Affect and Attention 1 Affective Pictures Influence Holding but not Drawing of Attention in a Modified Dot-Probe Paradigm Ulrich Schimmack University of Toronto, Mississauga RUNNING HEAD: Affect and Attention February 2002 about 4,000 words Author’s Note. Correspondence concerning this article should be addressed to Ulrich Schimmack, Department of Psychology, UTM, 3359 Mississauga Road North, Mississauga, Ontario, L5L 1C6, Canada, email: uli.schimmack@utoronto.ca Affect and Attention 2 Abstract The influence of affective picture on drawing and holding of attention in a modified dot-probe paradigm was examined. Pairs of pictures combining seven levels of valence ranging form strong displeasure to strong pleasure were presented for 200 ms in two out of four possible locations. The pictures were replaced by two circles. Participants had to determine whether a target circle was broken once or twice as fast as possible. Affective pictures in the location of the target circle had no influence on response latencies. Strong unpleasant pictures in the non-target location produced slower response latencies than neutral pictures in the target location. The results suggest that strong unpleasant pictures hold attention, but that affective pictures do not draw attention. Affect and Attention 3 Affective Pictures Influence Holding but not Drawing of Attention in a Modified Dot-Probe Paradigm The influence of affective stimuli on attention has been documented in numerous studies (e.g., Fox, Russo, Bowles, & Dutton, 2001; MacLeod, Mathews, & Tata, 1986; Mogg, McNamara, Powys, Rawlinson, Seiffer, & Bradley, 2000; Rozin & Royzman, 2001). The existing evidence demonstrates that affective stimuli attract more attention than neutral stimuli and that unpleasant stimuli tend to attract more attention than pleasant stimuli. This article presents the results of a modified dot-probe study with affective pictures as stimuli. The study contributes to research on affective influences on attention in three ways. First, most studies have used words as stimuli. In recent years, a few studies have examined the influence of facial expressions on attention (Bradley, Mogg, Falla, & Hamilton, 1998). However, very few studies have examined the influence of affective pictures on attention (Mogg et al., 2000). De Houwer and Hermans (1994) demonstrated that affective pictures have a stronger influence on attention than affective words. Research with pictures also has a higher ecological validity than research with words. Hence, more research with affective pictures is needed to understand how affective stimuli influence attention in everyday life. The second contribution of this study is the systematic manipulation of stimulus intensity. The only previous study found a negativity bias (i.e., unpleasant words attracted more attention than pleasant words). The negativity bias was not moderated by the extremity of the unpleasant words (Pratto & John, 1991). The present study examined whether the negativity bias would generalize to affective pictures as stimuli, and whether it would be consistent across different levels of intensity. Affect and Attention 4 The third contribution of this study concerns the differentiation between two types of attention (Fox et al., 2001). Over time, attention shifts between different stimuli. Affect can influence drawing and holding of attention (Fox et al., 2001). If affective stimuli draw attention, they facilitate responses to other stimuli in the location of the affective stimulus. If affective stimuli hold attention, they impede responses to other stimuli in other locations than the affective stimulus. Most studies of affect and attention have used one of two tasks, namely the Emotional Stroop task and the dot-probe paradigm. As noted by Fox et al. (2001), the Emotional Stroop task cannot differentiate between drawing and holding of attention because a single stimulus is presented in the focus of visual attention. Hence, the location of attention is predetermined by the task and it is impossible to determine whether affective stimuli draw attention to their location. The dot-probe paradigm is better suited for the examination of drawing and holding of attention. In this paradigm, an affective stimulus is presented with a neutral stimulus for a relatively short time (500ms). Afterwards, one of the two stimuli is replaced by a dot. Participants have to press a key as soon as they detect the dot. Participants are able to detect the dot faster, if they are already attending to the location where the dot appears (e.g., Mogg et al., 2000). Unfortunately, most studies employing the dot-probe paradigm could not separate drawing and holding of attention because they did not include a control condition with neutral stimuli. Without this condition, faster responses to dots in the location of the affective stimulus can be due to drawing or holding of attention, or both. That is, the affective stimulus may draw attention to its location so that participants already attend to the location of the dot. Alternatively, participants may randomly attend to either one of the two stimuli. When attention is directed to the affective stimulus, it holds attention in that location, which makes it harder to detect the dot in the other location. With a control condition, it would be possible Affect and Attention 5 to distinguish these two processes. If affective stimuli draw attention, they should facilitate responses to dots in the target location in comparison to the neutral control condition. If affective stimuli hold attention, they should delay responses