Human Movement Science 40 (2015) 298–314

Contents lists available at ScienceDirect

Human Movement Science journal homepage: www.elsevier.com/locate/humov

What determines the impact of context on sequential action? Marit F.L. Ruitenberg a,b,⇑, Willem B. Verwey b, Elger L. Abrahamse a a b

Department of Experimental Psychology, Ghent University, Henri Dunantlaan 2, B-9000 Ghent, Belgium Cognitive Psychology and Ergonomics, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands

a r t i c l e

i n f o

Article history: Available online 29 January 2015 PsycINFO classification: 2300 2330 2343 Keywords: Context-dependent learning Sequential action Motor chunks Motor preparation

a b s t r a c t In the current study we build on earlier observations that memorybased sequential action is better in the original learning context than in other contexts. We examined whether changes in the perceptual context have differential impact across distinct processing phases (preparation versus execution of a motor chunk) within an ongoing movement sequence. Participants were trained on two discrete keying sequences, each of which was systematically presented in its own unique color during a practice session with either limited or extended practice. In a subsequent test session, sequences were performed with the same, with reversed, and with completely novel sequence-specific colors. The results confirm context-dependence in sequential action, the relevance of practice for its development, and its selective expression for the preparation but not the execution of highly practiced motor chunks. As such, the current study provides novel insights into the determinants of context-dependent sequential action. We finish by outlining the overall status of context-dependence in sequential motor behavior, and specify a general working model. Ó 2015 Elsevier B.V. All rights reserved.

⇑ Corresponding author at: Department of Experimental Psychology, Faculty of Psychology and Educational Sciences, Ghent University, Henri Dunantlaan 2, 9000 Ghent, Belgium. Tel.: +32 9 264 64 31; fax: +32 9 264 64 96. E-mail address: [email protected] (M.F.L. Ruitenberg). http://dx.doi.org/10.1016/j.humov.2015.01.006 0167-9457/Ó 2015 Elsevier B.V. All rights reserved.

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

299

1. Introduction The ability to acquire and perform action sequences is essential for everyday behavior, as most complex actions that people perform (e.g., lacing a shoe or playing the piano) consist of a series of simple movements that are executed in a specific order. Research has shown that such sequential action can become context-dependent – that is, performance has been found to be better in the original acquisition environment, as compared to different environments (e.g., Abrahamse & Verwey, 2008; Anderson, Wright, & Immink, 1998; Ruitenberg, Abrahamse, De Kleine, & Verwey, 2012; Ruitenberg, De Kleine, Van der Lubbe, Verwey, & Abrahamse, 2012; Shea & Wright, 1995; Wright & Shea, 1991). Context here refers to those perceptual features of the task setting that are not formally required for successful task performance, yet that may influence performance with practice based on contingencies with task-relevant information. Indeed, context-dependent learning and related concepts such as procedural reinstatement or specificity of learning (Healy, Wohldmann, Parker, & Bourne, 2005) are typically related to the notion that context-features become associated with the task during acquisition and subsequently enhance performance by acting as a cue for memory retrieval processes (e.g., Healy et al., 2005; Wright & Shea, 1991). In recent studies (Ruitenberg, Abrahamse, et al., 2012; Ruitenberg, De Kleine, et al., 2012) we have focused specifically on context-effects in discrete sequential action – the execution of short and fixed series of simple movements (i.e., key presses) at high pace. Such effects are both surprising and important because they show the ongoing relevance of (perceptual) context even in predominantly motoric tasks. In one of our earlier studies (Ruitenberg, Abrahamse, et al., 2012) the overall notion of impaired memory-based sequencing performance upon contextual changes was confirmed, but more specific predictions on the distinct sensitivity to context-effects of preparation and execution phases (see below) in such sequential action were not supported. Here we reexamine this issue using an improved experimental design. 1.1. Discrete sequential action A task that is well-suited for studying the cognitive processes underlying sequential action is the discrete sequence production (DSP) task (Abrahamse, Ruitenberg, De Kleine, & Verwey, 2013; Verwey, 1999; Verwey, Abrahamse, & De Kleine, 2010). This task typically involves series of two to seven stimuli that are presented in a fixed order. Each single stimulus remains on the screen until participants respond to it by means of a spatially compatible key press, after which the next stimulus of the series immediately appears (response-to-stimulus interval of 0 ms). Two fixed series of stimuli (and corresponding responses) are practiced extensively in random order so that a discrete sequence skill develops for each sequence. Due to the relatively simple responses in the form of key presses, the motor control component in this task is minimized (e.g., little need for joint angle or force control) such that cognitive control mechanisms involved in sequencing performance can be optimally examined. Based on work with this task a Dual Processor Model (DPM) of discrete sequence production has been developed (Abrahamse et al., 2013; Verwey, 2001). According to the DPM, sequencing performance involves sequence retrieval and motor buffer loading by a cognitive processor (i.e., preparation processes), followed by the fast execution of the motor buffer content by a dedicated motor processor (i.e., execution processes). The precise content that is loaded into the motor buffer changes over practice. Initially, the cognitive processor loads each individual element – that is, key press – by translating each stimulus into the appropriate response, which is then directly executed by the motor processor. With practice, motor chunks develop: representations of a series of successive responses that can be retrieved and loaded as if they were a single response. The cognitive processor can thus select and load such a chunk into the motor buffer as a whole, after which the motor processor executes all elements within the chunk. This means that the response time on the first key press of a motor chunk reflects selection, retrieval and execution (i.e., preparation phase), while response times on later key presses primarily reflect execution processes (i.e., execution phase) because motor chunk selection and retrieval had already occurred. Importantly, if total sequence length exceeds the limited capacity of a single motor chunk (about four key presses; e.g., Bo & Seidler, 2009), then a sequence will be segmented into multiple successive

