A central neurobehavioral tenet asserts that the long-term retention of memories is subserved by two separate and distinct systems: a declarative system which retains singular experiences and memories of facts and events, and a procedural system which addresses repeated experiences and memories of skills and habits (Cohen and Squire, 1980; Brown and Robertson, 2007). Declarative and procedural memory processes interact closely during learning in everyday life. Procedural memory plays a major role during childhood when many new motor skills are acquired.

The cognitive processes and neural substrates that mediate our capacity to acquire and retain new skills have been studied extensively in recent years by following the time-dependent course of learning (for reviews see Robertson et al., 2004; Doyon and Benali, 2005; Censor et al., 2012). An extensive body of research has shown that in adults, the development of skilled performance often extends beyond the actual training experience. This has recently been shown to occur also in children (Dorfberger et al., 2007; Savion-Lemieux et al., 2009). Training-dependent gains in performance may appear hours after the termination of training, for example 24 h post-training. It has been proposed that these delayed (“offline”) gains in performance reflect memory consolidation neural processes within the processing stream that are involved in task performance, i.e., these processes are triggered by the training experience but require time to reach completion (Feldman, 2009; Xu et al., 2009; Caroni et al., 2012). The resulting gains are maintained for weeks (e.g., Korman et al., 2003; Dorfberger et al., 2007; but see Savion-Lemieux and Penhune, 2005).

In the previous decade, attention was devoted to the developmental difference that occurs from kindergarten to adulthood in gains accrued during a training session (Vinter and Perruchet, 2000; Ferrel-Chapus et al., 2002; Janacsek et al., 2012; Witt et al., 2013; Hodel et al., 2014). However, less attention was devoted to post-training processes in young children (Vinter and Perruchet, 2000; Wilhelm et al., 2008; Savion-Lemieux et al., 2009). It is not clear whether the generation of long-term procedural memory (i.e., the consolidation and retention phases) is similar among kindergarteners, older children, and young adults. Accumulating evidence supports the notion of faster memory consolidation during wakefulness in children (Dorfberger et al., 2007; Wilhelm et al., 2008; Ashtamker and Karni, 2013; Adi-Japha et al., 2014). However, it is unclear whether age-dependent differences emerge during retention. Furthermore, despite its importance, the retention of skills has not been studied in children younger than 8 years of age.

There exist several standardized tests of declarative memory for preschool children (e.g., within the Kaufman Assessment Battery for Children (K-ABC-II; Kaufman and Kaufman, 2004) and the children’s memory scale (Cohen, 1997, see also Adi-Japha, 2013). However, there are no standardized tests that assess procedural memory. One reason may be that available tasks for testing skill learning in young children are usually adaptations of grapho-motor (Vinter and Perruchet, 2000; Ferrel-Chapus et al., 2002), or motor-sequencing (Wilhelm et al., 2008; Savion-Lemieux et al., 2009; Janacsek et al., 2012; Hodel et al., 2014) tasks. These tasks generally require extensive training, and the learning activities involved can be very different from activities that young children learn to master. Task-specific accuracy demands and use of explicit strategies impede developmental studies of skill acquisition both within (Janacsek et al., 2012; Witt et al., 2013) and between (Wilhelm et al., 2008; Savion-Lemieux et al., 2009) training sessions. Young children may fail to improve in more complex tasks (e.g., Huyck and Wright, 2011; Vasudevan et al., 2011), or their results may depend on broader constructs, such as attention and task motivation (Hodel et al., 2014).

Furthermore, older children and young adults may reach a performance ceiling for accuracy (Vinter and Perruchet, 2000; Savion-Lemieux et al., 2009) or a plateau for speed (Wilhelm et al., 2008) during the training session, whereas young children may improve continuously. These performance differences at the end of the training session may further complicate the comparison of consolidation and retention processes across age groups. Children may exhibit consolidation gains in accuracy 24 h post-training in tasks in which adults exhibit a performance ceiling for accuracy during training (Savion-Lemieux et al., 2009). When a performance plateau for speed is observed only in adults, it is difficult to interpret differences in consolidation gains because these differences are related to whether the plateau was reached by the end of the training session (Hauptmann et al., 2005).

