Crossmodal to unimodal transfer of temporal perceptual learning

Subsecond temporal processing is crucial for activities requiring precise timing. Here, we investigated perceptual learning of crossmodal (auditory-visual or visual-auditory) temporal interval discrimination (TID) and its impacts on unimodal (visual or auditory) TID performance. The research purpose was to test whether learning is based on a more abstract and conceptual representation of subsecond time, which would predict crossmodal to unimodal learning transfer. The experiments revealed that learning to discriminate a 200-ms crossmodal temporal interval, defined by a pair of visual and auditory stimuli, significantly reduced crossmodal TID thresholds. Moreover, the crossmodal TID training also minimized unimodal TID thresholds with a pair of visual or auditory stimuli at the same interval, even if crossmodal TID thresholds are multiple times higher than unimodal TID thresholds. Subsequent training on unimodal TID failed to reduce unimodal TID thresholds further. These results indicate that learning of high-threshold crossmodal TID tasks can benefit low-threshold unimodal temporal processing, which may be achieved through training-induced improvement of a conceptual representation of subsecond time in the brain.

Shift in excitation-inhibition balance underlies perceptual learning of temporal discrimination

Temporal perceptual learning (TPL) constitutes a unique and profound demonstration of neural plasticity within the brain. Our understanding for the neurometabolic changes associated with TPL on the other hand has been limited in part by the use of traditional fMRI approaches. Since plasticity in the visual cortex has been shown to underlie perceptual learning of visual information, we tested the hypothesis that TPL of an auditory interval involves a similar change in plasticity of the auditory pathway and if so, whether these changes take place in a lower-order sensory-specific brain area such as the primary auditory cortex (A1), or a higher-order modality-independent brain area such as the inferior parietal cortex (IPC). This distinction will inform us of the mechanisms underlying perceptual learning as well as the locus of change as it relates to TPL. In the present study, we took advantage of a new technique: proton magnetic resonance spectroscopy (MRS) in combination with psychophysical measures to provide the first evidence of changes in neurometabolic processing following 5 days of temporal discrimination training. We measured the (E)xcitation-to-(I)nhibition ratio as an index of learning in the right IPC and left A1 while participants learned an auditory two-tone discrimination task. During the first day of training, we found a significant task-related increase in functional E/I ratio within the IPC. While the A1 exhibited the opposite pattern of neurochemical activity, this relationship did not reach statistical significance. After timing performance has reached a plateau, there were no further changes to functional E/I. These findings support the hypothesis that improvements in temporal discrimination relies on neuroplastic changes in the IPC, but it is possible that both areas work synergistically to acquire a temporal interval.

Cognition of Time and Thinking Beyond

A common research protocol in cognitive neuroscience is to train subjects to perform deliberately designed experiments while recording brain activity, with the aim of understanding the brain mechanisms underlying cognition. However, how the results of this protocol of research can be applied in technology is seldom discussed. Here, I review the studies on time processing of the brain as examples of this research protocol, as well as two main application areas of neuroscience (neuroengineering and brain-inspired artificial intelligence). Time processing is a fundamental dimension of cognition, and time is also an indispensable dimension of any real-world signal to be processed in technology. Therefore, one may expect that the studies of time processing in cognition profoundly influence brain-related technology. Surprisingly, I found that the results from cognitive studies on timing processing are hardly helpful in solving practical problems. This awkward situation may be due to the lack of generalizability of the results of cognitive studies, which are under well-controlled laboratory conditions, to real-life situations. This lack of generalizability may be rooted in the fundamental unknowability of the world (including cognition). Overall, this paper questions and criticizes the usefulness and prospect of the abovementioned research protocol of cognitive neuroscience. I then give three suggestions for future research. First, to improve the generalizability of research, it is better to study brain activity under real-life conditions instead of in well-controlled laboratory experiments. Second, to overcome the unknowability of the world, we can engineer an easily accessible surrogate of the object under investigation, so that we can predict the behavior of the object under investigation by experimenting on the surrogate. Third, the paper calls for technology-oriented research, with the aim of technology creation instead of knowledge discovery.

