(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



A double dissociation between savings and long-term memory in motor learning [1]

['Alkis M. Hadjiosif', 'John A. Paulson School Of Engineering', 'Applied Sciences', 'Harvard University', 'Cambridge', 'Massachusetts', 'United States Of America', 'J. Ryan Morehead', 'School Of Psychology', 'University Of Leeds']

Date: 2023-05

Memories are easier to relearn than learn from scratch. This advantage, known as savings, has been widely assumed to result from the reemergence of stable long-term memories. In fact, the presence of savings has often been used as a marker for whether a memory has been consolidated. However, recent findings have demonstrated that motor learning rates can be systematically controlled, providing a mechanistic alternative to the reemergence of a stable long-term memory. Moreover, recent work has reported conflicting results about whether implicit contributions to savings in motor learning are present, absent, or inverted, suggesting a limited understanding of the underlying mechanisms. To elucidate these mechanisms, we investigate the relationship between savings and long-term memory by experimentally dissecting the underlying memories based on short-term (60-s) temporal persistence. Components of motor memory that are temporally-persistent at 60 s might go on to contribute to stable, consolidated long-term memory, whereas temporally-volatile components that have already decayed away by 60 s cannot. Surprisingly, we find that temporally-volatile implicit learning leads to savings, whereas temporally-persistent learning does not, but that temporally-persistent learning leads to long-term memory at 24 h, whereas temporally-volatile learning does not. This double dissociation between the mechanisms for savings and long-term memory formation challenges widespread assumptions about the connection between savings and memory consolidation. Moreover, we find that temporally-persistent implicit learning not only fails to contribute to savings, but also that it produces an opposite, anti-savings effect, and that the interplay between this temporally-persistent anti-savings and temporally-volatile savings provides an explanation for several seemingly conflicting recent reports about whether implicit contributions to savings are present, absent, or inverted. Finally, the learning curves we observed for the acquisition of temporally-volatile and temporally-persistent implicit memories demonstrate the coexistence of implicit memories with distinct time courses, challenging the assertion that models of context-based learning and estimation should supplant models of adaptive processes with different learning rates. Together, these findings provide new insight into the mechanisms for savings and long-term memory formation.

Remarkably, we find that savings is driven not by the reemergence of temporally-persistent motor memories, but instead by faster relearning of temporally-volatile memories. We go on to find that these temporally-volatile memories responsible for savings in our paradigm represent implicit, rather than explicit, adaptation. When we measure the long-term retention of the temporally-persistent and temporally-volatile components, however, we find that it is temporally-persistent adaptation, not temporally-volatile adaptation, that leads to long-term memory. Together, these findings demonstrate a clear double dissociation between savings and long-term memory.

Here, we compare the mechanisms that lead to savings and those that lead to the formation of stable, long-term motor memories. We hypothesized that dissecting motor adaptation into specific memory components based on temporal persistence could shed light into these mechanisms. We use short 60-s time delays to dissect overall adaptation into 2 components: temporally-volatile adaptation, which would decay during this time delay, and temporally-persistent adaptation, which would survive the delay [ 56 – 59 ]. Since it decays away in just 60 s, temporally-volatile adaptation would not lead to long-term memory that is associated with stability across timescales that are orders of magnitude greater [ 9 , 25 , 60 ]. Hence, mechanisms leading to long-term memory would be contained within temporally-persistent adaptation, and thus, consolidation-dependent savings would predict faster learning solely for temporally-persistent adaptation. Conversely, if we found faster relearning solely for temporally-volatile adaptation, that would indicate a savings mechanism that is not driven by long-term memory.

What could explain this apparent discrepancy? A possibility is that implicit adaptation may not be monolithic. It may consist of distinct components that are relearned at different rates and differentially elicited in these different paradigms. In fact, an intriguing alternative to recall-based mechanisms is that savings arises from changes in learning rate [ 11 , 15 , 23 ], an idea reinforced by recent work which has demonstrated that the rate at which learning occurs is systematically modulated by specific characteristics of the learning environment. These characteristics include the amount of task-relevant variability present before learning [ 46 , 47 ], the balance between sensory uncertainty and uncertainty about state estimation [ 48 , 49 ], prior exposure to perturbations characterized by a similar covariance structure between learning parameters [ 50 , 51 ], prior exposure to similar motor errors [ 21 , 52 ], and the trial-to-trial consistency of the learning environment [ 53 , 54 ]. In particular, high-consistency environments, whereby perturbations tend to persist from one trial to the next and thus confer more predictability to the imposed perturbation, can strongly increase learning rates (up to 3×). This is a critical finding as far as the study of savings is concerned: The adaptation paradigms used to study savings usually consist of the same perturbation being active for a large number of trials (usually 60 to 120), resulting in highly consistent errors, which would in turn lead to strongly increased learning rates [ 52 – 55 ]. This mechanism could lead to savings by enabling the faster relearning of a short-term memory, as opposed to the reemergence of a long-term, stable memory. This is consistent with recent results indicating that prior experience with high-consistency errors, as opposed to the repetition of successful actions, leads to savings [ 21 ].

