Elsevier

Acta Psychologica

Volume 157, May 2015, Pages 122-130
Acta Psychologica

Contributions from associative and explicit sequence knowledge to the execution of discrete keying sequences

https://doi.org/10.1016/j.actpsy.2015.02.013Get rights and content

Highlights

  • Sequential motor skill is based on various representation types.

  • Explicit sequence knowledge is partly reconstructed from implicit sequence knowledge.

  • Explicit sequence knowledge increases execution rate only at moderate execution rates.

Abstract

Research has provided many indications that highly practiced 6-key sequences are carried out in a chunking mode in which key-specific stimuli past the first are largely ignored. When in such sequences a deviating stimulus occasionally occurs at an unpredictable location, participants fall back to responding to individual stimuli (Verwey & Abrahamse, 2012). The observation that in such a situation execution still benefits from prior practice has been attributed to the possibility to operate in an associative mode. To better understand the contribution to the execution of keying sequences of motor chunks, associative sequence knowledge and also of explicit sequence knowledge, the present study tested three alternative accounts for the earlier finding of an execution rate increase at the end of 6-key sequences performed in the associative mode. The results provide evidence that the earlier observed execution rate increase can be attributed to the use of explicit sequence knowledge. In the present experiment this benefit was limited to sequences that are executed at the moderately fast rates of the associative mode, and occurred at both the earlier and final elements of the sequences.

Introduction

The question how serial motor skills are represented in the brain has inspired researchers for many decades. Consistent with behavioristic thinking, the classic chaining theory claimed that serial motor skills are based on associating the response-produced feedback stimuli from each movement with the ensuing movement (Bain, 1868, James, 1890, Skinner, 1934). However, the insight developed that chaining theory cannot account for relationships between non-adjacent items in serially organized behavior (Lashley, 1951). With the advent of cognitive psychology in the late 1950s, it became clear that knowledge is based on a combination of sensorimotor and symbolic (e.g., verbal) representations that are hierarchically or linearly associated to make up more complex representations (Paivio, 1963). Keele (1968) used this idea in his proposal that muscle commands can be planned before movement begins using a representation called a motor program. Later, the well-known schema theory further worked out this motor programming idea by combining two types of representations, the abstract General Motor Program defining a class of movements, and the Recall Schema containing parameters like speed and size to scale the General Motor Program into an executable motor program (Schmidt, 1975). In his neuropsychological theory of motor skill learning, Willingham (1998) assumed that motor sequences rely on a skill-dependent mixture of egocentric representations (e.g., relative to the head, shoulder, or trunk) and allocentric spatial representations (i.e., relative to a particular object in the outside world). At about the same time, Hikosaka et al. (1999) inspired many serial motor studies with their neurocognitive proposal of a fast learning effector-independent system using allocentric and eye- and hand-centered egocentric spatial coordinates, and a slowly learning effector-dependent motor system (e.g., Seidler et al., 2012, Shea et al., 2011, Wühr and Heuer, 2014). Representations in the motor system would code joint angles (a type of coding that may well be responsible for indications that movement control occurs in terms of successive body postures, Rosenbaum, Meulenbroek, Vaughan, & Jansen, 2001).

A particularly fruitful sequence learning paradigm involves the serial reaction time (RT) task (Nissen & Bullemer, 1987; for reviews see Abrahamse et al., 2010, Keele et al., 2003). Research with this task demonstrated that sequential motor skills may be based on knowledge that is not accessible to consciousness, that is, on implicit sequence knowledge. Still, there always appeared to be a few participants who do develop awareness of the order of the movements, in that they can verbalize that order. These participants are said to possess explicit sequence knowledge. This knowledge is used when, for example, we are typing our PIN on the basis of an explicitly recalled number (Fendrich & Arengo, 2004).1 Explicit sequence knowledge may develop also because repeated execution of externally guided movement sequences allows participants to test hypotheses about the order of the sequence elements (Rünger & Frensch, 2008). Aware participants are often found to execute movement sequences a little faster than participants without awareness (Curran and Keele, 1993, Mayr, 1996, Rüsseler et al., 2003). The fact that an execution benefit of explicit knowledge is not always observed has been attributed to a lack of time to translate explicit knowledge into actual movement (Cleeremans and Sarrazin, 2007, Destrebecqz and Cleeremans, 2001).

