1
Conflict: Run! Reduced Stroop interference with Avoidance Responses Nathalie Schouppe1, Jan De Houwer2, K. Richard Ridderinkhof3,4, and Wim Notebaert1 1 2
Department of Experimental Psychology, Ghent University
Department of Experimental Clinical and Health Psychology, Ghent University 3
Department of Developmental Psychology, University of Amsterdam 4
Cognitive Science Center Amsterdam, University of Amsterdam
Running Head: Avoidance reduces Stroop interference Word count: 2573 Corresponding author: Nathalie Schouppe Department of Experimental Psychology Henri Dunantlaan 2 B – 9000 Ghent
[email protected] Phone: +32 9 264 64 31 Fax: +32 9 264 64 98
2
Abstract Conflict has been hypothesized to be aversive, triggering avoidance behaviour (Botvinick, 2007). To test this hypothesis, a standard Stroop task was modified such that avoiding was part of the response set. More precisely, participants were asked to move a manikin towards or away from Stroop stimuli, depending on the colour of the words. Results showed that the type of response (approach versus avoidance) modulated the Stroop congruency effect. Specifically, the reaction time analysis revealed that the stimulus congruency effect disappeared with avoidance responses, contrary to approach responses where a stimulus congruency effect was present. Moreover, the error data showed a reduction of the general congruency effect when avoiding. These results suggest that in the face of conflict, avoidance is the predominant response. Key words: approach-‐avoidance; conflict; stimulus interference; response interference; Stroop task
3
Introduction Conflict tasks have been extremely useful as a means to understand how we deal with conflicting response tendencies. One task widely used to study conflict processing is the Stroop task, in which participants have to name the ink colour of a colour word, while ignoring the irrelevant word information (Stroop, 1935). In general, participants are slower and less accurate on incongruent (the word RED written in green) than on congruent stimuli (the word RED written in red). This interference or conflict arises from competition between the relevant colour and the irrelevant word meaning at both stimulus encoding and response selection (De Houwer, 2003; van Veen & Carter, 2005). Recently, it has been hypothesized that the occurrence of conflict is aversive, generating a negative value (Botvinick, 2007). Using an affective priming paradigm, Dreisbach and Fischer (2012) clearly demonstrated this affective, negative nature of conflict. More specifically, participants were faster to evaluate the valence of negative targets, when these stimuli were preceded by incongruent Stroop primes, than when they were preceded by congruent trials. Furthermore, other recent studies confirm the aversiveness of conflict, showing a bias away from high-‐conflict situations (Kool, McGuire, Rosen, & Botvinick, 2010; Schouppe, Ridderinkhof, Verguts, & Notebaert, 2012). For instance, in our lab, we constructed a conflict selection task, in which participants had to choose between two alternatives associated with a different degree of Stroop conflict. As shown in the choice rates, participants systematically avoided the high-‐ conflict alternative (Schouppe et al., 2012). Additionally, Lynn, Riddle, and Morsella
4
(in press) found a higher urge to quit after incongruent trials, again supporting the hypothesis that conflict, with its negative quality, triggers avoidance behaviour. Interestingly, in a standard Stroop task naming the ink colour or pressing a response button associated with the ink colour could actually be qualified as an approach response. Yet, in the face of an aversive stimulus (i.e. incongruent conflict trial), avoidance is the more likely response. It is therefore plausible that the slowdown on incongruent trials reflects an incompatibility between the elicited avoidance tendency and the required approach response. This idea is in line with a recent study of Chajut, Mama, Levy, and Algom (2010), showing that the emotional Stroop effect reversed under avoidance behaviour. More specifically, participants were instructed to step forward or backward (Experiment 1) or push a joystick towards or away (Experiment 2) from emotional and neutral stimuli depending on the colour of the stimuli. When making an approach response, Chajut and colleagues found a normal emotional Stroop effect, however, when avoiding the stimuli, this emotional Stroop effect reversed, indicating that participants were faster in avoiding negative than neutral stimuli. The standard slowdown on negative stimuli thus completely vanished when avoidance was a permissible response. In the present study, we wanted to investigate whether a similar modulation by response type can be observed in a standard Stroop task. We used the Manikin task (De Houwer, Crombez, Baeyens, & Hermans, 2001) as approach-‐avoidance paradigm, since it has been proven to be a sensitive measure of approach-‐avoidance reactions (Krieglmeyer & Deutsch, 2010). In this task, participants are instructed to imagine being a manikin that has to step towards or away from Stroop stimuli. If
