A Double Dissociation Between Savings And Long-term Memory in Motor Learning Part 5

Mechanisms for learning rate modulation in temporally-volatile adaptation

What are the mechanisms behind learning rate increases in temporally-volatile adaptation? 

Time is a magical existence. It can heal the pain of the past and change the destiny of the future. In our daily lives, the passage of time often affects our memory. Sometimes we feel like we can't remember a specific moment in time for an event, or we may find that our memory is fading. Although these phenomena may sound worrying time fluctuations have many positive effects on our memory.

First, the passage of time helps us better process and store information. Sometimes we learn something new at some point but don't fully understand it at the time. However, over time, we digest this knowledge and store it in our brains. This process of processing and storing information is extremely important to our memory. In this process, we can gradually consolidate what we learn until it eventually becomes our common sense and skills.

Secondly, time fluctuations can also help us express and tell stories better. We often hear someone say: "I remember I did something interesting last week, but I really can't express it clearly." At this time, the passage of time will help us. Over time, the details of this event will become clear to us, and we can better understand it through continued practice. This way, we can express and tell the story better in future conversations.

Finally, time fluctuations in our memory can also help us better remember experiences and feelings. Some might say that over time, our feelings and experiences become blurred. But in fact, the passage of time also makes our experiences more profound. For example, we may often recall exciting or surprising moments in our past experiences and feelings. These moments will gradually be placed deep in our hearts and become our eternal memories.

In short, time fluctuations have a very close relationship with memory. Although the passage of time may have some effects on our memory, we should view it as a positive force. Over time, we get better at processing and storing information, better at expressing and telling stories, and better at remembering experiences and feelings. It is because of these positive effects that we can make better use of our powerful memory in our lives. It can be seen that we need to improve memory, and Cistanche deserticola can significantly improve memory because Cistanche deserticola is a traditional Chinese medicinal material that has many unique effects, one of which is to improve memory. The efficacy of minced meat comes from the various active ingredients it contains, including acid, polysaccharides, flavonoids, etc. These ingredients can promote brain health in various ways.

Click know ways to improve brain function

Recent work suggests that such a learning rate increase can be driven by learning environments with increased statistical consistency, defined as a positive correlation between successive perturbations (i.e., lag-1 autocorrelation) in environmental dynamics or errors from one trial to the next [52–54]. 

This consistency-driven effect is further enhanced when the environment repeats the same perturbation [54], with highly consistent, repetitious switching environments increasing learning rates up to 3× from baseline and highly inconsistent ones decreasing them up to 5×. 

Critically, the initial training periods in savings paradigms, including the current one, are characterized by both consistency and repetition, as they usually consist of a large number of trials with the same perturbation, leading to an increase in the learning rate during retraining.

Savings and explicit adaptation

Experiment 3 revealed that the temporally-volatile savings we observe arise from implicit adaptation. This adds to recent evidence [19,20] against the idea that savings are exclusively driven by explicit adaptation [18,43–45]. 

Although savings can occur because of explicit strategies, we did not observe this in our experiment. While this may, in part, reflect the considerable variability in the balance between implicit and explicit adaptation across individuals [74,86], the lack of explicit savings in our study was likely due to an experimental design that promoted implicit adaptation and minimized explicit strategy, and thus provided little power to detect explicit savings. 

The design elements included the lack of aiming instructions, the absence of workspace markers positioned to aid re-aiming, the use of point-to-point rather than shooting movements, and the engineering of low (approximately 25 ms) visual feedback latency for onscreen cursor motion, all of which may promote implicit learning [76,80–83]. 

In contrast, studies that reversed most of these design elements elicited primarily explicit adaptation and found clear explicit savings [44].

Taken together, the evidence now suggests that both implicit and explicit adaptation can show savings, albeit via different mechanisms. 

The current study shows that implicit savings are specifically driven by the faster relearning of a temporally-volatile memory, whereas previous work provides evidence that explicit savings are driven by temporally-persistent memory, as explicit savings are observable in multi-target paradigms that would minimize temporally-volatile memory [44,45]. 

As a consequence, for experimental paradigms that primarily elicit explicit learning, we would expect savings to primarily be driven by this temporally persistent adaptation, and for experimental paradigms that elicit a balance of implicit and explicit learning, we would expect the dichotomy between temporally persistent and temporally volatile contributions to savings to be blurred. 

However, because we did not examine these cases, we cannot know whether our expectations will be borne out or whether complex interactions between implicit and explicit learning might lead to different results that cannot be predicted from the current findings.

Parallels between temporally volatile/temporally persistent learning and the fast/slow learning processes of motor adaptation

Another line of work has dissected motor adaptation, not experimentally, but instead based on a computational model with 2 distinct adaptive processes: a fast adaptive process that learns rapidly and displays weak retention, and a slow adaptive process that learns slowly and displays strong retention [10]. 