300

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

motor chunks. This implies that a new motor chunk needs to be retrieved and initiated within an ongoing sequence, and this is assumed to be reflected in the typically observed unusually large response time somewhere halfway through the sequence – indicating a second wave of preparation (i.e., selection and retrieval). During the execution phase, the cognitive processor may support the motor processor in producing the motor chunk elements by means of direct stimulus–response (S–R) translations (Verwey, 2001, 2003b; Verwey & Abrahamse, 2012), so that the fastest possible responses are generated (i.e., statistical facilitation; e.g., Verwey, 2001). As S–R translations also involve preparation related to the selection and triggering of each individual response, in the typical DSP task all key presses in a sequence may include preparation processes of some kind. To better distinguish between preparation and execution processes during sequencing performance, De Kleine and Van der Lubbe (2011) developed the go/no-go DSP task. In this version of the DSP task, all key-specific stimuli of a sequence are displayed at a fixed pace before execution is started in response to a go-signal – participants thus perform the sequence from memory. As such, the cognitive processor cannot assist the motor processor by means of direct S–R translations as stimuli are absent during sequence execution. Consequently, preparation processes in the go/no-go DSP task are restricted to motor chunk initiation, which as explained above is reflected in the first response time of a sequence (T1) and in the transition from one motor chunk to the next within longer sequences (i.e., concatenation, TC; e.g., Bo & Seidler, 2009; Kennerley, Sakai, & Rushworth, 2004; Verwey et al., 2010). In line with this notion, previous studies demonstrated experimentally induced dissociations between T1 and TC on the one hand, and remaining T’s (reflecting mainly execution) on the other hand (for an overview see Abrahamse et al., 2013). For example, Ruitenberg, Verwey, Schutter, and Abrahamse (2014) showed that motor chunk preparation but not execution was slowed by rTMS stimulation of the pre-supplementary motor area, thus supporting the notion that responses reflecting preparation are distinct from those reflecting execution (see also Kennerley et al., 2004; Ruitenberg, Abrahamse, & Verwey, 2013; Verwey, 1999). 1.2. The present study In the current study we test a number of predictions of the DPM on context-effects in the go/no-go DSP task. In fact, these predictions were also central to our previous paper (Ruitenberg, Abrahamse, et al., 2012), but were not supported by our results (see below). In hindsight, we believe that the design in that study may not have been optimal for testing these predictions, and we therefore decided to reexamine them in an improved design. In anticipation of the current results, we now indeed confirm our predictions, but surprisingly also observed differences with our previous results beyond these predictions (we will elaborate on this in the Section 4). As context dependence logically seems experience-based, we predict that context manipulations will increasingly affect sequencing performance in the go/no-go DSP task with more practice, as the association between contextual features and the sequencing task – and thus the facilitating effect of a familiar context on sequence retrieval – grows stronger over time (e.g., Healy et al., 2005; Wright & Shea, 1991). This would imply that the impact of context on sequencing performance is larger after substantial as compared to relatively little practice, and thus that performance suffers more when some or all of these features are changed during testing after substantial practice. As sequencing performance in the DSP task becomes increasingly based on motor chunks with practice (e.g., Hikosaka et al., 1999; Verwey, 1999) another prediction for context-dependent retrieval can be specified. When practice is sufficiently extensive to allow for motor chunk formation, contexteffects should impact only those instances during performance that involve the selection and retrieval of these motor chunks, since only there context has facilitating potential. Hence, it follows that only the preparation phase but not the execution phase should be sensitive to context.1 In line with this notion, Magnuson, Wright, and Verwey (2004) found that search and retrieval processes used as part of response selection are facilitated by reinstatement of the learning context. 1 Besides such general motor chunk preparation, it could be predicted that the cognitive processor’s online S–R translations are affected by changes in context as well, but – as outlined above – these translations are prevented in the here employed go/no-go DSP task.

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

301

Overall, our main hypothesis therefore is that context-dependence after considerable practice will result in prolonged T1 and TC when the context is changed, since these responses reflect the start of new motor chunks and thus involve selection and memory retrieval processes. This prediction fits well with the notion of context-dependent memory retrieval observed before in other domains (e.g., Godden & Baddeley, 1975; Smith, 1985). In contrast to motor chunk preparation, we predict that execution remains relatively unaffected by our context manipulations. In our analyses below, we used T2 as an indicator for execution because it is least likely that this response involves the start of another motor chunk.2 In our previous study, neither of these predictions was confirmed as context-dependence was observed across all key presses within a movement sequence irrespective of the amount of practice (Ruitenberg, Abrahamse, et al., 2012). The aim of the current study therefore was to reexamine our predictions in an adapted design. We provided training (limited versus extended) to participants on two fixed sequences that each were systematically presented in a unique context (cf. Ruitenberg, Abrahamse, et al., 2012). The methodological changes compared to Ruitenberg, Abrahamse, et al. (2012) were as follows. First, instead of employing a between-subject design, each participant performed their sequences in all three test conditions: same context, switched context, and novel context conditions (see below for details). As such, the effect of context on individual chunking patterns could be examined, using a method for determining chunk points that has been used in various previous studies (see Section 3; cf. Bo & Seidler, 2009; Kennerley et al., 2004; Ruitenberg et al., 2014). This is crucial as several studies have shown individual differences in chunking patterns (e.g., Bo & Seidler, 2009; Kennerley et al., 2004; Verwey, 2003a; Verwey, Abrahamse, & Jiménez, 2009), and the between-participant comparison in our previous study may have concealed potential effects. Importantly, this design choice also substantially increased the power of the design. Second, to the latter purpose we additionally increased the number of trials in each test condition from 10 to 20 trials per sequence. Third, we instructed participants to use the four fingers of their left hand for responding – rather than two fingers of each hand in the previous study – as in our earlier study the spontaneous development of chunk points may have been confounded by slowing due to inter-manual transitions (Bruijnes, 2010; Jiménez, 2008; Koch & Hoffmann, 2000; Verwey et al., 2009). Excluding such transitions therefore renders a more valid determination of chunk points in the present study. As a final change, participants now performed both an unstructured and a prestructured sequence – that included a pause between the third and fourth stimulus – instead of performing two unstructured sequences as in Ruitenberg, Abrahamse, et al. (2012). Unlike the three methodological changes addressed above, this particular addition was not intended to more adequately suit the design to study the differential sensitivity to context manipulations of motor chunk preparation and execution, but rather to explore the development and transfer of motor chunk patterns in the go/no-go DSP task. In the typical DSP task, the inclusion of such a pause within a sequence has been shown to induce highly similar chunking patterns across participants (e.g., Verwey, 1996; Verwey & Eikelboom, 2003; Verwey et al., 2009), which does not seem to transfer to a concurrently practiced unstructured sequence (Verwey et al., 2009). We here examined whether prestructuring also results in uniform chunking patterns for sequences in the go/no-go DSP task, and whether transfer can be observed. This could provide novel insight into the strategies that participants adopt while performing sequences in the go/no-go DSP task. In anticipation of what follows, our results on this secondary theoretical interest of the current paper were not unambiguous and will therefore only be briefly addressed below. Yet, as elaborated on below, the sequence manipulation was observed to modulate context effects, thereby providing unexpected but interesting insights into the development of context-dependent sequence skill. As aforementioned, the present study involved three test conditions. In the same context condition, the combination of sequences and key-specific stimulus colors was identical to that in the practice session. In the reversed context condition, the colors of the key-specific stimuli were switched between the two sequences. Comparing the same and reversed context conditions allowed us to test whether the contextual color information would become directly associated with the (motor chunks within each) sequence such that it automatically cues retrieval. It was predicted that responses reflecting

2

We thank an anonymous reviewer for this suggestion.

