Acta Psychologica 66 (1987) 237-250 North-Holland
237
THE ORGANIZATION OF RAPID MOVEMENT AS A FUNCTION OF SEQUENCE LENGTH * Adela GARCIA-COLERA
SEQUENCES
and Andras SEMJEN
Centre National de la Recherche Scientifique, Marseille, France Accepted July 1987
An experiment was performed to test the predictions made by the subprogram retrieval model (Stemberg et al. 1978) for the production of rapid movement sequences, and to search for the maximum number of elements that can be planned in advance of sequence execution. Subjects performed rapid sequences of 1 to 8 finger taps under both simple and choice RT conditions. Increasing sequence length had no effect on choice RT, but caused simple RT to increase nonlinearly, with the greatest effect between 1 and 3 taps. Intertap intervals did not increase as a function of sequence length The sequences’ timing and force patterns suggested that sequences up to 8 taps were organized as single performance units. The results indicate a fundamental difference between activating a plan for a single tap and a sequence plan in which several elements must be coordinated and timed. Increasing the number of elements beyond 3 does not necessarily add processing steps in the selection and activation of the sequence plan, at least for sequences involving the repetition of a simple element.
Since Lashley (1951) postulated the existence of a central nervous mechanism that governs the production of action sequences, many efforts have been directed at understanding the functioning of such a mechanism. Motor plan or motor program have become the terms most widely used to refer to the central representation of a movement sequence. In the approach originally suggested by Henry and Rogers (1960), motor programs may be assessed by measuring the reaction time (RT) needed to begin sequences of varying complexity. From variations in the time to initiate a movement sequence, as a function of the characteristics of the whole sequence, one can draw an inference about some properties of the process by which a plan for the entire sequence is generated or activated. * Requests for reprints should be sent to A. Garcia-Colera, Aiguier, 13402 Marseille Ckdex 9, France.
CNRS-LNFl,
OOOl-6918/87/$3.50 0 1987, Elsevier Science Publishers B.V. (North-Holland)
31 Chemin Joseph-
238
A. Garcia-Colera, A. Semjen / RT and sequence length
The numerous studies inspired by the RT approach have yielded inconsistent results due to the choice of manipulated factors, to whether choice or simple RT procedures have been employed, and even to the difficulty in appropriately defining movement complexity (for reviews see Kerr (1978) and Marteniuk and MacKenzie (1980)). One of the variables most often used as an index of complexity has been sequence length. RT has been found to increase with the number of connected movement parts (Henry and Rogers 1960; Fischman 1984), syllables (Ericksen et al. 1970; Klapp et al. 1973) or stress groups (Stemberg et al. 1978) to be pronounced, button presses (Rosenbaum and Patashnik 1980; Rosenbaum et al. 1984a), and keystrokes (Stemberg et al. 1978). Sternberg et al. (1978) have proposed a model for the production of rapid movement sequences. According to the model, a program for the entire sequence, composed of a subprogram for each element (words to be recited or letters to be typed in their experiments), is loaded into a buffer in advance of execution. Prior to the execution of each element, the corresponding subprogram must be found in the nonshrinking buffer through a self-terminating sequential search. As the number of elements in the sequence increases, the number of subprograms in the buffer increases accordingly so that a longer search time is needed to find the appropriate subprogram. Therefore, increasing sequence length should result not only in a linear increase in RT, but also in a similar increase in the intervals between the successive elements. This model has received a great deal of attention and has been successfully tested in various experimental situations. However, its general applicability may be questioned on the basis of results such as (a) the dependence of the RT increase upon the type of response material (Knapp et al. 1979) and upon practice (Hulstijn and Van Galen 1983), (b) the attenuation or even the disappearance of the sequence length effect when length is varied within (choice RT) rather than between (simple RT) trial blocks (Klapp et al. 1979; Rosenbaum et al. 1984b), and (c) the lack of an effect of sequence length on interresponse times (Hulstijn and Van Galen 1983; Stelmach et al. 1984). One of the aims of the present study was to test the effects predicted by the model of Stemberg et al. (1978) in a repetitive finger-tapping task. Another purpose of the experiment was to search for the maximum number of elements that can be taken into account in the advance planning of a rapid sequence of movements. Accordingly,
A. Garcia-Colera, A. Semjen
/ RT and sequence length
239