Cross-modal implicit learning of random time patterns

Perception is sensitive to statistical regularities in the environment, including temporal characteristics of sensory inputs. Interestingly, implicit learning of temporal patterns in one modality can also improve their processing in another modality. However, it is unclear how cross-modal learning transfer affects neural responses to sensory stimuli. Here, we recorded neural activity of human volunteers using electroencephalography (EEG), while participants were exposed to brief sequences of randomly timed auditory or visual pulses. Some trials consisted of a repetition of the temporal pattern within the sequence, and subjects were tasked with detecting these trials. Unknown to the participants, some trials reappeared throughout the experiment across both modalities (Transfer) or only within a modality (Control), enabling implicit learning in one modality and its transfer. Using a novel method of analysis of single-trial EEG responses, we showed that learning temporal structures within and across modalities is reflected in neural learning curves. These putative neural correlates of learning transfer were similar both when temporal information learned in audition was transferred to visual stimuli and vice versa. The modality-specific mechanisms for learning of temporal information and general mechanisms which mediate learning transfer across modalities had distinct physiological signatures: temporal learning within modalities relied on modality-specific brain regions while learning transfer affected beta-band activity in frontal regions.

Musical Training Facilitates Exogenous Temporal Attention via Delta Phase Entrainment within a Sensorimotor Network

Temporal orienting of attention plays an important role in our day-to-day lives and can use timing information from exogenous or endogenous sources. Yet, it is unclear what neural mechanisms give rise to temporal attention, and it is debated whether both exogenous and endogenous forms of temporal attention share a common neural source. Here, older adult nonmusicians (N = 47, 24 female) were randomized to undergo 8 weeks of either rhythm training, which places demands on exogenous temporal attention, or word search training as a control. The goal was to assess (1) the neural basis of exogenous temporal attention and (2) whether training-induced improvements in exogenous temporal attention can transfer to enhanced endogenous temporal attention abilities, thereby providing support for a common neural mechanism of temporal attention. Before and after training, exogenous temporal attention was assessed using a rhythmic synchronization paradigm, whereas endogenous temporal attention was evaluated via a temporally cued visual discrimination task. Results showed that rhythm training improved performance on the exogenous temporal attention task, which was associated with increased intertrial coherence within the δ (1-4 Hz) band as assessed by EEG recordings. Source localization revealed increased δ-band intertrial coherence arose from a sensorimotor network, including premotor cortex, anterior cingulate cortex, postcentral gyrus, and the inferior parietal lobule. Despite these improvements in exogenous temporal attention, such benefits were not transferred to endogenous attentional ability. These results support the notion that exogenous and endogenous temporal attention uses independent neural sources, with exogenous temporal attention relying on the precise timing of δ band oscillations within a sensorimotor network.SIGNIFICANCE STATEMENT Allocating attention to specific points in time is known as temporal attention, and may arise from external (exogenous) or internal (endogenous) sources. Despite its importance to our daily lives, it is unclear how the brain gives rise to temporal attention and whether exogenous- or endogenous-based sources for temporal attention rely on shared brain regions. Here, we demonstrate that musical rhythm training improves exogenous temporal attention, which was associated with more consistent timing of neural activity in sensory and motor processing brain regions. However, these benefits did not extend to endogenous temporal attention, indicating that temporal attention relies on different brain regions depending on the source of timing information.

Temporal learning in the suprasecond range: insights from cognitive style

The acquisition of information on the timing of events or actions (temporal learning) occurs in both the subsecond and suprasecond range. However, although relevant differences between participants have been reported in temporal learning, the role of dimensions of individual variability in affecting performance in such tasks is still unclear. Here we investigated this issue, assessing the effect of field-dependent/independent cognitive style on temporal learning in the suprasecond range. Since different mechanisms mediate timing when a temporal representation is self-generated, and when it depends on an external referent, temporal learning was assessed in two conditions. Participants observed a stimulus across six repetitions and reproduced it. Unbeknownst to them, in an internally-based learning (IBL) condition, the stimulus duration was fixed within a trial, although the number of events defining it varied; in an externally-cued learning (ECL) condition, the stimulus was defined by the same number of events within each trial, although its duration varied. The effect of the reproduction modality was also assessed (motor vs. perceptual). Error scores were higher in IBL compared to ECL; the reverse was true for variability. Field-independent individuals performed better than field-dependent ones only in IBL, as further confirmed by correlation analyses. Findings provide evidence that differences in dimensions of variability in high-level cognitive functioning, such as field dependence/independence, significantly affect temporal learning in the suprasecond range, and that this effect depends on the type of temporal representation fostered by the specific task demands.

