This article was downloaded by: [University of Sussex Library]On: 09 September 2013, At: 09:42Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Journal of Motor BehaviorPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/vjmb20

On the Modularity of Sequence RepresentationSteven W. Keele a , Peggy Jennings a , Steven Jones a , David Caulton a & Asher Cohen aa Department of Psychology, University of OregonPublished online: 14 Jul 2010.

To cite this article: Steven W. Keele , Peggy Jennings , Steven Jones , David Caulton & Asher Cohen (1995) On the Modularityof Sequence Representation, Journal of Motor Behavior, 27:1, 17-30, DOI: 10.1080/00222895.1995.9941696

To link to this article: http://dx.doi.org/10.1080/00222895.1995.9941696

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in thepublications on our platform. However, Taylor & Francis, our agents, and our licensors make no representationsor warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Anyopinions and views expressed in this publication are the opinions and views of the authors, and are not theviews of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should beindependently verified with primary sources of information. Taylor and Francis shall not be liable for any losses,actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoevercaused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Journal of Motor Behavior, 1995, Vol. 27, No. 1, 17-30

On the Modularity of Sequence Representation Steven W. Keele Peggy Jennings Steven Jones David Caulton Asher Cohen Department of Psychology University of Oregon

ABSTRACT. A modular theory of motor control posits that the representation of an action sequence is independent of the effector (motor) system that implements the sequence. Three experiments tested this theory. Each used a variant of a method developed by Nissen and Bullemer (1987) in which subjects re- sponded to visual signals occupying different spatial positions by pressing a key corresponding to each signal position. Se- quence learning is indicated when reaction times to signals that follow a sequence become faster with practice than reaction times to random signals. The first experiment showed transfer of sequential learning of key pressing from the fingers to the arms, or vice versa. Similar transfer was found when a distrac- tion task was added that likely blocked an attentional form of learning (cf. Curran & Keele, 1993). In a third experiment, much but not all of the sequential learning transferred from a situation in which the response was a key press to one with a vocal response, suggesting that at least part of the sequential learning is embedded in a system that describes successive loca- tions of signals in space. These studies suggest that sequential representation resides in a module prior to the selection of effector systems to execute the movement.

Key words: effector independence, key pressing, modularity, motor program, sequence representation, transfer of learning

hen a sequence of activity is acquired, be it hand- W writing, typing, piano playing, dancing, or the like, the sequence typically is learned with a particular effector system. The hand and fingers are used in writing; the fingers, arms, and feet are used in piano playing; and so on. In the studies reported here, we explored the rela- tionship between the representation for sequences such as these and the motor system that executes them. We asked, Is the motor system the repository of sequence information or is the sequence representation indepen- dent of the effector system of initial learning? Consider handwriting. This skill may be learned predominantly with fingers and hand. Does learning occur anew when

the arm is used to write in large size, or does the sequen- tial information that drives writing transfer to different effectors?

Different theorists have proposed different answers to this question. One version of Jordan’s (in press-a) con- nectionist model of sequencing embeds sequential repre- sentation within the effector system that dictates move- ment. Such embedding, in his model, provides an account of coarticulatory phenomena in which nearby elements in a movement sequence influence each other’s production. In contrast, Berkinblit and Feldman (1988) separated sequence representation and effector specifi- cation: “We consider one possible scheme. There is a neuronal level that creates an abstract image (verbal or graphic) of the forthcoming movement (a circle, line, etc.) Then a combination of effectors and a coordinative structure is specified” (pp. 372-373).

According to the latter view, which we call modularity, the sequence representation encodes where successive re- sponses are to be directed, an effector-free spatial de- scription. The effector system itself simply selects a mus- cle combination that will get a desired effector to a location designated by the sequence representation. By this view, one system or module is responsible for com- puting successive locations. A second module is respon- sible for computing articulatory activities to achieve the locations.

The notion that different modules compute different functions is not a new one (Fodor, 1983). In the present study, however, we were concerned with an additional is-

Correspondence address: Steven W Keele, Department of Psychology, University of Oregon, Eugene, OR 97403. E-mail: SKEEL E@oregon. uoregon. edu

17

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

S. W. Keele, P. Jennings, S. Jones, D. Caulton, & A . Cohen

sue. The underlying question here was whether a module that computes a function of serial succession not only is independent of a module involved in response execution but can in fact be interfaced with different effectors. The more general question was whether in humans a particu- lar computation can be accessed by or output to a variety of different systems, an idea earlier articulated by Rozin ( 1976).

The issue of separability of sequence representation from the effector system has received little empirical at- tention. One approach to the problem examines similar- ity of output produced by different motor effectors. I t frequently has been noted that different writing samples from the same person are very similar even when pro- duced by different motor effectors-fingers and hand, arm, and even head, and mouth (Bernstein. 1947; Rai- bert, 1977). The interpretation of such similarity is prob- lematic because, although there is a surface similarity of writing style across effectors, some effectors are markedly less fluent than others, producing dissimilarities as well as similarities (Wright, 1990). How should dissimilarities be explained? Are they the result of different sequential representations for different effectors?

Another approach to the problem of modularity, again with respect to writing, involves the analysis of neurolog- ical patients. Margolin (1984) concluded from a review of such patients that an orthographic buffer specifies the serial order of graphemes. The same orthographic buffer can be used to drive either typing, writing, or oral spell- ing. Thus, sequence specification is independent of the effector system of output. Striking confirmation of the Margolin view was provided by Hillis and Caramazza (1988). They described two patients who made numerous spelling errors of unusual sorts, including sequential or- dering. Often letters would be reversed, such as rpiest for priest or chucrh for church. The same proportion of spelling errors were made regardless of whether the out- put was written or oral. Moreover, the types of spelling errors-letter reversals, letter omissions, and letter sub- stitutions-were similar in the two cases and of about equal proportion. These results all suggest that the serial order specification of letters is independent of the effector system involved in their execution.