sequence length was varied over a wider range than that used in most of the previous studies. Subjects produced sequences of 1 to 8 finger taps under both simple and choice RT conditions. The rate of sequence execution was also manipulated. In the Speed condition, the sequences had to be executed as fast as possible, like in most of the previously cited experiments. In the Cadence condition, the sequence elements had to be paced according to a prespecified rate. According to the model of Sternberg et al. (1978), both RT and the intertap intervals should increase as a function of the number of taps to be produced, up to a ceiling which should correspond to the maximum number of elements able to be comprised in the sequence plan. The upper limit of planning was also expected to be reflected in the timing of the individual taps, with more fluent timing of the taps that were planned in advance, followed by a slowing down of execution rate at the point at which control was switched from the previously established plan to the additional programming required for the remaining elements. In order to gain a more complete picture of the subjects’ motor performance, the force of the taps was also recorded. In previous work we found that the initial and terminal taps of 5-element finger-tapping sequences were stronger than the intermediate taps. Thus, the sequence boundaries were marked by a small, spontaneous stress (Semjen and Garcia-Colera 1986). In the present experiment, the produced force patterns were examined as a further potential source of information on whether long sequences were segmented into smaller subunits.
Method Task The subject sat at a table, facing a circular key (2 cm in diameter) on which the tapping sequences were performed, and a digital display unit on which the warning and reaction signals were presented. The subject’s forearm rested on the table with the tapping finger held just above the key, so that both wrist and finger movements contributed to the realization of the sequences. The finger used could be the index or middle of the preferred hand, but it had to remain the same throughout the experiment. The subjects were instructed to make very brief, ‘staccato’ taps, by just hitting the key and releasing it immediately afterwards. In the Cadence condition, each series of trials was preceded by the presentation to the subject, via headphones, of a string of clicks separated by a constant 160-msec
240
A. Garcia-Colera, A. Semjen / RT and sequence length
interval. This model indicated the rate at which the taps had to be executed in a sequence. In the Speed condition, the subjects were asked to execute the sequences as fast as possible. Nevertheless, the 160-msec interval model was also presented before each trial series as a reference speed to remind the subjects that they should attempt to tap faster than the model. Every trial started with a warning signal, which consisted of three horizontal lines, displayed for 500 msec. It was followed by a variable preparatory period of one of three equally likely values: 900, 1500, or 2100 msec. At the end of that period was presented a digit, used as reaction signal (RS), which indicated the number of taps to be executed in the sequence. The sequence had to be initiated as quickly as possible after presentation of the digit. During the intertrial interval, which lasted 4 set, the RT of the sequence just performed was displayed on a television screen in front of the subject. Measurements
Any contact between the finger and the surface of the key triggered an electronic circuit whose output served to identify the beginning and the end of each tap. The force of each tap was measured as the peak output voltage of a strain gauge that was incorporated into the key. All experimental events were automatically controlled by a PDP-12 computer. For each response sequence we recorded the force of each tap, the time during which the finger was in contact with the key (tap duration), the interval between the end of one tap and the beginning of the next (lift interval), and the RT. RT was measured as the delay between onset of the digit used as RS and onset of the first tap. A test was automatically performed on the measurements to determine whether each sequence had been executed correctly. If the number of taps was other than the number indicated by the RS, the trial was classified as an error. All the erroneous trials in a block were repeated once at the end of the block. Subjects
and design
The subjects were 8 paid volunteers. All subjects participated in three experimental sessions. Session 1 was devoted to training for all subjects. The Cadence condition was presented first. Fight series of trials were run following the simple RT (SRT) procedure. In the first, the digit 1 was presented in 9 successive trials. The following digits, in ascending order, were presented in the same manner. Afterwards, a choice RT (CRT) block was run. It comprised 72 trials in which each digit from 1 to 8 was presented 9 times in a pseudo-random order. After a rest period, the same SRT and CRT blocks were run in the Speed condition. Sessions 2 and 3 were each devoted to either the Cadence or the Speed condition. The subjects were divided into two equal groups which performed the two conditions in opposite order. The series of trials run were the same as in session 1, but there were two CRT blocks of 72 trials each. Half of the subjects in each group started with the two CRT blocks, followed by the eight SRT series, and the other half did the same in the
A. Garcia-Colera,A. Semjen / RT and sequence length
241
reverse order. The order of presentation of the eight SRT trial series was varied, according to a latin square design, for each of the subjects. Data analyses were conducted only on the data from sessions 2 and 3. The first trial from each SRT series and the first eight from each CRT block were considered as warm-up trials and, therefore, were not included in the computation of the individual means.