A Recurrent Neural Network Model Accounts for Both Timing and Working Memory Components of an Interval Discrimination Task

Interval discrimination is of fundamental importance to many forms of sensory processing, including speech and music. Standard interval discrimination tasks require comparing two intervals separated in time, and thus include both working memory (WM) and timing components. Models of interval discrimination invoke separate circuits for the timing and WM components. Here we examine if, in principle, the same recurrent neural network can implement both. Using human psychophysics, we first explored the role of the WM component by varying the interstimulus delay. Consistent with previous studies, discrimination was significantly worse for a 250 ms delay, compared to 750 and 1500 ms delays, suggesting that the first interval is stably stored in WM for longer delays. We next successfully trained a recurrent neural network (RNN) on the task, demonstrating that the same network can implement both the timing and WM components. Many units in the RNN were tuned to specific intervals during the sensory epoch, and others encoded the first interval during the delay period. Overall, the encoding strategy was consistent with the notion of mixed selectivity. Units generally encoded more interval information during the sensory epoch than in the delay period, reflecting categorical encoding of short versus long in WM, rather than encoding of the specific interval. Our results demonstrate that, in contrast to standard models of interval discrimination that invoke a separate memory module, the same network can, in principle, solve the timing, WM, and comparison components of an interval discrimination task.

A supramodal and conceptual representation of subsecond time revealed with perceptual learning of temporal interval discrimination

Subsecond time perception has been frequently attributed to modality-specific timing mechanisms that would predict no cross-modal transfer of temporal perceptual learning. In fact, perceptual learning of temporal interval discrimination (TID) reportedly shows either no cross-modal transfer, or asymmetric transfer from audition to vision, but not vice versa. However, here we demonstrate complete cross-modal transfer of auditory and visual TID learning using a double training paradigm. Specifically, visual TID learning transfers to and optimizes auditory TID when the participants also receive exposure to the auditory temporal interval by practicing a functionally orthogonal near-threshold tone frequency discrimination task at the same trained interval. Auditory TID learning also transfers to and optimizes visual TID with additional practice of an orthogonal near-threshold visual contrast discrimination task at the same trained interval. Practicing these functionally orthogonal tasks per se has no impact on TID thresholds. We interpret the transfer results as indications of a supramodal representation of subsecond time. Moreover, because TID learning shows complete transfer between modalities with vastly different temporal precisions, the sub-second time presentation must be conceptual. Double training may refine this supramodal and conceptual subsecond time representation and connect it to a new sense to improve time perception.

Temporal perceptual learning distinguishes between empty and filled intervals

Temporal perceptual learning (TPL) refers to improved temporal performance as a result of training with sub-second intervals. Most studies on TPL have focused on empty intervals (i.e. intervals marked by two brief stimuli); however, scholars have suggested that filled intervals (i.e. intervals presented as continuous sensory inputs) might have different underlying mechanisms. Therefore, the current study aimed to test whether empty and filled intervals yield similar TPL performance and whether such learning effects could transfer mutually. To this end, we trained two groups of participants with empty and filled intervals of 200 ms for four days, respectively. We found that the empty-interval group clearly improved their timing performances after training, and such an effect transferred to filled intervals of 200 ms. By contrast, the filled-interval group had neither learning nor transfer effect. Our results further shed light on the distinct mechanisms between empty and filled intervals in time perception while simultaneously replicating the classical findings on TPL involving empty intervals.

A novel Android app to evaluate and enhance auditory and tactile temporal thresholds. ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY. IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY. ANNUAL INTERNATIONAL CONFERENCE 2021;2021:5885-5888. [PMID: 34892458 DOI: 10.1109/embc46164.2021.9630028] [Citation(s) in RCA: 0] [Impact Index Per Article: 0]

