Generalization of procedural motor sequence learning after a single practice trial

Abstract

When humans begin learning new motor skills, they typically display early rapid performance improvements. It is not well understood how knowledge acquired during this early skill learning period generalizes to new, related skills. Here, we addressed this question by investigating factors influencing generalization of early learning from a skill A to a different, but related skill B. Early skill generalization was tested over four experiments (N = 2095). Subjects successively learned two related motor sequence skills (skills A and B) over different practice schedules. Skill A and B sequences shared ordinal (i.e., matching keypress locations), transitional (i.e., ordered keypress pairs), parsing rule (i.e., distinct sequence events like repeated keypresses that can be used as a breakpoint for segmenting the sequence into smaller units) structures, or possessed no structure similarities. Results showed generalization for shared parsing rule structure between skills A and B after only a single 10-second practice trial of skill A. Manipulating the initial practice exposure to skill A (1 to 12 trials) and inter-practice rest interval (0-30 s) between skills A and B had no impact on parsing rule structure generalization. Furthermore, this generalization was not explained by stronger sensorimotor mapping between individual keypress actions and their symbolic representations. In contrast, learning from skill A did not generalize to skill B during early learning when the sequences shared only ordinal or transitional structure features. These results document sequence structure that can be very rapidly generalized during initial learning to facilitate generalization of skill.

Fig. 1. Influence of length of training of skill A on generalization to skill B during early learning.

Experiment 1 evaluated the influence of length of training of a skill A on generalization to a skill B (n = 551; see Supplementary Table 1 for demographics). a Task: Participants were randomized to practice a skill A for 1, 2, 5 or 12 trials. Rest intervals between trials were 10 s duration. Practice of skill A (4-1-3-2-4) was followed in all groups by five testing trials of skill B (2-3-1-4-2). Skill was measured as the average inter-tap interval within correct sequences (tapping speed measured in keypresses/s). b Performance of the training groups (grey: skill A; cyan: skill B; mean ± s.e.m.). Training in skill A resulted in rapid performance improvements, consistent with previous work. Specifically, participants demonstrated rapid motor skill learning primarily during micro-offline periods which reached plateau by trial 11 in the 12 trials group (Supplementary Fig. 1). c Skill at the onset of skill B (red: 1 trial; blue: 2 trials; green: 5 trials; yellow: 12 trials). Following the end of practice on skill A, performance at the onset of skill B was comparable between experimental groups. d Change in skill from the end of skill A to the onset of skill B (i.e., ${\mathit{Gen}}_{B_0-A_f}$). All groups had significant changes in performance between the end of skill A to the onset of skill B. The group that practiced one trial of skill A was the only group that showed ${\mathit{Gen}}_{B_0-A_f}$, suggestive of a micro-offline contribution to generalization (shaded bar). *p < 0.05, where * over individual group plots indicate significant one-sample t-test within-group differences between B₀ and A_f and * between group plots indicate significant Kruskal-Wallis (pairwise comparison) between-group differences.

Fig. 2. Varying rest interval durations does not impact rapid generalization during early learning.

Experiment 2 evaluated the influence of varying rest duration between skill A and skill B on generalization (n = 795; see Supplementary Table 2 for demographics). a Task: Participants were randomized to practice 1 trial of skill A followed by five testing trials of skill B (2-3-1-4-2) with either 0, 2, 5, 10 or 30 s following the end of practice of the skill A trial. Subjects in all groups performed equal number of total trials (six). Performance is shown for the single trial of skill A and the first trial of skill B for each of the five groups (grey: 0 s; orange: 2 s; green: 5 s; yellow: 10 s; purple: 30 s; mean ± s.e.m.). Performance in all 5 trials of skill B are shown in Supplementary Fig. 5. b Skill at the onset of skill B. Note the similar performance at the onset of skill B regardless of interval durations. c Change in skill from the end of skill A to the onset of skill B (i.e., ${\mathit{Gen}}_{B_0-A_f}$). Note that all groups but the 0 s group experienced significant micro-offline (i.e., ${\mathit{Gen}}_{B_0-A_f}$) gains. *p < 0.05, where * over individual group plots indicate significant one-sample t-test within-group differences between B₀ and A_f.

