Action error and observation-based contagions in pointing movements: Testing kinematic prediction errors (Study 2)

The following study registration is a second study to follow up on a pattern of null results from Study 1 (pre-registration: https://osf.io/ptsbv). The rationale and predictions for this new study are similar to Study 1, and as such, the background information is repeated here. There are methodological differences in this new study and the rationale for these differences is detailed in the final paragraphs of the introduction. Watching the actions of others can elicit unintentional performance biases, with the direction depending on the instructions accompanying the demonstration (Ikegami et al., 2018). Watching presumed “correct”, yet different actions (such as someone putting to a different target), can cause performance biases in the same direction as the watched action (Gray & Beilock, 2011; Ikegami et al., 2018; Kilner et al., 2003; Peters et al., in review). These imitative contagions can occur both from watching another person’s body kinematics (e.g., Takeuchi et al., 2018) and also just by seeing the outcomes of their action, such as the flight of a ball or its landing position (e.g., Gray & Beilock, 2011). Unintentional imitation is thought to occur as a result of shared representations for action and observation of action (e.g., Prinz, 1997; Shin et al. 2010). The representation generated from watching a different action is thought to interfere with the representation used for action planning on subsequent execution (Brass & Heyes, 2005; Heyes, 2010, 2011). A different type of action observation contagion has been identified in response to seeing presumed errors. In the case when an actor is thought to be aiming to one target, but misses, subsequent performance of the observer has been shown to be biased away from the observed error (Ikegami et al., 2018; Ronchi et al., 2018). This “compensatory contagion” (Ikegami et al., 2018) is thought to be prediction error dependent. The difference between the expected and actual performance of the actor creates a prediction error, which is thought to drive updates to the observer’s action planning processes for their own actions (Ikegami et al., 2019; Ikegami & Ganesh, 2017). However, prediction and task errors are always confounded and therefore it is possible that the “prediction” itself is not a necessary component of the bias. Just seeing an errorful action (i.e., target miss) may be sufficient to drive compensations in execution following the observed error. In the absence of a prediction error, watching a “miss” might induce a more cognitive strategy to avoid making the same mistake. To replicate and extend these effects, we tested experienced golfers watching both “on-target” and errorful putts of an actor (pre-registration available at https://osf.io/p629c). Across four blocked conditions, we covaried the location of the actor’s putts on a nine-square target grid (centre, corner) and the instructions about the actor’s intended aiming location to manipulate the presence of an error. We expected to replicate imitative effects for the observer’s putting to a centre target after observing presumed target "hits" to the corners. We expected to find compensatory effects after observing presumed "misses" toward the corners. In our new condition, where the observer saw presumed “misses” to the centre target, we also expected compensatory biases in subsequent putting to the centre target, even though there was now no discrepancy between the representations generated through observation and action. In terms of our actual results, predicted imitative biases were present after watching the actor “correctly” putt to the corner, but not when those same corner putts were perceived as errors. Watching “misses” to both corner and centre squares did not lead to imitative biases as predicted, but contrary to our predictions, neither did it lead to compensatory biases (Peters et al., in review). There are several explanations as to why we did not show compensatory biases from error observation in our previous experiment. One explanation might be due to task differences between our task (golf putting) and the aiming and throwing tasks where compensatory biases have previously been shown (Ikegami et al., 2018; Ronchi et al., 2011). There is evidence that expert observers are attuned to the kinematic features of the actions of others and that action outcomes can be anticipated from body kinematics (for reviews see Torricelli et al., 2023; Müller & Abernethy, 2012). In past work (Ikegami et al., 2018; Ronchi et al., 2011), the actor’s kinematics may have divulged more information about the upcoming target error, making kinematic prediction errors (i.e., differences between the expected and observed kinematics) more salient and ultimately driving compensation biases. This saliency may be due to either the increased forces (and hence arm swing) needed to throw a baseball and/or the direct aiming (without an implement) to