By manipulating the training duration to elicit different amounts of fast and slow learning, a subsequent study found that 24-hour retention was specifically predicted by the amount of slow learning, rather than by the amount of fast learning or overall adaptation [60]. 

Interestingly, this model-based dissection mirrors our temporal-stability– based dissection as the slow process, like temporally-persistent adaptation, leads to 24-h retention, whereas the fast process, like temporally-volatile adaptation, does not. 

There is a remarkable correspondence between the Joiner and colleagues study, which found that 49 ± 6% (95% confidence) of slow learning on day 1 is retained after 24 h, and Experiment 4 in the current study, which found that 46 ± 9% of persistent learning on day 1 is retained after 24 h. Moreover, the trial-to-trial learning characteristics of the fast and slow processes mirror the ones for volatile and persistent adaptation, respectively. In particular, slow adaptation displays slower learning and better retention than fast adaptation, just as temporally persistent adaptation displays slower learning and better retention than temporally volatile adaptation (see Figs 3C, 3D, and 2A, respectively). 

These parallels argue, although speculatively so, that the temporally volatile and temporally persistent learning from our dissection of adaptation corresponds to the implicit components of fast and slow processes from the two-state model. This possibility challenges 2 prominent ideas from the recent literature. First, the possible correspondence between the fast process and implicit temporally volatile learning challenges the idea that fast-process learning is synonymous with explicit adaptation [18,43,87]. 

Second, the possibility of a measurable instantiation of fast and slow process learning from the two-state model challenges the assertion that models of context-based learning and the switching between should supplant models of adaptive processes with different learning rates [41].

The coexistence of temporally-volatile and temporally-persistent memories provides a mechanism for contextual interference

The possible mapping of temporally volatile adaptation onto a fast learning/low retention process and of temporally persistent adaptation onto a slow learning/high retention process provides an intriguing potential explanation for previous work on contextual interference in both motor and cognitive tasks. Contextual interference refers to the phenomenon that memories formed in high-interference environments, where the task being performed is randomly switched from one trial to the next, are learned more slowly but show higher retention than memories formed in low-interference environments, where a single task is serially practiced [88–93]. 

If, as in the VMR adaptation task studied here, both temporally volatile and temporally persistent memories contribute to learning in tasks where contextual interference has been observed, then contextual interference effects can be predicted based solely on the temporal spacing inherent in the paradigms that elicit it. 

The idea here is that the high-interference condition in which tasks are randomly intermingled from one trial to the next would necessarily increase the temporal spacing between the trials within each task compared to the low-interference condition in which tasks are serially practiced. This increase in temporal spacing would allow temporally-volatile memories to decay, at least partially, and thus reduce the amount of temporally-volatile learning, which would slow the overall learning and promote increased temporally persistent learning. 

Moreover, the resulting reduction in temporally-volatile learning and increase in temporally-persistent learning would both act to increase the proportion of learning that is temporally-persistent in the high-interference condition, which would, in turn, increase long-term retention according to the current findings. The slowed overall learning and increased retention predicted here for the high-interference condition are, in fact, the defining features of contextual interference. 

Correspondingly, the converse, faster overall learning but reduced retention, would be predicted for the low-interference condition where serially practiced tasks with reduced temporal spacing would allow temporally volatile memories to rapidly build during training to improve performance, but would decay before a retention or transfer test, resulting in poor retention. 

Thus, the coexistence of temporally volatile and temporally persistent memories explains contextual interference that does not require any interference itself. Further work will be required to determine the fraction of observed contextual interference effects that stem from this mechanism.

Materials and methods

Ethics statement

This study was approved by the Harvard University Committee on the Use of Human Subjects (CUHS). Participants were naïve concerning the purpose of the experiments and provided written informed consent by CUHS policies.

Participants

A total of 118 subjects (48 men, age 22.6 ± 4.7, 13 left-handed) participated in the present study (20 each in Experiments 1 and 2, 12 in Experiment S1, 41 in Experiment 3, and 25 in Experiment 4).

Apparatus

We used the same experimental setup as the one used in recent work [62,63]. Subjects sat in front of an apparatus consisting of a 200 Hz digitizing tablet (Wacom Intuos 3 12" × 19", resolution of position data: 0.005 mm; accuracy: 0.25 mm) positioned below a 23" 120 Hz LCD monitor. 

During the experiment, subjects moved a custom-made handle, which contained a stylus, on top of the tablet allowing us to record hand position. Vision of the hand was occluded by the monitor and subjects instead observed their movement on the screen through a white cursor representing hand position.

For more information:1950477648nn@gmail.com