The experiments here present another, more quantita- tive approach to the issue of sequence modularity. The approach we used enabled us to examine transfer of learning of sequential knowledge from one motor system to another. We adapted a paradigm extensively explored by Nissen and her colleagues (e.g., Nissen & Bullemer, 1987; Nissen, Knopman, & Schacter, 1987). Visual sig- nals appeared at various locations on a screen. De- pending on the experiment, subjects responded to the signals by pressing a corresponding key on a response board or by responding verbally. On some blocks of trials the signals appeared in repeating sequences of locations. The subjects were never informed that the signals would occur in a repeated sequence, but, nonetheless, their re-

action times improved with practice. After several learn- ing blocks, transfer occurred. At the point of transfer, some subjects changed the manner of response, from finger movement to arm movement or the reverse, or from manual to verbal responding. The question asked was whether knowledge of sequence transfers to the changed effector.

The assessment of transfer of knowledge involved a comparison of performance when signals occurred in the same order as in pretransfer training with performance when signals occurred at random. Upon beginning the transfer phase, subjects received two blocks of trials in which signals occurred at random. Such practice with random events prevented any new learning of sequence while providing familiarization with mode of responding for those who changed effector. Immediately following the two random blocks, there was an additional block of tri- als in which the previously experienced sequence was re- introduced. That block was followed by one more ran- dom block. The extent during transfer to which reaction time to signals in sequence is faster than reaction time to signals at random is an index of the degree to which the sequence is learned. The issue examined was whether the advantage of the repeated sequence over random events is equivalent for subjects who change effectors and those who maintain the same effectors. If such is the case, it would suggest that all sequence knowledge gained with one effector transfers to another.

I t is possible that during the transfer phase, the initial two blocks with random events might result in loss of sequence knowledge. Any test-phase advantage of the se- quence block that is sandwiched between random blocks might simply reflect new learning in that block alone rather than transfer of previous knowledge. To preclude this possibility in the studies reported here, we used a control group that experienced only randomly occurring stimuli in the pretransfer ‘‘learning’’ phase. Their first ex- perience with stimuli in sequence occurred in the critical sequence block during the transfer phase. The perfor- mance of that group on the sequence block, compared with the surrounding random blocks, allowed a test of whether any sequence knowledge was a result of new learning.

Three studies are reported. In the first, we examined transfer of sequential knowledge from fingers to arm, or vice versa, under conditions in which learning occurred without distraction. In the second study, a secondary task was performed concurrently with the primary reac- tion time task (see also Nissen & Bullemer, 1987). Curran and Keele ( 1993) have argued for two forms of sequential memory, one heavily dependent on attention (i.e., free- dom from distraction) and the other not. One might speculate that attentionally based knowledge is indepen- dent of response mode, whereas attention-free knowl- edge is bound to the response system of acquisition. The contrast between Experiments 1 and 2 allowed us to ana- lyze this issue.

18 Journal of Motor Behavior

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

Sequence Modularity

The third study extended the second. It too involved learning under distraction, but responding was trans- ferred from manual to verbal. When responding is trans- ferred between fingers and arm, not only do signals occur in the same order, but responses are directed to the same response keys. Thus, response-key order is maintained during transfer. Only the effector changes. When re- sponding is transferred from manual to verbal responses, however, response-key order is not maintained, because response keys are no longer used. In this case, signals occur in the same order as prior to transfer but both mo- tor effector and the nature of the response changes. If transfer occurs between manual and verbal responding, then some of the knowledge must reside in a stimulus- based representation.

These studies have conceptual similarity to earlier ones by MacKay (1982) and by Stadler (1989), who also used transfer paradigms and whose findings suggested that sequential knowledge is independent of the mode of response. MacKay (1982) showed that when German- English bilinguals practiced a novel sentence, improve- ments in production time transferred in their entirety to a sentence that expressed the same content but in a different language. This suggests that the learned sequen- tial representation is more abstract than the production system, which of course differs between German and En- glish. In a procedure similar to one developed by Lew- icki, Hill, and Bizot (1988), Stadler’s subjects searched for letters among distractors, striking one of four keys that corresponded to the display quadrant in which a tar- get was found. The seventh key press in a series could be predicted from preceding key presses, and Stadler found that with extensive practice, reaction times to the predict- able position were facilitated even though subjects ex- pressed no awareness of the sequence. Moreover, the learning transferred to a different keyboard and from the use of two fingers on one hand to four fingers, two on each hand. Such transfer suggests the representation was based on stimulus position rather than on response type.

Our paradigms explored similar issues. We manipu- lated distraction across experiments to determine whether learning with full versus impaired attention affects the nature of the representation. We changed more radically the nature of the motor response in one experiment of transfer. The paradigm we used also con- stitutes a model task for the acquisition of motor se- quences, such as keyboard tasks.

EXPERIMENT 1

Method

Subjects Thirty-six college-aged subjects from a psychology de-

partment pool were paid $5 each for participation in the experiment. Twelve participants were assigned to each of three groups.

Tasks and Design

Subjects responded to X marks that appeared at one of three horizontal locations on a computer screen by pressing the corresponding button on a keyboard placed below the screen. Response onset terminated the current signal, and the next signal appeared 200 ms later. Succes- sive signals appeared in one of two arrangements, ran- dom or sequential. In the former case, the successive sig- nals appeared at random, with the restriction that the same signal position could not occur in immediate suc- cession. In the sequential case, the signals appeared cycli- cally in a fixed sequence of five elements. There were six possible fixed sequences; two individuals from each of two sequential conditions were assigned to each se- quence. If the signal positions are designated 1,2, and 3 from left to right, the sequences were 12323, 13232, 21313, 23131, 31212, and 32121. Each of the sequences can be characterized as having one unique signal position and two repeated positions. The purpose of different se- quences was to average out any peculiarities caused by particular sequences.