Fig. 3. Content of generalization during early skill learning.

Experiment 3 evaluated the content of generalization of skill A to four different skill Bs (n = 537; see Supplementary Table 3 for demographics). a Task: Each skill B shared with skill A different features: a similar parsing cue structure (sequence started and ended with the same number, i.e., 2-3-1-4-2; PARSING), similar transitions (i.e., 2-4-1-4-3, transitions 2-4 and 4-1; TRANSITION), a combination of similar transitional/ordinal features (i.e., 2-4-3-2-4; O + T) or a repetition of skill A (i.e., 4-1-3-2-4; SAME_A). b Performance in the first trial of skill A and testing of the four groups performing different types of skill B (grey: PARSING; green: TRANSITION; yellow: O + T; purple: SAME_A; mean ± s.e.m.). Note the similar performance in skill A across groups. Performance in all 5 trials of skill B are shown in Supplementary Fig. 7. c Skill at the onset of skill B. Onset of B was highest in the group that continued performance of skill A (SAME_A group). Note that skill at the onset of skill B in the SAME_A and PARSING groups are significantly larger than the TRANSITION group. d Change in skill from the end of skill A to the onset of skill B (i.e., ${\mathrm{Gen}}_{{\mathrm{B}}_0-{\mathrm{A}}_{\mathrm{f}}}$). ${\mathrm{Gen}}_{{\mathrm{B}}_0-{\mathrm{A}}_{\mathrm{f}}}$ was highest in the SAME_A group, being significantly greater than the TRANSITION and O + T groups. Note that the SAME_A and PARSING groups showed significant micro-offline gains (i.e., ${\mathrm{Gen}}_{{\mathrm{B}}_0-{\mathrm{A}}_{\mathrm{f}}}$), while the TRANSITION and O + T groups did not. *p < 0.05, where * over individual group plots indicate significant one-sample t-test within-group differences between B₀ and A_f and * between group plots indicate significant Kruskal-Wallis test (pairwise comparison) between-groups differences (Fig. 3c) or one-way ANOVA (pairwise comparison) between-group differences (Fig. 3d).

Fig. 4. Contribution of similar parsing cue structure to generalization during very early learning.

Experiment 4 attempted to dissociate the relative contributions of parsing cue and sensorimotor mapping structure to generalization during very early learning (additional sample added onto that of Experiment 3: n = 212; see Supplementary Table 4 for demographics). a Task: Experiment 4 added two groups: in one, subjects performed a single trial of random order keypresses (replacing skill A) followed by 5 test trials of skill B (2-1-3-4-2; RANDOM) and the second group trained over one trial of skill B (2-1-3-4-2) followed by 5 test trials of the identical skill B (SAME_B). Performance in all 5 trials of skill B for the SAME_B and RANDOM groups are shown in Supplementary Fig. 8. b Skill at the onset of skill B (mean ± s.e.m). Onset of B in the PARSING, SAME_A, and SAME_B groups were all greater than in the RANDOM group, demonstrating generalization of content beyond simple sensorimotor transformations observed in the RANDOM group. Conversely, onset of B was comparable in the RANDOM, TRANSITION, and O + T groups. Thus, ordinal and transitional information contribution to generalization was negligible, largely comparable to that of typing random keypresses. c Onset of B was greater in the PARSING group than in the RANDOM group, providing an estimate for the relative contribution of knowledge of the parsing cue to generalization during very early skill learning. *p < 0.05, where * between group plots indicate significant Kruskal-Wallis (pairwise comparison) between-group differences.

Fig. 5. Measures of generalization of motor skill.

Generalization from a skill A to a skill B can be measured as: raw performance at Onset of B or the change in performance from the end of A to onset of B (${\mathrm{Gen}}_{{\mathrm{B}}_0-{\mathrm{A}}_{\mathrm{f}}}$).