302

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

preparation (T1 and TC) would be predominantly slowed by this context manipulation (cf. Magnuson et al., 2004), as other (within-chunk) responses are assumed to be entirely driven by the motor processor. More precisely, this was predicted more for extended than for limited practice as contextdependence takes time to develop. A novel context condition was also included, in which key-specific stimuli of both sequences were presented in the same, new, color. Comparing the same and novel context conditions allowed us to test whether the context information was used to begin with, for example to distinguish between the sequences (i.e., sequence selection). If sequence-specific context information is not used at all, performance will be similar in all context conditions. However, if the context is used for distinguishing between the sequences, but does not automatically trigger retrieval based on strong color-sequence associations, performance declines only in the novel context but not in the reversed context (as compared to the same context). 2. Method 2.1. Participants The participants in this study were 48 students (38 women and 10 men) of the University of Twente, aged 17–32 years (M = 22 years, SD = 2.8). According to Annett’s (1970) Handedness Inventory 46 participants were right handed and 2 were ambidextrous. Participants reported not to suffer from color-blindness, dyslexia or ADHD, and they had no problems with their sight (glasses or contact lenses were allowed). All participants gave their written informed consent. The study was approved by the ethics committee of the Faculty of Behavioral Sciences of the University of Twente. 2.2. Apparatus E-PrimeÓ 2.0 was used for stimulus presentation and data registration. The program ran on a Pentium IV class PC. Stimuli were presented on a 17-in Philips 107 T5 display. 2.3. Task and procedure Participants placed the fingers of their left hand on the c, v, b and n keys of a computer keyboard. A fixation cross was presented in the center of the screen, along with four horizontally aligned squares (see Fig. 1). They were drawn with a silver color line on a black background. The four stimulus squares spatially corresponded with the four response keys (e.g., the left most square corresponded with the ‘c’ key). After 1000 ms, one square was filled for 750 ms, after which immediately a second square was filled (i.e., inter-stimulus-interval of 0 ms), and so on until all six stimuli of the sequence had been presented. The default screen then reappeared for 1500 ms, after which the fixation cross was colored either red or green. The red fixation cross, presented for 3000 ms, indicated that sequence execution had to be withheld (a no-go trial). No-go trials were included to prevent participants from simply learning to start performing the sequence after the 1500 ms preparation interval had ended. Instead, they had to await the (no-)go signal before deciding whether or not to execute the sequence. The green fixation cross, presented for 100 ms, indicated that participants had to repeat the presented sequence by pressing the appropriate keys in the correct order (a go trial). After each completed sequence, the default screen appeared for 1000 ms before the first stimulus of the next sequence was presented. Participants were instructed to produce the sequences as fast and accurately as possible. An error message appeared when a participant reacted before the go/no-go signal. In addition, feedback was provided upon completion of a response sequence to indicate whether errors had been made. During the practice session, participants practiced two 6-key sequences that were each constructed by concatenating two 3-key series (cbv, vnc, bcn or nvb). Whereas fixed sequences were performed by each participant, the possible combinations of these 3-key series were counterbalanced across participants (see also below). For each participant, one of the two sequences involved a repetition of a 3-key series and included an inter-stimulus-interval of either 200, 400 or 600 ms between stimulus 3 and 4

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

303

Fig. 1. An overview of stimulus presentation for an unstructured sequence in the go/no-go DSP task. Each key-specific cue was presented for 750 ms. After a 1500 ms preparation interval, a go or no-go signal was presented. The go signal consisted of a green ‘+’ and a no-go signal consisted of a red ‘+’ (color figure online). Note that the 200–600 ms pause between the presentation of the third and fourth stimuli of the prestructured sequence is not depicted.

(e.g., bcn–bcn, with ‘’ indicating the pause) – with all other inter-stimulus-intervals being 0 ms as mentioned above. This sequence is hereafter referred to as the prestructured sequence. The three inter-stimulus-intervals were used to prevent participants from learning a fixed rhythm, and the exact duration was determined randomly per trial. Importantly, in the test session all intervals were 0. The second, unstructured sequence involved a combination of two different 3-key series and did not include an enlarged inter-stimulus-interval (e.g., bcnvnc). In order to prevent finger-specific effects on individual response times, the sequences were counterbalanced across participants. For each participant, the prestructured and unstructured sequence never started with the same key and combinations resulting in repetitions (e.g., cbvvnc) were discarded. One of the sequences was always presented in yellow, while the other sequence was always presented in blue (counterbalanced across participants and prestructured vs. unstructured sequences). A practice block included 50 go-trials per sequence, and 8 nogo trials. The sequences were presented in random order. Half of the participants completed one practice block (i.e., limited practice group), the other half completed six practice blocks and thus practiced 300 trials per sequence (i.e., extended practice group). There was a short 30-s break halfway through each practice block and a 2-min break in between blocks. After completion of the final practice block, there was another 2-min break before the test session was started. During the test session, participants performed their sequences in three context conditions. In the same context condition, each sequence was presented in the same color as during practice. In the reversed context condition, the colors of the two sequences were switched. In the novel context condition both sequences were presented in red. Each test block consisted of 20 randomly presented trials per sequence (plus 4 no-go trials) and the order of the three blocks was counterbalanced across participants in each practice group.