All participants began the experiment with two blocks of 100 signals each in which the signals appeared at ran- dom. For one group, the random control, the next eight blocks continued to involve only randomly occurring sig- nals, again with 100 signals per block. For two other groups of subjects, Blocks 3-10 involved 100 signals that occurred in fixed sequential order. Because the sequences all had five elements, a sequence cycled 20 times per block. The first signal in a block started at a random point within the sequence, and there was no mark or pause at the end of cycle and the return to its beginning. Subjects were told nothing about the sequence and were only instructed to respond rapidly and accurately in pressing the key that corresponded to the current signal. This phase of the experiment was called the training phase.

Following training, subjects transferred to a test phase of four blocks of 100 trials each. The first two blocks involved random signals. The third block involved signals in sequence. For those subjects who had had fixed se- quences during training, the sequence remained the same during the test. For the group of subjects who previously had had only random events, the sequence they experi- enced was, of course, new. In no case were subjects in- structed of the presence of a sequence. Following the test block with sequence, there was one more block at ran- dom. The first random block during the test phase was considered familiarization, and the primary interest was in the RT difference between the single-sequence block of the test phase and the average of the random blocks on each side.

Subjects responded to the signals in one of two re- sponse modes, fingers or arm. In the former, subjects rested three fingers of their dominant hand on the keys and responded with finger presses. In the latter, subjects

March 1995, Vol. 27, No. 1 19

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

S. W. Keele, P. Jennings, S. Jones, D. Caulton, & A. &hen

moved their dominant arm back and forth and struck the keys with their index finger. Note that in this case it was the arm and not the index finger that did the moving. Because we wish to emphasize that different muscles are used in the two cases, we call the one response modefin- gers and the other arm. In the initial training, half the subjects in each group used their fingers throughout and the other half used their arm. During the transfer phase, half the subjects who had received sequence training con- tinued with the same effector used during training, whereas the other half switched to the other effector. These groups are called same effector and diflerent effector. respectively. All the random control subjects switched effector from the training to the transfer phase.

Subjects were given short rest breaks after each block of signals but were given no feedback on reaction time or accuracy. At no time during the training and test phases of the experiment were subjects informed of the sequential or random arrangement of successive stimu- lus locations.

Following the final test phase of the experiment, sub- jects were given a questionnaire that assessed their aware- ness of the sequence. Subjects were shown numerical rep- resentations of the six possible sequences used in the experiment. They were told that often the signals ap-

peared in a repeating pattern, and they gave a confidence rating on a scale of 0 to 100 regarding whether a particu- lar sequence had been seen.

Results and Discussion Median reaction times of each trial block were calcu-

lated for each subject and submitted to an analysis of variance (ANOVA). There was no significant difference as a function of arm versus fingers and no interactions with that variable. Therefore, the data portrayed in Fig- ure 1 were averaged over the two effectors and further analyses ignored that variable. Although the data of Fig- ure 1 show RTs on the first two blocks of trials, those blocks were random and were included to give subjects familiarity with procedures and also to obscure the fact that, later on, signals would occur in sequence. They are not of interest in further analyses.

For two groups of subjects, Trial Blocks 3-10 involved presentations of the signals in sequence. During these pretransfer blocks, there was no treatment difference be- tween the two sequence groups, so, for statistical pur- poses, they were combined. For the control group, the signals occurred at random. An analysis of variance showed a significant effect of block, indicating that, over- all, subjects improved with practice, F(7, 210) = 4.18, p

b

Practice t t 4 t T t

Learning Phase b b Effector Switch b Phase b b b

-m- Sequence - No Switch b b b b

-*-Random - Switch t I I R R : S s s S S S S S : R R S R

1 2 ; 3 4 5 6 7 8 9 10 : 11 12 13 14

Block Number and Condition (Random or Sequence)

FIGURE 1. Experiment 1 (single-task) reaction times. Data from the critical transfer blocks are enclosed in the dotted outline. During Blocks 12 and 14, events occurred at random (R). During block 13, events occurred in sequence (S). For the sequence- no-switch group, the same motor effectors, fingers or arm, were used as during training on a sequence. For the sequence- switch group, the sequence of training was maintained, but the responding effector was switched during transfer from arm to fingers, or vice versa. For the random group, all training prior to Block 13 involved only random events.

20 Journal of Motor Behavior

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

Sequence Modularity

< .001. Some of the practice effect, however, would be EXPERIMENT 2 nonspecific and unrelated to learning the sequence per se. The difference between random and sequence groups was significant, F(1, 30) = 3 1 . 0 , ~ < .001, but more im- portant, the Group X Block interaction was significant, F(7, 210) = 10.0, p < .001, indicating that the sequence advantage grew over time, as would be expected by grad- ual learning.

The critical issue concerns how sequence learning transfers from one effector to the other. We examined this issue in the last three blocks of the transfer phase. Here there were two blocks in which events occurred at random, with one sequential block sandwiched between. The difference in reaction times to events in sequence versus at random served as a measure of the degree of sequence learning. It must be noted that for the groups that changed effectors following pretransfer training, the transfer sequence block constituted the first opportunity for the effectors used during transfer to experience the sequence. Figure 1 indicates that the improvement of re- action time from random to sequence was nearly identi- cal for the two sequence groups, although one group changed effector and the other did not. The random con- trol group also showed improved performance on the se- quence block, however, indicating some sequence learn- ing on that block alone. However, the advantage of sequence over random for the control group was not as large as it was for the two groups that had had prior se- quence experience.

These impressions were all confirmed by statistical analysis. A series of planned and directional compari- sons were conducted. For some of these comparisons, we wanted a liberal test that would maximize the chance of rejecting a null hypothesis. For that reason, one-tailed t tests were used. The first test showed no difference be- tween the group that changed effector from training to transfer in comparison with the group that maintained effector, t(l1) = .85, p > .40. Although the control group showed a significant sequence effect, suggesting some se- quence learning during the critical test block, the control showed less of a sequential effect than either group with previous sequential learning. The test comparing the control to the sequence group with no change in effector was significant, t(l1) = 3.65, p < .005; the test compar- ing the control with the sequence training group that changed effector also was significant, t(l1) = 4.53, p < .001.