There is considerable consensus now that serial motor skill relies on a task-, age-, and practice-dependent mixture of verbal, spatial, and motor representations (Panzer, Gruetzmacher, Ellenbürger, & Shea, 2014). This redundant way of coding movement allows flexible adjustment of these serial motor skills to a variety of situations (Shea et al., 2011). This view was recently worked out in the Cognitive framework for Sequential Motor Behavior (C-SMB; Verwey, Shea, & Wright, 2015).

In our own empirical work, we have focused especially on the cognitive system that processes the various representations that underlie sequential movement series. Following a number of theoretical approaches that all assume that central processing and motor processing involve independent cognitive systems (Allport, 1980, Keele et al., 2003, MacKay, 1982, Pew, 1966, Schmidt, 1975), we argued that with the execution of relatively short (i.e., discrete) movement sequences a distinction can be made between a central and a motor processor (Abrahamse et al., 2013, Verwey, 2001). According to this Dual Processor Model, the central processor selects and loads movement representations into a short-term motor buffer, either one by one or on the basis of an integrated representation (Verwey, 1996). The motor processor then reads the individual movements from this motor buffer and executes each of them. If the central processor is not occupied by another task it can speed up execution by triggering, in parallel to the motor processor, the individual responses on the basis of stimuli (Verwey, 2001, Verwey, 2003b), or cognitive sequence representations (Ruitenberg, Abrahamse, De Kleine, & Verwey, 2012).

The Dual Processor Model assumes that the various representations underlying serial motor skills are processed in different ways. First, a distinction can be made between external and internal execution modes (Verwey, 2001). External control involves guidance of movement sequences by the central processor reacting to element-specific stimuli (Hikosaka et al., 1999, Tubau et al., 2007). This external control encompasses two execution modes (Verwey & Abrahamse, 2012). In the so-called reaction mode, a movement sequence is carried out by merely reacting to successive stimuli. However, when a movement series is carried out over and over again the sequence may continue to involve reacting to stimuli while processes and representations required for each response are gradually primed by the processes used to produce earlier responses. This priming of individual response movements probably occurs at all processing levels, and involves perceptual, egocentric and allocentric spatial, symbolic, and motor representations (Abrahamse et al., 2010, Goschke and Bolte, 2012). This way of producing movement sequences has recently been named the associative mode and is believed to be responsible for improvement in the serial RT task (Verwey & Abrahamse, 2012).

In the case of internal sequence control, each individual movement is assumed to be selected on the basis of a memory representation called a motor chunk (Rhodes et al., 2004, Sakai et al., 2003, Verwey, 1996). The use of these motor chunks is referred to as the chunking mode (Verwey, 2003b, Verwey et al., 2011). The above discussion suggests that in principle these motor chunks may include a mixture of motor and various spatial codes. However, the motor component in the motor chunks is likely to become dominant over the slower spatial representations because the motor representations can be more rapidly applied by the motor system (Verwey et al., 2015).

