Contextual effects in sensorimotor adaptation adhere to associative learning rules

Read the full article See related articles


Traditional associative learning tasks focus on the formation of associations between salient events and arbitrary stimuli that predict those events. This is exemplified in cerebellar-dependent delay eyeblink conditioning, where arbitrary cues such as a light or tone act as conditioning stimuli that predict aversive sensations at the cornea. Here we ask if a similar framework could be applied to another type of cerebellar-dependent sensorimotor learning – sensorimotor adaptation. Models of sensorimotor adaptation posit that the introduction of an environmental perturbation results in an error signal that is used to update an internal model of a sensorimotor map for motor planning. Here we take a step towards an integrative account of these two forms of cerebellar-dependent learning, examining the relevance of core concepts from associative learning for sensorimotor adaptation. Using a visuomotor adaptation reaching task, we paired movement-related feedback with neutral auditory or visual contextual cues that served as conditioning stimuli (CSs). Trial-by trial changes in feedforward movement kinematics exhibited three key signatures of associative learning: Differential conditioning, sensitivity to the CS-US interval, and compound conditioning. Moreover, after compound conditioning, a robust negative correlation was observed between responses to the two elemental CSs of the compound (i.e., overshadowing), consistent with the additivity principle posited by theories of associative learning. The existence of associative learning effects in sensorimotor adaptation provides a proof-of-concept for linking cerebellar- dependent learning paradigms within a common theoretical framework.

Article activity feed

  1. Evaluation Summary:

    This paper is of interest to scientists within the field of motor control and learning. The experiments provide novel insight into the potential role of associative learning in sensorimotor adaptation. The results are compelling, although further data are required to support several key conclusions.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)

  2. Reviewer #1 (Public Review):

    Here using visuomotor rotation task the authors examine if associative learning applied to sensorimotor learning/ adaptation and see if integrative account of these two forms of cerebellar-dependent learning applied. There is a useful introduction to the article referencing past work on contextual effects seen in motor learning.

    Relating motor control experiments to classical conditioning approaches to associative learning is certainly interesting.

    The authors paired movement-related feedback with neutral auditory or visual contextual cues that served as conditioning stimuli. The authors found trial-by-trial changes in feedforward movement kinematics and the results seem convincing.

    I found the methods really quite hard to follow. I also feel the overall presentation of results and the discussion could be clearer and more concise.

  3. Reviewer #2 (Public Review):

    This paper provides a compelling account of the role associative learning plays in sensorimotor adaptation. In contrast to previous work, the authors show that arbitrary stimuli (light/sound) can be used as contextual cues to differentiate between sensorimotor perturbations/states. The authors describe these results in the context of associative learning, and across 2 experiments provide evidence for 2 key signatures of associative learning: differential conditioning and compound conditioning. The authors believe the key parameter for observing such associative learning is ensuring a strict temporal relationship between the cues and their associated movement-related sensory outcomes. This is clearly written article that uses well-designed experiments to not only provide new insight into a phenomenon that has received a lot of previous attention (i.e., the inability to use arbitrary stimuli as contextual cues in adaptation) but also places these results in a novel mechanistic context (associative learning). Whilst sensorimotor adaptation (and the tasks used to assess it) feel heavily overused, this work is important in trying to align 2 key cerebellar processes which often feel difficult to integrate. I can envisage the concepts proposed (and the tasks developed) within this article being the foundation of a rich new line of inquiry into the role of associative learning in adaptation.

    Despite many positives, there is a key issue with the current paper which make some of the author's conclusions unjustified. Specifically, the authors predict & conclude that the key parameter for observing associative learning when using arbitrary stimuli as contextual cues is the temporal relationship of these cues to the feedback. They provide compelling evidence of associative learning however at present provide no evidence of the importance of this temporal relationship. Considering this is assumed to be the key difference between the current results and previous work who have failed to find an effect, this is a key omission. Other issues include:

    Exp 1 probe phase results: Whilst not significant (n-1 x n interaction), there seemed to be a clear difference between CS+ trials within the different trial n-1 contexts (Fig 1D). In fact, this difference seems bigger (and as consistent) as the meaningful/significant differences which are focused on. Interestingly, the RW model predicts a clear n-1 x n interaction however it is not discussed why (Fig 1F). To me it seems that the behaviour (at least partially) and model reflect an interaction between trial n-1 and n during probe trials however this is currently not discussed.

    Exp 1 magnitude of effect: Supplementary figure 1 (raw heading data) reveals that the contextual effects observed in Figure 1 were relatively small. Specifically, the differences between CS+ vs CS- (1-degree) are approx. 6% of the total adaptation that occurred (15-degrees). Whilst this is referred to within the final exp 1 analysis (fig 2B), the magnitude of these contextual (associative learning) are not explicitly discussed. Although I believe the results are important, the article reads as if these conditioning effects were large when in fact other people might conclude that conditioning had little impact on adaptation (as similar adaptation was observed across both contexts (assumed as this data is not currently provided) and performance looks very similar to exp 2 where there was no 0-degree context).

    Exp 2 results: It was unclear why the RW model would predict a negative heading angle within the single CS conditions (Fig 3C)? Whilst a weaker conditioning response would be expected due to compound conditioning, would you not expect this still to be positive? Why would the model predict extinction to occur and why is this seen in the behaviour? The details of this result (and the predictions of the model) are currently not discussed.

    What is the CS (confusion between results & discussion)? Between lines 395-410 the authors describe the primary CS as being the heading angle ('the movement plan itself, rather than the target cue, that constitutes the primary CS'), however in the results (lines 74-96) they describe the CS as being the arbitrary cue ('When considered through the lens of classical conditioning, the arbitrary cues are the conditioned stimuli (CSs)') and the CR being the heading angle ('the conditioned response (CR) would be the movement heading angle elicited by a CS'). As a result of this discrepancy, this section of the discussion was confusing (lines 387-410 & then again from line 433). Are the authors saying that sometimes the heading angle/plan is the CS and other times it is the CR...? How does this all align?