Reaction time
The group means for SRT and CRT in the Cadence and Speed conditions are shown in fig. 1 as a function of the number of taps in the sequence. The CRT means are presented separately for each of the two CRT blocks. The individual mean CRTs were subjected to a 2 x 2 X 8 analysis of variance (ANOVA) with the factors Timing Condition (Cadence or Speed), Block (First and Second), and Number of Taps. The effect of Block was significant, F&7) = 23.34, p < 0.01, indicating that CRT was significantly shorter in the second than in the first block. No other effect was found significant at the 0.05 probability level or better. The RT values observed in the SRT condition and those from the second block in the CRT condition were subjected to a 2 x 2 x 8 ANOVA which tested the effects of Timing Condition, Block (SRT vs. CRT) and Number of Taps. The only significant main effect was that of Block, I;(1,7) = 58.01,
REACTION
TIME
CADENCE
SPEED
,p.~ ...O--Q ..,,
o..o..o...o
250
p”‘0”
9 .‘b
.:’
SRT
O..d 6
i 200
_
12345676
12345670
NUMBER
Fig. 1. Mean RT in the each timing condition.
OF TAPS
SRT and the two CRT blocks, as a function of the number of taps, for
242
A. Garcia-Colera, A. Semjen / RT and sequence length
p < 0.001, indicating that the CRT was significantly longer than the SRT. Significant interactions were found between Timing Condition and Block, F(1,7) = 16.03, p < 0.01, and between Number of Taps and Block, F(7,49) = 5.16, p < 0.001. The latter interaction denotes the fact that Number of Taps reached the significance level only in the SRT condition, F(7,49) = 4.21, p < 0.01 Tests of linearity were performed on the SRT data for each timing condition separately. The linear component was nonsignificant in the Speed condition, but it did reach significance in the Cadence condition, F(1,7) = 5.96, p < 0.05. A closer inspection of the data suggested that the significant effect of sequence length in the SRT condition was mainly due to the difference between one-tap and two-tap sequences in the Speed condition, and between two- and three-tap sequences in the Cadence condition. In the Cadence condition, SRT increased 17 msec from the two- to the three-tap sequences, but only 11 msec from the three- to the eight-tap sequences. This 11 msec increase failed to reach significance, as shown by an additional test of linearity (F&7) = 1.61). In the Speed condition, SRT increased 29 msec between the one-tap and two-tap sequences, but only 14 msec from the two- to the seven-tap sequences. Again, the linear trend in this 14 msec increase proved to be nonsignificant (F(1,7) = 1.93). Sequence timing The mean duration of the interval between the onset of a tap and the onset of a subsequent tap in the sequence was calculated for each individual subject. Thus, these interval measurements include the time during which the finger was in contact with the key to produce a given tap (tap duration), plus the interval during which the finger was raised between the end of a tap and the onset of the following one (lift interval). These interval data for sequences from 2 to 8 taps were analyzed with seven separate ANOVAs, one for each sequence length. The differences between the intervals produced in the SRT and CRT blocks proved to be nonsignificant in all of the analyses. On the contrary, the effect of Timing Condition was significant in all of them. The difference between the mean interval durations produced in the Cadence and Speed conditions can be seen in fig. 2. Here the interval values are averaged over the SRT and CRT blocks, and presented as a function of the number of taps in the sequence. The mean interval in the Cadence condition barely increases as the number of taps in the sequence increases, but a more pronounced augmentation is observed in the Speed condition. This observation could seem to go in the direction predicted by the Stemberg et al. model. However, a closer look at the origin of this effect affords a more complete picture of the timing structure of these sequences that is not compatible with such a model. In table 1 are presented the mean interval durations as a function of the number of taps in the sequence, and of the serial position of the intervals in each sequence. For all sequences longer than three taps, the first interval was always longer than the second and the last interval was always the longest. In the Cadence condition, the duration of the intermediate intervals was very similar, with a trend for the interval next to the last to be slightly longer than the intermediate ones. A different pattern appeared in the Speed condition in which the intervals displayed a progressive lengthening from the