Given that participants may also be using verbal sequence knowledge to produce movement sequences, it is important to realize that research on serial verbal learning demonstrated that participants initially learn the first and last items of a series after which their explicit knowledge of the series gradually extends to the items in the middle (if at all) (for an overview see, Johnson, 1991). For instance, in an eight word list error rate increased from 5% at Position 1 to 18% at Positions 5 and 6, and then reduced to 10% at Position 8 (Figure 1 in Johnson, 1991). This phenomenon appears to hold for any task in which participants are required to give a response to a stimulus in a list, and it therefore is likely that this occurs with the development of explicit sequence knowledge in discrete sequence production (DSP) sequences too. A recent DSP task study did indeed show that explicit knowledge of two 6-key sequences was stronger for the initial two and last two responses than for the third and fourth responses (Verwey & Wright, 2014). However, this awareness correlated with execution rate only in the first, relatively slow, practice block. In unfamiliar DSP task sequences, too, the second and last responses have been reported to be relatively fast (De Kleine and Van der Lubbe, 2011, De Kleine and Verwey, 2009, Verwey, 2010, Verwey et al., 2010, Verwey et al., 2009). The fast second response in DSP sequences further appears to be relatively vulnerable to conditions that require cognitive processing, like concatenating sequences in a new order, and reversing stimulus-sequence mappings (Verwey, 2001), suggesting a high cognitive contribution to especially that response. Taken together, these findings suggest that in DSP sequences participants develop explicit knowledge especially of the first and last key presses, but that this speeds up keying sequences only when they are carried out at a moderate execution rate because translating explicit knowledge takes too long to speed up execution that already is based on the rapidly executed motor representations.

In line with the possibility to produce movement sequences in different execution modes, several studies demonstrated that participants can intentionally switch between these modes. One of those studies reported that when participants were responding to an unfamiliar stimulus series to produce 6-key sequences, and that sequence unexpectedly appeared to be familiar, execution rate increased (Verwey, 2003b). The finding that the response time distributions of the moderately fast key presses included two or three peaks strongly suggested that participants switched from the reaction to the associative and/or the chunking mode. More recently, Jiménez, Méndez, Pasquali, Abrahamse, and Verwey (2011) found indications that color coding in a serial RT task induced application of motor chunks in that short key pressing segments were carried out in bursts. Yet, when the colors were removed execution occurred again in the associative mode instead of in the chunking mode. Furthermore, Verwey and Abrahamse (2012) first had participants execute keying sequences in the chunking mode. Then these participants were exposed to a mixed-familiar condition in which 75% of the sequences included two deviating stimuli (called deviants) in an otherwise familiar sequence, while the remaining 25% of the sequences did not include deviants. The results suggested that when the participants expected these deviants the sequences were executed again in the associative mode (for replications, see Ganor-Stern et al., 2013, Ruitenberg et al., 2014b). Moreover, comparison with a mixed-unfamiliar condition suggested that after the first deviant in these mixed-familiar sequences participants switched from the associative to the reaction mode. These indications that the chunking and the associative execution modes can be intentionally switched on and off are consistent with theories that assume that there is a class of automatic processes that is active only after having strategically been prepared (Cohen and Poldrack, 2008, Dehaene and Naccache, 2001, Kiefer, 2012, Neumann, 1990, Pashler et al., 2001). Recent research indicated that participants had a tendency to use the execution mode that they had most experience with in the DSP task, and that experienced video-gamers were more likely to use the chunking mode (Verwey & Wright, 2014).

In conclusion, a familiar movement sequence practiced in the DSP task can be executed in the chunking, the associative, and the reaction modes, and participants can intentionally switch between these modes. Which mode is used probably depends on variables like the properties of the task, the participant's level of experience and age, and individual abilities and preferences (Verwey et al., 2015). It should be kept in mind that execution in both associative and chunking modes may in principle involve a mixture of explicit (e.g., verbal and/or spatial) and implicit (various spatial and/or motor) representations, but that it is not yet clear exactly which representations are being used in a particular execution mode.

To improve our understanding of the representations and execution modes people use when executing a movement sequence, in the present study I tested three accounts for a recent finding with the earlier mentioned DSP task study (Verwey & Abrahamse, 2012). The finding was that when a familiar 6-key sequence was executed in the abovementioned mixed-familiar condition – in which deviants could occur – execution rate increased toward the end (i.e., R5 and R6) of the 25% of the sequences without a deviant.

