The Effect of Different Combinations of Practice Schedules on Motor Response Stability during Practice

Abstract Many results in motor learning have indicated that relative and absolute timing dimensions are modulated by factors that modify response stability among trials. One of these factors is the combination of constant and variable practices. Although many researchers have investigated the combination of practice schedules, these researchers have used measurements that do not assess performance and motor response separately. This study aimed to investigate the effect of different combinations of practice schedules on motor response stability during practice. Participants performed a sequential key-pressing task with two goals: (1) to learn the relative timing dimension and (2) the absolute timing dimension. Participants were assigned to one of two groups: constant-variable or variable-constant. Our findings indicate an influence of the increase in variability over the practice in the constant-variable group. Precisely, the increase in variability of total time in the second half (constant-variable group) of practice was followed by the maintenance of the same level of cross-correlate between absolute timing error and variability of total time. Finally, our findings support the hypothesis that practicing in a constant schedule favors the relative timing dimension of learning regardless of the order in which the constant practice is provided.


Introduction
I n a series of studies, Wulf and colleagues (Wulf, 1992;Wulf et al., 1994;Wulf & Schmidt, 1989;Wulf et al., 1993), and Lai and colleagues (Lai et al., 2000;Lai & Shea, 1998, 1999Shea et al., 2001) showed the manipulation of variables over the learning process could affect different aspects of a motor skill. The manipulation of variables might favor either the relative timing dimension learning or the absolute timing dimension learning of a motor skill (Lai et al., 2000). The relative timing dimension relates to the spatio-temporal pattern (Wulf et al., 1994) and the absolute timing dimension refers to the subject's ability to adjust to a spatio-temporal pattern (Lai & Shea, 1999). These aspects are critical for skilled behavior (Glencross et al., 1994). A skilled behavior requires the ability to maintain the spatio-temporal pattern of movement while it adapts to the demands of the environment (Glencross et al., 1994). For example, when we observe a sequence of volleyball serves, the spatio-temporal pattern of the movement is maintained throughout the sequence (that is, the proper technique is assured). However, the spatio-temporal pattern can be adapted to the dynamics of the environment (e.g., decrease the speed of execution) without being mischaracterized. Finding variables that favor learning of both relative and absolute dimensions has been a constant challenge for researchers in motor learning Czy_ z et al., 2019).
In addition to the aforementioned reasons, Wulf and colleagues (Wulf, 1992;Wulf et al., 1994;Wulf & Schmidt, 1989;Wulf et al., 1993) and Lai and colleagues (Lai et al., 2000;Lai & Shea, 1998, 1999Shea et al., 2001) address two other interesting facts. First, the method in these studies enables the systematic assessment of the relative timing dimension and the absolute timing dimension of a motor task through two dependent measures: the relative timing error and the absolute timing error (Wulf & Schmidt, 1989). The relative timing error infers the relative timing dimension learning (Wulf et al., 1994). On the other hand, the absolute timing error infers the absolute timing dimension learning (Lai & Shea, 1999). In general, these two measures have been made possible through tasks that enable the subject to learn each one apart (Lai & Shea, 1998), such as the sequential key-pressing task (Nogueira et al., 2020). The sequential key-pressing task requires the learner to learn two goals: (a) the relative timing (or partial time) between the keys and (b) the absolute timing (or total time) between the first and the last key of the sequence (Ferreira et al., 2019;Lage et al., 2007;Lelis-Torres et al., 2017).
Second, the result in these studies consistently showed that motor response stability is a determinant factor in learning the relative and absolute timing dimensions (Lai & Shea, 1998). The motor response stability refers to the adjustments made at each trial (Lai & Shea, 1998 motor response stability increases, while the adjustments decrease (Lai et al., 2000). On the other hand, the motor response stability decreases, while the adjustments increase (Lai & Shea, 1999).