In line with this idea, a recent but influential view has proposed that savings in motor adaptation specifically results from the recall of an explicit strategy [ 18 , 43 , 44 ], whereas the implicit component of visuomotor learning does not contribute to savings [ 44 ]. These studies provided clear evidence for explicit savings, but the paradigms they used elicited little implicit adaptation, limiting the power to assess implicit savings. More recent studies, which elicited greater implicit adaptation, have led to disparate findings, concluding that implicit adaptation is either faster during relearning [ 19 , 20 ] or slower [ 45 ]. The Avraham [ 45 ] and Albert [ 20 ] studies are of particular interest, as they were both designed to isolate implicit adaptation (albeit using different paradigms) and yet reached opposite conclusions.

Previous research has generally maintained that savings results from the recall of a previously consolidated long-term memory [ 14 , 24 , 25 ]. The ability to form enduring long-term memories is one of the most remarkable biological abilities, with some memories lasting a lifetime despite incessant neural plasticity. Elderly Danes recalled mundane details—such as the weather—surrounding the news of their country’s invasion and liberation during WWII more than 50 years later [ 26 ], and similar examples of memories surrounding shocking and/or consequential events abound [ 27 – 29 ]. People recognize high school classmates decades later [ 30 ], tacitly use synthetic grammar rules years after training [ 31 ], and identify specific images months [ 32 ] or even years [ 33 ] later. Long-term memories emerge through consolidation, a process driven by molecular mechanisms that facilitate synaptic activity [ 34 , 35 ] and can be mediated through the hippocampus [ 36 , 37 ]. In motor tasks, the presence or absence of savings itself has often been taken as a litmus test for whether a previously trained memory has been consolidated, at timescales ranging from hours to months [ 14 , 24 , 25 ]. Consequently, in terms of underlying mechanisms, savings has been further suggested to result from the following: (1) the unmasking of a slower-learning, strong-retention process in a multi-rate learning model [ 10 ]; (2) context- or relevance-based switching between such multiple slow processes, each specific to a different memory [ 38 – 41 ]; or (3) reverting to the memory of a previously learned motor plan that was reinforced by success or mere repetition [ 16 , 42 ]. All of these proposed mechanisms focus on savings as the manifestation of a latent, stable, consolidated motor memory that is robust to both interference and the passage of time.

Memories, both declarative and procedural, are easier to relearn than to learn from scratch. This advantage, known as savings, was first appreciated in Hermann Ebbinghaus’s seminal work [ 1 ], in which he observed that relearning a forgotten list of words was faster than learning a novel list. Savings has since been demonstrated in a plethora of different paradigms, including cognitive tasks in humans [ 2 , 3 ], operant conditioning in animals [ 4 – 6 ], and motor tasks in humans such as saccade adaptation [ 7 ], force-field adaptation [ 8 – 11 ], visuomotor adaptation [ 12 – 21 ], and gait adaptation [ 22 , 23 ].

Results

We designed a set of experiments to elucidate the mechanisms for savings and long-term memory and investigate the relationship between them. We began by investigating whether savings, the faster relearning of a previously learned adaptation, is driven by the reemergence of a previously consolidated temporally-persistent memory, or by a propensity for faster acquisition of a transient, temporally-volatile memory. In particular, we created a paradigm to dissect initial adaptation, the washout of adaptation, and savings in readaptation into temporally-persistent and temporally-volatile components. We first investigated the dynamics by which temporally-persistent and temporally-volatile memories decay during a washout period following initial adaptation, as savings can arise from the incomplete washout of a component of adaptation [10,61]. This allowed us to compare the rates of unlearning for temporally-persistent and temporally-volatile adaptation during washout, and critically, to measure the initial value of both temporally-persistent and temporally-volatile memories prior to readaptation, so that savings could be accurately assessed for both. We next examined how savings depends on temporally-persistent and temporally-volatile memories by measuring savings separately for these 2 components of motor adaptation, allowing us to determine whether one of these memories is specifically responsible for savings. We then investigated whether long-term memory, measured as the retention of a previously trained adaptation 24 h later, is associated with the temporally-persistent or the temporally-volatile component of adaptation.