A first account for the increasing execution rate in the Verwey and Abrahamse (2012) study is based on the observation of a similar rate increase when participants were uttering learned word series (Fowler, 1981, Huggins, 1978, Lehiste, 1970). MacKay (1982) attributed this finding to activation accumulating across successive words in an associative memory network, and argued that accumulating activation is a common mechanism for controlling serial behavior in general. Earlier research with the DSP task did not find support for accumulating activation (Verwey, 1994, Verwey, 1999, Verwey and Eikelboom, 2003), but those sequences were carried out in the chunking mode in which an additional rate increase may not have been possible due to a ceiling effect. In the Verwey and Abrahamse (2012) study the rate increase may have emerged because sequences were carried out at the slower rate typical for the associative mode (Cleeremans & Sarrazin, 2007). This reasoning suggests that an increasing execution rate should appear in sequences carried out in the associative mode irrespective of whether one or two deviants are expected because accumulating activation is assumed to be independent of strategic control. In addition, accumulating activation should lead to a stronger response conflict for deviants that occur later in the sequence. These two effects should not correlate with the participants' awareness.

It is possible also that participants in the Verwey and Abrahamse (2012) study had developed sufficient explicit sequence knowledge to know that after the fourth response no deviants would occur anymore. This may have allowed them to strategically switch to another execution mode. In that case, one would expect that the fast last responses occur only with participants having explicit knowledge of those responses. The present experiment included only one deviant. Therefore, this explicit knowledge based mode switching could in sequences with no deviant be expected only at the last response because only there participants could be sure that no deviants would occur anymore.

Given this possibility that participants switch to another mode at the end of the sequence, a second account is that participants switched to the chunking mode. This account can explain the rate increase with familiar sequences for which motor chunks had developed, but does not predict such an increase with unfamiliar sequences. A third account is that participants switched to using explicit sequence knowledge. We do know that, like in the serial RT task, aware participants in the DSP task are often a little faster than those without awareness (Ruitenberg et al., 2012, Verwey and Abrahamse, 2012, Verwey et al., 2010, Verwey et al., 2011; however, Verwey & Wright, 2014 showed that explicit knowledge correlated with performance only early in practice when execution was still slow). This benefit of awareness might well be stronger when execution is in the associative mode due to its relatively low execution rate (see Cleeremans and Sarrazin, 2007, Destrebecqz and Cleeremans, 2001). As partial explicit knowledge (of the second and last responses) develops quite fast, one can expect that explicit knowledge also speeds up unfamiliar sequences in the mixed test condition. In addition, one can expect that the execution rate of individual responses correlates with explicit knowledge of those responses.

In the present study, participants executed familiar keying sequences in a test phase condition in which 75% of the sequences included one deviant at one of the Positions 2 to 6. The remaining 25% of the sequences did not include a deviant. This differed from the Verwey and Abrahamse (2012) study in which sequences included either 2 or no deviants. As explained earlier, practice was assumed to induce motor chunks, associative learning and explicit sequence knowledge, and also that participants are able to switch between the three execution modes. On the basis of these assumptions the present experiment was, thus, aimed at testing whether the increasing execution rate in the mixed-familiar keying sequences, reported in the Verwey and Abrahamse (2012) study, can be attributed to (a) accumulating activity across successive response representations, or to a strategic switch to (b) using implicit motor chunks, or (c) using explicit sequence knowledge.

Section snippets

Participants

Twenty-four undergraduate students took part in this experiment (average age of 20.9, range 18–26 years; 12 women) in exchange for course credits. The study was approved by the ethics committee of Faculty of Behavioral Sciences of the University of Twente.

Apparatus

Stimulus presentation, timing, and data collection were achieved using the E-prime© 2.0 experimental software package on a standard Pentium© IV Windows XP© PC. Unnecessary Windows services were shut down to improve RT measurement accuracy.