to non-affective stimuli in the opposite location compared to response latencies in the control condition. Fox et al. (2001) used a slightly different approach for the same purpose. The authors modified the exogenous cuing paradigm (Posner, Inhoff, Friedrich, & Cohen, 1987). In this paradigm, participants also have to detect a target (a dot or a square), which is displayed in one of two locations. Before the appearance of the dot, a cue appears either in the target location or in the opposite location. On valid trials the target appears in the location of the cue (80%); on invalid trials the target appears in the opposite location (20%). Neutral cues (a flashing light) lead to faster responses on valid trials and slower responses on invalid trials. Fox et al. (2001) modified this paradigm by using three different types of cues (neutral, positive, threatening). If affective stimuli influence drawing of attention, affective cues should produce faster responses on valid trials than neutral cues. If affective stimuli influence holding of attention, they should produce slower responses on invalid trials than neutral cues. The authors conducted a series of studies with words and angry facial expressions as stimuli. The general result was that none of the affective stimuli influenced drawing of attention. Threatening stimuli influenced holding of attention for high-anxious participants but not for low anxious participants. Yiend and Mathews (2001) used the exogenous cuing paradigm with gray-scaled affective pictures, which were either threatening (mutilations, dangerous animals) or not threatening (landscapes). The authors found the same pattern of results. That is, threatening pictures did not produce faster responses on valid trails than non-threatening pictures. Threatening pictures produced slower responses on invalid trails than non-threatening pictures, but this Affect and Attention 6 effect was only observed for high-anxious individuals. These studies suggest that affective stimuli influence holding rather than drawing of attention. The major difference between the affective exogenous cue paradigm (Fox et al., 2001) and the dot-probe paradigm with a control condition is the presentation of one or two pictures before the appearance of the target. I used the dot-probe paradigm because the presentation of two stimuli has two advantages over paradigms with a single stimulus. First, the presentation of two stimuli has higher ecological validity for situations, in which people have to detect affective stimuli in an array of multiple stimuli (e.g., see the snake between flowers, rocks, etc.). Second, the presentation of a single stimulus may attenuate effects of affective stimuli on drawing of attention (cf. Fox et al., 2001). The reason is that suddenly occurring stimuli already draw attention (Yantis, 1996). Hence, the appearance of a single stimulus may be sufficient to draw attention to that location independent of the stimulus’s affective valence. However, if two stimuli appear suddenly on a screen, sudden onset cannot determine which location draws attention first. Affective stimuli may be more likely to draw attention to their location in this situation. To summarize, this study examined the influence of affective pictures that varied systematically in valence and intensity on the drawing and holding of attention in a modified dot-probe paradigm. I expected the following results. (a) Affective pictures should influence the holding of attention and not the drawing of attention. This prediction is based on the only study that differentiated drawing and holding of attention (Fox et al., 2001). Although the authors did not study affective pictures as stimuli and used a slightly different paradigm, it is most parsimonious to assume that their findings generalize to other stimuli and a slightly different task. Affect and Attention 7 (b) Unpleasant pictures have a stronger influence on attention than pleasant pictures because negative stimuli have shown to attract more attention than positive stimuli (Pratto & John, 1991; Rozin et al., 2001). (c) Negativity dominance should be modified by affect intensity. This prediction is based on the only study with affective pictures that manipulated the intensity of unpleasant pictures, and found that strong unpleasant pictures attracted more attention than mild unpleasant pictures (Mogg et al., 2000). Method Participants Forty-three Introductory psychology students (29 female, 14 male) at the University of Toronto, Mississauga, participated in this study for course credit. Materials The pictures for this study were taken from the International Affective Picture System (IAPS; Lang, Bradley, & Cuthbert, 1995) and the Internet. The pictures were selected from pilot studies with participants from the same population to represent seven levels of valence (on a 7-point bipolar scale ranging from –3 to +3): strong displeasure (M = -1.91), moderate displeasure (M = -1.25), mild displeasure (M = -0.69), neutral (M = 0.10), mild pleasure (M = 0.60, moderate pleasure (M = 1.20), and strong pleasure (M = 1.91). The stimulus set did not include pictures with erotic content or mutilations. The picture set included 98 pictures; 14 pictures for each level of valence, which allowed the complete pairing of all levels of intensity without repeated presentations of the same picture. The size of the pictures was 180 x 270 pixels on a screen with a 1024 x 768 pixel resolution. Procedure The design of the study was a complete 7 x 7 design, in which each level of valence was paired with all levels of valence. All