A grapho-motor learning task, suitable for kindergarten children, was recently introduced (Adi-Japha et al., 2011). The task requires the reproduction of a novel pattern during training, 24 h post-training and 2-weeks post-training. On each block of this task, the participants are asked to repeatedly connect three dots to form an “invented letter.” Identical patterns are arranged in a spaced writing array. The participant connects the dots that form the first pattern, and then connects the dots that form the following pattern, until the end of the block. Progression across the block follows the writing direction, and is composed of movement on the page when the participant is producing the pattern, and movement off-page (in the air) between patterns. The processes of forming the patterns across a block resemble handwriting production. The task is typical of kindergarteners’ activities and was designed to minimize accuracy demands. The skill learned in this task is the production of the pattern. The learning protocol of the skill confirms with the definition of procedural learning that leads to a formation of long-term memory for the learned skill (procedural memory, dealing with memories for skills). The individual is aware of acquiring the skill (defined as explicit procedural learning, Robertson et al., 2004). The to-be-learned pattern resembles shapes commonly found in developmental studies of visual-motor skills (e.g., Beery et al., 1997; Meisels et al., 1997). Copying such shapes in kindergarten was found to predict academic achievement later in school (Grissmer et al., 2010; Cameron et al., 2012).

The invented letter task (ILT) was introduced in order to study procedural learning in kindergarteners with language impairment and in typically developing peers who had comparable visual-motor integration skills (Adi-Japha et al., 2011). The findings indicated that both groups maintained a low error rate throughout the 2-weeks period. The children’s speed (i.e., the time taken to complete a block) improved during training, but only the typically developing kindergarteners further increased their speed at 24 h post-training. Their performance level was retained 2 weeks post-training.

The aim of the current study was to compare the learning, consolidation, and retention of the grapho-motor pattern (i.e., the ILT) in kindergarteners, second graders and young adults. Based on the previous findings in kindergarteners (Adi-Japha et al., 2011), we hypothesized that in addition to learning during the training phase, all three age groups would exhibit consolidation gains and retention of this simple task in terms of overall speed, while maintaining accuracy. Specifically, we hypothesized that improvement in speed would not come at the expense of accuracy (no speed-accuracy tradeoff). Gradual improvement in task performance across multiple sessions without speed-accuracy tradeoff is recognized as a characteristic of skill acquisition (procedural learning) in both motor and perceptual domains (Censor et al., 2012). Such learning of the task across the three age groups would suggest that the procedural learning processes that underlie learning in adults are present in early childhood, and can be tested using the ILT. This would enable future developmental studies of different factors associated with skill learning.

Of the 36 kindergarteners who began the study, three children did not participate in the 2 weeks post-training session. One child chose not to participate, one child exhibited reduced compliance, and one child was not available. One second grader and one adult could not attend the 24 h post-training session. The data from these participants were not included in the main analyses across the study period. However, their data were included in the correlation analyses that pertained to the training phase.

Two levels of analysis were performed. First, the overall production time and error rate were analyzed, in order to enable comparisons with other studies of skill learning that report speed and accuracy (Vinter and Perruchet, 2000; Dorfberger et al., 2007; Wilhelm et al., 2008; Savion-Lemieux et al., 2009). In light of the studies of drawing and handwriting by children (Adi-Japha and Freeman, 2001; Rosenblum et al., 2006), these were followed by analyses of the time spent while producing the shapes, the off-page time (i.e., while moving between the shapes and shape segments), the distance of the drawn line from the midpoint, and the number of strokes used to produce the shapes.

Performance was averaged across the blocks that constitute the four time points: initial training (blocks 1–4 on day 1, TP₁); end-training (blocks 9–12 on day 1, TP₂); 24 h post-training (four blocks, 24 h post-training, TP₃); and 2 weeks post-training (four blocks, assessed 2 weeks after day 1, TP₄). The data were subjected to a 3 (age group) × 4 (time points: initial training, end-training, 24 h post-training, and 2 weeks post-training) repeated measures ANOVA.

