Task Complexity Modulates Sleep-Related Offline Learning in Sequential Motor Skills

Recently, a number of authors have advocated the introduction of gross motor tasks into research on sleep-related motor offline learning. Such tasks are often designed to be more complex than traditional key-pressing tasks. However, until now, little effort has been undertaken to scrutinize the role of task complexity in any systematic way. Therefore, the effect of task complexity on the consolidation of gross motor sequence memory was examined by our group in a series of three experiments. Criterion tasks always required participants to produce unrestrained arm movement sequences by successively fitting a small peg into target holes on a pegboard. The sequences always followed a certain spatial pattern in the horizontal plane. The targets were visualized prior to each transport movement on a computer screen. The tasks differed with respect to sequence length and structural complexity. In each experiment, half of the participants initially learned the task in the morning and were retested 12 h later following a wake retention interval. The other half of the subjects underwent practice in the evening and was retested 12 h later following a night of sleep. The dependent variables were the error rate and total sequence execution time (inverse to the sequence execution speed). Performance generally improved during acquisition. The error rate was always low and remained stable during retention. The sequence execution time significantly decreased again following sleep but not after waking when the sequence length was long and structural complexity was high. However, sleep-related offline improvements were absent when the sequence length was short or when subjects performed a highly regular movement pattern. It is assumed that the occurrence of sleep-related offline performance improvements in sequential motor tasks is associated with a sufficient amount of motor task complexity.


3
In the present study, we wanted to determine, whether any significant reductions in participants' TET (i.e. 4 improvement in execution speed) at retest(s) in our experiments was a result of any offline enhancement 5 consolidation, or just a consequence of further practice. One way to assess enhancement consolidation 6 considering continued learning at retention, involves extrapolation of each subject's respective initial training 7 data by means of power law functions. Power law functions can be used to mathematically model practice-8 dependent changes in performance in the course of skill acquisition (Newell and Rosenbloom, 1981). By 9 extrapolation of this practice data fit participants' performance at retest(s) can be estimated as if practice had 10 just continued without any pause and under the same conditions that were present during training. Suchlike 11 predicted retention data can then be used in conjunction with the individuals' observed performance on the 12 retest trials. If the observed performance is better (i.e. TET lower) than the predicted performance, offline 13 facilitation is assumed to have occurred.
14 15 In the present study, predicted TET-Retest-measures were provided as follows: based on each single 16 subject's TET-acquisition data (means per trial block), for each individual a power function of the type y = 17 kn -c was calculated and used to obtain an estimate for that individual's performance for the additional three 18 trial blocks during Retest. Here k is the mean value of TET on the first practice trial block, c is the learning 19 rate (with c < 0), and n is the number of trial blocks. Predicted TET-data for each individual then were 20 collapsed across blocks, thus providing mean predicted TET-performance at Retest for each one subject.

21
Thus, if sleep (but not wake) had indeed enhanced memory consolidation, observed TET should turn out 22 significantly shorter as compared to predicted TET in the EM-groups when tested right after the sleep-filled 23 12 hrs retention interval, but not in the ME-groups when tested following the 12 hrs wake retention interval.

24
Special consideration has also been given to the possibility that the first Retest block (i.e. block11) might

46
Contrary to any á-priori expectations, observed sequence execution time in both experimental groups of experiment 2 proved to be slightly, but significantly longer at Retest than the predicted one. This was shown 48 by a significant factor Data-Type (F (1, 22) = 19.977, p < .001, η 2 p = .476), while neither the factor Group (p 49 = .756), nor the interaction term (p = .723) did reach statistical significance. The surprising result of 50 observed TET-data being longer at Retest than the predicted ones was also corroborated by paired t-tests 51 calculated for each group separately, comparing observed and predicted TET-data at Retest. With a decrement present in the first of the three trial blocks at Retest (i.e. in block 11) in both groups of experiment 1 2. This may have caused the average observed TET-data at Retest being slightly longer than the predicted 2 ones.

4
To control for this possible warm-up effect, observed and predicted TET-data at Retest were compared once 5 more for each group separately without trial block 11. To this end for each participant the respective 6 performance measures were averaged only across trial blocks 12 and 13. According to the 7 respective t-tests, in both groups observed and predicted TET-data at Retest statistically did not differ any 8 more (EM-group (t (11) = 2.04, p [two-tailed] = .067; ME-group (t (11) = 1.17, p [two-tailed] = .265). Thus, it seems 9 that participants have compensated the warm-up decrement evident in block 11 by achieving TET-values in 10 block 12 and in block 13 that about matched the predicted ones.