Cond^sub 2^) and Lag experiments are submitted to a standard ANOVA in which Lag and Cond are treated as crossed factors (sometimes both within-subjects). In such situations, there could be a main effect of the factor Cond, one curve being significantly and systematically lower than the second. However, it is more frequent that a significant interaction between the factor of theoretical interest and Lag (Cond × Lag) occurs.

Figure 2 shows some possible results. For example, the position of the curve could change along the horizontal axis, resulting in a pattern of results labeled Lag1 sparing (Visser, Bischof, & DiLollo, 1999). This difference could indicate how quickly processing of T^sub 1^ begins to have an effect on processing of T^sub 2^ (Figure 2, top left panel). The width of the depressed region could vary, reflecting the duration of the effect (Figure 2, bottom left). The severity of the AB effect could vary, which would be reflected in the difference between the asymptotic performance at long lags and the deepest portion of the curve at shorter lags (Figure 2, top right). Finally, the overall difficulty of the task could change (Figure 2, bottom right). Three of these types of changes (Lag-1 sparing, width, amplitude) generate a Cond × Lag interaction in an ANOVA.

The difficulty with the ANOVA technique is that it does not, in itself, allow investigators to characterize in which ways two AB curves differ. In this article, we describe a method for summarizing and analyzing the results of AB experiments that allows investigators to distinguish Lag-1 sparing differences, amplitude differences, width differences, and minimum accuracy differences. The approach is designed to capture aspects of the data that are most likely to be of theoretical importance. In this sense, the approach is driven from an interest in advancing theoretical research on the AB. On the other hand, the specifics of the present approach are atheoretical, in the sense that it does not ascribe any particular cause to the changes observed in parameter variations. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.