A. Garcia-Colera, A. Semjen / RT and sequence length
O-O
243
.0-.&o
.0-o
CADENCE
k
5 0
160 J
A
a
SPEED
I 234
567
NUMBER
6
OF TAPS
Fig. 2. Mean duration of the interval between taps’ onsets in sequences of 2 to 8 taps, for timing condition.
Table 1 Mean interval duration (msec) in each timing condition as a function of the number of taps and the serial position of the interval in the sequence. Number of taps
Serial position 1
2
3
4
5
6
I
Cadence 2 3 4 5 6 I 8
182 181 185 186 185 186 186
183 180 178 182 181 181
189 181 184 183 183
190 183 182 183
196 185 183
194 187
196
Speed 2 3 4 5 6 7 8
145 152 154 156 155 156 157
144 150 148 151 149 150
154 153 154 153 154
156 158 155 155
161 157 156
162 157
162
A. Garcia-Colera, A. Semjen / RT and sequence length
244
Table 2 Mean tap duration (msec) in each timing condition as a function of the number of taps and the serial position of the taps in the sequence. Number of taps Cadence 1 2 3 4 5 6 I 8 Speed 1 2 3 4 5 6 I 8
Serial position 1
2
3
4
5
6
7
8
14 65 66 63 65 61 62 61
61 60 54 53 54 53 51
69 55 55 54 54 53
63 55 53 53 51
60 55 54 52
59 54 52
58 51
55
65 54 50 49 49 50 48
65 51 50 49 48 48
58 48 49 48 41
51 48 48 41
55 41 48
52 48
54
IO
56 56 54 51 54 56 55
to the last. It must be noted that, whereas the difference between the second and the last interval increased with sequence length, the variations in the duration of the intervals in the same serial position were very small and did not show a systematic increase with sequence length. These observations indicate that the very slight increase in mean interval duration, seen in fig. 2 for the Cadence condition, is mainly due to variations in the lengthening of the last and of the next-to-last intervals. The sharper increase in mean interval duration with sequence length, observed for the Speed condition, simply reflects the progressive deceleration over the successive intervals within a sequence. Hence, in neither condition did we observe an overall lengthening of each intertap interval as a function of increasing sequence length, as should have occurred according to the Stemberg et al. model. To further look for possible effects of sequence length on the timing of the successive elements, tap duration and lift intervals were also analyzed separately. The mean tap duration values are presented in table 2 as a function of the number of taps and of the taps’ serial position in each sequence. Overall, tap duration was a little longer in the Cadence condition than in the Speed condition. In both conditions and for all sequence lengths, the first and last taps were longer than the intermediate ones. The duration of the intermediate taps was remarkably stable across sequence lengths and serial positions. Given this pattern of tap durations, the lift intervals differed from second
245
A. Garcia-Colera, A. Semjen / RT and sequence length
the whole intervals as presented above in that the first lift interval was not longer than the intermediate ones. The lift intervals displayed, therefore, a more continuous decelerating pattern from the first to the last tap, but, again, those sharing the same serial position were not systematically lengthened across sequences of different length. Force and error patterns The mean force of the taps is presented in table 3 as a function of the number of taps and of each tap’s serial position in a sequence. Overall, the taps were stronger in the Speed condition than in the Cadence condition. In both conditions, the strongest tap was the last one and the next strongest was the first one. In the Cadence condition, force decreased from the first to the second tap, showed a slight increasing trend over the successive taps, and a more abrupt increase from the next-to-last to the last tap. In the Speed condition, the force level of the fist tap was always higher than the level of the first tap in the Cadence condition. It decreased gradually from the first to the third tap, where it approached the level of the third tap in the Cadence condition, then it showed a small but progressive increase over the successive taps, and a much larger increase on the last tap. The jump in force level from the next-to-last to the last tap was much greater in the Speed than in the Cadence condition.
