Perception and action as viewed from the Theory of Event Coding: a multi-lab replication and effect size estimation of common experimental designs

Abstract

The Theory of Event Coding (TEC) has influenced research on action and perception across the past two decades. It integrates several seminal empirical phenomena and it has continued to stimulate novel experimental approaches on the representational foundations of action control and perceptual experience. Yet, many of the most notable results surrounding TEC originate from an era of psychological research that relied on rather small sample sizes as judged by today's standards. This state hampers future research aiming to build on previous phenomena. We, therefore, provide a multi-lab re-assessment of the following six classical observations: response-effect compatibility, action-induced blindness, response-effect learning, stimulus-response binding, code occupation, and short-term response-effect binding. Our major goal is to provide precise estimates of corresponding effect sizes to facilitate future scientific endeavors. These effect sizes turned out to be considerably smaller than in the original reports, thus allowing for informed decisions on how to address each phenomenon in future work. Of note, the most relevant results of the original observations were consistently obtained in the present experiments as well.

Fig. 1

Percentages of affirmative responses to the question on which empirical phenomena and approaches are related particularly closely to TEC, either by lending considerable support to its theoretical notions, or by being immediately stimulated by TEC. Percentages are relative to the total number of the 49 expert responses

Fig. 2

Overview of the effect sizes in the original studies and in the present experiments

Fig. 3

Design and results of the response-effect (R-E) compatibility experiment (Exp. 1). A Participants responded with horizontally aligned response keys to a colored target circle. Each keypress response predictably triggered a visual effect on the computer screen, either at a corresponding spatial location (compatible condition) or at a non-corresponding spatial location (incompatible condition). R-E compatibility conditions were manipulated between experimental halves. B Response times (RTs) in milliseconds (ms) as a function of RT quintile and R-E compatibility mapping, accompanied by the resulting R-E compatibility effects (ΔRT) and corresponding standard errors of the mean (SE_M). C Distribution of individual R-E compatibility effects (ΔRT; shown as kernel density estimate) in RTs together with means and standard errors of the four samples (blue dots) and the pooled data (black diamond)

Fig. 4

Design and results of the action-induced blindness experiment (Exp. 2). A Participants prepared a left or right keypress response as indicated by an arrow cue. While holding this response active, but before actually executing it, they were briefly presented with a masked arrow target. The main variable of interest is the percentage of correct target identifications (identification performance) as a function of whether the target arrow was compatible or incompatible to the prepared response. B Identification performance as a function of response-target compatibility. Error bars indicate standard errors of paired differences (SE_PD; Pfister & Janczyk, 2013). C Distribution of individual action-induced blindness effects (ΔIdentification performance, computed as the difference between the compatible and incompatible condition; shown as kernel density estimate) together with means and standard errors of the four samples (blue dots) and the pooled data (black diamond)

Fig. 5

Design and results of the response-effect learning experiment (Exp. 3). A The experiment consisted of an acquisition phase followed by a test phase. In the acquisition phase, participants freely chose between pressing a left or a right key, and each key press triggered a contingent effect tone of distinct pitch. The test phase was similar except that the previous effect tones now also prompted each choice, whereas a third tone indicated a no-go trial to discourage response preparation ahead of the trial. B Response frequency for consistent and inconsistent choices. Error bars indicate standard errors of paired differences (SE_PD; Pfister & Janczyk, 2013). C Distribution of individual percentages of consistent choices (shown as kernel density estimate) together with means and standard errors of the four samples (blue dots) and the pooled data (black diamond)

Fig. 6

Design and results of the stimulus–response binding experiment (Exp. 4). A Participants prepared a first response (R1) as indicated by a cue and they executed this response upon the onset of a colored line that was oriented either horizontally or vertically (S1). A second line stimulus followed shortly thereafter, prompting a response depending on line orientation (S2). Trials were constructed so that all possible response and stimulus sequences were varied orthogonally, allowing to assess the integration of stimulus and response features into event files via partial repetition costs (computed as the difference in repetition benefits for features between response repetitions and response alternations; see text for details). B Stimulus feature repetition benefits (positive) and costs (negative) calculated as RT_alternation − RT_repetition for stimulus form, color, and location, as a function of response sequence (repetition vs. alternation). RT_alternation corresponds to the mean of all conditions involving an alternation of the respective stimulus feature while RT_repetition corresponds to the mean of all conditions with repetition of that feature. Error bars indicate standard errors of paired differences (SE_PD; Pfister & Janczyk, 2013). C Distribution of partial repetition costs (ΔRepetition benefit) for the task-relevant stimulus feature form (shown as kernel density estimate) with means and standard errors of the four samples (blue dots) and the pooled data (black diamond)

Fig. 7

Design and results of the code occupation experiment (Exp. 5). A Participants prepared a response as indicated by a cue (R_A) and they executed another response (R_B) while holding the first response in working memory. Response times (RT_Bs) of this latter response were assessed as a function of whether this response overlapped or did not overlap with the action plan held in working memory to probe for occupation of the corresponding feature. B RT_Bs as a function of feature overlap between the current response and the action plan held in working memory. Error bars indicate standard errors of paired differences (SE_PD; Pfister & Janczyk, 2013). C Distribution of overlap costs (shown as kernel density estimate) together with means and standard errors of the four samples (blue dots) and the pooled data (black diamond)

Fig. 8

Design and results of the response-effect binding experiment (Exp. 6). A Each trial comprised two parts. In the first part, a freely chosen response triggered one of two possible effect tones at random. This effect tone was either repeated as the stimulus for the second part (effect-stimulus repetition) or the alternative tone was used as the stimulus (effect-stimulus switch). Short-term binding of response and following effects would be evident in higher response repetition rates across trial parts for effect-stimulus repetitions than for effect-stimulus switches. B Response repetitions as a function of effect-stimulus relation. Error bars indicate standard errors of paired differences (SE_PD; Pfister & Janczyk, 2013). C Distribution of binding effects, measured as the difference in response repetitions between conditions (shown as kernel density estimate) together with means and standard errors of the four samples (blue dots) and the pooled data (black diamond)

Fig. 9

Widths of different confidence intervals for standardized effect sizes as a function of the underlying population effect size and the sample size of an empirical study. All estimates were computed with help of the MBESS package for R (Kelley, 2007)