In general, learning in the relative timing dimension has been enhanced by factors that promote an increase in motor response stability (Shea et al., 2001). For instance, increased motor response stability has been enabled by the practice schedule that requires the execution of a single goal execution (constant practice) (Czy_ z et al., 2019). Practicing a single goal execution decreases the demand for adjustments at each trial favoring the maintenance of the spatio-temporal pattern (relative timing dimension) (Lage et al., 2016). On the other hand, learning the absolute timing dimension has been enhanced by factors that promote a decrease in motor response stability (Lai & Shea, 1999). For instance, decreased motor response stability has been enabled by the practice schedule that requires the execution of a skill based on two or more goals (variable practice) (Czy_ z, 2021). In a variable practice, the frequent trial-to-trial goal modification increases the demand for adjustments at each trial (Czy_ z, 2021). The decrease in the motor response stability favors the adjustment to a spatio-temporal pattern (absolute timing dimension learning) (Apolin ario- . Although the results of several studies support the notion that the level of motor response stability influences the learning of absolute and relative dimensions in opposite ways (Lage et al., 2007;Lai et al., 2000), the measures utilized did not detach the assessment of motor response from performance. Performance can be framed as the difference between the action-outcome and task goal (Pan & Bjork, 2020;Salmoni et al., 1984). Motor response is the motor action resulting from internal processes produced in an environment context (Schmidt et al., 2019). Assessing these two pieces of information detached is crucial to ensure that the motor response stability was measured. Failure to separate performance and motor response are potentially problematic, especially in a variable practice schedule, where the learner may maintain a stable motor response during trials while showing variations in performance ( Figure 1). Figure 1 illustrates two hypothetical conditions (A and B). Condition A shows no adjustment among trials in motor response (left upper panel). Condition B shows adjustments among trials in motor response (left bottom panel). The adjustment among trials (trial n -trial n-1 ) in motor response can be seen by the differences between the total times over the trials. Despite having different motor responses (higher motor response stability in condition A), both conditions show similar motor response stabilities when performance is evaluated (right panel). The adjustment among trials (trial n -trial n-1 ) in performance can be seen by the differences between the total times produced over the trials and the total time goals. Thus, it remains to be investigated whether the notion of motor response stability would be supported when performance and motor response were assessed separately.
Furthermore, these studies have indicated that the order in the combination of practice schedules affects the learner's performance along the practice phase (Lage et al., 2007;Lai et al., 2000). Constant-variable practice combination tends to maintain the same level of performance when learners switch from practicing in a constant to a variable condition (Lai et al., 2000). On the other hand, the variable-constant practice combination indicates a decrease in performance when learners move from practice in a variable to a constant condition (Lage et al., 2007). This performance decrease in the variable-constant combination holds for both relative and absolute time errors (Lage et al., 2007). Regardless of whether the constant practice is performed before or after variable practice (constant-variable or variable-constant), constant practice favors the relative time dimension learning (Lai et al., 2000). Nonetheless, performing the constant practice before the variable practice (constant-variable) is more advantageous for learning the absolute timing dimension than the other way around (variable-constant) (Lage et al., 2007).
As aforementioned, although the relation between motor response stability and absolute and relative dimensions has been addressed, no specific measure was used to assess the motor response stability. The lack of specific measures to assess the motor response stability indicates that its actual behavior in practice remains unknown. Perhaps the increase in motor response stability provided by the constant practice, in the constantvariable combination, had contributed to learners nearly achieving the task goals early in the practice phase. This probably generated a ceiling effect in performance (Hirano et al., 2020), not allowing for an improvement in performance in the second half of practice. In opposition, it may be possible that the variable practice in the variable-constant combination provided less motor response stability, requiring learners to test different motor responses to achieve the tasks goals (Czy_ z, 2021). This might not generate a ceiling effect, supporting the reported reduction in performance in the second half of practice. To clarify these suppositions, it is required to assess performance and motor response separately.
This study first aimed to investigate the effect of different combinations of practice schedules on motor response stability during practice. We hypothesize that the constant-variable combination will show higher motor response stability in the first half of practice (1-60 trials) than the variable-constant combination. In contrast, in the second half of the practice (61-120 trials), the constant-variable combination will show less motor response stability than the variable-constant combination. Secondly, this study aimed to investigate how motor response stability relates to performance in absolute and relative timing dimensions during the practice phase. We hypothesize that the stability created by the constant-variable combination will be more related to performance over the practice. Thirdly, we aimed to confirm that introducing goal variations late in the learning process is more effective for learning the absolute timing dimension.