5
conflict trials are truly aversive, we would predict a modulation of the Stroop effect by the type of response (avoidance or approach). Given the dissociable effects of stimulus and response conflict (e.g. Notebaert & Verguts, 2006; Verbruggen, Notebaert, Liefooghe, & Vandierendonck, 2006), a second, more exploratory aim of this study was to investigate whether these two types of conflict would be differentially influenced by an approach versus avoidance response. By assigning two colours (of the four colours used) to the same response (approach vs. avoidance), we could distinguish between three trial types: (1) congruent (CO) trials in which the colour of the word was identical to the word meaning; (2) stimulus incongruent (SI) trials, in which the colour of the word was different from its meaning, but resulted in the same response; and (3) response incongruent (RI) trials for which colour and meaning differed and also mapped onto a different response. Typically, reaction times tend to increase from CO trials to SI trials to RI trials, thus showing a contribution of both stimulus and response conflict to the general interference effect (De Houwer, 2003). Method Participants Forty students at Ghent University (range: 18-‐25 years of age; 33 right-‐ handed; 34 female) participated in the study. They provided written informed consent and were paid 6 euro for their participation. The study procedures were
6
approved by a local ethics committee and complied with relevant laws and institutional guidelines. Materials and Procedure The Stroop stimuli consisted of four colour words (‘BLUE’, ‘PURPLE’, ‘YELLOW’, AND ‘BROWN’), presented on a black background in blue, purple, yellow or brown ink
colour1. The manikin was a picture of a stick figure, approximately 2.5 cm in length. Participants were instructed to move this manikin as fast and accurately as possible towards or away from a centrally presented word, depending on its colour. Two colours were mapped onto the same response (e.g. “if the colour of the word is blue or purple, move the manikin towards the word; if the colour of the word is yellow or brown, move the manikin away from the word”). The different colour-‐to-‐response combinations were counterbalanced across participants. Following the procedure of Krieglmeyer, Deutsch, De Houwer, and De Raedt (2010), a trial started with the picture of the manikin, presented at the upper or lower part of the computer screen. After 750 ms, the Stroop stimulus was displayed in the centre of the screen. The stimulus and the manikin remained on the screen until a response was given. Participants had to press the ‘8’ or ‘2’ key on the numeric keyboard three times to move the manikin three steps upwards or downwards respectively. Consequently, the manikin would either stop near the centrally presented word or at the edge of the screen. Reaction times were measured as the time between the onset of the Stroop stimulus and the first key press. After an incorrect response, error feedback was given (the Dutch word ‘FOUT!’, meaning
7
‘WRONG!’ was displayed in the middle of the screen for 500 ms). The inter-‐trial-‐ interval was 1000 ms. Participants performed ten blocks of 64 trials. Trials were randomly presented, with 50% CO trials, 25% SI trials and the remaining 25% RI trials. In half of the trials, the picture of the manikin appeared above the word, in the other half the manikin appeared below the word. Results The data of one participant were excluded from the analyses because performance was at chance level (error rate of 47%). Mean reaction times and mean error rates were submitted to a repeated-‐measures ANOVA with response type (approach; avoidance) and congruency (CO; SI; RI) as within-‐subjects factors. Errors and outliers (3 standard deviations above and below the mean, calculated for each participant and condition) were excluded from the reaction time analysis. Note that Greenhouse-‐Geisser corrections to the degrees of freedom and p-‐values are used when the sphericity assumption was violated, but uncorrected degrees of freedom are reported for ease of reading. To interpret the interaction between response type and congruency, post-‐ hoc paired sample t-‐tests were carried out, testing each level of the two factors against each other. Also, to disentangle the effect of response type on stimulus conflict and response conflict separately, three additional t-‐tests were computed comparing the overall congruency effect (RI-‐CO), the stimulus congruency effect (SI-‐ CO) and the response congruency effect (RI-‐SI) between response types. The
8