Table 3 Mean force (arbitrary units) in each timing condition as a function of the number of taps and the serial position of the tap in the sequence. Number of taps
Serial position 1
2
3
4
5
6
1
8
141 128 126 124 129 122 128 121
142 119 115 113 112 109 107
143 123 120 111 115 114
150 127 111 115 110
154 117 119 112
140 120 110
142 114
136
162 149 139 141 145 143 147 138
162 137 130 127 130 132 125
149 125 117 121 119 108
173 127 124 125 117
178 130 129 120
182 139 125
172 126
176
Cadence 1 2 3 4 5 6 7 8
Speed 1 2 3 4 5 6 7 8
246
A. Garcia-Colera, A. Semjen
/ RT and sequence length
Table 4 Number of errors in each timing condition and trial block as a function of number of taps in the sequence. Number of taps
Total
1
2
3
4
5
6
I
8
Cadence TRS TRCl TRC2
5 1 2
3 2 1
10 5 8
13 5 4
5 3 10
5 10 9
3 11 6
4 12 15
48 49 55
Speed TRS TRCl TRC2
2 2 8
6 3 7
9 I 8
10 1 10
9 5 10
16 12 11
10 13 18
12 8 14
14 51 86
The subjects performed sequences with fewer or more taps than required in 9.9% of the total number of trials. Over 80% of the errors were due to the execution of one tap more or one tap less than the number required. The number of errors as a function of Timing Condition, Trial Blocks and Number of taps, is presented in table 4. More errors were committed in the Speed than in the Cadence condition. Most importantly, the errors in the SRT condition did not increase systematically as the required number of taps increased from 3 to 7 or 8. Therefore, the lack of a significant linear trend in the SRT increase over this range of taps cannot be accounted for by a trade-off between speed of sequence initiation and precision of sequence execution.
Discussion One of the aims of this experiment was to test whether the effects of sequence length on RT and intertap intervals were those predicted by the model of Sternberg et al. (1978). Differences were observed in the timing of the successive taps as a function of the taps’ serial position. However, for any interval in a given serial position no systematic increase was observed as a function of sequence length. This observation agrees with the previously cited results of Hulstijn and Van Galen (1983) and Stelmach et al. (1984). Rosenbaum et al. (1984a) also failed to observe an effect of sequence length upon the timing of the responses beyond the first one. The RT increased as a function of sequence length in the SRT condition. However, the greatest increase occurred between 1 and 3 taps. Beyond 3 taps, the increase was very small and did not show a
A. Garcia-Colera, A. Semjen / RT and sequence length
241
significant linear trend. Although this result is contrary to the model proposed by Sternberg et al. (1978), it is very close to their own one-hand typing results which showed a large increase in RT when the letters to be typed increased from 1 to 2, and a nonlinear and very small further RT increase when the number of keystrokes augmented from 2 to 5. The present results also resemble those obtained by Inhoff et al. (1984: exp. 1) and Inhoff (1986: exp. l), who found a nonlinear increase in RT when the number of elements increased from 1 to 3. It must be noted that in several other studies that have reported an RT lengthening as a function of the number of elements, the number tested did not exceed 2 or 3 elements (Rosenbaum and Patashnik 1980; Rosenbaum et al. 1984a,b). This seems therefore to be the range within which the RT variations as a function of the number of elements are strongest and can be most reliably reproduced. The second aim of this experiment was to search for the maximum number of elements that could be comprised in the advance planning of sequences of repeated finger taps. At first glance, the breaking-point observed in the SRT function when the number of taps equaled 2 or 3 could be considered as indicative of the upper limit of planning that we had looked for. However, other studies have shown that the advance planning of rapid movement sequences may extend up to 5 or 6 elements. In experiments in which subjects had to choose between movement sequences that