Following the final block of reaction time trials, sub- jects answered an awareness questionnaire. These data will be considered more fully when we discuss Experi- ment 2, but the analysis basically indicated that although subjects in the sequence condition showed more aware- ness of a sequence than the control subjects who had not seen a sequence, such differences were not reliable. In general, subjects had little ability to correctly identify the sequence they had experienced, even though reaction times had indicated learning.

In Experiment 1, sequence benefit transferred more or less entirely from the effector system that had practiced the sequence to an effector that had no sequence experi- ence. Thus, sequence knowledge appeared to be indepen- dent of the effector system. This result occurred, how- ever, under conditions in which no distraction task was superimposed on the primary sequencing task. Although it appeared that subjects in this particular experiment had, at best, limited awareness of the sequence, in other experiments without distraction some subjects clearly be- came aware. This finding raised the possibility that the results were due to a declarative form of memory (e.g., Schacter, 1987; Willingham, Nissen, & Bullemer, 1989) in which subjects can describe their knowledge or make explicit predictions of succeeding stimuli. In a somewhat related vein, Curran and Keele (1993) argued that two different representational systems are available for learn- ing sequences. One system demands attention, and the other needs little or no attention to the relationship be- tween adjacent events. It is possible that the attentionally based system builds a representation that is transferrable to different effector systems, whereas the nonattentional system builds an effector-specific representation.

In Experiment 2, we investigated this issue by using an almost identical paradigm to that of Experiment 1, ex- cept for the addition of a distraction task intended to degrade attention to the relationship between successive events in the reaction time task. In each 200-ms interval between a response and the next reaction time signal, ei- ther a high- or low-pitched tone was inserted. After a block of trials, subjects reported how many of the tones were high pitched. Our previous studies (Cohen, Ivry, & Keele, 1990) showed that although such a distraction task blocks awareness, learning of the sequence types used here is nonetheless possible. The question is whether such nonattentional learning is independent of the effector system that executes the sequence during the learning phase.

Method

Subjects Altogether, 40 college-aged subjects completed this

study, of which 36 were retained to complete the design.

Tasks and Design The design of the experiment was the same as in Ex-

periment I , except for the addition of a secondary task. As in Experiment I , a 200-ms interval transpired be- tween each response and the next signal on the reaction time task. In this experiment, a high- or a low-pitched tone was inserted in each interval, with onset either 40, 80, or 120 ms into the interval. Between 50 and 75 of the tones were high pitched, and at the end of a block of trials subjects reported their number.

March 1995, Vol. 27, No. 1 21

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

S. W. Keele, P. Jennings, S. Jones, D. Caulton, & A. Cohen

As in Experiment 1 , no feedback was given on reaction times, but after judging the number of tones on each block, subjects were informed of the actual number of target tones. The purpose of this feedback was to moti- vate the subject to attend to the distraction. Subjects were never told of the presence of a sequence. On the basis of our past experiments with this paradigm (Cohen et a].. 1989), one would expect few if any subjects to be- come aware of the sequence. The awareness question- naire of Experiment 1 was employed here as well.

Results and Discussion

Because the tone task was used only for distraction and yielded no information regarding the primary issues, it will not be discussed.

Reaction times under dual-task conditions were er- ratic. To reduce variance, we calculated the median reac- tion time on each block of trials for each subject and we then calculated means of medians for all further data summaries. Despite our use of medians, we observed that some subjects still found the dual-task conditions very difficult, and their extreme reaction times added consid- erable experimental noise. Any subject whose median re- action time exceeded 700 ms on either of the first two blocks of random trials was replaced until there were 12

subjects in each group. This method prevented selective bias in the different groups. Altogether, 4 subjects were replaced. We also examined the accuracy of tone count- ing with the intention of replacing any subjects who ap- peared to neglect the secondary task. In this experiment, however, all subjects met the criterion of being within ~3 of the number of target tones in at least one of the first two blocks of trials.

Reaction times are shown in Figure 2. Because there were no significant interactions involving arm versus finger in initial training, all data were collapsed over that variable, and it was not further considered.

Subjects were slower than in Experiment 1 because of the distracting influence of the secondary task. As in Ex- periment 1, the two sequence groups were treated identi- cally in Blocks 3-10 and were therefore combined for comparison with the group that received only random signals in that period. An analysis of variance (ANOVA) on Blocks 3-10 showed a significant effect of block, F(7, 210) = 3.94, p < .001, and a significant difference be- tween the sequence-trained and the random group, F( 1, 30) = 4.92, p < .05. However, the interaction Block X Condition (sequence versus random) was not significant, F < 1 . Such an interaction would indicate learning of the sequence over time. The main effect of sequence over random appeared to be the result of an anomaly, because

800

700 € 6 Practice , t

Learning Phase b Effector Switch b

b Phase b b

b I I

b b

b

t I -m- Sequence - No Switch b I I 1 b \,,,,,/

I .& Sequence - Switch

-+-Random - Switch b b b I +

R R : S s s S s s S S : R R S R I I t I I I 1 1 1 I : :- I I b I I I I I I I

1 2 : 3 4 5 6 7 8 9 10 : 11 12 13 14

Block Number and Condition (Random or Sequence)

FIGURE 2. Experiment 2 (dual-task) reaction times. During Blocks 12 and 14, events occurred at random (R). During Block 13, events occurred in sequence (S). For the sequence-no-switch group, the same motor effectors, fingers or arm, were used as had been used during training on a sequence. For the sequence-switch group, the sequence of training was maintained, but the responding effector was switched from arm to fingers, or vice versa. For the random group, all training previous to block 13 involved only random events.

22 Journal of Motor Behavior

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

Sequence Modularity

a difference between the groups was found even on the first two blocks of trials when all events were at random. Although the expected interaction failed to materialize, there is limited sensitivity in this kind of between-group comparison. The really critical comparisons occurred during the transfer period, in which the design was more sensitive.