304

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

Finally, participants completed a questionnaire that measured their awareness of the sequences. They were asked to recall both of the sequences they had performed during the experiment, by writing down the letters of the keys in the correct order. Then, they were asked to recognize their two learned sequences from a list of twelve alternatives. 2.4. Data analysis Response time (RT) was defined as the time between onset of the go-signal and depression of the first key and as the time between two consecutive key presses within a sequence. Sequences in which one or more errors were made were excluded from the analyses. In addition, sequences were omitted from the analyses when the mean RT per key press within the sequence exceeded the group mean in a particular (practice or test) block by more than 2.5 standard deviations. This was done separately for the prestructured and unstructured sequences and resulted in the removal of 9% of the data in the practice session and .20), suggesting that performance of the practice groups in this block did not differ. Performance was also analyzed in terms of accuracy by using ANOVAs on error percentages including the same independent variables as described above. For the limited practice group, errors differed across key position within the sequences, F(5, 115) = 30.65, p < .001, g2p ¼ :57. Errors increased from key 1 to key 5 (3.8% to 11.8%), then slightly decreased to key 6 (9.9%). Participants made fewer errors in the prestructured sequence than in the unstructured sequence (6.2% vs. 9.8%), F(1, 23) = 7.22, p < .05, g2p ¼ :24. However, a Sequence  Key interaction suggested that this differed per key position, F(5, 115) = 3.20, p < .05, g2p ¼ :12. There was no difference in error rates between prestructured and unstructured sequences on keys 1–3 (4.0% vs. 5.8%, p = .51), but probably due to the repetition of the 3-key segment, participants made fewer errors on keys 4–5 of the prestructured sequence than on the same keys of the unstructured sequence (8.4% vs. 13.7%), F(2, 46) = 6.44, p < .05, g2p ¼ :22.

Fig. 3. Mean RTs (ms) per key position in the prestructured and unstructured sequences in the test session. Error bars represent standard errors of the mean.

306

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

For the extended practice group, error percentages decreased from 7.9% in the first practice block, to on average 3.3% in blocks 2–6, F(5, 115) = 10.23, p < .001, g2p ¼ :31 (the effect of Block was absent when removing the first practice block from the analysis). Results further showed that errors increased with key position in the sequence (2.1% on key 1–6.0% on key 6), F(5, 115) = 27.69, p < .001, g2p ¼ :55, but this difference reduced with practice, F(25, 575) = 3.65, p < .01, g2p ¼ :14. There was no difference in errors between prestructured and unstructured sequences (p = .13). Finally, in line with the RT data, an ANOVA on errors in the first practice block with Key (6), Sequence (2) and Practice group (2) showed no indications for performance differences between the groups (ps > .45).

3.2. Test session Below we first report an omnibus ANOVA which is aimed at testing the general assumption that practice affects the degree to which memory-based sequencing performance in the go/no-go DSP task is context-dependent. Second, we address our main research question and analyze whether preparation and execution processes involved in well-learned movement sequences are differently affected by our context manipulations, with motor chunk preparation involving both the first key press of the sequence and possible chunk points somewhere halfway through an ongoing sequence (see above) and execution involving the second key press of the sequence.3 This issue is explored via a separate analysis because it logically zooms in on the performance of participants in the extended practice group only (since motor chunks are known to develop only with sufficient practice; e.g., Bo & Seidler, 2009; Verwey, 1999).

3.2.1. Effects of practice on context-dependence Participants’ performance in the three context conditions was analyzed via a mixed ANOVA on RTs with Context (3; same vs. reversed vs. novel), Sequence (2) and Key (6) as within-subject variables and Practice (2) as a between-subject variable. The prestructured sequence was performed faster than the unstructured sequence (269 vs. 292 ms), F(1, 46) = 12.73, p < .01, g2p ¼ :22, while RT patterns across keys differed for the two sequences, F(5, 230) = 2.92, p < .05, g2p ¼ :06 (see Fig. 3). Results showed that performance across the various context conditions differed, F(2, 92) = 3.31, p < .05, g2p ¼ :07. Moreover, a tendency toward a Context  Practice interaction suggested that practice modulated the effect of the context manipulations, F(2, 92) = 2.97, p = .057, g2p ¼ :06. This two-way interaction was further qualified by a Context  Practice  Sequence interaction, suggesting that the prestructured and unstructured sequences were differently affected by the context manipulations after limited and extended practice, F(2, 92) = 3.89, p < .05, g2p ¼ :08. To investigate in more detail the significant three-way interaction, an ANOVA on RTs with Context (3), Sequence (2) and Key (6) was performed for each practice group. No main or interaction effects of context were obtained after limited practice (ps > .27), but for the extended practice group results showed a main effect of Context, F(2, 46) = 7.73, p < .01, g2p ¼ :25, as well as a Context  Sequence interaction, F(2, 46) = 4.81, p < .05, g2p ¼ :17. Although both the prestructured and the unstructured sequences were sensitive to the context manipulations, Fs(2, 46)>5.57, ps < .01, g2p s ¼ :19, the effects of context differed for the two sequences. Specifically, performance of the prestructured sequence dropped in both the reversed and novel context conditions (266 and 262 ms, respectively) compared to the same context condition (238 ms), Fs(1, 23)>9.47, ps < .01, g2p s ¼ :29. In contrast, performance of the unstructured sequence was only impaired in the novel context condition compared to the same context condition (285 vs. 257 ms), F(1, 23) = 11.27, p < .01, g2p ¼ :33, but was unaffected in the reversed context condition (262 ms; p = .34). In sum, memory-based sequencing performance was affected by the context manipulations after extended practice, but not after limited practice. 3 As discussed above, we believe that T2 entails the cleanest indicator of pure execution processes because it is rather unlikely that a new motor chunk is selected and loaded at this point in the sequence. Importantly, related analyses that considered T1/TC vs. other Ts, T1 vs. T2, or TC vs. T2 all provided consistent findings similar to what is reported here.