Participants
Twenty-two right-handed participants (11 men and 11 women, counterbalanced by groups), ranging from 18 to 35 years old (mean 25.81 ± 3.91), participated in this experiment. All participants had normal or corrected-tonormal visual acuity in both eyes and had no prior experience with the motor task. The sample size was defined using the software G Ã Power (version 3.1) using data regarding absolute timing error from a pilot study. The software indicated a sample size of 22 subjects (critical F ¼ 4.35), setting the power of the tests to 85%. Participants' handedness was confirmed by the Edinburgh Handedness Inventory (Oldfield, 1971), in which they should achieve a coefficient higher than 80 to classify as right-handed.
A local ethics committee approved the study (CAAE 24116513.2.0000.5149). All participants provided written informed consent after receiving a full explanation about the study.

Apparatus
A computer, color monitor, and numeric keypad were placed on a standard table in the lab room. A custommade software program (https://github.com/edftercio/ pressing_sequential_keys) was used to control the experimental task in the LabVIEW software (National Instruments, Texas, EUA). Participants were asked to sit on a chair in front of the computer monitor and to adjust the numeric keypad position to use it comfortably. They used it with their right hand.

Task
Participants were asked to sequentially press four keys (2, 8, 6, and 4) on the numeric keypad using the index finger of their right hand. The participants had to achieve two goals in each trial: the total time and the partial time between each key. The total time was 900 ms for the constant practice condition and 700, 900, and 1,100 ms for the variable practice condition ( Figure 2). Variable practice conditions were classified as random variable practice with all the movements belonging to the same class of movements (for more details, see Czy_ z, 2021).
The partial time between each key was (1) 22.2% (key 2 to 8), (2) 44.4% (key 8 to 6), and (3) 33.3% (key 6 to 4) in all experimental phases. The total time represents the absolute timing dimension of the task. The partial time between each pair of keys represents the relative timing dimension of the task (Lai & Shea, 1998;Wulf, 1992). After each trial, the knowledge of results was displayed on the screen immediately. The knowledge of results included the partial time between each key, the relative timing error (highlighted in Figure 2), and the total time performed (Figure 2 B).

Procedures and Experimental Design
The exact instructions were given to each participant about the information displayed on the computer screen.
All subjects were required to be as accurate as possible regarding partial and total time goals. Before each trial, partial and total time goals were displayed on the computer screen. After the task execution, the results were presented on the screen for six seconds (Figure 2 B). Participants were instructed to spend the time they needed to compare their results with the goals. After six seconds, a sign requiring the initiation of the next trial was presented. Participants were instructed to start whenever they wanted the next trial after receiving the sign to start. When a participant pressed the keys incorrectly, a warning was displayed, and the trial was repeated. No participant pressed the keys incorrectly.
Participants were randomly assigned to two practice groups: constant-variable or variable-constant. The experiment consisted of three parts: practice phase, retention test, and transfer test. During the practice phase, all participants performed 120 trials at the same partial time goals, 22.2% (key 2 to 8), 44.4% (key 8 to 6), and (3) 33.3% (key 6 to 4). However, the total time goal was different between groups. The constant-variable group performed the same total time goal (900 ms) in the first 60 trials (constant practice) and three different ones (700, 900, and 1,100 ms) in the last 60 later trials After approximately 24 hours of the end of the practice phase, participants performed the retention test. The retention test consisted of practicing 12 trials at the total time goal of 900 ms and the same partial time goal practiced during the practice phase (22.2% key 2 to 8, 44.4% key 8 to 6, and 33.3% key 6 to 4) (Figure 2(A)). Immediately after the retention test, participants performed the transfer test in which they completed 12 trials at a new total time (1,300 ms) (Figure 2(A)). The partial time goal was kept the same. No knowledge of results was provided in retention and transfer tests.