The last three blocks of transfer consisted of two ran- dom blocks with a sequence block between. The differ- ence between random and sequence was much reduced from the single-task conditions of Experiment 1, but still, for the groups that had previously learned a sequence, a sequence advantage persisted. Moreover, the magnitude of the difference when effector was switched during transfer was practically identical to that for the un- changed effector group, indicating transfer of all se- quence learning. In contrast, the control group that had had practice only with random signals prior to transfer showed no sequence learning, suggesting that the se- quence effects for the other two groups were not the re- sult of new sequence learning on the critical transfer block.

Again, we used planned, one-tailed t tests because all hypotheses were directional in nature. Moreover, some comparisons called for a sensitive test to determine if the null hypothesis could be rejected. Given that the control subjects, experiencing sequence for the first time during transfer, showed slightly slower reaction times to sequen- tial than random events, there was no significant effect of sequence versus random conditions. The sequence- trained group that maintained the same effector during transfer showed a larger sequence effect than did the con- trol group, t( 11) = 2.088, p < .025, as did the sequence- trained group that changed effector, r( 11) = 2.27, p < .025. Note, however, that the sequence-trained groups did not differ from each other, r( 1 I ) = .99, p > .33.

The awareness questionnaire confirmed that the sec- ondary task not only provided distraction but prevented sequence awareness. Subjects’ mean confidence ratings that they had seen the sequence they actually experienced in this experiment were numerically lower than the rat- ings for ones they had not experienced. In Experiment 1, under single-task conditions, the mean ratings for actu- ally experienced sequences were slightly higher than the ones not experienced. However, an ANOVA on confi- dence ratings for experienced versus inexperienced se- quences, random versus sequence training, and dual ver- sus single task, showed no significant effects. Thus, in general, subjects seemed unaware of sequences.

The results of Experiment 2 were consistent with those of the first experiment in suggesting that all of the se- quence knowledge is independent of the effector system used to execute the responses. They were consistent de- spite the addition of a distraction task in the second ex- periment. Regardless of whether the sequence knowledge is stored in an attentionally based or nonattentional rep- resentation system (Curran & Keele, 1993), sequential

knowledge appears to reside in a module separate from the effector system.

Similar results were reported in studies by Keele, Jen- nings, Jones, and Cohen (1992). Those studies were very similar, except that random versus sequence comparisons were between-group rather than within-group compari- sons, and there was no control group to rule out the pos- sibility that sequence learning occurred anew during the transfer phase of the experiment.

EXPERIMENT 3

The nature of a response must be distinguished from an effector that executes the response. In the two preced- ing experiments, subjects pressed response keys with ei- ther finger or arm. If the response is defined by the key that is struck, then when transfer occurs from finger to arm, or vice versa, the responses nonetheless stay the same. Although the first two experiments suggested that the knowledge of sequence is not effector specific, they did not clarify whether the knowledge is bound to the succession of locations at which stimuli occur or to the succession of keys to be pressed. Moreover, despite the fact that arm and fingers clearly are operated by different muscles, it might be argued that there is some intrinsic similarity of their motor codes. In the third ex- periment, we attempted to determine whether the se- quential knowledge obtained in these experiments ac- crues to a system that represents the stimuli or to a response system. The issue was addressed by examining transfer from a situation in which signals are responded to with key presses to one in which verbal labels are given for the signal positions. If sequential knowledge trans- fers, it would suggest that the sequential representation is a description of successive stimuli in space.

Previously Willingham et al. (1989) used a paradigm, from which we have borrowed many features, to address a similar issue. In one of their studies, the signals were colors that occurred in a specific order. Subjects re- sponded with a different key press for each color. Then, in a transfer phase, the subjects responded to the loca- tions of stimuli, again with key-press responses, and the repeating order was the same as it had been for colors. If the sequential code is embedded in terms of response or- der, then transfer of knowledge should have occurred. However, no transfer was found, suggesting not only that the sequential code is not effector specific, but that it is not response specific either. Willingham and coworkers suggested that the code for sequential learning is neither stimulus based nor response based, but is based instead on condition-action pairs. That is, subjects learn the or- der of actions linked to specific stimuli. Although their experiment did not find support for a response-based code, the ossibility of a stimulus-based code as an alter- native to [ ondition-action pairs has not been strongly ex- amined.

March 1995, Vol. 27, No. 1 23

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

S. W. Keele, P. Jennings, S. Jones, D. Caulton, i3 A. a h e n

This experiment, though similar in design to the pre- ceding ones, was run in two parts, with each part involv- ing two groups of participants. The first part compared two groups that learned a sequence. During transfer, one group changed response system; the other did not. In this portion of the experiment, we made an explicit compari- son of whether all sequential knowledge transfers from one response system to another. In the second part of the experiment, with new subjects, we compared one group that experienced the sequence in the pretransfer phase with another that during pretransfer had only random events. Both groups changed response system during transfer. This portion allowed us to examine whether the sequential knowledge exhibited by a group with changed response modality actually transfers previous learning or only learns the sequence anew during the transfer phase. The methodology of the first part of the experiment is described fully, and the second part is incorporated in the context of reporting results of the first part.

Subjects Subjects recruited from introductory psychology

courses were assigned to each of two groups. The initial two blocks of trials involved random orders of stimuli and verbal responses and were in all respects identical for the two groups. Three subjects were slower than a criterion mean reaction time of 700 ms on the first two blocks and were replaced, leaving 14 acceptable subjects in each group.

Tasks The tasks were similar to the tasks of Experiment 2,

but with some notable differences. The signals, as before, were X marks that appeared at one of three stimulus lo- cations, and as before, they occurred either at random or in a fixed sequence. The same six sequences from the preceding experiments were used; 2 subjects in each con- dition were assigned to four of the sequences, and 3 sub- jects to the other two. There were two different ways of responding to the visual signals. In the manual condition, the subjects pressed keys with their fingers in the same manner as in the preceding two experiments. In the ver- bal condition, subjects responded to the position of the stimulus with the words lest. middle, and right.