Measuring temporally-volatile and temporally-persistent contributions to savings For Experiments 1 and 2, we recruited N = 40 subjects and trained them on a 30° visuomotor rotation (VMR) [12,13,15–18,62–64] (Fig 1A and 1B). After 80 trials of initial training, subjects were tested for savings after either a short (40-trial) washout period, which was previously reported to be sufficient for the washout of overall VMR adaptation [15], or a longer (800-trial) washout period we employed to effect a more definitive washout. We also used the data from this 800-trial washout to trace out the time course of unlearning for both the temporally-persistent and temporally-volatile components of adaptation; this unlearning would encompass both active unlearning (i.e., relearning the baseline behavior) and natural trial-to-trial decay of adaptation [17,65–68]. Each subject experienced both types of washout duration following training (see Fig 1C). In Experiment 1 (N = 20), the short washout period was presented first and the long washout period second. In Experiment 2 (N = 20), this order was flipped (Fig 1C, for a detailed description see Materials and methods). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Experiment setup and training paradigm. (a) Experiment setup. Subjects made point-to-point reaching movements on a digitizing tablet and received continuous visual feedback on a screen mounted above it. (b) VMR training. During baseline (left), the cursor follows the hand motion, whereas during training, cursor motion is skewed by 30° from the hand motion (in this example, counter-clockwise), resulting in a 30° error before adaptation (middle). If full adaptation is achieved, hand motion must completely counter the imposed rotation, corresponding to a 30° clockwise hand motion in this example (right). (c) Top: experiment schedule and raw data for Experiment 1. There were 3 phases: a baseline period followed by the initial 80-trial VMR training (average adaptation level shown in gray); a short, 40-trial washout period followed by retraining (green); and a long, 800-trial washout period followed by another 80-trial retraining session (blue). Red dashed vertical lines indicate trials conducted after 60-s delays to isolate temporally-persistent adaptation. Brown dashed vertical lines indicate trials following rest breaks. Note that, during the washout periods, adaptation peaks on these delay and rest break trials, illustrating a slower washout for temporally-persistent vs. temporally-volatile adaptation. Bottom: same but for Experiment 2, where the long, 800-trial washout period came first. Error bars indicate SEM. Underlying data supporting this panel can be found in file Exp_1_2_data.mat. (d) Dissection of adaptation into temporally-persistent and temporally-volatile components. Throughout the experiment, we used 60-s delays to allow temporally-volatile adaptation to decay. The amount of adaptation on the trial following such a delay (open gray circle) was taken as a measure of temporally-persistent adaptation, whereas the average amount of adaptation 2 trials before and 2 trials after this post-delay trial (filled gray circles) was taken as a measure of overall adaptation. Temporally-volatile adaptation was operationally defined as the difference between overall and persistent adaptation. https://doi.org/10.1371/journal.pbio.3001799.g001 Throughout these experiments—during both learning and washout—we occasionally inserted 1-min delays which would allow for temporally-volatile adaptation to decay. Since the 1-min delays we imposed amount to 2.5 to 4× the time constant for decay of temporally-volatile adaptation [56–58], approximately 95% decay of volatile adaptation would be expected, effectively isolating temporally-persistent adaptation. We operationally defined temporally-persistent adaptation as the adaptation measured during the post-delay trial and temporally-volatile adaptation as the difference between overall adaptation (itself taken as the average adaptation in the 2 preceding and 2 proceeding non-delay trials) and temporally-persistent adaptation (Fig 1D, also see Materials and methods).