targets, both different to golf putting. In golf putting, the adjustments to the swing of the club between targets of different distances is subtle therefore, kinematic prediction errors would be suppressed. There is evidence that these kinematic prediction errors are important in driving compensatory biases from error observation (Ikegami et al., 2014; Saito et al., 2024). When kinematic prediction errors were suppressed by removing the goal belief about the actor’s aiming location (negating a prediction), interference from observation on action was reduced (Ikegami et al., 2014). However, in this manipulation, task error (i.e., success or failure) was also suppressed along with kinematic errors and therefore the role of kinematic prediction errors is still somewhat unclear. There is evidence that movement outcomes, either ball flight or ball landing position, without observation of action kinematics, are sufficient to drive imitative biases (Gray & Beilock, 2011). Biases from observed outcomes are thought to be driven by strong relationships (e.g., Catmur & Heyes, 2019; Heyes, 2011) or perhaps a shared common code (e.g., Prinz, 1997; Shin et al., 2010) for actions and their effects. Conversely, Ikegami and Ganesh (2014; 2017) suggest that seeing the actor’s kinematics is a necessary source of information to drive prediction errors and biases as a result of error observation. This suggestion was based on a study where kinematic prediction errors and outcome prediction errors were isolated and shown independently when two groups of skilled participants watched the same videos of a novice throwing to a dart board (Ikegami & Ganesh, 2014). The videos were spatially occluded to show the movement kinematics but not the dart flight or outcome, and participants predicted where the dart would land. To suppress outcome prediction errors and show only kinematic prediction errors, participants were told that the actor was aiming to a centre target and were shown the body kinematics associated with target “misses”. After predicting the throw’s outcome, no outcome feedback was provided. To suppress and isolate kinematic prediction errors, a second group was told that the actor was aiming to various targets and only correct attempts were shown, such that there would be no prediction about what the actor’s kinematics should look like. After making the prediction about the throw’s landing position, participants received outcome feedback about where the dart landed to elicit an outcome prediction error. Only kinematic prediction errors impacted the observer’s subsequent performance (and self-error estimations). However, limiting the interpretation of these results, task-related errors (i.e., “hit” or “miss”) were not controlled across the two groups. Kinematic prediction errors were always associated with “misses” while outcome prediction errors were always associated with “hits”. To address concerns from previous studies, with respect to the task and the salience of the kinematic prediction errors, we conducted a replication (Study 1; https://osf.io/ptsbv) of the hand-to-target aiming study by Ikegami et al (2018, Experiment 2). In this Ikegami et al study, there were imitative and compensatory biases after watching others. Our replication therefore, also involved target-directed reaching on a touchscreen computer. We expected that kinematic prediction errors would be more salient than those observed in golf putting, due to the direct relationship between hand movements and target success. Aiming movements were also expected to be less variable than putting, affording more power to see the small, expected effects, if present. Different to Ikegami et al (2018), we also manipulated the type of information shown in the videos to show the whole movement (i.e., kinematics and outcomes, replicating Ikegami et al.), or only target outcomes. Through this manipulation we were able to test if kinematic prediction errors were necessary to drive any potential compensatory biases independently from trials which are seen as outcome ‘hits’ (correct) or ‘misses’ (errors). Contrary to predictions, there was no evidence of imitative biases after watching hits, nor was there evidence of compensation after watching errors, regardless of whether participants watched the actor’s kinematics or outcomes (all ps > .76). Despite replicating the task and instructions given by Ikegami and colleagues in Study 1, one potential reason as to why we did not show the same effects was that our targets were displayed on a smaller screen, which may have masked kinematic prediction errors in the videos and subsequently observation-induced biases. Therefore in this new Study 2, we will again replicate the design and methods of Ikegami et al (2018), but with greater spacing between the targets, hence resulting in more salient kinematic differences between target hits and misses. Two groups will be compared in a repeated measures’ design, where one group will only watc

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.