The nature of the secondary task was changed from Experiment 2 so that the primary and secondary tasks would differ in response mode and so the difficulty of the two versions of the secondary task would be about equal. A high or a low tone occurred in each 200-ms interval between one primary task response and the next stimu- lus. When the primary task involved manual responding, subjects said “high” if the tone was high pitched and re- frained from response when the tone was low pitched. When the primary task involved verbal responding, the secondary task involved pressing a key for high tones and

refraining from response for low tones. Pilot work indi- cated that responding to both high- and low-pitched tones made the task too difficult when paired with the primary task.

Design and Procedure Subjects were divided into two groups. The initial two

blocks of signals (100 signals each) were identical for both groups and involved random presentations of x marks in the different signal positions and verbal re- sponses to the signal positions. For the eight blocks fol- lowing the initial random blocks, the X marks appeared in sequence. One of the groups continued to respond with verbal responses on the primary reaction time task and manual responses to tones. The other group made manual responses on the primary task and verbal re- sponses on the secondary one. After the training blocks, all subjects again gave verbal responses on the primary reaction time task. First there were two blocks with ran- dom orders. The third block in the transfer phase in- volved the same sequential order of stimuli that the sub- ject had received earlier. The fourth transfer block once again involved random events. The last three blocks of transfer permitted us to perform a within-subject evalua- tion of sequence learning.

Note that in this design one group used verbal re- sponses throughout. The other group used manual re- sponses to practice the sequence and then was given a test of sequential knowledge in which verbal responses were used. The latter group had never experienced the sequence in the context of verbal responses until the criti- cal test block during transfer.

Following the final transfer block, subjects were given a questionnaire that assessed their awareness of the se- quence. The questionnaire differed from that used in the first two experiments. Subjects were first asked to make a forced choice between whether the visual signals ap- peared at random or in a repeating pattern. If they an- swered, “repeating pattern,” they checked what percent- age of the pattern they recalled (0, 10, 25, 50, 75, 90, or 100%) and then described as much of the pattern as they could.

Results Reaction times were based on means of the medians

derived from each block of trials and are shown in Figure 3. Blocks 3-10 showed progressive improvement in reac- tion times during the phase in which events occurred in sequence. Subjects in the manual group were consider- ably faster in this phase, presumably either because the manual task was easier or because manual responses typ- ically have shorter reaction times (see data summarized in Fitts & Posner, 1967, or Keele, 1986). Of primary inter- est, however, are the data of the transfer blocks-Blocks 12-14. Reaction times over these three blocks were slower for the group that had been trained on manual responses but used verbal responses during Blocks 12-

24 Journal of Motor Behavior

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

Sequence Modularity

7 0 0 -

600

500 a 2 - 400

t- c 300

.i 5 m d 200

-- -- --

-- --

Practice ; b b Q b

Learning Phase Effector Switch b Phase b b b b

_9_. Verbal/Verbal I

R R R : S s s s s s S S i R R S 0 :

I I . . I I 1 I I I I . I I I I I . I I I I I I 1 m . . I I I 1 2 : 3 4 5 6 7 8 9 10 : 11 12 13 14

Block Number and Condition (Random or Sequence)

FIGURE 3. Experiment 3 reaction times. In Blocks I and 2, signals occurred at random (R) and subjects made verbal responses. In Blocks 3-10, signals occurred in sequence (S); half of the subjects responded with manual key presses and the other half with verbal responses. During the critical transfer blocks, 11-14, enclosed in a dotted outline. all subjects made ver- bal responses. Blocks 11, 12, and 14 were random, and Block 13 was with sequence.

14. The slowed reaction times presumably reflected less practice with verbal responding. The amount of sequen- tial learning was indexed, however, by the difference in reaction times between the two random blocks-Blocks 12 and 14-and the sequence block-Block 13. Indexed in this way, the amount of sequence benefit for the group with verbal training throughout was 100 ms. For the group trained manually and transferred to verbal, se- quence benefit was 34 ms. The 34-ms effect was signifi- cantly less than the 100-ms effect. as shown by an interac- tion of Blocks 12, 13, and 14 with condition, F(2, 52) = 8.84, p < .001. A subsequent test showed, however, that the 34 ms of transfer was significantly greater than zero, t(13) = 3 . 6 , ~ < .005.

It is possible that the sequential knowledge exhibited on Block 13 by the group that transferred from manual to verbal responding reflected not the transfer of knowl- edge but new learning that occurred on Block 13 itself. To evaluate this possibility, we conducted a second part of the experiment with two new groups of subjects. One group was trained and tested identically to the manual group of the main experiment. That is, following two blocks of dual-task performance on the verbal task but with random order, there were eight blocks of manual dual-task learning with sequence and then transfer to verbal responding. Blocks 1 1, 12, and 14 were again ran-

dom, and Block 13 involved the practiced sequence but with change to verbal response. The other group, after the two initial verbal, random blocks, had eight addi- tional blocks of manual responding, but in this case the signals always occurred in random order. For this group there was, therefore, no opportunity to learn the se- quence in the training period. Then the subjects trans- ferred to the verbal condition, again with two blocks of random, one of sequence, and a final with random. The reaction time results are shown in Figure 4.

The group that had never experienced a sequence until Block 13 showed no appreciable benefit from the se- quence on that block: t( 13) = .9, ns. Thus, they were un- able to learn the sequence within that single block. On the other hand, the group that had experienced the se- quence manually showed a 47-ms advantage on Block 13 compared with the average of Random Blocks 12 and 14, t( 13) = 3.4, p C .005. The 47-ms effect here was compa- rable with the 34-ms effect of the main experiment, which involved an identical condition. The 47-ms value suggests transfer of sequential benefit from the manual system in an amount nearly 50% of that found following verbally based learning, rather than the 34% estimated from the main experiment. These different values likely reflected subject differences, and they should be viewed as con- firming one another.