Temporally-persistent adaptation washes out more slowly than overall adaptation The data from the long, 800-trial washout period allowed us to carefully examine the time course of unlearning for both the overall adaptation and for the temporally-persistent component of it. Analysis of the washout curves revealed that overall adaptation displayed rapid unlearning; however, persistent adaptation (circles in Fig 2A) was unlearned much more slowly. We found that by trials 16 to 25, labeled as “early washout” in Fig 2A, overall adaptation had already dropped below 10% of the pre-washout asymptotic adaptation level, whereas about 40% of pre-washout persistent learning remained. By trials 51 to 150, labeled as “mid washout,” overall adaptation had dropped below 3%, whereas about 20% of persistent adaptation still remained (Fig 2A, see inset). Correspondingly, we found the retention of persistent learning to be significantly greater than overall learning in both early washout (t(23) = 4.8, p = 6.9 × 10−5 and mid washout periods (t(39) = 8.5, p = 1.9 × 10−10). To quantify the rate of unlearning during washout for both overall and persistent adaptation, we fit single exponential decay functions to the washout data (see Materials and methods). This revealed the time constants for unlearning to be 6-fold slower for temporally-persistent adaptation than for overall adaptation (median time constant estimated using bootstrap: 106.0 trials, interquartile range (IQR) [92.6 to 121.9] versus 17.4 trials, IQR [15.1 to 20.1], p < 10−4, Fig 2A), in line with the higher retention we observed in the early- and mid-washout data. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Temporally-persistent adaptation washes out more slowly than overall adaptation. (a) Washout curves for the overall adaptation (shading indicates mean ± SEM) and temporally-persistent adaptation (circles) for both Experiment 1 (blue) and Experiment 2 (light blue), illustrating the contrast between rapid washout for overall adaptation and slower washout for temporally-persistent adaptation. The thick dashed or dotted lines indicate exponential fits. Inset: Three different measures of washout for overall vs. persistent adaptation. Retention expressed as a percentage of asymptote adaptation after 16–25 trials (left) or 51–150 trials (center) indicates slower washout for temporally-persistent compared to overall adaptation. Time constants for the washout curves (right) also show slower washout for temporally-persistent adaptation. *p < 0.05, **p < 0.01, ***p < 0.001. (b) Residual adaptation before initial training (gray, left bar in each cluster), at the end of the 40-trial washout period (green, middle bar) and at the end of the 800-trial washout period (blue, right bar). The data show that the 40-trial washout period leaves a significant amount of temporally-persistent adaptation and a smaller but also significant amount of overall adaptation. Consequently, retraining after only 40 washout trials starts from a nonzero baseline. *p < 0.05, **p < 0.01, ***p < 0.001. Underlying data supporting this figure can be found in file Exp_1_2_data.mat. https://doi.org/10.1371/journal.pbio.3001799.g002 As a minor point, we also noticed that unlearning curves for temporally-persistent adaptation display somewhat greater retention during the early and mid-washout in Experiment 1 compared to Experiment 2 (Fig 2A). This might reflect the difference in the amount of training between the 2 conditions, as 2 training blocks (160 trials in total) preceded this washout period in Experiment 1, whereas only a single training block (80 trials) preceded this washout in Experiment 2, due to the condition balancing (see Fig 1). In summary, we found that temporally-persistent adaptation is unlearned at a considerably slower rate than overall adaptation.

Residual adaptation prior to the onset of retraining One consequence of the slower unlearning of persistent compared to overall adaptation is that, while the washout of overall adaptation can appear complete after a short 40 to 100 trial per direction washout period [15,16,18,24], substantial temporally-persistent adaptation can nevertheless remain. This suggests that longer washout periods may be required to examine savings independent of the effect of residual temporally-persistent adaptation and that measuring this residual adaptation prior to relearning may facilitate a better understanding of relearning behavior. When we measured the residual overall and persistent adaptation before the onset of retraining, we found significant levels of both overall and temporally-persistent adaptation for the 40-trial washout but no significant residuals of the previous adaptation following the 800-trial washout period. In particular, we found small but significant residuals for overall adaptation at the end of the 40-trial washout periods, around 5% of pre-washout levels (1.34 ± 0.37°, t(39) = 3.6, p = 0.00087 for Experiments 1 and 2 combined, with positive values indicating adaptation in the direction of the previously imposed VMR, Fig 2B). Note that these residuals were related to previous adaptation rather than a movement direction bias because both Experiments 1 and 2 were balanced, with 10 participants trained with clockwise, and 10 with counter-clockwise VMRs for each experiment. The residuals were even larger for temporally-persistent adaptation, in line with the substantially slower unlearning of temporally-persistent adaptation compared to overall adaptation we observed. In particular, we found that the residual persistent adaptation before the end of the 40-trial washout was around 25% of pre-washout persistent adaptation (4.91 ± 0.49° for Experiments 1 and 2 combined, t(39) = 10.0, p = 2.7 × 10−12). In contrast, the 800-trial washout period was sufficient to bring both overall adaptation and temporally-persistent adaptation back to baseline, with measured residuals of only 0% to 2% of pre-washout levels on average. These residuals were not consistently in the direction of the pre-washout adaptation and were not statistically significant. Overall adaptation at the end of the 800-trial washout was −0.00 ± 0.18° (t(39) = −0.01, p = 0.9955), whereas temporally-persistent adaptation was 0.13 ± 0.32° (t(39) = 0.44, p = 0.6772), as shown in Fig 2B. These results show that a prolonged washout period is required to eliminate residual temporally-persistent adaptation. As washout periods in previous experimental work on savings [15,16,18,24,64] are typically much shorter than the 800-trial washout period we examined, it is likely that the savings observed in these studies is, at least in part, driven by interactions between different components of adaptation that were not fully washed out prior to retraining—apparent savings—as suggested in Smith and colleagues [10]. In order to examine faster relearning that is not contaminated by such interactions—that is, examine true savings—one should ideally eliminate residual levels of overall, temporally-persistent, and temporally-volatile adaptation, or, at least, take these residual levels into account.