differed in the serial ordering of their constituents, RT was influenced by the characteristics of the last two responses in a sequence of 5 or 6 button presses or taps (Inhoff et al. 1984: exp. 3; Semjen and Garcia-Colera 1986: exp. 4). In rapid speech production, the capacity of the ‘program buffer’ has been estimated at 7 or 8 items, a value similar to the memory span (Monsell 1984). It would be surprising if the storage capacity for simple repeated finger taps was limited to only 2 or 3 items. Instead, the sharp SRT increase between one and two or three taps may be interpreted as an index of the basic difference between preparing for a single movement or having to activate a sequence plan in which a beginning, an intermediate, and a terminating element are to be coordinated and timed. The lack of a further significant increase in RT for longer sequences would indicate that adding intermediate elements to such a simple sequence does not require additional processing operations before sequence initiation. In this experiment, the difference between CRT and SRT was relatively small for an 8-choice task, which may reflect the relative
248
A. Garcia-Colera, A. Semjen / RT and sequence length
simplicity of the plan selection process in the case of sequences of repeated elements. The most striking aspect of the CRT data was, however, the complete absence of any sequence length effect. Other investigators have also reported a decrease or a total absence of sequence length effects under the CRT paradigm, as compared to the SRT paradigm (Klapp et al. 1979; Rosenbaum et al. 1984b; Inhoff 1986). The fact that in this study the CRT of a single tap was as long as that of sequences of any length suggests that, in the context of preparation for sequence production, the selection of a plan for a single tap involves the same operations as the selection of any other plan. Furthermore, the lack of an effect of sequence length over the whole range of lengths tested suggests that the representation of the sequence size was accessed in a single step rather than structured in a serial fashion. Otherwise, longer sequences should have had longer CRT. The timing and force data collected in this experiment lend support to the hypothesis that the sequences were planned in advance, rather than organized step-by-step in the course of their execution. A possible division of the longer sequences into subunits was expected to appear as a disruption in the timing of the taps at the transition from one subunit to the next. Instead, a very stable timing pattern was observed across sequences of different lengths. This pattern was characterized by a lengthening of the first and last intervals with respect to the intermediate ones, and a trend for progressive deceleration from the second to the last interval, particularly in the Speed condition. Such deceleration can be interpreted as a preparation for the stop, which probably poses a greater difficulty in this faster condition than in the Cadence condition. This is a further indication that the sequence end was anticipated at least several steps before its actual execution. The lengthening of the intervals at the beginning and the end of movement sequences has been systematically observed by us in sequences of 4 to 5 taps (Semjen et al. 1984; Semjen and Garcia-Colera 1986), as well as by other authors in different experimental situations (Povel 1977; Shaffer et al. 1985). It has been interpreted as a characteristic of ‘coherent sequences’ that constitute a perceptual-motor unit. In the present experiment, the lengthening of the intervals at the sequence boundaries was also accompanied by a longer duration of the first and last taps, and a stronger force on these same taps. The fact that none of these ‘boundary markers’ was observed for any of the intermediate taps, even in sequences of up to eight taps, gives further support to the
A. Garcia-Colera, A. Semjen
/ RT and sequence length
249