With this work, we introduce a novel Android app designed to monitor and enhance auditory and tactile temporal sensitivity. To assess the app's reliability, we tested its technical performance evaluating stimuli production's accuracy (i.e., onset, offset, and duration of stimulation). To validate the app with participants we generated temporal intervals, using either sounds or vibratory stimuli, by implementing two versions of a Two-Alternative Forced-Choice (2AFC) task. Auditory and tactile temporal sensitivity of 12 participants was evaluated using this procedure. To investigate whether temporal abilities could be enhanced using the app, participants were then divided into two groups: one group was trained for four days on the auditory temporal task, while the other was trained for four days on the tactile temporal task. Results suggest that the app can i) effectively measure auditory and tactile temporal thresholds and ii) be used to enhance temporal abilities through perceptual learning. The accessibility of the experimental protocols, combined with our findings, fosters the app's involvement in rehabilitation programs, for example, with a specific focus on sensory disabilities that are associated with temporal deficits (e.g., deafness and Parkinson).Clinical Relevance- The current work introduces a novel app that can be used to monitor and improve temporal abilities, in both the auditory and the tactile modalities.

Minimal interplay between explicit knowledge, dynamics of learning and temporal expectations in different, complex uni- and multisensory contexts

While temporal expectations (TE) generally improve reactions to temporally predictable events, it remains unknown how the learning of temporal regularities (one time point more likely than another time point) and explicit knowledge about temporal regularities contribute to performance improvements; and whether any contributions generalise across modalities. Here, participants discriminated the frequency of diverging auditory, visual or audio-visual targets embedded in auditory, visual or audio-visual distractor sequences. Temporal regularities were manipulated run-wise (early vs. late target within sequence). Behavioural performance (accuracy, RT) plus measures from a computational learning model all suggest that learning of temporal regularities occurred but did not generalise across modalities, and that dynamics of learning (size of TE effect across runs) and explicit knowledge have little to no effect on the strength of TE. Remarkably, explicit knowledge affects performance-if at all-in a context-dependent manner: Only under complex task regimes (here, unknown target modality) might it partially help to resolve response conflict while it is lowering performance in less complex environments.

Effects of stimulus and task structure on temporal perceptual learning

Our ability to discriminate temporal intervals can be improved with practice. This learning is generally thought to reflect an enhancement in the representation of a trained interval, which leads to interval-specific improvements in temporal discrimination. In the present study, we asked whether temporal learning is further constrained by context-specific factors dictated through the trained stimulus and task structure. Two groups of participants were trained using a single-interval auditory discrimination task over 5 days. Training intervals were either one of eight predetermined values (FI group), or random from trial to trial (RI group). Before and after the training period, we measured discrimination performance using an untrained two-interval temporal comparison task. Our results revealed a selective improvement in the FI group, but not the RI group. However, this learning did not generalize between the trained and untrained tasks. These results highlight the sensitivity of TPL to stimulus and task structure, suggesting that mechanisms of temporal learning rely on processes beyond changes in interval representation.

Modality-specific and multisensory mechanisms of spatial attention and expectation

In our natural environment, the brain needs to combine signals from multiple sensory modalities into a coherent percept. Whereas spatial attention guides perceptual decisions by prioritizing processing of signals that are task-relevant, spatial expectations encode the probability of signals over space. Previous studies have shown that behavioral effects of spatial attention generalize across sensory modalities. However, because they manipulated spatial attention as signal probability over space, these studies could not dissociate attention and expectation or assess their interaction. In two experiments, we orthogonally manipulated spatial attention (i.e., task-relevance) and expectation (i.e., signal probability) selectively in one sensory modality (i.e., primary modality) (experiment 1: audition, experiment 2: vision) and assessed their effects on primary and secondary sensory modalities in which attention and expectation were held constant. Our results show behavioral effects of spatial attention that are comparable for audition and vision as primary modalities; however, signal probabilities were learned more slowly in audition, so that spatial expectations were formed later in audition than vision. Critically, when these differences in learning between audition and vision were accounted for, both spatial attention and expectation affected responses more strongly in the primary modality in which they were manipulated and generalized to the secondary modality only in an attenuated fashion. Collectively, our results suggest that both spatial attention and expectation rely on modality-specific and multisensory mechanisms.