Savings arises from the faster acquisition of temporally-volatile memories Remarkably, we found that the savings observed in overall adaptation was due to temporally-volatile learning. We calculated savings for temporally-volatile adaptation (bottom row in Fig 3E and 3F) based on normalized volatile adaptation (bottom row of Fig 3C and 3D), which was computed as the difference between the normalized overall and normalized persistent adaptation (top row of Fig 3C and 3D). We found that volatile adaptation during early training (trial 10) was 2- to 3-fold faster for retraining than for initial training after both the short and long washout periods in both experiments (as shown in the bottom row of Fig 3C and 3D). Specifically, we found that trial 10 volatile readaptation was 52.2 ± 3.8% of the ideal adaptation to the 30° VMR with data after both types of washout combined versus 22.0 ± 4.2% for initial adaptation, t(38) = 6.2, p = 1.4 × 10−7 (readaptation for 40-trial washout data: 51.3 ± 4.5%, t(37) = 5.7, p = 9.0 × 10−7 for savings; readaptation for 800-trial washout data: 52.1 ± 4.1%, t(38) = 5.6, p = 9.9 × 10−7 for savings). This indicates substantial, statistically significant savings in temporally-volatile adaptation as illustrated in the bottom row of Fig 3E and 3F.

Savings does not arise from the rapid reemergence of temporally-persistent memories Intriguingly, the clear pattern of savings we found in the learning curves for overall and temporally-volatile adaptation was absent for temporally-persistent adaptation. In only 1 of the 4 conditions in Experiments 1 and 2 (readaptation after a 40-trial washout in Experiment 1) was the unnormalized temporally-persistent adaptation even nominally higher during relearning than initial adaptation, and in that condition the readaptation built upon a substantially higher pretraining level than the corresponding initial training condition (Fig 3A). When pretraining levels of persistent adaptation were taken into account by normalizing the learning curves, we found that relearning for temporally-persistent adaptation was nominally slower, rather than faster, than initial learning in all 4 conditions as shown in Fig 3C and 3D. Specifically, early (trial 10) savings were, on average −10.0 ± 4.3% of the ideal persistent adaptation, t(38) = −2.3, p = 0.99 for savings (40-trial washout data: −7.3 ± 4.4%, t(37) = −1.7, p = 0.95; 800-trial washout data: −10.3 ± 4.5%, t(38) = −2.3, p = 0.99) as shown in Fig 3E and 3F. The temporally-persistent adaptation measured 40 and 70 trials into the training period for the combined 40-trial and 800-trial washout data displays results similar to trial 10 adaptation, with a tendency towards anti-savings (slower readaptation) (t(39) = −3.4, p = 1.00 for trial 40 and t(38) = −2.2, p = 0.98 for trial 70). The absence of savings in temporally-persistent adaptation stands in stark contrast to the high levels of savings observed in temporally-volatile adaptation, suggesting that overall savings arises from the former, but not the latter. Thus, our result indicates that savings arises from the faster relearning of volatile memories, rather than the re-manifestation of persistent memories.

