Title: Arousal and emotional valence interact in written word recognition
Abstract: AbstractBehavioural and neurophysiological studies reveal a prioritisation for emotional material during different cognitive tasks. Although emotion comprises two dimensions, i.e., valence and arousal, previous research has mostly focused on the former. This study aimed to investigate the effects of valence and arousal on lexical decision (LD) by manipulating both dimensions, while controlling correlated psycholinguistic variables (e.g., word length, frequency, imageability). Results showed that valence and arousal affect word recognition in an interactive way: LD latencies are slower for positive high-arousal and negative low-arousal words compared to positive low-arousal and negative high-arousal words, in line with an approach-withdrawal tendency model. Furthermore, principal component and regression analyses revealed a unique contribution of a cluster of emotion variables, independent of lexico-semantic variables, to explaining LD latencies. We conclude that emotional valence and arousal both need to be taken into account in studies of word processing as they show an interactive relationship.Keywords: word recognitionvalencearousalemotionlexical decisiongender AcknowledgementsF.M.M.C. would like to thank Cristina Burani for her advice on the data analysis. This work is part of F.M.M.C.'s doctoral dissertation, funded by a Graduate Teaching Assistant scholarship from the University of Sussex, UK. Parts of this study were presented at the conferences of the Architecture and Mechanisms of Language Processing in 2010 as well as at the meetings of the British Psychology Society and the Society for Psychophysiological Research in 2009.Notes1. Sixty-nine participants (64 women), aged 18–34 years (M = 19.64; SD = 2.19) were instructed to rate "how much does each of the following words describe yourself" on a scale from 1 (not at all) to 7 (very much). All 150 words and some fillers were rated.2. Ratings for 50 more items were additionally collected and then included in the PCA because increasing the number of items is beneficial for this analysis.3. T-values and degrees of freedom for non-homogeneous variance are reported.4. When imageability is not controlled in the analyses, a main effect of arousal (confounded with imageability) is obtained in the RTs (F1(1,42) = 29.41, p < .001; F2(1,99) = 8.33, p = .005), whereby highly arousing and imageable words are responded to faster, in line with the typical imageability effect (e.g., Bird, Franklin, & Howard, Citation2001; Paivio, Yuille, & Madigan, Citation1968). Please refer to Table 2 for the descriptive statistics. A significant interaction is found in the analysis by participant only (F1(1,42) = 5.75, p = .021; F2(1,99) = 1.25, ns) whereby the arousal/imageability effect is larger for negative words. No effect of valence is observed. Accuracy rates show significantly higher accuracy for positive words (F1(1,42) = 16.08, p < .001; F2(1,99) = 6.15, p = .015) and for highly arousing/imageable words only in the analysis by participant (F1(1,42) = 5.07, p = .030; F2(1,99) = 2.21, ns), but no interaction (F1(1,42) = 1.02, ns; F2(1,99) = 0.43, ns). Overall, these results show typical imageability effects and no apparent influence of imageability on valence, as previously reported by Kanske and Kotz (Citation2007). We therefore conclude that the control of this variable in the main analyses helped us to partial out possibly confounding effects without hiding any interesting combined effects between emotional variables and imageability.5. Self-reference could not be included in the PCA as ratings for this variable were collected post-hoc only for the 150 items used in the study. Therefore, we have no values for the 350 items on which the PCA was based.