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

307

Fig. 4. Mean RTs (ms) for the preparation and execution phase of the prestructured sequence (A, top) and unstructured sequence (B, bottom) for participants in the extended practice group as a function of context condition in the test session. Across sequences, preparation was slower than execution. In the prestructured sequence, context manipulations affected preparation but not execution. Specifically, preparation was impaired in both the reversed and novel context conditions compared to the same context condition. Performance of the unstructured sequence was only impaired in the novel context condition, but this effect did not differ between the preparation and execution phases. Error bars represent standard errors of the mean. (⁄⁄⁄ p < .001, ⁄⁄ p < .01, ⁄ p < .05)

308

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

3.2.2. Context-dependence of preparation vs. execution To test the hypothesis that motor chunk preparation would be differently affected by the context manipulations than execution (i.e., T2), we evaluated sequencing performance of participants in the extended practice group and determined for both sequences the mean RTs of key presses that reflect preparation processes. We followed a procedure that has been previously employed for determining participants’ chunking patterns (e.g., Bo, Borza, & Seidler, 2009; Bo & Seidler, 2009; Kennerley et al., 2004; Ruitenberg, Abrahamse, et al., 2012; Ruitenberg et al., 2013, 2014). The first key press of each sequence was assumed to always reflect preparation of the first chunk, and was thus defined as a preparation point (i.e., T1). To examine whether other preparation points (TC) had developed within the sequence during practice, paired t-tests were performed (p < .05, one tailed) per participant on RTs of the third, fourth and fifth key press of the sequence in the final practice block to evaluate whether the RT of a given key press was significantly longer than the RT of both its preceding and succeeding key presses. The second and sixth key presses were not evaluated as such, because T2 was assumed to always be included in the first chunk and T 6 was not succeeded by a key press. This procedure yielded additional preparation points (i.e., TC) for 18 of the 24 participants with respect to the prestructured sequence,4 and 14 of the 24 participants with respect to the unstructured sequence. Mean RTs of participants in the extended practice group were then subjected to an ANOVA with Context (3; same vs. reversed vs. novel), Sequence (2; prestructured vs. unstructured) and Phase (2; preparation5 vs. execution) as within-subject variables. Results again showed the aforementioned main effect of context, F(2, 46) = 5.11, p < .05, g2p ¼ :18, and in addition showed that key presses reflecting preparation (i.e., T1 and TC) were slower than the key press reflecting execution (i.e., T2; 352 vs. 253 ms; cf. Fig. 4), F(1, 23) = 530.62, p < .001, g2p ¼ :57. Most interestingly, the analysis revealed a three-way interaction between Context, Phase and Sequence, F(2, 46) = 5.03, p < .05, g2p ¼ :18. To investigate this interaction, we ran separate ANOVAs for the prestructured and unstructured sequences. For the prestructured sequence a Context  Phase interaction was observed, F(2, 46) = 4.40, p < .05, g2p ¼ :16. Specifically, as illustrated in Fig. 4 (panel A), results showed that context affected preparation, F(2, 46) = 5.88, p < .05, g2p ¼ :20, but not execution (p = .31). Paired t-tests showed that preparation was slowed in both the reversed, t(23) = 2.79, p < .05, and novel context conditions, t(23) = 3.04, p < .01, compared to the same context condition. Preparation in the reversed and novel context conditions did not differ significantly (p = .10). Results showed no indications that context differently affected preparation and execution in the unstructured sequence (p = .99), but the main effect of context remained significant, F(2, 46) = 4.47, p < .05, g2p ¼ :16 (Fig. 4, panel B). Finally, to explore whether T1 and TC were differently affected by the context manipulations, we ran separate ANOVAs on RTs for the prestructured and unstructured sequences with Context (3) and Key press (2; T1 vs. TC) as within-subject variables. Sequence was not included as a variable as this would have resulted in the loss of data from participants who did not segment one or both of their sequences. For both sequences, results showed main effects of Context (Fs > .3.54, ps < .05, g2p s ¼ :20) as well as Key press (Fs > .9.96, ps < .01, g2p s ¼ :37, with T1 being slower than TC; 372 vs. 274 ms in the prestructured sequence and 418 vs. 260 ms in the unstructured sequence). There was no significant interaction effect for either of the sequences (ps > .23), suggesting that context had a similar impact on T1 and TC. In summary, results showed that after extended practice, both the preparation and execution phases of the unstructured sequence were impaired in the novel context condition, but unaffected in the reversed context condition. In the prestructured sequence performance of key presses reflecting preparation – but not execution – of motor chunks was slowed in both the reversed and novel context conditions.

4 Detailed inspection of the results showed that the location of chunk points within the prestructured sequence differed across participants and thus did not uniformly develop at the location of the pause within the sequence, in contrast to what has been observed for sequences in the typical DSP task. Furthermore, we found no indications for transfer of chunking patterns, as only six participants showed similar patterns for the prestructured and unstructured sequences. We will not further discuss these issues here. 5 As outlined in the Introduction, key presses that reflect motor chunk preparation logically also include execution, but for the sake of clarity we here opt for a binary description.

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

309

3.2.3. Accuracy The percentage of erroneously executed key presses was calculated per participant for each sequence in each context condition and subjected to an ANOVA with Context (3), Sequence (2) and Key (6) as within-subject variables and Practice (2) as a between-subject variable. The mean error percentages for the same, reversed and novel context conditions were, respectively, 2.8% vs. 4.1% vs. 3.5% (p = .20). Error percentages increased with key position within the sequences, F(5, 230) = 14.47, p < .001, g2p ¼ :24. In addition, a Context  Sequence  Key interaction suggested that this was dependent on both context condition and sequence, F(10, 460) = 2.67, p < .05, g2p ¼ :06. In the same context, error percentages differed between the keys presses of the two sequences, F(5, 230) = 2.72, p < .05, g2p ¼ :06. Specifically, there was no difference in errors between the prestructured and unstructured sequences on key presses 1–3 (1.7% vs. 1.5%, p = .73), but participants made fewer errors on key presses 4–6 of the prestructured sequence than the unstructured sequence (2.8% vs. 5.0%), F(1, 46) = 5.02, p < .05, g2p ¼ :09. Results showed no differences in accuracy between the key presses within the sequences for the other context conditions (ps > .12). No further effects were observed. 3.3. Explicit sequence knowledge Finally, we determined the number of correctly recalled and recognized sequences for each participant. With respect to recall, the prestructured sequence was correctly reproduced by 40 participants (83%) and the unstructured sequence by 47 participants (98%). The prestructured sequence was correctly recognized by 47 participants (98%) and the unstructured sequence by 45 participants (94%). Results of the awareness questionnaire showed no differences in recall or recognition of the prestructured and unstructured sequences between the limited and extended practice groups (v2s(1) < 3.2, ps > .07). This indicates that the observed modulating effect of practice on context-dependence cannot be attributed to group differences with regard to explicit sequence knowledge. The large number of participants that correctly recalled and recognized their sequences in this go/no-go DSP task did not allow for a meaningful analysis of the relationship between explicit sequence knowledge and behavioral measures (ceiling effect). 4. Discussion The present study strengthens the notion that performing discrete movement sequences is susceptible to, and partly dependent on, the context in which the sequences were acquired (cf. Ruitenberg, Abrahamse, et al., 2012). It is typically assumed that sequential action is predominantly represented in the brain in terms of motor codes. Therefore, results of the current and previous studies are interesting as they demonstrate that (a change in) perceptual context impacts the execution of even well-learned movement sequences (e.g., Abrahamse & Verwey, 2008; Ruitenberg, Abrahamse, et al., 2012; Ruitenberg, De Kleine, et al., 2012). Whereas in the studies of Abrahamse and Verwey (2008) and Ruitenberg, De Kleine, et al. (2012) such context-dependence was shown in tasks (respectively the SRT and typical DSP task) that continue to provide perceptual information during execution, the impact of context in the current study is even more surprising because stimuli were no longer present at the time of actual performance (go/no-go DSP task). The temporal parameters in this latter task, which we adopted from De Kleine and Van der Lubbe (2011), allow participants sufficient time to build up a representation and prepare the to-be-performed sequence – yet we still observed a clear impact of context on performance. The current study thus strengthens the claim from our previous study that even motor representations include perceptual information (Ruitenberg, Abrahamse, et al., 2012), and shows that the context-dependence of memory-based sequencing performance can be replicated. In line with the notion that context-dependent retrieval involves the development of associations between context and task features, which strengthen with practice and can facilitate memory retrieval processes (Healy et al., 2005; Wright & Shea, 1991), the here observed modulation by practice (limited versus extensive) shows that contextual dependencies need time to develop; for the specific task employed here, 50 trials vs. 300 trials of training is sufficient to produce a detectable modulating effect of practice on the impact of contextual changes.