Temporally-persistent memories display anti-savings Based on recent work that reported anti-savings for implicit motor adaptation [45], we asked, in a post-hoc analysis, whether temporally-persistent adaptation consistently displayed the slowed relearning that would constitute anti-savings. This analysis revealed that, relearning for persistent adaptation was, in fact, significantly slower than initial learning (t(38) = −2.3, p = 0.0247, 2-tailed paired t test), based on the trial 10 the data from both washout periods combined. This anti-savings was most clear in the long 800-trial washout data, which allowed us to examine savings without any effects of residual temporally-persistent adaptation (t(38) = −2.3, p = 0.0276, 2-tailed paired t test). Savings at trial 10 after the short incomplete washout was also nominally negative but, in this case, not significantly so (t(37) = −1.7, p = 0.1073, 2-tailed paired t test). Analysis of temporally-persistent adaptation at trials 40 and 70 provides consistent results, with statistically significant anti-savings observed for the combined data from the 40 and 800-trial washout periods (t(39) = −3.4, p = 0.0017 at trial 40; t(38) = −2.2, p = 0.0359 at trial 70, 2-tailed paired t tests) and also for the 800-trial washout data analyzed in isolation (t(39) = −3.3, p = 0.0022 at trial 40; t(38) = −3.1, p = 0.0035 at trial 70, 2-tailed paired t tests). Accordingly, the 40-trial washout data analyzed in isolation showed mixed results at these individual time points (t(39) = −2.1, p = 0.0417 at trial 40; t(38) = −0.7, p = 0.5053 at trial 70, 2-tailed paired t tests). However, when the data combined across all time points are used, we find individually significant anti-savings for both washout periods (t(39) = −5.0, p = 0.000012, for the 800-trial data; t(39) = −2.6, p = 0.0131 for the 40-trial data). In sum, the negative savings results we observe in the 800-trial and 40-trial washout data are similar, but it appears that the 800-trial result is somewhat clearer, possibly because the 40-trial washout data suffer from incomplete washout of the initial adaptation before relearning. Overall, our data show a conspicuous absence of savings in the relearning of temporally-persistent adaptation in all conditions we examined, instead showing anti-savings despite robust savings in the relearning of temporally-volatile adaptation. Although a small effect, we found it interesting that anti-savings was somewhat more consistently observed following the 800-trial washout condition than the 40-trial condition, suggesting that the prolonged repeated execution of the same no-rotation trials that constitute the washout period might make anti-savings more consistent. Indeed, the strengthening of an action following repeated execution, often termed use-dependent learning (UDL) [69,70], can manifest in reaching in the form of a directional bias toward its direction [70–72]. Interestingly, the expected bias toward the baseline no-rotation movement direction following washout would oppose the movement direction changes associated with VMR relearning, and thus act in the direction of anti-savings to reduce relearning. However, it is critical to note that (1) because we are examining savings, the slowed relearning that constitutes anti-savings refers to slower than initial learning; and (2) that the initial learning period in our experiments was also preceded by a prolonged period of the execution of repeated no-rotation trials, which would likewise elicit a UDL effect. The key question would not, therefore, be whether a UDL effect might slow relearning following the 800-trial washout period, but whether such an effect would show a meaningful size increase between the 220-trial duration of the no-rotation baseline period that precedes initial learning and the 800-trial no-rotation washout period that precedes the 800-trial relearning condition? However, the available literature on how UDL effects increase with the number of repeated trials suggests that this is unlikely. Studies examining UDL effects in reaching movements showed effects after only 1 to 15 trials ([71], Exp 3; [70]), and the one study that looked at the time course for UDL effects beyond 15 trials, found effects that asymptoted between 50 and 150 trials ([71], Fig 4), which is smaller than the duration of the 220-trial baseline that preceded initial learning in our experiment. This suggests that UDL effects, if they indeed affect VMR training in our study, would do so equally for both initial learning, which was preceded by 220 no-rotation trials, and relearning following long-washout, which was preceded by 800 no-rotation trials. Consequently, UDL effects should have little effect on the difference between these learning curves and thus on the savings we measure after 800 washout trials, and are, therefore, unlikely to explain the temporally-persistent anti-savings we observed. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. Temporally-volatile savings are due to an implicit adaptation component. (a) Learning curves for Experiment 3, showing overall adaptation (solid lines), overall implicit adaptation (square) and persistent adaptation (circles). Gray denotes initial learning, blue denotes relearning. Shading indicates SEM. (b) Diagram illustrating measurements of specific components of implicit and explicit learning for Experiment 3, based on a zoomed-in view of the blue retraining data from panel a. Instructions 1 trial before and 1 trial after the 1-min delay (trials 9 and 10) allow the direct measurement of overall-implicit and implicit-persistent components, respectively. This allows measurement of implicit-volatile as the difference between them, and overall-explicit adaptation as the difference between overall adaptation and overall-implicit, where overall adaptation is measured as the average amount of adaptation 2 trials before the first instruction trial and 2 trials after the last persistent-only trial (i.e., trials 7, 8, 12, and 13; blue filled circles). Both instruction trials have no visual feedback to avoid per-trial learning leading to posttrial recovery of adaptation. The following trial (trial 11) thus allows direct measurement of combined implicit and explicit persistent adaptation, therefore, the difference between trial 11 and trial 10 measures explicit-persistent. Trial 11 also reintroduces visual feedback, allowing adaptation to recover by the next trial. (c) Savings in overall, temporally-persistent, and temporally-volatile adaptation in Experiment 3, both for combined implicit and explicit adaptation (left column) and broken into implicit and explicit components. In line with Experiments 1 and 2, data show overall and volatile, but not persistent, savings for combined (implicit + explicit) adaptation. Further dissociation into implicit and explicit components reveals this savings is due to implicit, temporally-volatile adaptation. Error bars indicate SEM. *p < 0.05; **p < 0.01. Underlying data supporting this figure can be found in file Exp_3_data.mat. https://doi.org/10.1371/journal.pbio.3001799.g004 However, the one paper that looked at the effect of repeat duration [71] did not study UDL training periods as long as 800 trials (540 trials was their maximum) and investigated UDL outside the context of VMR adaptation. We thus performed an additional experiment (Experiment S1) to determine whether the differences between the number of no-rotation reaches before initial learning (220 trials) versus before the long-washout relearning (800 trials) might explain the anti-savings we observe. Experiment S1 examined initial learning after a baseline of 800 rather than 220 trials to match the duration of the action selection history of the 800-trial washout before relearning. If the reduction in temporally-persistent relearning with respect to initial learning we observed were indeed due to the longer 800 trial movement/action selection history, we would expect the initial temporally-persistent learning in this new dataset to match the slowed temporally-persistent relearning from the 800-trial washout condition rather than the initial temporally-persistent learning that we previously observed after 220 baseline trials. However, the results from Experiment S1 instead show that the initial temporally-persistent learning following 800 baseline trials in this new dataset was a closer match to the initial temporally-persistent learning that was previously observed after 220 baseline trials, as this dataset did not show the slower initial temporally-persistent learning that would be predicted by increased UDL following 800 rather than 220 trials (trial 10 learning: 47.1 ± 9.6% for initial learning in the new data versus 40.1% ± 4.2% for initial learning in the previous data, t(49) = −0.8, p = 0.45, see S1 Fig). Instead, as shown in S1 Fig, the data are in line with the Verstynen and Sabes data [71] whereby UDL effects asymptote before 220 trials. In line with this prediction, the temporally-persistent relearning following 800-trial washout trials observed in the previous data was also significantly slowed compared to this new 800-trial baseline data with significant anti-savings observed when the data from trials 10, 40, and 70 were averaged together (t(50) = −3.2, p = 0.0014) and also when the data from these time points were analyzed separately (t(50) = −2.2, p = 0.017 at trial 10; t(50) = −2.5, p = 0.0086 at trial 40; t(50) = −2.2, p = 0.0162 at trial 70). These findings suggest the UDL effects cannot explain the 800-trial temporally-persistent anti-savings we observe. As a side note, if UDL effects were, on the other hand, somewhat lower after 40 trials (the short washout period duration) compared to 220 trials (the baseline period duration) [71], it would lead to a reduction in the amount of UDL-induced slowing for relearning after 40 washout trials compared to the initial learning. This would suggest that, if anything, we underestimated temporally-persistent anti-savings in the 40-trial washout data rather than overestimating it in the 800-trial washout data, perhaps contributing to the less consistent results when this condition was considered in isolation.