Measurements
Relative timing error (RE) and absolute timing error (AE) were used as task performance measures. RE was Combinations of Practice on Motor Learning used to measure the proficiency in learning the relative timing dimension, while AE was used as a measure of proficiency in learning the absolute timing dimension (Wulf & Schmidt, 1989). The RE was determined as the sum of differences between the partial time performed and the partial time goal for each segment. It was computed as the following: REðiÞ ¼ ðjS1 i À22:22jÞ þ ðjS2 i À44:44jÞ þ ðjS3 i À33:33jÞ where S1, S2, and S3 are the values in each segment (S1 key 2 to 8, S2 key 8 to 6, and S3 key 6 to 4) performed in each trial (i). The values of each segment were relativized by total time, as the following: where ts is the time (in ms) in each segment and T is the total time (in ms) in each trial.
The AE was computed as the difference between the total time performed (time spent between press keys 2 and 4) and the total time goal.
Two other dependent variables were used to detach performance from the motor response: the variability of partial time and the variability of total time. Here we use variability (variability of partial time and the variability of total time) as a measure that infers the motor response stability. The increase in variability indicates a decrease in motor response stability. First, the average (in percentage) of the differences among the three task segments (S1, S2, and S3) was calculated for each trial. Then, the average between consecutive trials over the practice was calculated as the following: where S1, S2, and S3 are the values in each segment (S1 key 2 to 8, S2 key 8 to 6, and S3 key 6 to 4) performed in each trial (i). VPT is the variability of partial time.
The variability of total time was obtained through differences among consecutive trials over the practice, as the following: where x(i) are the values assumed in the trials (i). VTT is the variability of total time. Figure 3 exemplifies and summarizes how the calculation of the performance measures (Figure 3(A)) and variability (Figure 3(B)) was performed. In this example, the partial times 205.87 ms, 375.41 ms, and 629.72 ms were performed between keys 2 to 8, 8 to 6, and 6 to 4, respectively. The total time performed between pressing keys 2 and 4 was 1,211 ms. Calculating the partial times performed as a function of the total time performed results in 17%, 31%, and 52% between key 2 to 8, key 8 to 6, and key 6 to 4, respectively. The absolute and relative errors calculation is highlighted in "the measures of performance" board. (B) The variability example is given, considering only four trials. The subsequent trial value subtracts the value of the current trials and so on. K2 ¼ key 2; K8 ¼ key 8; K6 ¼ key 6; K4 ¼ key 4. Moreover, the relative retention (Schmidt et al., 2019) and the switching cost (Apolin ario-Souza et al., 2021) were calculated for RE and AE. The relative retention compares the change in performance in the retention test with the change over the practice phase. This measure represents how much the performance increased from the practice phase to the retention test (Schmidt et al., 2019). The relative retention was calculated as the following: where bl1, bl10, and R represent the values of block 1, block 10, and retention test, respectively. The switching cost was used to indicate the amount of transfer to a new context (Apolin ario-Souza et al., 2021). The switching cost for all variables was calculated as the following: ðTÀRÞ Ã 100 R where T and R represent the values of the transfer test and retention test, respectively. Furthermore, cross-correlations were conducted to investigate how stability relates to performance in absolute and relative timing dimensions during practice. The cross-correlations were conducted between AE and variability of total time, AE and variability of partial time, RE and variability of total time, and RE and variability of partial time.