multiple comparison p-‐values were corrected following Benjamini and Hochberg (1995). Mean Response Times Results showed main effects of response type, F(1, 38) = 63.4, p < .001, and congruency, F(2, 76) = 25.0, p < .001. Most important, the interaction between response type and congruency was significant, F(2, 76) = 3.5, p < 0.05. As is depicted in Table 1, the overall congruency effect (RI-‐CO: t(38) = 5.3, p < .001), just as the stimulus congruency effect (SI-‐CO: t(38) = 4.1, p < .001) and the response congruency effect (RI-‐SI: t(38) = 2.3, p < .05) were significant for approach responses. For avoidance responses, the overall congruency effect was significant (RI-‐CO: t(38) = 4.5, p < .001), as was the response congruency effect (RI-‐SI: t(38) = 3.6, p < .01). However, the stimulus congruency effect was not significant (SI-‐CO: t(38) = 0.8, p > .1). The remaining comparisons of Table 1 all indicate slower avoidance responses than approach responses, thus reflecting the significant main effect of response type. As illustrated by Figure 1 and supported by the above-‐mentioned statistics, a stimulus interference effect (SI-‐CO) was evident for approach responses (M = 26 ms; SD = 39 ms), but not for avoidance responses (M = 2.9 ms; SD = 21 ms). This difference in stimulus interference between approach and avoidance responses was significant, t(38) = 3.0, p < .01. Response type had no effect on the difference between CO and RI trials, t(38) = 1.2, p > .1, and SI and RI trials, t(38) = 1.3, p > .1.
9
(Figure 1 and Table 1 about here) Error Rates Concerning error rates, the results showed no effect of response type, F(1, 38) < 1. However, the effect of congruency, F(2, 76) = 15.9, p < .001, and the interaction between congruency and response type were significant, F(2, 76) = 4.4, p < .05. As depicted in Table 2, for approach responses, the overall congruency effect and the response congruency effect were significant (RI-‐CO: t(38) = 5.5, p < .001, RI-‐ SI: t(38) = 3.9, p < .001), however, the stimulus congruency effect was not (SI-‐CO: t(38) = 0.79, p > .1). For avoidance responses, neither the overall congruency effect (RI-‐CO: t(38) = 1.3, p > .1), nor the stimulus congruency effect were significant (SI-‐CO: t(38) = 1.4, p > .1), but there was a small response congruency effect (RI-‐SI: t(38) = 2.5, p < .05). Results from the remaining comparisons (see Table 2) showed that significantly more errors were made on RI approach trials, than on CO, SI and RI avoidance trials. Also, more errors were made on RI avoidance trials than on CO approach trials. Importantly, only the overall congruency effect (RI-‐CO) differed between approach and avoidance responses, t(38) = 3.1, p < .05. Thus, participants made more errors on RI trials than on CO trials, but only when an approach response had to be carried out (see Figure 1). Response type had no effect on the stimulus congruency effect, t(38) = 1.5, p > .1, and on the response congruency effect, t(38) = 1.5, p > .1.
10
(Table 2 about here) Discussion This study is the first to report a modulation of the Stroop congruency effect by response type. We observed both in the reaction time and error rate data a reduction of the Stroop effect when participants gave an avoidance response instead of a more typical approach response. More specifically, the reaction time analysis revealed the typical pattern of increasing reaction times from CO to SI to RI trials when participants approached the stimuli. However, under avoidance responses the difference between CO and SI trials (i.e. stimulus congruency effect) was abolished. Furthermore, the accuracy data on approach responses showed more performance errors on RI trials than on CO and SI trials, with no difference between the latter two (see De Houwer, 2003; van Veen & Carter, 2005 for comparable results). It is not surprising that stimulus conflict is not evident in error rates since on SI trials the relevant and irrelevant features of the stimulus indicate the same response. More interestingly, when investigating the avoidance responses, we found that the accuracy difference between RI and CO trials was reduced, thus clearly indicating a modulation of the general congruency effect by response type. The reaction time pattern indicates that stimulus conflict disappears with avoidance responses while response conflict is unaffected. In the original conflict monitoring model (Botvinick, Braver, Barch, Carter, & Cohen, 2001), conflict has been conceptualized as the competition between two simultaneously active
11
