Comparison of two reaction-time-based and one foraging-based behavioral approach-avoidance tasks in relation to interindividual differences and their reliability

Abstract

Approaching rewards and avoiding punishments is a fundamental aspect of behavior, yet individuals differ in the extent of these behavioral tendencies. One popular method to assess differences in approach-avoidance tendencies and even modify them, is using behavioral tasks in which spontaneous responses to differently valenced stimuli are assessed (e.g., the visual joystick and the manikin task). Understanding whether these reaction-time-based tasks map onto the same underlying constructs, how they predict interindividual differences in theoretically related constructs and how reliable they are, seems vital to make informed judgements about current findings and future studies. In this preregistered study, 168 participants (81 self-identified men, 87 women) completed emotional face versions of these tasks as well as an alternative, foraging-based paradigm, the approach-avoidance-conflict task, and answered self-report questionnaires regarding anxiety, aggression, depressive symptoms, behavioral inhibition and activation. Importantly, approach-avoidance outcome measures of the two reaction-time-based tasks were unrelated with each other, showed little relation to self-reported interindividual differences and had subpar internal consistencies. In contrast, the approach-avoidance-conflict task was related to behavioral inhibition and aggression, and had good internal consistencies. Our study highlights the need for more research into optimizing behavioral approach-avoidance measures when using task-based approach-avoidance measures to assess interindividual differences.

Figure 1

Examples of the three tasks, the (a) visual joystick, (b) manikin, and (c) AACT, utilized in this study. The representations of the visual joystick and manikin task are schematic and deviate in size and background color from the tasks performed by the participants. Greyscaled versions of the images with the codes AF21ANS, AF22HAS, AM28ANS and BM32HAS from Lundqvist, et al. are shown in the schematic representations of the visual joystick and manikin tasks and can be requested from kdef.se for non-commercial scientific research purposes. The figure has been adapted with permission from Fricke and Vogel.

Figure 2

Correlations of visual joystick and manikin task measures (a) overall reaction time (r₁₅₆ = .53, p < .001), (b) overall accuracy (r₁₅₆ = .41, p < .001), (c) global congruency (r₁₅₆ = .12, p = .135), (d) valence-unspecific bias (r₁₅₆ = .02, p = .844), (e) happy faces bias (r₁₅₆ = .05, p = .551), (f) angry faces bias (r₁₅₆ = .08, p = .339), (g) happy faces D-Score (r₁₅₆ = -.02, p = .796) and (h) angry faces D-Score (r₁₅₆ = .05, p = .495). Line indicates linear regression over all data points with 95% confidence interval.

Figure 3

Correlation matrix of all questionnaire scores with task outcome measures. Color intensity indicates strength of correlation, color and shade indicate direction of effect (blue/no shade = positively correlated; red/ shaded = negatively correlated). p-values are uncorrected for multiple comparisons. Significance codes: **** p < .0001, *** p < .001, ** p < .01, * p < .05.

Figure 4

Internal consistency estimates for all outcome measures sorted by task. Bars indicate the confidence interval based on the 25th and 975th value based on thousand randomly permutated datasets (for restrictions see “Methods” section).