Statistical Analysis
The outlier analyses (intra-group, trial-by-trial) were conducted for relative error, absolute error, partial time, and total time. Any datum lying outside the range 3rd quartile þ 1.5 Ã interquartile range and 1st quartile À 1.5 Ã interquartile range was considered an outlier. If outlier data were detected, the outlier would first be removed. A new data average would be calculated, and finally, the outlier value would be replaced by the latest average value. No outlier was found.
RE and AE data were organized in blocks of 12 trials (data online access: 10.17632/zvwn9kcd7c). A Two-way ANOVA (2 groups X 10 blocks) with repeated measures in the second factor was conducted in the practice phase. Post-hoc analyses were performed using Tukey's test. Retention and transfer were analyzed using Student's ttests for independent samples. The variability of partial time and total time were organized in blocks of 12 trials. A two-way ANOVA (2 groups X 10 blocks) with repeated measures in the second factor was conducted for both variables in the practice phase. Post-hoc analyses were performed using Tukey's test.
Cross-correlation was organized into blocks of 60 trials. A two-way ANOVA (2 groups X 2 blocks) with repeated measures in the second factor was conducted for each cross-correlation comparison in the practice phase. Post-hoc analyses were performed using Tukey's test. Relative retention and switching cost were analyzed using Student's t-tests for independent samples.
The descriptive analysis is given in mean and standard error. A significant difference at the level of a ¼ 0.05 was adopted for all statistical analyses. Effect sizes were calculated using partial eta-squared (gp 2 ) in ANOVA and Cohen's d in t-tests.

Relative Retention and Switching Cost
There was a significant difference between groups in the relative retention in RE [t(20) ¼ 2.77, p ¼ 0.01, d ¼ 1.18]. The constant-variable group showed a higher level of relative retention than the variable-constant group (Figure 4(C)). There was no difference between groups in switching cost [t(20) ¼ 1.11, p ¼ 0.27, d ¼ 0.47] (Figure 4(D)).

Retention and Transfer Tests
Although there was no difference between groups in retention test [t(20) ¼ 0.03, p ¼ 0.97, d ¼ 0.01], there was a significant difference between groups in transfer The constantvariable group was more accurate than the variable-constant group (Figure 5(A)).

Relative Retention and Switching Cost
Inferential analyses did not detect differences between groups in the relative retention in AE [t(20) ¼ 0.35, p ¼ 0.72, d ¼ 0.15] (Figure 5(C)). However, there was a significant difference between groups in the level of switching cost [t(20) ¼ 2.88, p < 0.01, d ¼ 1.22]. The constant-variable group showed less switching cost than the variable-constant group (Figure 5(D)).

Variability of Partial and Total Time
Descriptive analyses of variability of partial and total time are shown in Figure 6. There were no significant   (Figure 6(B)). Although there was a statistically significant block effect [F(9,180) ¼ 4.98, p < 0.01, gp 2 ¼ 0.19], the post hoc analysis did not detect a difference among blocks. There was a significant group x block interaction in the variability of total time [F(9,180) ¼ 2.10, p ¼ 0.03, gp 2 ¼ 0.09] (Figure 6(D)). The post hoc analysis indicated that the constant-variable group showed lower variability of total time in block 3 and block 4 than in block 8 (p ¼ 0.02 and p < 0.01 respectively).

Cross-Correlation
Descriptive analyses of cross-correlation are shown in Figure 7. There was a significant group x block interaction in the cross-correlation between variability of partial time and RE [F(1,20) ¼ 5.67, p ¼ 0.02, gp 2 ¼ 0.22] (Figure 7(A)). However, the post hoc detected no significant difference (p > 0.05 for all interactions). There was a significant group x block interaction in the cross-correlation between variability of partial time and AE [F(1,20) ¼ 12.07, p < 0.01, gp 2 ¼ 0.37] (Figure 7(B)). However, the post hoc analysis detected no significant difference (p > 0.05 for all interactions). There were no significant main group condition effect [F(1,20)  The post hoc analysis indicated the level of crosscorrelation between variability of the total time and AE was lower in the first half of practice (1-60 trial) than in the second half of practice (61-120) in the variable-constant group (p ¼ 0.02) (Figure 7(D)).