responses, implying a major role for response conflict in conflict detection and thereby suggesting that our interaction with response type should have been primarily driven by this type of conflict. However, previous studies (Notebaert & Verguts, 2006; Verbruggen et al., 2006) report adaptation effects after stimulus conflict and response conflict, suggesting that stimulus conflict triggers conflict adaptation (as stimulus conflict is also present in response incongruent trials). Notebaert and Verguts (2006) argued that response conflict mainly affects reaction times, while conflict adaptation is triggered by stimulus conflict. Given the theoretical formulation that the negative quality of conflict triggers behavioural adjustments (Botvinick, 2007; see also van Steenbergen, Band, & Hommel, 2009), this would suggest that stimulus conflict is an aversive signal. Also, in the affective priming study of Dreisbach and Fischer (2012), participants did not have to respond to the Stroop stimuli, indicating that mere stimulus conflict can already lead to negative affect, consistent with our findings. Thus, there is some support for the idea that stimulus conflict is aversive, and perhaps more aversive than response conflict. Quite speculative, one could argue that correctly responding to response incongruent stimuli means that the conflict was successfully resolved, resulting in positive affect. However, we should not forget that the error rates show that the overall congruency effect (RI-‐CO) is reduced with avoidance responses. The reaction time results also showed a main effect of response type, indicating that approach responses were generally faster than avoidance responses. It could well be that a natural propensity to orient towards the source of stimulation (i.e. Stroop stimulus) drives this effect. This is particularly applicable to this type of task as participants saw the picture of the manikin walk towards or away from the
12
stimulus. Alternatively, the main effect of response type can be explained by assuming that response type represents polar oppositions (i.e. positive/negative – approach/avoidance). It has been shown that participants process +polar targets faster than –polar targets (Proctor & Cho, 2006). Accordingly and as predicted, a compatibility benefit should thus appear when processing polar equivalents (i.e. incongruent avoidance trials). In conclusion, our results show that the congruency level of the stimulus interacts with response type, thereby clearly supporting the conflict avoidance hypothesis (Botvinick, 2007). These findings suggest that conflict is negative (Dreisbach & Fischer, 2012) and therefore more likely to be avoided than approached. Moreover, this argument is in line with several studies on automatic stimulus evaluation, showing that perceiving positive or negative stimuli evokes approach or withdrawal tendencies respectively (e.g. Chen & Bargh, 1999; Krieglmeyer et al., 2010). Our results thus suggest that, when confronted with conflict, avoiding is the predominant response. References Benjamini, Y., & Hochberg, Y. (1995). Controlling the False Discovery Rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society, Series B, 57, 289-‐300.
13
Botvinick, M. M. (2007). Conflict monitoring and decision making: Reconciling two perspectives on anterior cingulate function. Cognitive, Affective, & Behavioral Neuroscience, 7(4), 356-‐366. Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D. (2001). Conflict monitoring and cognitive control. Psychological Review, 108(3), 624-‐ 652. Chajut, E., Mama, Y., Levy, L., & Algom, D. (2010). Avoiding the approach trap: A response bias theory of the emotional Stroop effect. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36(6), 1567-‐1572. Chen, M., & Bargh, J. A. (1999). Consequences of automatic evaluation: Immediate behavioral predispositions to approach or avoid the stimulus. Personality and Social Psychology Bulletin, 25, 215-‐224. De Houwer, J. (2003). On the role of stimulus-‐response and stimulus-‐stimulus compatibility in the Stroop effect. Memory & Cognition, 31(3), 353-‐359. De Houwer, J., Crombez, G., Baeyens, F., & Hermans, D. (2001). On the generality of the affective Simon effect. Cognition and Emotion, 15(2), 189-‐206. Dreisbach, G., & Fischer, R. (2012). Conflicts as aversive signals. Brain and Cognition. doi:10.1016/j.bandc.2011.12.003
14
Elliot, A. J., Maier, M. A., Moller, A. C., Friedman, R., & Meinhardt, J. (2007). Color and psychological functioning: The effect of red on performance attainment. Journal of Experimental Psychology: General, 136(1), 154-‐168. Kool, W., McGuire, J. T., Rosen, Z. B., & Botvinick, M. M. (2010). Decision making and the avoidance of cognitive demand. Journal of Experimental Psychology: General, 139, 665-‐682. Krieglmeyer, R., & Deutsch, R. (2010). Comparing measures of approach avoidance behaviour: The manikin task vs. two versions of the joystick task. Cognition and Emotion, 24(5), 810-‐828. Krieglmeyer, R., Deutsch, R., De Houwer, J., & De Raedt, R. (2010). Being moved: Valence activates approach-‐avoidance behavior independently of evaluation and approach-‐avoidance intentions. Psychological Science, 21(4), 607-‐613. Lynn, M. T., Riddle, T. A., & Morsella, E. (in press). The phenomenology of quitting: Effects from repetition and cognitive effort. Journal of Cognitive Science. Notebaert, W., & Verguts, T. (2006). Stimulus conflict predicts conflict adaptation in a numerical flanker task. Psychonomic Bulletin & Review, 13(6), 1078-‐1084.
