Generation of Rapid Sequences by Motor Cortex

Abstract

1 Rapid execution of motor sequences is believed to depend upon the fusing of movement elements into 2 cohesive units that are executed holistically. We sought to determine the contribution of motor cortex 3 activity to this ability. Two monkeys performed highly practiced two-reach sequences, interleaved with 4 matched reaches performed alone or separated by a delay. We partitioned neural population activity 5 into components pertaining to preparation, initiation, and execution. The hypothesis that movement 6 elements fuse makes specific predictions regarding all three forms of activity. We observed none of 7 these predicted effects. Instead, two-reach sequences involved the same set of neural events as 8 individual reaches, but with a remarkable temporal compression: preparation for the second reach 9 occurred as the first was in flight. Thus, at the level of motor cortex, skillfully executing a rapid sequence 10 depends not on fusing elements, but on the ability to perform two computations at the same time. 65first reach and did so without disrupting the first reach. Thus, at the level of motor and premotor cortex, 66 skilled performance depends not on fusing elements, but upon the ability to prepare one element while 67 executing another. 68 Results 69Task and Behavior 70We trained two rhesus macaques (monkeys B and H) to perform a modified delayed-reach task ( Fig. 71 1a,b). All trials began with a randomized (0-1000 ms) instructed delay period. Each trial required the 72 monkey to make either a single reach, two reaches separated by an instructed pause (delayed double-73 reach), or two reaches with no pause between them (compound reach). Target color indicated which 74 should be acquired first. A salient visual cue (the diameter of the colored portion of the first target) 75 indicated whether a pause was required. For monkey H, the pause between delayed double-reaches was 76 always 600 ms. For monkey B it was variable (100, 300, or 600 ms; the last is used for most analyses) and 77 indicated by the visual cue. Thus, all key information -target locations and any instructed pause -was 78 given during the instructed delay. Reach paths are illustrated ( Fig. 1a) for all single reaches, for three 79 compound reaches that began down-and-right, and for three compound reaches that began down-and-80 left (Extended Data Fig. 1 shows paths for all compound reaches). 81Compound reaches were performed briskly; the hand stayed on the first target only briefly before 82 moving to the second target. Median dwell times on the first target were 119 ms (monkey B) and 137 83 ms (monkey H). The median duration for the full two-reach sequence was 561 ms (monkey B) and 645 84 ms (monkey H). This rapid pace resulted from extensive training over months, with each sequence 85 performed tens of thousands of times. This pace is even faster than that reported in other motor 86 sequence tasks performed by non-human primates, where dwell times are typically >200 ms 10,11,29 . 87To enable comparisons at the neural level, every compound r...

Motor cortex activity across movement speeds is predicted by network-level strategies for generating muscle activity

“…At the other end of the spectrum, sequences of multiple movements are generated similarly by motor cortex regardless of overall sequence speed, which is presumably controlled by upstream / areas 38 . Thus, the broader point is not that there is any single strategy for speed-control across all types of behaviors, but rather that the network-dynamics perspective can often predict and/or explain the strategies employed by biological networks.…”

Combining repetition suppression and pattern analysis provides new insights into the role of M1 and parietal areas in skilled sequential actions

“…The finding is also consistent with results from neurophysiology (31) and magneto-encephalography (MEG; 32), showing the first action in a sequence is most highly activated in premotor and motor areas during the preparatory period. Since M1 does not contain representation of sequence identity, continuous signals from higher-order regions provide it with the sequential context needed throughout the sequence production (8,33).…”

“…However, this study did not show differential activation for different trained sequences, thus no sequence representation as defined here. Moreover, recent electrophysiological studies have also shown that M1 does not represent the sequential context (Russo et al, 2020; Zimnik and Churchland, 2021). Altogether, this makes it unlikely that the RS observed in M1 reflects sequence dependency.…”

“…This study, however, only reflected an overall difference between trained and random sequences, but not M1's differential activation for different trained sequences, thus no sequence representation as defined here. More recent electrophysiological studies have shown that M1 only represents the ongoing movement, and does not represent the sequential context (8,33).…”

The planning horizon for movement sequences

“…Therefore, a more likely scenario is that planning and execution occur in overlapping neuronal populations, but occupy orthogonal subspaces of the multidimensional neuronal code (Fig. 8B), such that planning activity does not trigger motor output (Kaufman et al, 2014; Lara et al, 2018; Zimnik and Churchland, 2020). Finally, an intriguing possibility that should be investigated in future studies is that multiple future movements could be planned as an integrated packet (i.e., as a movement chunk), such that there is only one, shared, planning subspace, with specialized code for specific transitions (i.e., 2-3) in movement sequences.…”