Musical expertise generalizes to superior temporal scaling in a Morse code tapping task

A key feature of the brain’s ability to tell time and generate complex temporal patterns is its capacity to produce similar temporal patterns at different speeds. For example, humans can tie a shoe, type, or play an instrument at different speeds or tempi—a phenomenon referred to as temporal scaling. While it is well established that training improves timing precision and accuracy, it is not known whether expertise improves temporal scaling, and if so, whether it generalizes across skill domains. We quantified temporal scaling and timing precision in musicians and non-musicians as they learned to tap a Morse code sequence. We found that non-musicians improved significantly over the course of days of training at the standard speed. In contrast, musicians exhibited a high level of temporal precision on the first day, which did not improve significantly with training. Although there was no significant difference in performance at the end of training at the standard speed, musicians were significantly better at temporal scaling—i.e., at reproducing the learned Morse code pattern at faster and slower speeds. Interestingly, both musicians and non-musicians exhibited a Weber-speed effect, where temporal precision at the same absolute time was higher when producing patterns at the faster speed. These results are the first to establish that the ability to generate the same motor patterns at different speeds improves with extensive training and generalizes to non-musical domains.

The Neural Basis of Timing: Distributed Mechanisms for Diverse Functions

Timing is critical to most forms of learning, behavior, and sensory-motor processing. Converging evidence supports the notion that, precisely because of its importance across a wide range of brain functions, timing relies on intrinsic and general properties of neurons and neural circuits; that is, the brain uses its natural cellular and network dynamics to solve a diversity of temporal computations. Many circuits have been shown to encode elapsed time in dynamically changing patterns of neural activity-so-called population clocks. But temporal processing encompasses a wide range of different computations, and just as there are different circuits and mechanisms underlying computations about space, there are a multitude of circuits and mechanisms underlying the ability to tell time and generate temporal patterns.

The duration aftereffect does not reflect adaptation to perceived duration

Recent studies have provided evidence for a role of duration-tuned channels in the encoding of duration. Duration encoding in these channels is thought to reflect the time between responses to the onset and offset of an event. This notion is in apparent conflict with studies that demonstrate that the perceived duration of an event can vary independently from the time separating its perceived onset and offset. Instead, these studies suggest that duration encoding is sensitive to other temporal aspects of a sensory event. In the current study, we investigated whether duration-tuned channels encode duration based on the time between the on- and offset of an event (onset-offset duration), or if they encode a duration corresponding to the perceived duration of that event. We used a duration illusion to dissociate onset-offset duration and perceived duration and measured whether repeated exposure to illusion-inducing stimuli caused adaptation to the onset-offset duration or the perceived duration of these illusion-inducing stimuli. We report clear evidence for adaptation to the onset-offset duration of illusion-inducing stimuli. This finding supports the notion that duration-tuned mechanisms respond to the time between the onset and offset of an event, without necessarily reflecting the duration perceived, and eventually reported by the participant. Implications for the duration channel model and the mechanisms underlying duration illusions are discussed.

A model of temporal scaling correctly predicts that motor timing improves with speed

Timing is fundamental to complex motor behaviors: from tying a knot to playing the piano. A general feature of motor timing is temporal scaling: the ability to produce motor patterns at different speeds. One theory of temporal processing proposes that the brain encodes time in dynamic patterns of neural activity (population clocks), here we first examine whether recurrent neural network (RNN) models can account for temporal scaling. Appropriately trained RNNs exhibit temporal scaling over a range similar to that of humans and capture a signature of motor timing, Weber's law, but predict that temporal precision improves at faster speeds. Human psychophysics experiments confirm this prediction: the variability of responses in absolute time are lower at faster speeds. These results establish that RNNs can account for temporal scaling and suggest a novel psychophysical principle: the Weber-Speed effect.

Short-Term Synaptic Plasticity as a Mechanism for Sensory Timing

The ability to detect time intervals and temporal patterns is critical to some of the most fundamental computations the brain performs, including the ability to communicate and appraise a dynamically changing environment. Many of these computations take place on the scale of tens to hundreds of milliseconds. Electrophysiological evidence shows that some neurons respond selectively to duration, interval, rate, or order. Because the time constants of many time-varying neural and synaptic properties, including short-term synaptic plasticity (STP), are also in the range of tens to hundreds of milliseconds, they are strong candidates to underlie the formation of temporally selective neurons. Neurophysiological studies indicate that STP is indeed one of the mechanisms that contributes to temporal selectivity, and computational models demonstrate that neurons embedded in local microcircuits exhibit temporal selectivity if their synapses undergo STP. Converging evidence suggests that some forms of temporal selectivity emerge from the dynamic changes in the balance of excitation and inhibition imposed by STP.