310

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

The main goal of our study was to test the hypothesis that context affects key presses that involve preparation processes – but not key presses that involve only execution processes. Results of the prestructured sequence were in line with this notion. Results of the unstructured sequence showed a different pattern, though, which we elaborate on below. In our previous study (Ruitenberg, Abrahamse, et al., 2012), modulating effects of processing phase and practice on context-dependence were not confirmed, yet the present study was successful in reexamining and confirming predictions from the DPM in an adapted experimental design. 4.1. The impact of our methodological changes As outlined in the Introduction, the present study involved various methodological changes compared to our prior work (Ruitenberg, Abrahamse, et al., 2012). In this section, we discuss how these changes may have contributed to the current findings. First, the within-subject manipulation of context and increased number of trials in the test session are assumed to have increased the power of the present study compared to the previous study. More importantly, as aforementioned, using a withinsubject design allowed us to examine the effect of our context manipulations on individual chunking patterns – rather than comparing chunk patterns of different participants, which may in fact have concealed potential effects in our previous study as other work has demonstrated individual differences in chunking patterns (e.g., Bo & Seidler, 2009; Kennerley et al., 2004; Verwey, 2003a; Verwey et al., 2009). We feel that these changes render the current study design to be stronger than our previous one in terms of reliability of the observed effects. Another change concerns the use of different hand settings for responding to the key-specific stimuli: whereas participants in the Ruitenberg, Abrahamse, et al. (2012) study were instructed to use the index and middle fingers of both hands, participants in the present study were instructed to use the four fingers of their left hand. As previous findings indicated that inter-manual transitions may significantly slow responses (Jiménez, 2008; Koch & Hoffmann, 2000; Verwey et al., 2009), we considered it critical to avoid this possible confound regarding the development and determination of chunk points as defined in the present study. Finally, there also seems to be an unexpected but important role for sequence structure with respect to the impact of context. We observed that the first key press of a motor chunk (i.e., T1 and TC) in the prestructured sequence was slowed in both the reversed and novel context compared to the same context, while execution was not significantly affected by the context manipulations. These observations are fully in line with the idea that preparation processes – including the selection and retrieval of motor chunks – are sensitive to perceptual changes, and suggest that context may facilitate the search for and retrieval of motor chunks (cf. Magnuson et al., 2004) because of its integration in the motor chunk representations (Healy et al., 2005; Wright & Shea, 1991). It should be noted, though, that slowing in the reversed and novel context conditions may relate to different causes: while preparation in the reversed context condition may have been more difficult because the stimulus color primed preparation of the alternative overall sequence and/or specific motor chunks (cf. Ruitenberg, Abrahamse, et al., 2012), it may have been slowed in the novel context condition because the absence of sequence-specific stimulus colors hindered retrieval. At the same time, results of the unstructured sequence showed a very different picture, indicating that – unexpectedly – the sequence structure also determines the impact of context, rendering dependence on the perceptual context in sequential action itself being sensitive to the overall task context (including level of practice and sequence structure). Specifically, for the unstructured sequence it appeared that contextual changes affected both preparation and execution phases to a similar extent. Moreover, performance was impaired in the novel context condition, but unaffected by the context reversal, suggesting that participants did not develop associations between the contextual features and the sequences. Rather, it seems that motor chunk selection was facilitated by the use of sequence-specific stimulus colors in both the same and reversed context conditions, but impaired in the novel context condition as selection could then only be based on the spatial stimulus information. This would explain slowing of key presses that reflect preparation processes. The observation that the second key press of the unstructured sequence was slowed in the novel context condition may be even harder to explain – it seems unlikely that the motor processor