participants first completed a Affect and Attention 8 different task, in which they saw the same pictures and rated their affective reactions to the pictures. The results of this task are not relevant for the present study. However, it is important to note that participants were at least vaguely familiar with the pictures. Before the dot-probe task, participants received extensive instructions about the task. They clicked a button on the screen to start the dot-probe task. The computer first presented four practice trials, which were the same for all participants and used eight pictures that were not included in the experimental picture set. Then the computer presented two repeated blocks of the 49 experimental trials. In each block, individual pictures from the seven sets were randomly assigned to the 49 pairs. As a result, different participants saw different combinations of pictures and each participant saw different combinations of the pictures in the two blocks. Each trial started with a fixation cue (a red square) in the center of the screen. The fixation cue was shown for 250ms. Afterwards, two pictures appeared in two of four locations (upper left, upper right, lower left, lower right). The pictures remained on the screen for 200ms. Afterwards, the pictures were replaced by two circles. One of the two circles (the target circle) was broken either once (on the top) or twice (on the top and the bottom). The circles remained on the screen for 2000ms or until participants responded. Participants had to indicate how often the target circle was broken. They responded by pressing one of two mouse buttons (left button = once broken; right button = twice broken). This modification of the dot-probe task was used for two reasons (see Bradley et al., 1998). The presentation of two stimuli was used to inhibit the influence of sudden-onset on attention. A single stimulus that suddenly appears automatically attracts attention (Yantis, 1996). Hence, people can quickly detect targets even if their attention was not in that location. By presenting the target circle with a non-target circle, the target does not automatically attract attention due to its sudden onset. The broken circle was used to ensure that Affect and Attention 9 participants had to attend to the target stimulus to provide a correct response. I hoped that this task would produce larger effects than the typical dot-probe paradigm, which often produces very small differences in response latencies due to affect (MacLeod et al., 1986). The interval of 200ms was chosen to allow for enough time for an eye-movement to the location of an affective stimulus, although eye-movements in the dot-probe paradigm are rare (cf. Fox et al., 2001). This interval also happened to be similar to the interval in Fox et al.’s (2001) studies (150ms). Results I used HLM5 (Raudenbush, Bryk, Cheong, & Congdon, 2000) to analyze the response latencies. HLM is useful for the investigation of unbalanced designs. Although the present design was balanced, missing data in response latencies render the design essentially unbalanced. HLM can estimate group averages from data with missing data, whereas Analysis of Variance requires a balanced data structure. This statistical approach was particularly necessary in the present study, which relied on relatively few trials. Participants responded correctly on 89% of the trials. Trials with incorrect responses were excluded from the analysis of the response latencies. The relatively low hit rates indicate that this task was more difficult than the mere detection of a probe (Fox et al., 2001). Extreme response latencies were eliminated following the guidelines by Fox et al. (2001). First, response latencies that were faster than 300ms (N = 4, 0.1%) and responses that were slower than 1500ms (N = 63, 1.7%) were eliminated. Next, response latencies that were 2.5 standard deviations slower or faster than individuals’ mean latencies were also excluded (N = 42, 1.1%). The final analysis is based on the response latencies of 86% of the trials. The response latencies were still slightly skewed (skewness = .47), which was eliminated by Affect and Attention 10 logarithmic transformation (skewness = .04). Hence, logarithmically transformed response latencies were used as the dependent variable. Type of picture in the target location and type of picture in the non-target location was dummy coded with neutral pictures as the contrast for all other types of pictures. In addition, block was included as a predictor. Table 1 shows the fixed effects with robust standard errors. The results reveal trivial effects for the Intercept (i.e., participants needed time to respond) and block (i.e., participants became faster over time). Importantly, there were no significant effects for type of picture in the target location. This finding supports the hypothesis that affective stimuli do not draw attention. However, type of picture in the non-target location produced a significant effect for strong unpleasant pictures. The positive sign of this effect reveals that participants needed more time to respond to the target circle when a strong unpleasant picture was presented in the non-target location than when a neutral picture was presented in the non-target location. Two effects showed a tendency towards significance, namely moderate unpleasant pictures in the non-target location, and strong pleasant pictures in the target location. The finding that strong negative pictures were the only pictures that produced a significant effect supports the second hypothesis of a negativity bias. The next