DISCUSSION
This study first aimed to investigate the effect of the different combinations of practice schedules on motor response stability during practice. We hypothesized the constant-variable combination would show higher motor response stability in the first half of practice (1-60 trial) than the variable-constant combination. In contrast, in the second half of the practice (61-120 trials), the constant-variable combination would show lower motor response stability than the variable-constant combination. Our results partially confirmed our hypothesis. Motor response stability in the absolute timing dimension (variability of total time results) was similar between groups in the first half of practice. However, it was different in the second half of practice. The constant-variable group decreased motor response stability in the absolute timing dimension in the second half of practice (see the increase in variability of total time in Figure 6(D)). In contrast, the variable-constant group maintained the same level ( Figure 6(D)).
This result is reasonable since the goal variation was inserted at that moment for the constant-variable group and was not for the variable-constant group. Nonetheless, The variable-constant group showed lower variability of total time in block 3 and block 4 than in block 8 than the constant-variable practice. Data are presented in mean and standard error. S1 ¼ relative timing of key 2 to 8; S2 ¼ relative timing of key 8 to 6; S3 ¼ relative timing of key 6 to 4; bl ¼ blocks; R ¼ retention test; T ¼ transfer test; Ã ¼ p < 0.05.

Combinations of Practice on Motor Learning
when inserted in the first half of practice, there was no decrease in motor response stability in the variable-constant group. The insertion of variation before the learner has been able to develop stability (via constant practice) does not decrease motor response stability (Chamberlin & Magill, 1992). This result supports the hierarchy of learning proposition (Roth, 1988). The hierarchy of learning claims that a stable, well-defined relative timing movement pattern must be developed early in the learning process to permit an effective change in the spatialtemporal pattern (Lee et al., 1987). Our results show that the lack of development of stability in the first half of the variable-constant group did not allow for a decrease in motor response stability.
This study secondly aimed to investigate how stability relates to performance in absolute and relative timing dimensions during practice. We hypothesized the stability created by constant-variable combination would be more related to performance over the practice. Our results partially confirmed this hypothesis. Both groups did not differ in relative and absolute timing errors in the variability of partial time. However, the constantvariable combination showed the same level of cross-correlation between absolute error and variability of total time between the first and second halves of practice. Nonetheless, the variable-constant combination increased cross-correlation between the first and second halves of practice.
Our results show that in the constant-variable group, the decrease in motor response stability in absolute timing in the second half of practice (Figure 6(D)) was followed by the maintenance of the same level of crosscorrelation between variability of total time and absolute timing error (Figure 7(D)). This result indicates that motor response stability is related to performance. Lai et al. (2000) reported the level of performance achieved over the first half was maintained in the second half of practice in the constant-variable group. These results might indicate the decrease in motor response stability enabled learners to produce different motor responses to get closer to the different task goals required in variable practice. Recently findings have shown that variability can be handled according to the context (Dhawale et al., 2017;Dhawale et al., 2019;Wilde et al., 2005). For instance, songbirds can reduce or increase the variability of their songs to cater to the potential partner preference (Woolley & Kao, 2015). In motor learning, Dhawale et al. (2019) reported a decrease in motor response stability when performance was poor and an increase in motor response stability when performance was good. Wilde et al. (2005) reported that learners practicing a sequential key-pressing task adopted a uniform sequential FIGURE 7. Results of Cross-correlation. (A) Cross-correlation between variability of partial time and relative timing error during practice phase (2 blocks of 60 trials each). (B) Cross-correlation between variability of partial time and absolute timing error during practice phase (2 blocks of 60 trials each). (C) Cross-correlation between variability of total time and relative timing error during practice phase (2 blocks of 60 trials each). (D) Cross-correlation between variability of total time and absolute timing error during practice phase (2 blocks of 60 trials each). Data are presented in mean and standard error. 1-60 ¼ first block of 60 trials; 61-120 ¼ second block of 60 trials; PT ¼ partial time; TT ¼ total time; RE ¼ relative timing error; AE ¼ absolute timing error; Ã ¼ p < 0.05.