The results of the ANOVA analyses are presented in Table 1. Table 1 further reports the analyses across the three study phases, defined as the difference between successive time points: training (= end-of-training – initial training; TP₂–TP₁ in Table 1), consolidation (= 24 h post-training – end-training; TP₃–TP₂)_, and retention (= 2 weeks post-training – 24 h post-training; TP₄–TP₃)_.

In the current study, 5- to 6-years-old kindergarteners, 7- to 8-years-old second graders and young adults learned to produce a grapho-motor pattern. Our results indicate that training on the graphic symbol task resulted in significant gains concurrent with the training experience in performance time in all three age groups. Larger gains occurred for younger participants and individuals who initially performed more poorly. Between-session gains, as expressed at 24 h after the termination of the training session, were also exhibited by the three age groups. Gains in production time did not occur at the expense of an increase in the error rate, which was maintained. This demonstrates that major characteristics of skill acquisition, previously defined in adults (Censor et al., 2012), are typical of children’s learning. Our findings are similar to previous results of a developmental skill acquisition study in 9-years-old children and adults that used a more complex task (Dorfberger et al., 2007). Effective skill learning in young children can therefore be studied using this simple daily task, which does not require a great deal of attentional resources or declarative elements. Nevertheless, our findings suggest that an age advantage may exist in the long-term retention of skills.

Various studies of skill learning that compared children with young adults have reported differences in performance gains accrued during a given training experience. In different studies, either the adults (Thomas et al., 2004; Huyck and Wright, 2011; Vasudevan et al., 2011; Lejeune et al., 2013) or the children (Fischer et al., 2007; Bishop et al., 2012; Janacsek et al., 2012) exhibit an advantage. The results have differed even for the same task under different training lengths (Thomas et al., 2004; Fischer et al., 2007), and may depend on the method of comparing age groups (Janacsek et al., 2012), with adults showing better learning in shorter training lengths and in terms of normalized gains. Advantages in retention may further depend on the interval of delay studied, with children showing an advantageous retention after a delay of a few hours as opposed to days (Bishop et al., 2012; Ashtamker and Karni, 2013). Another line of research suggests that procedural-implicit skill learning is age-invariant (Reber, 1993; Meulemans et al., 1998; Vinter and Perruchet, 2000; Karatekin et al., 2007). Recent studies have attempted to address the disparity in previous findings (Lejeune et al., 2013; Nemeth et al., 2013; Witt et al., 2013; Lukacs and Kemeny, 2014). These authors suggest an additional account by which skill learning per se is effective at young ages. However, participants in different age groups use different strategies. For example, older participants rely more on executive attention control, sensorimotor integration, and the ability to build complex internal models and representational units (Ferrel-Chapus et al., 2002; Nemeth et al., 2013; Witt et al., 2013; Hodel et al., 2014). As a result, young adults have an apparent advantage over children in skill learning. The findings of the current study indicated higher training gains in younger children. This finding remained after improvement was considered proportionally to the younger children’s lower initial performance. Our data thus support the notion of childhood advantage or the assumption of age-invariance in gains accrued during the training of a novel task. Our data do not indicate qualitative differences in temporal or spatial features of the task acquisition between age groups. It should be noted, however, that skill learning studies have focused primarily on tasks that are acquired implicitly (e.g., statistical learning), whereas the current task was acquired explicitly (procedural-explicit learning, Robertson et al., 2004). Furthermore, from an educational point of view, our observation that children with a lower initial performance speed improved more than their peers underscores the benefits of training in the improvement of procedural skills (e.g., handwriting proficiency) and potentially other academic skills that can benefit from repeated practice (e.g., reading).