model tested whether strong and moderate unpleasant pictures differed significantly from each other. This model constrained the gammas for these two picture types in the non-target location to be equal. The fit of the constrained model was not significantly worse than the fit of the unconstrained model, χ2 (df = 1) = 1.67, p = .19. However, a model that constrained the gammas of all levels of unpleasant pictures in the non-target location did not fit the data as well as the unconstrained model, χ2 (df = 2) = 6.30, p < .05. This finding supports the hypothesis that the influence of negative pictures on attention is moderated by the extremity of the negative pictures. Although the difference between strong and Affect and Attention 11 moderate pictures did not reach significance, the difference between strong and mild negative pictures was significant (see Figure 1 for the retransformed response latencies as a function of picture type in the non-target location). In a final model, I included an interaction term between block and the dummy variable coding for strong unpleasant pictures in the non-target location. The interaction was not significant, beta = 0.00, p = .84, indicating that the effect of strong unpleasant pictures did not diminish with repetition. Discussion I examined the ability of affective pictures to draw and to hold attention. The results supported three predictions. First, the present study failed to find significant effects of affect on the drawing of attention to the location of affective stimuli. This finding replicates previous findings with the exogenous cuing paradigm (Fox et al., 2001; Yiend & Mathews, 2001). As noted earlier, one concern with the findings in these studies could be the presentation of a single stimulus. It is possible that the sudden onset of a single stimulus is sufficient to draw attention, precluding any additional effects of affect. This concern is alleviated in the present study that presented two pictures at a time. Still, affective stimuli did not facilitate responses when the appeared in the location of the target stimulus. This finding provides additional evidence that affective stimuli do not draw attention. The results also supported the predictions that negative pictures have a stronger influence on the holding of attention than positive pictures, and that this effect is moderated by the extremity of negative pictures. Indeed, only strong negative pictures produced a significant effect on the holding of attention, although moderate negative pictures showed a trend in the same direction, and the difference between strong and moderate pictures was not statistically significant. Affect and Attention 12 Limitations The greatest limitation of the study is the reliance on ready-made pictures for the manipulation of affect. Although the use of these stimuli is common practice in affect research, it prevents a causal interpretation of the results. It is possible that the effects were due to other features of the pictures than their affective value. I tried to address this problem as best as I could by sampling pictures from a large set of affective pictures that have been used extensively in affect research (Lang et al., 1995). It seems unlikely that the significant effect for the strong unpleasant pictures is due to unique perceptual features of this heterogeneous set of pictures (e.g., a dead cat, messy toilet, airplane crash). Salient perceptual features (e.g., high contrast to background) should also have influenced drawing of attention (Yantis, 1996). One plausible alternative explanation of the effect is in terms of novelty. Unpleasant pictures are more unusual than pleasant pictures. However, participants were already familiar with the pictures from a previous task, and there the ability of strong negative pictures to hold attention did not decrease over repeated blocks. Hence, it seems unlikely that the effect is due to novelty. Another alternative explanation is arousal. Strong unpleasant pictures are more arousing than strong pleasant pictures when erotic pictures are excluded (Bradley, Codispoti, Cuthbert, & Lang, 2001; Schimmack, Colcombe, & Crites, 2001). Future research needs to disentangle the influences of displeasure and arousal on attention. The present study also does not address the possibility that specific unpleasant stimuli (snakes, spider, threatening stimuli) influence attention differently than other strong unpleasant pictures. However, Fox et al.’s (2001) found that angry facial expressions, an evolutionarily significant threat stimulus, did not draw attention. Affect and Attention 13 Another limitation of this study was the neglect of individual differences. Many studies have demonstrated that high-anxious individuals show stronger effects of negative stimuli on attention than low-anxious individuals (Fox et al., 2001; Matthews et al., 1986; Mogg et al., 2000; Yiend & Mathews, 2001). The present study could not examine this question because studies of individual differences require many trials to eliminate random error. In future research it will be important to examine whether the present findings are moderated by individual differences. Implications Despite its limitations, the present study contributes to the literature on affect and attention in several ways. First, it is one of the few studies that systematically varied the intensity of stimuli along the valence dimension (Pratto & John, 1991). Affect words showed a categorical effect on attention. However, the present study demonstrated a significant effect of stimulus intensity. This finding