Results

Trials (i.e., sequences) with an error and the first two trials of each sub-block were excluded from the analyses because they are likely to involve another execution mode than the later sequences in that sub-block (Ruitenberg, Abrahamse, De Kleine, & Verwey, 2014a). In addition, trials were removed from the analyses when total execution time was greater than the average in that block across participants, plus 3 times the standard deviation across all participants. This eliminated 2.8% of the

Discussion

The present study aimed at clarifying why an earlier study had shown an increasing execution rate toward the end of discrete keying sequences that were most likely executed in the associative mode (Verwey & Abrahamse, 2012). Clarifying this was considered important because this was expected to provide a better understanding as to how various representations contribute to the skilled execution of movement sequences. In the Introduction, I distinguished three possible explanations for this

Conclusions

The present findings support earlier indications that practicing familiar movement sequences induces concurrent development of motor chunks, associations between successively used representations, and – to an individually different degree – explicit sequence knowledge (Abrahamse et al., 2013, Verwey et al., 2015). The results further support the earlier suggestion that when execution rate is high explicit knowledge has little effect on sequence execution rate (Verwey & Wright, 2014). Applying

Acknowledgments

This experiment was carried out by students in my laboratory classes. I would like to thank Rafaela Zigoli, Kamila Queitsch, and Sarah Schäffer for running the experiment, and David Wright and Jonathan Barnhoorn for reviewing a draft version of this paper.

References (69)

  • J. Rüsseler et al.

    Human error monitoring during implicit and explicit learning of a sensorimotor sequence

    Neuroscience Research

    (2003)
  • W.B. Verwey

    Effects of extended practice in a one-finger keypressing task

    Acta Psychologica

    (1993)
  • W.B. Verwey

    Evidence for the development of concurrent processing in a sequential keypressing task

    Acta Psychologica

    (1994)
  • W.B. Verwey

    Concatenating familiar movement sequences: The versatile cognitive processor

    Acta Psychologica

    (2001)
  • W.B. Verwey

    Diminished motor skill development in elderly: Indications for limited motor chunk use

    Acta Psychologica

    (2010)
  • W.B. Verwey et al.

    Distinct modes of executing movement sequences: Reacting, associating, and chunking

    Acta Psychologica

    (2012)
  • W.B. Verwey et al.

    Segmentation of short keying sequences does not spontaneously transfer to other sequences

    Human Movement Science

    (2009)
  • W.B. Verwey et al.

    Learning a keying sequence you never executed: Evidence for independent associative and motor chunk learning

    Acta Psychologica

    (2014)
  • E.L. Abrahamse et al.

    Representing serial action and perception

    Psychonomic Bulletin & Review

    (2010)
  • E.L. Abrahamse et al.

    Control of automated behaviour: Insights from the discrete sequence production task

    Frontiers in Human Neuroscience

    (2013)
  • P.L. Ackerman

    Determinants of individual differences during skill acquisition: Cognitive abilities and information processing

    Journal of Experimental Psychology: General

    (1988)
  • D.A. Allport

    Patterns and actions: Cognitive mechanisms are content-specific

  • A. Bain

    The sense and the intellect

    (1868)
  • J. Bo et al.

    Visuospatial working memory capacity predicts the organization of acquired explicit motor sequences

    Journal of Neurophysiology

    (2009)
  • J. Cohen et al.

    Automaticity in motor sequence learning does not impair response inhibition

    Psychonomic Bulletin & Review

    (2008)
  • T. Curran et al.

    Attentional and nonattentional forms of sequence learning

    Journal of Experimental Psychology: Learning, Memory, and Cognition

    (1993)
  • E. De Kleine et al.

    Representations underlying skill in the discrete sequence production task: Effect of hand used and hand position

    Psychological Research

    (2009)
  • A. Destrebecqz et al.

    Can sequence learning be implicit? New evidence with the process dissociation procedure

    Psychonomic Bulletin & Review

    (2001)
  • Z. Dienes et al.

    Measuring unconscious knowledge: Distinguishing structural knowledge and judgment knowledge

    Psychological Research

    (2005)
  • D. Fendrich et al.

    The influence of string length and repetition on chunking of digit strings

    Psychological Research

    (2004)
  • E.A. Fleishman et al.

    The relation between abilities and improvement with practice in a visual discrimination reaction task

    Journal of Experimental Psychology

    (1955)
  • C.A. Fowler

    A relationship between coarticulation and compensatory shortening

    Phonetica

    (1981)
  • A.W.F. Huggins

    Speech timing and intelligibility

  • W. James

    The principles of psychology

    (1890)
  • Cited by (0)

    View full text