Is the emotional modulation of the attentional blink driven by response bias?

Abstract

Several studies have shown that the attentional blink (AB; Raymond, Shapiro, & Arnell, 1992) is diminished for highly arousing T2 stimuli (e.g., Anderson, 2005). Whereas this effect is most often interpreted as evidence for a more efficient processing of arousing information, it could be due also to a bias to report more arousing stimuli than neutral stimuli. We introduce a paradigm that allows one to control for such a response bias. Using this paradigm, we obtained evidence that the diminished AB for taboo words cannot be explained by a response bias. This supports the idea that the emotional modulation of the AB is caused by attentional processes.

Keywords:

• Attentional blink
• Arousal
• Emotion
• Attention
• Response bias

Acknowledgements

Helen Tibboel is a Research Assistant of the Flemish Research Foundation (FWO – Vlaanderen). The preparation of this manuscript was supported by Grant BOF/GOA2006/001 of Ghent University

Notes

¹Scores of 0 and 1 needed to be adjusted because these values cannot be normalised. When the proportion of hits for a specific stimulus was 0, this stimulus was never identified correctly. When the proportion of false alarms was 0, this stimulus was never incorrectly chosen. When the proportion of hits was 1, this stimulus was always identified correctly. When the proportion of false alarms was 1, this stimulus was always incorrectly chosen when in fact another stimulus was presented. For the false alarms there were no scores of 1. At Lag 2, the percentage of scores of 0 was 13.22 for the neutral words (SD=9.91), and 26.44 (SD=7.12) for the taboo words. At Lag 8, the percentage of 0 scores was considerably higher: 78.16% for the neutral words (SD=4.18) and 79.89% for the taboo words (SD=5.93). For the hits, the percentage of scores of 1 at Lag 2 was 12.64 (SD=12.08) for neutral words, and 19.54 for taboo words (SD=13.21). At Lag 8, this percentage was higher: 73.56 (SD=8.06) and 84.48 (SD=3.62) respectively. The percentage of 0 scores at Lag 2 was 14.37 (SD=6.69) for the neutral words, and 5.17 (SD=1.89) for the taboo words. There were no scores of 0 at Lag 8. Note that both scores of 0 and scores of 1 are valid data points, that were replaced solely for the purpose of normalisation. As Wickens (2002) pointed out, it is important to realise that the addition or subtraction of a value like ¹/₁₀₀₀ is arbitrary. But as we substitute these values for both the neutral and the taboo hits and false alarms, the effect of the substitutions is equal for both categories and thus can not explain our effects. Furthermore, we also performed the same ANOVAs on the data after adding (subtracting) ¹/₁₀₀ to (from) the 0 (1) scores. Results were virtually the same.

²Our results show that for both neutral and taboo T2 stimuli, participants are biased to select an alternative other than the target. We must note that log β is usually used in tasks where there are only two response alternatives. However, in our study there were 12 response alternatives. This decreases the proportion of false alarms per stimulus. If a participant makes an incorrect response on a two-alternative forced choice task, the one alternative response is chosen. In our task, there is one correct and 11 incorrect answers. If a participant is incorrect, and each alternative response has an equal chance of being chosen, there is a false alarm for each stimulus in ¹/₁₁ of trials. Because the low number of false alarms distorts log β, we will focus on the difference in bias between neutral and taboo T2 stimuli instead of focusing on absolute values.

³An ANOVA with block as an additional within-subjects factor yielded the same results.

⁴Note that an “apple” false alarm could occur both when a neutral word was presented (i.e., “delta”) and when a taboo word (i.e., “anal”) was presented. In these analyses we do not make a distinction between these two situations. However, we analysed the false alarms also with an ANOVA with three within-subjects factors: Response Type (neutral or taboo), Target Type (neutral or taboo), and Lag (2 or 8). We found only a main effect of Target Type, F(1, 26)=21.34, p<.001, showing that there were more false alarms when a neutral target was presented. This mirrors the main effect of T2 type in the analyses of hits (i.e., more incorrect answers when a neutral T2 was presented). Target Type did not have a main effect and did not affect any of the higher order interactions, Fs<1.