15
Proctor, R. W., & Cho, Y. S. (2006). Polarity correspondence: A general principle for performance of speeded binary classification tasks. Psychological Bulletin, 132(3), 416-‐442. Schouppe, N., Ridderinkhof, K. R., Verguts, T., Notebaert, W. (2012). The aversive nature of conflict revealed in choice and switch rates. Manuscript submitted for publication. Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643-‐662. van Steenbergen, H., Band, G. P. H., & Hommel, B. (2009). Reward counteracts conflict adaptation: evidence for a role of affect in executive control. Psychological Science, 20(12), 1473-‐1477. van Veen, V., & Carter, C. S. (2005). Separating semantic conflict and response conflict in the Stroop task: A functional MRI study. NeuroImage, 27, 497-‐504. Verbruggen, F., Notebaert, W., Liefooghe, B., & Vandierendonck, A. (2006). Stimulus-‐ and response-‐conflict-‐induced cognitive control in the flanker task. Psychonomic Bulletin & Review, 13(2), 328-‐333.
16
Footnote 1
Note that we did not use the colours green or red. In daily life, these colours are
typically associated with go and no-‐go responses respectively (cfr. traffic lights). Recent findings indeed confirm that the colour red induces an avoidance tendency (Elliot, Maier, Moller, Friedman, & Meinhardt, 2007).
17
Author Note The research reported in this article was supported by grant no. 3F011209 of Research Foundation -‐ Flanders. Jan De Houwer is supported by grants BOF/GOA2006/001 and BOF/01M00209 from Ghent University.
18
Figure 1. Mean reaction times (left panel) and error rates (right panel) for congruency and response type. Vertical bars represent ± 1 standard error of the mean. 6
approach avoidance
700
Error rate (in %)
ReacCon Cme (in ms)
740
660 620 580 540
2
approach avoidance
0
CO
4
SI
RI
CO
SI
RI
19
Table 1. Reaction time differences (in ms) between the different levels of congruency and response type. Difference scores were calculated by subtracting values of the row conditions from values of the column conditions. Grey cells indicate a significant result for the comparison of the difference scores between approach and avoidance responses.
CO approach
SI approach
RI approach
CO avoidance
SI avoidance
RI avoidance
CO approach
x
26 **
43.2 **
62.9 **
65.8 **
94.3 **
SI approach
x
17.1 *
36.9 **
39.7 **
68.3 **
RI approach
x
19.7 *
22.6 *
51.2 **
CO avoidance
x
2.9
31.4 **
SI avoidance
x
28.5 **
RI avoidance ** p < .01 * p < .05, multiple comparisons p-‐values were corrected following Benjamini and Hochberg (1995)
x
20
Table 2. Error rate differences (in %) between the different levels of congruency and response type. Difference scores were calculated by subtracting values of the row conditions from values of the column conditions. Grey cells indicate a significant result for the comparison of the difference scores between approach and avoidance responses.
CO approach
SI approach
RI approach
CO avoidance
SI avoidance
RI avoidance
CO approach
x
0.3
2.6 **
0.7
0.1
1.4 *
SI approach
X
2.3 **
0.4
-‐0.2
1.0
RI approach
x
-‐1.9 **
-‐2.5 **
-‐1.3 *
CO avoidance
x
-‐0.6
0.6
SI avoidance
x
1.2 *
RI avoidance ** p < .01 * p < .05, multiple comparisons p-‐values were corrected following Benjamini and Hochberg (1995)
x