conclusion that the tapping sequences were organized as single units of performance. This conclusion, together with the failure to observe the main effects predicted by the subprogram retrieval model (Stemberg et al. 1978), lead us to question the general validity of this model. Even from a logical standpoint, it is not clear why a buffer should be searched to find the subprogram corresponding to the element to be performed next, when such element is the same throughout the sequence. References Ericksen, C.W., M.D. Pollack and W.E. Montague, 1970. Implicit speech: mechanism in perceptual encoding? Journal of Experimental Psychology 84, 502-507. Fischman, M.G., 1984. Programming time as a function of number of movement parts and changes in movement direction. Journal of Motor Behavior 16, 405-423. Henry, F.M. and D.E. Rogers, 1960. Increased response latency for complicated movements and a ‘memory drum’ theory of neuromotor reaction. Research Quarterly 31, 448-458. Hulstijn, W. and G.P. van Galen, 1983. Programming in handwriting: reaction time and movement time as a function of sequence length. Acta Psychologica 54, 23-49. Inhoff, A.W., 1986. Preparing sequences of saccades under choice reaction condition: effects of sequence length and context. Acta Psychologica 61, 211-228. Inhoff, A.W., D.A. Rosenbaum, A.M. Gordon and J.A. Campbell, 1984. Stimulus-response compatibility and motor programmin g of manual response sequences. Journal of Experimental Psychology: Human Perception and Performance 10, 724-733. Kerr, B., 1978. ‘Task factors that influence selection and preparation for voluntary movements’. In: G.E. Stelmach (ed.), Information processing in motor control and learning. New York: Academic Press. Klapp, S., J. Abbott, K. Coffman, D. Greim, R. Snider and F. Young, 1979. Simple and choice reaction time methods in the study of motor programming. Journal of Motor Behavior 11, 91-101. Klapp, S.T., W.C. Anderson and R.W. Berrian, 1973. Implicit speech in reading, reconsidered. Journal of Experimental Psychology 100, 368-374. Lashley, KS., 1951. ‘The problem of serial order in behavior’. In: L.A. Jeffres (ed.), Cerebral mechanisms in behavior. New York: Wiley. Marteniuk, R.G. and C.L. MacKenzie, 1980. ‘Information processing in movement organization and execution’. In: R.S. Nickerson (ed.), Attention and performance VIII. Hillsdale, NJ: Erlbaum. Monsell, S., 1984. ‘Components of working memory underlying verbal skills: a “distributed capacity” view’. In: H. Bouma and Don G. Bouwhuis (eds.), Attention and performance X. Control of language processes. Hillsdale, NJ: Erlbaum. Povel, D.J., 1977. Temporal structure of performed music. Some preliminary observations. Acta Psychologica 52, 107-123. Rosenbaum, D.A. and 0. Patashnik, 1980. ‘Time to time in the human motor system’. In: R.S. Nickerson (ed.), Attention and performance VIII. Hillsdale, NJ: Erlbaum. Rosenbaum, D.A., A.W. Inhoff and A.W. Gordon, 1984a. Choosing between movement sequences: a hierarchical editor model. Journal of Experimental Psychology: General 113, 372-393.
250
A. Garcia-Colera, A. Semjen / RT and sequence length
Rosenbaum, D.A., E. Sal&man and A. Kingman, 1984b. ‘Choosing between movement sequences’. In: S. Komblum and J. Requin (eds.), Preparatory states and processes. Hillsdale, NJ: Erlbaum. Semjen, A. and A. Garcia-Colera, 1986. Planning and timing of finger tapping sequences with a stressed element. Journal of Motor Behavior 18, 287-322. Semjen, A., A. Garcia-Colera and J. Requin, 1984. On controlling force and time in rhythmic movement sequences: the effect of stress location. Annals of the New York Academy of Sciences 423, 168-182. Shaffer, L.H., E.F. Clarke and N.P. Todd, 1985. Metre and rhythm in piano playing. Cognition 20, 61-77. Stelmach, G.E., P.A. Mullins and H.L. Teulings, 1984. Motor programming and temporal patterns in handwriting. Annals of the New York Academy of Sciences 423, 144-157. Stemberg, S., S. Monsell, R.L. Knoll and C.E. Wright, 1978. ‘The latency and duration of rapid movement sequences: comparisons of speech and typewriting’. In: G.E. Stelmach (ed.), Information processing in motor control and learning. New York: Academic Press.