Temporally-volatile savings arise from implicit adaptation Previous research associated savings in visuomotor adaptation with the rapid recall of explicit strategies, rather than faster implicit adaptation [18,43,44]. This led us to investigate the contributions of implicit and explicit processes in the temporally-volatile savings we observed in our paradigm. We thus ran Experiment 3 (N = 40), which consisted of two 80-trial learning episodes separated by 800 washout trials. We dissected savings into implicit and explicit components using special instruction trials that prompted participants to disengage any explicit strategy by aiming their hand directly to the target [44,73–78]. These instructions were presented immediately before and after the first (trial 10) 60-s time delay following the onset of the VMR in both initial learning and relearning and allowed us to dissect adaptation into 4 subcomponents: implicit-persistent, implicit-volatile, explicit-persistent, and explicit-volatile (Fig 4B, see Materials and methods for details). In line with our findings in Experiments 1 and 2, we found savings for overall and volatile adaptation (14.3 ± 3.6%, t(39) = 4.0, p = 0.00014 and 11.2 ± 4.8%, t(37) = 2.4, p = 0.0119, correspondingly) but not persistent adaptation (4.0 ± 4.9%, t(38) = 0.8, p = 0.21). Dissection of savings into explicit and implicit components revealed savings for both overall implicit and implicit-volatile adaptation (14.1 ± 5.7%, t(38) = 2.5, p = 0.0088 and 13.3 ± 5.5%, t(37) = 2.4, p = 0.0104, correspondingly) but not explicit-volatile adaptation (−2.1 ± 6.2%, t(37) = −0.3, p = 0.63) or any of the persistent subcomponents (implicit-persistent: 3.8 ± 4.7%, t(38) = 0.8, p = 0.21; explicit-persistent: 0.2 ± 4.4%, t(38) = 0.1, p = 0.48). This finding suggests that overall savings were driven by the implicit and temporally-volatile component of adaptation, in turn suggesting that the temporally-volatile savings we observed in Experiments 1 and 2 predominantly reflect an implicit process rather than an explicit strategy. That the volatile component observed in Experiments 1 and 2 is primarily implicit is not surprising: First, it is unclear why an explicit strategy could be temporally-volatile to the point of being largely or completely forgotten after a short 1-min delay. In fact, our recent work indicates that explicit adaptation displays essentially no temporal volatility, with over 95% stability across 1-min delays [79]. Second, our paradigm elicited scant explicit adaptation (likely due to elements of our experiment design aimed at inducing implicit learning such as the use of point-to-point (rather than shooting) movements, the lack of aiming instructions, the lack of markers that could aid off-target aiming, and the presence of low-latency online feedback [76,80–83]) and without substantial explicit adaptation we lacked power for measuring explicit savings.