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

311

(assumed to execute motor chunks based on motoric codes) would be sensitive to perceptual changes. A potential explanation is related to the construction of the sequences. Although both sequences involved a combination of two 3-key segments, the prestructured sequence involved a repetition of two identical segments (e.g., bcn–bcn), whereas the unstructured sequence involved two different segments (e.g., bcnvnc). It could therefore be argued that – besides the pause – the two sequences also differed in complexity. Previous studies have shown that context-dependence increases as the sequencing task becomes more complex (Anderson et al., 1998; Wright & Shea, 1991). Assuming that the unstructured sequence in the present study would be more complex than the prestructured sequence, it is thus surprising that performance of the unstructured sequence was in fact less broadly affected by the contextual changes (i.e., only in the novel context, but not the reversed context condition). Another – perhaps even more influential – consequence of the here employed sequence construction, is that the two identical segments in the prestructured sequence were practiced twice as often as the two different segments in the unstructured sequence. It could therefore be speculated that the associations between the context and specific motor chunks in the unstructured sequence were weaker than was the case for the prestructured sequence. Although the cognitive processor was clearly unable to assist the motor processor by means of S–R translations to trigger responses, its efforts may possibly have been affected at an earlier stage: as motor chunk retrieval for the unstructured sequence was more difficult in the novel context condition (where sequence-specific colors were absent), this may have resulted in poorer quality of the information that was eventually loaded into the motor buffer at the time that execution started. This may tentatively explain why performance across all key presses was impaired. Without enabling a clear-cut interpretation on the effect of sequence structure at this stage, the current study clearly demonstrates how multiple factors are at work in determining the overall context-dependence of sequential action. Specifically, we showed how the amount of practice, the processing phase within a sequence, and sequence structure all modulate context-effects. The fact that the current results differed somewhat from those in our previous study (Ruitenberg, Abrahamse, et al., 2012) and the surprising result for the unstructured sequence in the current study, indicate that context-effects on sequencing performance – though systematically observed – are complex and difficult to grasp. Future work should therefore systematically explore the role of sequence structure for a more comprehensive understanding of experimental factors affecting and mechanisms underlying context-dependence. Nonetheless, the present and previous studies provide important insights into the determinants of context-dependent sequential action, which we will outline in the next section. Finally, we want to briefly discuss the processes underlying preparation of motor chunks beyond the first chunk within a movement sequence (i.e., concatenation) – which in the present study was observed for more than half of the participants after extended practice (see above). Recently, indications have been reported that under very specific conditions (i.e., substantial practice of >500 trials per sequence and a dual task context) such concatenation may partly come under automatic control, as if the first motor chunk of a sequence is associated with and therefore primes a subsequent motor chunk (Verwey, Abrahamse, De Kleine, & Ruitenberg, 2014). Depending on the choice of terminology, this would render preparation of the first motor chunk in a sequence qualitatively different from preparation of later motor chunks. In the current study we can neither make any claims on the presence of such motor chunk priming to begin with (since we here do not employ a dual task paradigm and provide less practice than in the study by Verwey et al. (2014)), nor on the potential distinct impact of context on automatic versus cognitively controlled selection and retrieval of motor chunks. Future research should address this latter interesting issue. 4.2. The current state of affairs in context-dependent sequential action There is now abundant empirical support for the general notion of context-dependencies in sequential action. Here we list the various relevant studies, and outline a working model for the effect of contextual changes on sequencing performance. As stated in the Introduction, context refers to those features of the task setting that are not formally relevant for successful execution of the task. When exploring the impact of such features on sequencing performance, the foremost important determinant of context-dependence seems to be whether or not the irrelevant information interferes

312

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

with performance on the basis of dimensional overlap with task-relevant information. In a previous study, we showed that participants can learn to ignore irrelevant but interfering information to optimize performance by developing what we referred to as a context-specific filter (Ruitenberg, De Kleine, et al., 2012). Specifically, participants were required to respond to the location of a target that was presented in one of four fixed placeholders that were horizontally aligned. At each trial, a distracter (which differed from the target in color) was presented within another placeholder, creating stimulus–response interference. Because of a systematic relationship between target and distracter location, participants were able to learn to ignore (i.e., filter out) the distracter. Indeed, when changing the relationship between target and distracter locations (i.e., the context), performance declined. Importantly, this effect of context decreases with practice, as sequencing performance shifts from being stimulus-driven to being representation-driven (Hikosaka et al., 1999; Verwey, 1999) so that performance is increasingly based on internal representations (e.g., motor chunks) in response to just the first stimulus of a sequence. Consequently, subsequent stimuli do no longer need to be processed so that both task-relevant and -irrelevant information can be ignored (Ruitenberg, De Kleine, et al., 2012). In the absence of dimensional overlap between the task-relevant information and task-irrelevant context features, context may still affect performance when the context is contingent with any task-relevant dimension (cf. Jiménez & Méndez, 1999). If this is the case, then sequencing performance has the potential to become context-dependent as a result of associative learning (Anderson et al., 1998; Wright & Shea, 1991). That is, continuous pairing of context features with sequential information during sequence acquisition causes associations between task-relevant and -irrelevant information to develop – just like continuous pairing of two stimuli or of a stimulus and a response can result in associations (see Abrahamse, Jiménez, Verwey, & Clegg, 2010). When performing the sequence in the familiar context, the acquired associations can facilitate memory retrieval processes (Healy et al., 2005; Wright & Shea, 1991). Upon a contextual change, however, the associations can no longer aid such retrieval and performance drops. Notably, this is only observed when the associations are sufficiently strong (such as after extended practice in the present study) and the effect of context thus increases with practice. Furthermore, the present study and the study of Magnuson et al. (2004) suggest that contextual changes have the largest impact on those phases within a sequence that are assumed to include retrieval processes – and not on more motor-based execution phases. Finally, we propose that when the task-irrelevant perceptual features are not correlated with taskrelevant features and thus cannot be associated with a specific sequence or any specific process related to its execution, contextual changes will not affect performance. This notion fits the results of an unpublished study, in which we found no indications of context-dependent discrete sequence learning when the context was static during training (i.e., all practiced sequences were trained within the same context, and then tested in another), and could therefore not be used for the retrieval of motor chunks (Ruitenberg, Abrahamse, & Verwey, unpublished study). 5. Conclusions Overall, the present study replicates earlier findings that memory-based sequencing performance is susceptible to contextual changes (cf. Magnuson et al., 2004; Ruitenberg, Abrahamse, et al., 2012). This suggests that motor chunk representations can include perceptual information, and that perceptual processing takes place when preparing and/or executing sequences from memory, even though perceptual information is not present during actual execution of sequences. The present study demonstrates that practice affects the extent to which contextual changes impact performance. Importantly, it seems that – whereas context-dependent filtering has been shown to become less sensitive to contextual changes with practice (Ruitenberg, De Kleine, et al., 2012) – context-dependent retrieval becomes increasingly sensitive to such changes with practice (present study). Moreover, the current results offer support for the DPM’s prediction that preparation of highly practiced motor chunks, rather than execution of these chunks, would be sensitive to perceptual changes. Finally, in addition to practice and processing phase, it seems that the specific sequence structure further determines the impact of context on sequential action, but it remained unclear from the current study why this is the case.