replicates Mogg et al.’s (2000) finding that strong unpleasant pictures influenced attention more than milder unpleasant pictures. Hence, the findings suggest that affective pictures do not produce a categorical negativity bias. This finding casts doubt on a hard-wired detection mechanism that equally draws attention to all unpleasant stimuli. As noted by Yiend and Mathews (2001), such a mechanism would not be adaptive because it is often important to ignore negative stimuli (e.g., pain in a battle; minor discomfort while pursuing an important task). However, strong negative stimuli are more likely to pose a threat to survival and should demand more attention than mild negative stimuli. The present study also contributes to the distinction between affective influences on drawing and holding of attention (Fox et al., 2001; Yiend & Mathews, 2001). Consistent with previous studies, affect was only related to holding of attention. The failure to demonstrate effects on drawing of attention has Affect and Attention 14 important theoretical implications for models of the interplay between cognitive and affective processes. The Emotional Stroop task has often been interpreted in terms of fast and automatic appraisal processes. However, this paradigm shows stimuli in the focus of attention with instructions to evaluate these stimuli. This situation is quite different from the everyday situation, in which people are confronted with dozens of stimuli at the same time. In this situation affective stimuli do not seem to automatically draw attention to their location. One objection to this argument could be the short stimulus-onsetasynchronies. Maybe 200 ms are too short for the appraisal process to draw attention. Future studies need to examine this question. However, 200 ms are sufficient to draw attention by means of other hard-wired processes such as the sudden onset of a novel stimulus (Yantis, 1996), or the matching of a stimulus to representations in working memory (Downing, 2000). More research is needed to examine the effects of different affective stimuli and different stimulus onset asynchronies using a variety of paradigms to examine the ability of affective stimuli to draw attention. Affect and Attention 15 References Bradley, B. P., Mogg, K., Falla, S. J., & Hamilton, L. R. (1998). Attentional bias for threatening facial expressions in anxiety: Manipulation of stimulus duration. Cognition and Emotion, 12(6), 737-753. Bradley, M. M., Codispoti, M., Cuthbert, B. N., & Lang, P. J. (2001). Emotion and motivation I: Defensive and appetitive reactions in picture processing. Emotion, 1(3), 300-319. De Houwer, J., & Hermans, D. (1994). Differences in the affective processing of words and pictures. Cognition and Emotion, 8(1), 1-20. Fox, E., Russo, R., Bowles, R., & Dutton, K. (2001). Do threatening stimuli draw or hold visual attention in subclinical anxiety? Journal of Experimental Psychology: General, 130(4), 681-700. Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1995). International affective picture system: Technical manual and affective ratings. Gainesville, FL: University of Florida. MacLeod, C., Mathews, A., & Tata, P. (1986). Attentional bias in emotional disorders. Journal of Abnormal Psychology, 95(1), 15-20. Mogg, K., McNamara, J., Powys, M., Rawlinson, H., Seiffer, A., & Bradley, B. P. (2000). Selective attention to threat: A test of two cognitive models of anxiety. Cognition and Emotion, 14(3), 375-399. Posner, M. I., Inhoff, A. W., Friedrich, F. J., & Cohen, A. (1987). Isolating attentional systems: A cognitive-anatomical analysis. Psychobiology, 15(2), 107121. Pratto, F., & John, O. P. (1991). Automatic vigilance: The attention-grabbing power of negative social information. Journal of Personality and Social Psychology, 61(3), 380-391. Affect and Attention 16 Raudenbush, S., Bryk, A., Cheong, Y. F., Congdon, R. (2000). HLM 5: Hierarchical linear and nonlinear modeling. Lincolnwood, IL: SSI. Rozin, P., & Royzman, E. B. (2001). Negativity bias, negativity dominance, and contagion. Personality and Social Psychology Review, 5(4), 296-320. Schimmack, U., Colcombe, S., & Crites, S. L. Jr. (2001). Affective reactions to conflicting stimuli: Examining appealingness and appallingness appraisals. Manuscript submitted for publication. Yantis, S. (1996). Attentional capture in vision. In A. F. Kramer, M. G. H. Coles (Eds.), Converging operations in the study of visual selective attention (pp. 45-76Washington, DC, US: American Psychological Association. Yiend, J., & Mathews, A. (2001). Anxiety and attention to threatening pictures. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 54A(3), 665-681. Affect and Attention Table 1 Coefficients of HLM model predicting log(RT). Effect Coefficient Standard Error T-ratio Intercept 6.85 .03 245.64* Block -.06 .01 5.88* Strong Negative .00 .01 0.15 Moderate Negative .00 .01 0.10 Mild Negative -.01 .01 0.55 Mild Positive -.01 .01 0.67 Moderate Positive -.01 .01 0.64 Strong Positive -.02 .01 1.77 Strong Negative .03 .01 2.71* Moderate Negative .02 .01 1.73 Mild Negative .00 .01 0.37 Mild Positive .00 .01 0.31 Moderate Positive .00 .01 0.17 Strong Positive .01 .01 1.00 Target Location Dummy Non-Target Location Dummy * p < .05 17 Affect and Attention 18 Figure Caption Figure 1 Response Latencies as a Function of Picture Valence in the Non-Target Location Note. –3 = strong displeasure, -2 = moderate displeasure, 1 = mild displeasure, 0 = neutral, +1 = mild pleasure, +2 = moderate pleasure, +3 = strong pleasure Affect and Attention Response Latency (ms) 980 970 960 950 940 -3 -2 -1 0 +1 +2 +3 19