Age-related differences emerged in retention testing 2 weeks post-practice day. Although the overall level of performance was retained in all three age groups, a more detailed analysis revealed that the kindergarteners spent more time off-page and used slightly more segments for producing the shapes on the retention testing compared with the 24 h post-training assessment. Longer off-page periods and a more extensive use of pen strokes characterize the writing of children with low grapho-motor skills (Rosenblum et al., 2006) or low-attention (Adi-Japha et al., 2007). The study of handwriting skills suggests that the off-page time is associated not only with a move between writing elements but also with the planning of the next writing segment (Rosenblum et al., 2003, 2006). The number of strokes in grapho-motor production is a feature of fluent movement. In the current study, it is also a feature of accuracy. Accuracy is considered to be related to attention control and declarative task elements (Savion-Lemieux and Penhune, 2005; Janacsek et al., 2012). Thus, it may be suggested that with the passage of time, kindergarteners find it more difficult to retrieve the more explicit-, attentive-, or planning-related task elements that affect the shape of the trajectory. This is in contrast to lower-level motor execution processes that are expressed in the mean duration of shape performance (Sosnik et al., 2014) and were well-retained. A difficulty in retrieval of explicit task elements following an 8-weeks retention period has been identified in adults (Savion-Lemieux and Penhune, 2005). This finding may suggest that these less efficient retrieval behaviors are not unique to children. The finding of an age advantage during retention testing is consistent with a previous study by Dorfberger et al. (2007), who reported that adults exhibited improved performance on a finger-to-thumb opposition movement sequence task following 6 weeks of retention, whereas children maintained their performance level.

Previous developmental studies of skill learning have not identified associations between short-term or working memory performance assessed with the digit-span tasks and measures of learning (Savion-Lemieux et al., 2009; Lejeune et al., 2013). In the current study, however, kindergarteners’ changes in performance were associated with a motor measure of sequential short-term memory. This difference may have occurred because the current study used a task-specific (i.e., visual-motor) measure. A lower short-term motor memory span was associated with less on-page time improvement during training, and with an increase in the number of segments used on retention testing. Theories of skill learning, as well as imaging studies in adults, suggest that the fast learning phase (i.e., the training phase before a performance plateau is reached) is characterized by processes such as trial and error and the adaptation of performance solutions, as well as a more controlled execution in general (Anderson, 1982; Logan, 1988; Chein and Schneider, 2005; Doyon and Benali, 2005). These processes are related to interactions between the motor and pre-motor cortical regions and the prefrontal regions of the brain (Dayan and Cohen, 2011). The later phases of learning (i.e., consolidation or later learning) are characterized by less involvement of attentional and executive resources, and therefore correlation with short-term memory is not expected. The finding that kindergarten children with lower sequential short-term motor memory increased the number of segments used at retention testing support the hypothesis that this increase is related to planning and attentive task elements. Associations with short-term memory were restricted to kindergarteners, possibly because for second-graders the task was easy to perform and retain. As in previous studies (Savion-Lemieux et al., 2009), the verbal sequential short-term memory test was not associated with measures of learning. Digit-span tasks are considered short-term verbal/phonological memory tests (e.g., Gray, 2006) and are therefore less related to motor skill acquisition than motor sequential tasks.

Although the adults outperformed the kindergartners and second graders in terms of accuracy, the accuracy levels were maintained throughout the task in all age groups. Interestingly, the kindergarteners and second graders had a similar number of errors. However, the second graders were faster than the kindergarteners. These data suggest a different developmental pattern for speed and accuracy and are consistent with findings of other developmental studies of skill acquisition that used different paradigms (Savion-Lemieux et al., 2009; Janacsek et al., 2012). In addition, these findings strengthen the model proposed by Hikosaka et al. (2002), that motor skills are acquired and retained in two independent but parallel forms: speed and accuracy.

When considering the conclusions of this study, it should be taken into account that this is a small-scale study, and other developmental factors not examined in this study (e.g., attention) may have affected the results. The current study focused on younger children. However, more age groups should be tested in order to develop a comprehensive developmental understanding. Nevertheless, the current study extends the findings of previous researches which suggest that the stages of skill acquisition observed in adults are present not only in school-aged children but also in early childhood. These stages are present despite the numerous developmental differences in learning-related skills, such as memory span, working memory, attention, and planning. While the current study supports several associations between sequential motor short-term memory and sequential motor learning, more elaborate studies should be conducted to examine whether and how basic cognitive functions interact with skill learning (e.g., Fox et al., 2014).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.