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

313

Acknowledgements We would like to thank Marita Espey for her assistance in running the experiment. MR was supported by the Netherlands Organisation for Scientific Research (NWO) under contract number 40007-097 and by the Research Foundation – Flanders (FWO) as a Pegasus Marie Curie Fellow under grant number 1262214N. EA was supported by the NWO under contract number 446-10-025 and by the FWO under contract number 12C4712N. References Abrahamse, E. L., Jiménez, L., Verwey, W. B., & Clegg, B. (2010). Representing serial action and perception. Psychonomic Bulletin & Review, 17, 603–623. Abrahamse, E. L., Ruitenberg, M. F. L., De Kleine, E., & Verwey, W. B. (2013). Control of automated behaviour: Insights from the Discrete Sequence Production task. Frontiers in Human Neuroscience, 7, 82. Abrahamse, E. L., & Verwey, W. B. (2008). Context dependent learning in the serial RT task. Psychological Research, 72, 397–404. Anderson, T., Wright, D. L., & Immink, M. A. (1998). Contextual dependencies during perceptual-motor skill performance: Influence of task difficulty. Memory, 6, 207–221. Annett, M. (1970). A classification of hand preference by association analysis. British Journal of Psychology, 61, 303–321. Bo, J., Borza, V., & Seidler, R. D. (2009). Age-related declines in visuospatial working memory correlate with deficits in explicit motor sequence learning. Journal of Neurophysiology, 102, 2744–2754. Bo, J., & Seidler, R. D. (2009). Visuospatial working memory capacity predicts the organization of acquired explicit motor sequences. Journal of Neurophysiology, 101, 3116–3125. Bruijnes, M. (2010). Causes for segmentation of keying sequences: A predictive model. Enschede: University of Twente, Unpublished Bachelor thesis. De Kleine, E., & Van der Lubbe, R. H. J. (2011). Decreased load on general motor preparation and visual-working memory while preparing familiar as compared to unfamiliar movement sequences. Brain and Cognition, 75, 126–134. Godden, D. R., & Baddeley, A. D. (1975). Context-dependent memory in two natural environments: On land and underwater. British Journal of Psychology, 66, 325–331. Healy, A. F., Wohldmann, E. L., Parker, J. T., & Bourne, L. E. Jr., (2005). Skill training, retention, and transfer: The effects of a concurrent secondary task. Memory & Cognition, 33, 1457–1470. Hikosaka, O., Nakahara, H., Rand, M. K., Sakai, K., Lu, X. F., Nakamura, K., et al (1999). Parallel neural networks for learning sequential procedures. Trends in Neurosciences, 22, 464–471. Jiménez, L. (2008). Taking patterns for chunks: Is there any evidence of chunk learning in continuous serial reaction-time tasks? Psychological Research, 72, 387–396. Jiménez, L., & Méndez, C. (1999). Which attention is needed for implicit sequence learning? Journal of Experimental Psychology: Learning, Memory, and Cognition, 25, 236–259. Kennerley, S. W., Sakai, K., & Rushworth, M. F. S. (2004). Organization of action sequences and the role of the pre-SMA. Journal of Neurophysiology, 91, 978–993. Koch, I., & Hoffmann, J. (2000). Patterns, chunks and hierarchies in serial reaction-time tasks. Psychological Research, 63, 22–35. Magnuson, C. E., Wright, D. L., & Verwey, W. B. (2004). Changes in the incidental context impacts search but not loading of the motor buffer. Quarterly Journal of Experimental Psychology, 57A, 935–951. Ruitenberg, M. F. L., Abrahamse, E. L., De Kleine, E., & Verwey, W. B. (2012). Context-dependent motor skill: Perceptual processing in memory-based sequence production. Experimental Brain Research, 222, 31–40. Ruitenberg, M. F. L., Abrahamse, E. L., & Verwey, W. B. (2013). Sequential motor skill in preadolescent children: The development of automaticity. Journal of Experimental Child Psychology, 115, 607–623. Ruitenberg, M. F. L., De Kleine, E., Van der Lubbe, R. H. J., Verwey, W. B., & Abrahamse, E. L. (2012). Context-dependent motor skill and the role of practice. Psychological Research, 76, 812–820. Ruitenberg, M. F. L., Verwey, W. B., Schutter, D. J. L. G., & Abrahamse, E. L. (2014). Cognitive and neural foundations of discrete sequence skill: A TMS study. Neuropsychologia, 56, 229–238. Shea, C. H., & Wright, D. L. (1995). Contextual dependencies: Influence on response latency. Memory, 3, 81–95. Smith, S. M. (1985). Background music and context-dependent memory. The American Journal of Psychology, 98, 591–603. Verwey, W. B. (1996). Buffer loading and chunking in sequential keypressing. Journal of Experimental Psychology: Human Perception and Performance, 22, 544–562. Verwey, W. B. (1999). Evidence for a multi-stage model of practice in a sequential movement task. Journal of Experimental Psychology: Human Perception and Performance, 25, 1693–1708. Verwey, W. B. (2001). Concatenating familiar movement sequences: The versatile cognitive processor. Acta Psychologica, 106, 69–95. Verwey, W. B. (2003a). Effect of sequence length on executing familiar keying sequences: Lasting segmentation and preparation? Journal of Motor Behavior, 35, 343–354. Verwey, W. B. (2003b). Processing modes and parallel processors in producing familiar keying sequences. Psychological Research, 67, 106–122. Verwey, W. B., & Abrahamse, E. L. (2012). Distinct modes of executing movement sequences: Reacting, associating, and chunking. Acta Psychologica, 140, 274–282. Verwey, W. B., Abrahamse, E. L., & De Kleine, E. (2010). Cognitive processing in new and practiced discrete keying sequences. Frontiers in Psychology, 1, 1–13. Verwey, W. B., Abrahamse, E. L., De Kleine, E., & Ruitenberg, M. F. L. (2014). Evidence for graded central processing resources in a sequential movement task. Psychological Research, 78, 70–83.

314

M.F.L. Ruitenberg et al. / Human Movement Science 40 (2015) 298–314

Verwey, W. B., Abrahamse, E. L., & Jiménez, L. (2009). Segmentation of relatively short keying sequences does not transfer to other sequences. Human Movement Science, 28, 348–361. Verwey, W. B., & Eikelboom, T. (2003). Evidence for lasting sequence segmentation in the discrete sequence-production task. Journal of Motor Behavior, 35, 171–181. Wright, D. L., & Shea, C. H. (1991). Contextual dependencies in motor skills. Memory & Cognition, 19, 361–370.