Goal-directed movements are subject to continuous adaptive control to maintain their accuracy despite changes in the body or the environment (Krakauer et al., 2019), a process known as sensorimotor adaptation. Among the many processes that influence sensorimotor adaptation are savings and anterograde interference (AI; Krakauer et al., 2019). While savings can be defined as faster re-adaptation upon re-exposure to the same perturbation (A →A), AI refers to the phenomenon whereby initial exposure to perturbation A impairs adaptation to a subsequent perturbation B (A →B; Krakauer et al., 2019). Interestingly, recent work has shown that conditions that should yield faster re-adaptation (A →A; e.g., Haith et al., 2015, Morehead et al., 2015) can also trigger AI (Avraham et al., 2021, Hamel et al., 2021, 2022). For example, Avraham et al. (2021) used a clamped visuomotor adaptation protocol and found that exposure to the same perturbation twice reduced bias in both adaptation and aftereffects, suggesting impaired adaptation upon re-exposure. Similarly, Hamel et al. (2021) found that adapting to the same gradually introduced visual deviation twice (A →A) with a short interval between sessions (2 min) led to smaller aftereffects, interpreted as reduced retention and thus AI (Hamel et al., 2022 for an extension of these findings). The results also hinted toward a dose-dependent relationship, as adding a third learning session in close temporal succession further reduced aftereffects (Hamel et al., 2021). As of now, it remains unclear what causes AI in such A →A contexts.

Insights on this issue can be gained by considering alternative frameworks. Relevant to the present question is the sliding threshold model (STM; Keck, Hübener, et al., 2017), a model of homeostatic plasticity that posits that prior neuronal population activity levels dictate their subsequent capacity to undergo long-term potentiation (LTP), and by extension, learn and form memories (Tononi & Cirelli, 2014). To maintain metabolic homeostasis, STM postulates that high prior activity increases the threshold for LTP induction, whereas low prior activity decreases it. As such, prior activity would determine if neuronal activity falls above the synaptic modification range, and, by extension, can support memory formation through synaptic modifications triggered by LTP induction (Lee & Kirkwood, 2019). The STM stresses the importance of the learning history in conditioning subsequent learning and memory capabilities. Assuming that re-exposure to the same visual deviation recruits similar populations of neurons (Landi et al., 2011), and that initial learning increases net activity levels of these neurons (Karbowski, 2019), one tentative hypothesis is that AI arises upon re-exposure when the learning history is extended, due to saturated network-specific LTP capabilities (Rioult-Pedotti et al., 2000, Rioult-Pedotti et al., 2007). Hence, a functional implication of the STM is that learning history would mediate performance upon re-exposure to the same perturbation (A →A), and thus determine the degree of AI that will be observed.

The specific aim of this work was to test this hypothesis. In a fully within-subject and counterbalanced design, participants (n = 24) adapted to the same gradually introduced -20° visual deviation twice over two distinct sessions separated by a 2-min interval (Figure 1). In the first Session, the number of trials in which the visual deviation remained constant at -20° was parameterized across three conditions to manipulate the learning history. Namely, participants performed either 40 trials (Short), 160 trials (Moderate), or 320 trials (Long) compensating for the deviation at asymptotic performance levels. A fourth condition (Jagged) was conducted to determine whether reaching performance asymptote was necessary to trigger AI (Alhussein et al., 2019). In the latter, although the mean perturbation was also -20°, the visual deviation changed continuously for 160 trials to prevent performance stabilization at asymptote. The second Session, which was identical for the 4 conditions, also consisted in a gradually introduced visual deviation of -20°. This session served to evaluate the effect of the first Session on re-learning capabilities. Based on the STM (Keck, Hübener, et al., 2017; Lee & Kirkwood, 2019), it was expected that AI would be increasingly present in the Moderate, Long, and Jagged conditions as compared to the Short one due to the extended learning history. Similarly, the Long condition was expected to present similar levels of AI as compared to the Moderate condition, presumably because LTP capabilities would have already reached saturation (Rioult-Pedotti et al., 2000, Rioult-Pedotti et al., 2007).