Dissecting long-term memory in visuomotor adaptation We next investigated whether the ability to dissect motor learning into temporally-persistent and temporally-volatile components could shed light on the mechanisms for the formation of long-term memories. To accomplish this, we examined the relationship between the levels of temporally-persistent and temporally-volatile learning observed after initial training and the amount of retention observed 24 h later (Experiment 4). After a baseline period, we trained 25 participants on a 30° VMR for 120 trials. After this initial training, they were tested for temporally-persistent adaptation as present after a rest break (average break duration: 125 ± 8 s, which would let >99% of temporally-volatile adaptation decay based on a time constant of approximately 20 s). The above measurements were then repeated, with participants retrained for 60 trials, and retested for temporally-persistent adaptation (the average of these 2 measurement sessions was used to quantify temporally-persistent adaptation for each individual). Participants then returned the following day to be tested for retention (Fig 5A, see Materials and methods). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 5. Measuring temporally-volatile and temporally-persistent components of adaptation and subsequent long-term retention. (a) Experiment schedule and raw data for Experiment 4. After a baseline period, subjects were trained with a 30° VMR for 120 trials, and were then tested, after a break, for temporally-persistent adaptation, retrained for 60 trials, and then retested after another break. Subjects returned the following day when they were tested for 24-h retention (orange circle). Note that 24-h retention is lower than overall or temporally-persistent adaptation but higher than zero. TP: temporally-persistent; TV: temporally-volatile. Yellow background indicates trials used to measure overall adaptation. Error bars and shading indicate SEM. (b) Comparison of overall, temporally-persistent, and temporally-volatile adaptation from Experiments 1, 2, and 4 with the 24-h retention from Experiment 4. Experiments 1, 2, and 4 display similar levels of persistent and volatile adaptation. Underlying data supporting this figure can be found in file Exp_4_data.mat. https://doi.org/10.1371/journal.pbio.3001799.g005 We found that, the overall adaptation measured late in training (the last 20 trials) in Experiment 4 was similar to that observed in Experiments 1 and 2 (27.4 ± 0.3° for Experiment 4 versus 26.4 ± 0.6° and 27.7 ± 0.5° for Experiments 1 and 2, see Fig 5B). Similarly, the persistent component of adaptation was also similar across the 3 experiments (16.7 ± 1.0° for Experiment 4 versus 18.1 ± 0.7° and 20.4 ± 1.0° for Experiments 1 and 2, see Fig 5B), suggesting that the somewhat longer training duration in Experiment 4 had little effect on either overall or temporally-persistent adaptation. When examining long-term memory, retained 24 h after training, we found that participants retained 8.9 ± 1.1° of the trained 30° rotation (orange bar in Fig 5B). This corresponded to 32.4 ± 4.2% of the overall learning and 52.7 ± 5.2% of the temporally-persistent learning from day 1.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001799

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.