March 1995, Vol. 27, No. 1 25

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

S. W. Keele, P. Jennings, S. Jones, D. Caulton, & A. &hen

Practice I I I

Learning Phase I Effector Switch I Phase I

* I t R R : S S S S S S S : R

0 4 I I I I I I I I , I I

S , I I

1 I 7 1 I 1 I I I I - I 1 I I R S R

0 1 I

2 - 3 4 5 6 7 8 9 I

10 . 11 12 13 14

Block Number and Condition (Random or Sequence) FIGURE 4. Reaction times from the supplement to Experiment 3. During training Blocks 3-10, one group responded manual- ly only to random signals and the other to signals in sequence. During transfer, responding was verbal, Blocks 1 1 , 12, and 14 were random, and Block 13 was sequence.

Why is it that only about 50% of sequential benefit transferred from manual to verbal? One possibility is that, for unknown reasons, sequential learning under the manual conditions was simply less than under verbal conditions, and hence less was available to transfer. We are unable to evaluate this hypothesis. Alternatively, some sequential knowledge may be represented in the or- der in which response keys are pressed, and that informa- tion is lost when transferring from key pressing to verbal responses. Although the findings of Experiments 1 and 2 suggested that the effector system does not contain the sequential knowledge, and although the findings of Ex- periment 3 directly suggested that some sequential knowledge is stimulus based, we did not directly test whether some knowledge might be response based. This latter possibility will be further discussed in the Conclu- sions section. Moreover, as Willingham et al. (1989) suggested, some of the knowledge might be inseparable stimulus-response bindings.

The existing transfer from manual to verbal re- sponding did not seem to depend on awareness of the sequence. The secondary task appeared effective in lim- iting awareness for almost all subjects. In the first part of the main experiment, all 28 subjects experienced a se- quence during the training phase. Sixteen of them checked on the questionnaire that the response signals

appeared in sequence, but only 1 of these was able to describe the sequence. No other subjects indicated they had knowledge of as much as 50% of the sequence, and the knowledge expressed was confined to the observation that some signals seemed to occur more frequently than others. In the supplementary experiment, only 4 of 14 subjects who had experienced the sequence during train- ing affirmed that there was a sequence, and none of these could describe any part of it. It is notable that 3 of the 14 subjects trained with a random condition still thought the signals occurred in a repeating order, although, in fact, such had been the case only on one of the 14 test blocks. They could not describe the pattern. Thus, some reports of a repeating sequence appeared to be guesses.

The results of Experiment 3 suggest, therefore, that at least a sizable proportion of sequential knowledge is tied to the order of the stimuli. That knowledge is free to be shared with effector systems as different as manual and verbal. Moreover, given the general lack of expressed awareness about the sequence, although such representa- tion can be shared among different output systems, it does not seem part of a declarative memory system. We were unable to determine, however, whether additional sequential knowledge was tied to the mode of re- sponding, though not to effector system per se, or to stimulus-response combinations, or whether the appar-

26 Journal of Motor Behavior

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

Sequence Modularity

ent reduced transfer from manual to verbal was actually a result of less sequential learning in the manual con- dition.

CONCLUSIONS

In the current studies, we investigated whether the rep- resentation of a sequence of action is independent of the effector systems that implement movement. By and large, independence was supported. Under conditions in which people have attention available to learn the sequence of action, the learned sequence transfers more or less en- tirely from arm to fingers or vice versa. Similar results were reported by Stadler (1989). When attention is dis- tracted by a secondary task, not only making it unlikely that subjects will become aware of the relationship be- tween successive elements in a sequence but also forcing subjects to rely on a nonattentional form of learning (Curran & Keele, 1993), certain kinds of sequences can still be learned. Even in this circumstance, sequential knowledge is transferred, again more or less in its en- tirety, to other effector systems. Even when the re- sponding effector is changed as drastically as from fin- gers to voice, some knowledge transfers about the order of stimulus positions to which one responds.

Our results are consistent with a theory of modularity, such as one proposed by Berkinblit and Feldman (1 988), in which a spatial system specifies a trajectory of move- ment in space. A separate system specifies the combina- tion of muscles to use to achieve a position in space at one point in time, but that system does not itself contain the knowledge of the subsequent positions in space. One might say the fingers or arm know how to get to a speci- fied position in space, but they do not know what the next position will be. The next position is specified by a prior module. A similar theory of modularity has also been proposed by MacKay (1982) regarding the indepen- dence of sequential specification of speech from the mo- tor system that actually produces the speech.

Such a conclusion is in accord with neuropsychologi- cal findings regarding apraxia. Heilman, Rothi, and Va- lenstein (1982) and Gonzales and Heilman (1985) found that ideomotor apraxia, in which patients have difficulty performing gestures such as salutes, come in two varie- ties. One variety stems from anterior cortical lesions, pre- sumably affecting premotor areas. The other variety stems from lesions in posterior areas of parietal cortex. Why might a “motor” disorder of apraxia arise from pos- terior damage in a cortical area thought to deal with spa- tial representation? An important clue is provided by an additional fact. Whereas patients with anterior lesions have difficulty forming an appropriate gesture, they can recognize good ones made by others. Patients with le- sions of the posterior cortex not only have difficulty exe- cuting a gesture, but they have difficulty as well in recog- nizing one. One interpretation of the results follows from the analysis presented here. One system, presumably a

posterior cortical spatial system, specifies the spatial lay- out that will direct movement. We suppose that that sys- tem not only can be accessed by motor systems that then follow the dictated spatial directions, but that the same spatial representation also forms the basis of recognition. If the spatial representation is damaged, both motor per- formance and perceptual judgment suffer. If, however, the implementation system is damaged, motor perfor- mance suffers but recognition is intact.

A second neurological finding consistent with the con- clusion of a common sequential representation for different motor implementations is that misordering of letters during spelling, as exhibited by some patients, is manifested not only in writing but also in oral spelling (Hillis & Caramazza, 1988; see also Margolin, 1984).

