Task-irrelevant auditory metre shapes visuomotor sequential learning

The ability to learn and reproduce sequences is fundamental to every-day life, and deficits in sequential learning are associated with developmental disorders such as specific language impairment. Individual differences in sequential learning are usually investigated using the serial reaction time task (SRTT), wherein a participant responds to a series of regularly timed, seemingly random visual cues that in fact follow a repeating deterministic structure. Although manipulating inter-cue interval timing has been shown to adversely affect sequential learning, the role of metre (the patterning of salience across time) remains unexplored within the regularly timed, visual SRTT. The current experiment consists of an SRTT adapted to include task-irrelevant auditory rhythms conferring a sense of metre. We predicted that (1) participants’ (n = 41) reaction times would reflect the auditory metric structure; (2) that disrupting the correspondence between the learned visual sequence and auditory metre would impede performance; and (3) that individual differences in sensitivity to rhythm would predict the magnitude of these effects. Altering the relationship via a phase shift between the trained visual sequence and auditory metre slowed reaction times. Sensitivity to rhythm was predictive of reaction times over all. In an exploratory analysis, we, moreover, found that approximately half of participants made systematically different responses to visual cues on the basis of the cues’ position within the auditory metre. We demonstrate the influence of auditory temporal structures on visuomotor sequential learning in a widely used task where metre and timing are rarely considered. The current results indicate sensitivity to metre as a possible latent factor underpinning individual differences in SRTT performance. Supplementary Information The online version contains supplementary material available at 10.1007/s00426-022-01690-y.

54.81). Comparing RT within Block 9, those in the High Rhythm group generally increased speed within the test block (Mean = -15.06, 95% CI [-33.44,3.32], SD = 38.14), but this comparison was again more variable for participants with lower Rhythm Scores (Mean = +0.33, 95% CI [-25.30,25.97], SD = 57.82), who did not show a within-block improvement on average. Hence, although High Rhythm participants were proportionally more affected by the Phase-Shifted Metre Test, it does not appear that higher sensitivity to Rhythm was an impediment for long beyond the initial change between the visual sequence and auditory metre, and High Rhythm participants may have been able to adapt and regain some speed. This localised pattern of results may help to explain the reduced effect, when viewed at the level of Block, of the test and of Rhythm Score for participants in the 4/4 Metre, given they appear to have first slowed, then quickly rebounded, within a few cycles of the sequence.

A.3 New Metre Test Block
For RT near the boundaries of Blocks 10 and 11, responses to the New Metre test are variable, both for those with lower Rhythm Scores (Mean = -14.92, 95% CI [-51.97,22.14], SD = 83.57), and those with higher Rhythm Scores (Mean = +4.68, 95% CI [-54.67,64.03], SD = 123.14). In the planned modelling on the level of Block, there had been a statistically significant interaction between Block and Metre, suggesting that participants in the 4/4 Metre may have actually sped in RT from Block 10 to 11 (p < 0.001). Breaking down the local difference across blocks by Metre group, however, shows that participants in the 3/4 metre (Mean = -1.30, 95% CI [-44.38,41.79], SD = 83.80) and 4/4 metre (Mean = -9.05, 95% CI [-58.12,40.02], SD = 116.20) both produced a mix of faster and slower RT immediately following the onset of Block 11.
We had also modelled a three-way interaction with Block, Metre, and Rhythm Score. Low Rhythm 3/4 participants tended towards faster RT after the onset of the unfamiliar 4/4 metre (Mean = -49.59, 95% CI [-106.43,7.25

A.4 Discussion
During Learning, participants in the median-split higher Rhythm Score group tended to slow at the beginning of new blocks. These High Rhythm responses show a staircase-like pattern with a steady, decreasing trend in RT, as can be seen in Figure A1. By contrast, the responses of participants in the Low Rhythm group followed a flatter and much more variable trajectory throughout the SRTT. We cannot be sure whether the staircase pattern is due to sensitivity to rhythm per se. For example, it could be that people with higher rhythm sensitivity also have more experience with playing music or computer gaming, leading to an enhanced stability of motor responses for prolonged periods (i.e., an entire block of 120 responses). This does not in itself explain why the High Rhythm group begin each Learning block slower than the previous block ended, but it is possible that such a pattern could reflect a within-block ceiling effect in performance. These descriptive insights indicate possible betweengroup differences in behaviour across block boundaries that, whether due to differences in motor performance or rhythm sensitivity specifically, should be formally confirmed in a new data set.
In the analysis of between-block boundaries between Block 8 and the Phase-Shifted Metre test, it appears that the true effect on RT is actually dampened when modelling full blocks, perhaps because participants rapidly adjust to the new correspondence between the metre and visual sequence after a brief, initial disruption. On the other hand, we also observed the aforementioned staircase pattern in High Rhythm participants, with slowing at the beginning of almost every new block, which throws the effect of the test block into question. We note, however, the slowing of RT in transition to Phase-Shifted Metre for High Rhythm participants (Mean = 41.16,95% CI [14.34,67.99]) is much greater in magnitude in comparison to during Learning blocks (Mean = 10.74, 95% CI [-2.48,23.97]). Moreover, the reinstatement of the familiar, learned pairing between visual sequence and auditory metre in Block 10 is the only block where High Rhythm participants show a slight trend towards faster betweenblock differences in RT. Finally, with regards to the New Metre Test, neither breaking down performance by group, nor by responses at the beginnings and ends of blocks 10 and 11, revealed any further insights. We therefore have higher confidence that the null effect of the New Metre Test observed in the planned contrasts was not the result of pooling by Block. sampled as only the first and last two cycles (n = 24 reaction times) for each block. Responses are summarised by median-split Rhythm Group. Group means are shown by dots, with smaller lines representing 95% confidence intervals of the mean.

New Visual Sequence Test
Immediately after completing the main task, visual sequential learning was confirmed with an additional SRTT block that combined the learned auditory metre with an unfamiliar series of visual cues, corresponding to the canonical test block in the standard SRTT. RT in this task were compared with those from Blocks 8 and 10 in the main SRTT as a sanity check that the new visual sequence was performed more slowly than the learned visual sequence.

Explicit Recognition Test
Next, a secondary task was administered to gauge explicit knowledge of the SRTT learned visual sequence. Note that this task was introduced for only one of the two testing sites, and hence only a subset of participants who performed the main task took part (n = 26). The purpose of the task was to ensure that the cross-modal aspects of our modified SRTT did not result in increased explicit awareness, in comparison to the standard SRTT. The explicit recognition task followed a similar format to the SRTT in terms of visual presentation; however, the trials were self-paced, rather than strictly timed, and there was no auditory component. In each trial, the first two or three elements of a visual sequence were shown before the participant was prompted to guess the next element by responding with its key. After guessing, they were asked "Did you know what the next location would be?". There were forty-eight trials in total, of which 25% (12) contained truly familiar segments from the learned visual sequence. Of these, 50% (6) were presented in a grouping that reflected the participant's own SRTT metre condition (i.e., a group of either three or four). Only data from trials containing the truly familiar segments were retained, and the dependent variable was correct/incorrect responses. We modelled likelihood of a Correct Response in the Explicit Recognition Task data as a binomial distribution using the logit link function. Possible fixed effect terms were Metre, length of segment (i.e., 3 or 4 elements), and self-reported familiarity with the segment as factors; and Rhythm Score as a covariate.