key-pressing structure when the context induced a decrease in motor response stability. However, when the context induced an increase in motor response stability, learners exploited other ways to deliver the key-pressing sequence.
Another aim of the study was to confirm that introducing variation late in the learning process would be more effective for learning the absolute timing dimension. Our results showed the constant-variable group was better than the variable-constant group in learning the absolute timing dimension ( Figure 5). This result was prominent when learners were required to achieve a new task goal in the transfer test. These results are consistent with other findings (Lage et al., 2007;Lai et al., 2000). In the present study, learners were required to adjust the absolute timing dimension both in the practice phase and in the transfer test.
Interestingly, our results found significant differences only in measures related to the dimension handled during variable practice (cross-correlation between total time and absolute timing error (Figure 7(D)) and variability of total time (Figure (6D)). There is a particular specificity between what is requested in variable practice and what is learned (Czy_ z, 2021; Lage et al., 2016). We speculate that the variation in the absolute timing dimension leads learners' attention to the absolute timing dimension, favoring absolute timing learning. Bicalho et al. (2019) showed the variation in absolute timing dimension during variable practice led to increased visual scanning of absolute timing information.
As expected, the constant-variable group did not show significant differences in learning tests in relative timing error (Figure 4(A)). These results are consistent with other studies' results (Lage et al., 2007;Lai et al., 2000). Surprisingly, the relative retention of the relative timing error in the constant-variable group was more significant than in the variable-constant group (Figure 4(C)). This result means that the improvement during practice (online learning) in the constant-variable group was even more enhanced over the retention interval (offline learning) than in the variable-constant group (Robertson et al., 2005). Although this improvement in relative retention did not influence the performance results in learning tests (retention or transfer), it might indicate that the memory process involved in learning differs between the different practice schedule combinations (Wright et al., 2016). Kantak and Winstein (2012) proposed a framework that integrates memory processes and the motor learning paradigm. The memory processes of encoding, consolidation, and retrieval might correspond to the practice phase, the interval between the end of practice and the retention test, respectively . According to this framework, we speculate that the memory consolidation differed between the practice schedule combinations. The stability generated by the order in which each practice schedule was undertaken in the combinations can influence the consolidation memory process (Robertson et al., 2005). End-of-practice stability is critical for memory process consolidation (Mosha & Robertson, 2016). A decrease in stability at the end of the practice favors learning, as a less stable memory can better interact with other memories (Cohen & Robertson, 2011). Sharing elements provided by this interaction activates common neural resources, strengthening memory. Mosha and Robertson (2016) showed that the level of stability at the end of practice relates to the performance level in transfer tests. Thus, we suggest future studies to investigate the relation between stability generated by practice schedules and the level of learning.

Conclusion
Our results support the premise that motor response stability is a determinant factor in learning the relative and absolute timing dimensions. Furthermore, our performance and motor response assessment showed that these two measures are related, supporting the hypothesis of motor response stability. Interestingly, our results found significant differences in measures related to the dimension manipulated during variable practice, indicating a possible specificity between what is requested in variable practice and what is learned. Finally, our findings suggest that the memory process involved in learning differs between practice schedule combinations since the constant-variable group showed a better improvement between practice and retention.

ACKNOWLEDGMENTS
We thank Joana Andrade Ramalho Pinto for assistance with grammatical corrections.