An investigation of the spatial selectivity of the duration after-effect

Adaptation to the duration of a visual stimulus causes the perceived duration of a subsequently presented stimulus with a slightly different duration to be skewed away from the adapted duration. This pattern of repulsion following adaptation is similar to that observed for other visual properties, such as orientation, and is considered evidence for the involvement of duration-selective mechanisms in duration encoding. Here, we investigated whether the encoding of duration - by duration-selective mechanisms - occurs early on in the visual processing hierarchy. To this end, we investigated the spatial specificity of the duration after-effect in two experiments. We measured the duration after-effect at adapter-test distances ranging between 0 and 15° of visual angle and for within- and between-hemifield presentations. We replicated the duration after-effect: the test stimulus was perceived to have a longer duration following adaptation to a shorter duration, and a shorter duration following adaptation to a longer duration. Importantly, this duration after-effect occurred at all measured distances, with no evidence for a decrease in the magnitude of the after-effect at larger distances or across hemifields. This shows that adaptation to duration does not result from adaptation occurring early on in the visual processing hierarchy. Instead, it seems likely that duration information is a high-level stimulus property that is encoded later on in the visual processing hierarchy.

Temporal Processing by Intrinsic Neural Network Dynamics

It is becoming more apparent that there are rich contributions to temporal processing across the brain. Temporal dynamics have been found from lower brain structures all the way to cortical regions. Specifically,in vitrocortical preparations have been extremely useful in understanding how local circuits can time. While many of these results depict vastly different processing than a traditional central clock metaphor they still leave questions as to how this information is integrated. We therefore review evidence to place the results pertaining to local circuit timers into the larger context of temporal perception and generalization.

Multiple channels of visual time perception

The proposal that the processing of visual time might rely on a network of distributed mechanisms that are vision-specific and timescale-specific stands in contrast to the classical view of time perception as the product of a single supramodal clock. Evidence showing that some of these mechanisms have a sensory component that can be locally adapted is at odds with another traditional assumption, namely that time is completely divorced from space. Recent evidence suggests that multiple timing mechanisms exist across and within sensory modalities and that they operate in various neural regions. The current review summarizes this evidence and frames it into the broader scope of models for time perception in the visual domain.

Mental Summation of Temporal Duration within and across Senses

Perceiving, memorizing, and estimating temporal durations are key cognitive functions in everyday life. In this study, a duration summation paradigm was used to examine whether summation of temporal durations introduces an underestimation or overestimation bias, and whether this bias is common to visual and auditory modalities. Two within- or across-modality stimuli were presented sequentially for variable durations. Participants were asked to reproduce the sum of the two durations (0.6-1.1 s). We found that the sum of two durations was overestimated regardless of stimulus modalities. A subsequent control experiment indicated that the overestimation bias arose from the summation process, not perceptual or memory processes. Furthermore, we observed strong positive correlations between the overestimation bias for different sensory modalities within participants. These results suggest that the sum of two durations is overestimated, and that supra-modal processes may be responsible for this overestimation bias.

Networks that learn the precise timing of event sequences

Neuronal circuits can learn and replay firing patterns evoked by sequences of sensory stimuli. After training, a brief cue can trigger a spatiotemporal pattern of neural activity similar to that evoked by a learned stimulus sequence. Network models show that such sequence learning can occur through the shaping of feedforward excitatory connectivity via long term plasticity. Previous models describe how event order can be learned, but they typically do not explain how precise timing can be recalled. We propose a mechanism for learning both the order and precise timing of event sequences. In our recurrent network model, long term plasticity leads to the learning of the sequence, while short term facilitation enables temporally precise replay of events. Learned synaptic weights between populations determine the time necessary for one population to activate another. Long term plasticity adjusts these weights so that the trained event times are matched during playback. While we chose short term facilitation as a time-tracking process, we also demonstrate that other mechanisms, such as spike rate adaptation, can fulfill this role. We also analyze the impact of trial-to-trial variability, showing how observational errors as well as neuronal noise result in variability in learned event times. The dynamics of the playback process determines how stochasticity is inherited in learned sequence timings. Future experiments that characterize such variability can therefore shed light on the neural mechanisms of sequence learning.