The finding that in humans a single representation can be interfaced with distinctly different effectors is consis- tent with a more general idea that a single computation can be shared by diverse tasks, all of which need the same computation. For example, Ivry, Keele, and Diener (1 989) and Keele and Ivry (1990) presented evidence that a mechanism housed in the cerebellum computes time. The timing system can be addressed not only by various motor systems but also by perception. In a similar vein, Abrams and Landgraf (1990) and Abrams, Meyer, and Kornblum (1990) have presented evidence that the same visual representations of space underlie both movement and perception of space. To these examples of modular- ity can be added sequential representation.

Although the results of the present studies are consis- tent with a weak form of modularity that posits a sequen- tial representation that can be shared by different output systems, it remains an open question regarding a stronger form of modularity. A stronger form posits that sequential representation resides in only one module and does not reside in a module that specifies either responses or the effector combinations used to achieve a response. In the present studies, we found no evidence that sequen- tial representation resides in a system that specifies effector. But we did not analyze whether some sequential representation resides in a response specification system as well as in a stimulus specification system. This issue also remained unresolved in a recent study of typing (Jor- dan, in press-b).

The logic of modular theory applied to production of letter strings is that an abstract graphemic or letter code specifies the sequential order of letters. That module can then be accessed by a variety of output systems-typing, writing, oral spelling, and so forth. The abstract module specifies the next letter; the effector module simply exe- cutes the specified item. We have mentioned some evi- dence for this view from neurological patients. With such a theory in mind, Jordan rearranged the positions of four letters on the typewriter keyboard, switching a with s and h withj. Expert touch typists were given initial practice on the new arrangement, but only by typing individual letters, hzt words. Subsequently, a typing test was given

27 March 1995, Vol. 27, No. 1

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

S. W. Keele, P. Jennings, S. Jones, D. Caulton, & A. Cohen

both on prose and on nonsense constructed by scram- bling letters in a prose passage.

The expectation from modular theory is that changes in the keyboard would slow typing of the rearranged let- ters, but the slowing should be equal for prose and non- sense. Let us clarify. Specification of the next letter should occur before the mapping to keyboard. I t should take longer to specify the next letter in a nonsense con- text than for words in the context of prose. Following letter specification, the subject then must choose the key to be struck, and presumably such choice would take longer for switched letters than for unswitched letters. Because key specification follows letter specification, the effects of nonsense versus prose and of switched versus unswitched should be additive.

Jordan (in press-b) found, however, that the deficit from keyboard rearrangement was greater for prose than for nonsense. This and other more detailed results sug- gest that at least some aspects of sequence specification are specific to typing and do not occur in a module prior to attachment of a typing response to letter specification. That is. not all sequential specification occurs prior to interface with the output system. It is as though the fin- gers know their order for well-practiced words, and. therefore, when key assignments are changed, the result is more detrimental for words than for nonsense for which the fingers do not know the order.

An issue that can be raised about Jordan's (in press- b) results, though. is whether the sequential information specific to the act of typing is embedded in the effector system or in a response system. Our first two experiments found no evidence for effector-specific sequential repre- sentation, because sequential knowledge transferred en- tirely from finger to arm and vice versa. In the third ex- periment, we found that a portion of sequential information transferred from manual to verbal re- sponses, suggesting that some sequential representation is bound to the stimulus. However, not all information transferred. One possibility is that a portion of the se- quential representation may reside in the order in which keys are activated. A response-based code offers an alter- native to an effector-based code for Jordan's experiment.

There are two reasons for wishing to inaintain the pos- sibility of response-based sequential codes in addition to stimulus-based codes. One reason comes from a consid- eration of the nature of responses. In keyboard tasks, movements are directed to keys in space. The succession of keys in space differs little, conceptually, from a succes- sion of signals in space. As we have argued elsewhere that sequential order of spatial events can be learned under distraction (Cohen et al., 1990; Curran & Keele, 1993), it seems likely that key orders could also be learned, even though subjects might not be attending to key order. A second reason has to do with a consideration of se- quences other than the sort learned in our studies. One can certainly learn a sequence of key presses (or of other movements) in circumstances in which there is no succes-

sion of input signals. For example, one could simply be told to press three keys in a particular order, such as Key 1, then Key 3 . and then Key 2, and to learn that order. Thus, one would guess that a sequential representation based on key order is permissible in principle.

Some credence to the idea that sequence information can be response based was provided by Fendrich, Healy, and Bourne (1991). Subjects tapped short digit sequences into a keyboard arranged in either calculator format or touch-tone telephone format. On a calculator layout, the top row. left to right, consists of digits 7, 8, and 9. On the telephone, digits I , 2, and 3 appear on the top row. After practice with particular digit sequences and one key- board, subjects transferred to the other keyboard. Some savings occurred on the changed keyboard when the pre- viously practiced digit sequences were entered, consistent with the view that at least partial sequence representation accrues to the stimuli. In addition, however, savings oc- curred when the sequence of digits to be keyed was new but involved the same order of key presses as on the pre- vious keyboard (for example, typing 1687 on a calculator keyboard involves the same sequence of key pressing as 7621 on a telephone keyboard). This finding indicates that at least partial sequence representation accrues both to response order or to effector order. Unfortunately, Fendrich and colleagues did not include conditions that would separate a response-based representation from an effector-based representation. If their study is considered in combination with the current one, however, i t would appear that sequence knowledge accrues to response or- der as well as to stimulus order.

Although we raise the argument that some sequential knowledge, even in Jordan's (in press-b) study, may reside in response order and not in the effector system per se. it is possible that this may be too strong a conclusion. Although the proposition is difficult to test, it might be supposed that with extreme levels of practice. such as oc- cur with accomplished touch typists, some sequential knowledge becomes embedded in the effector system, perhaps as stroke combinations. Preliminary work and conceptions by Wright and Lindemann (1993) might prove useful here. In examining how writing by the non- dominant hand improves with extensive practice, they concluded that the production of strokes of letters is effector specific. Sequential organization above the level of stroke was shared by hands. In general, therefore, i t might be the case that relatively global aspects of a hier- archic representational system are independent of the effector system. There may be a stage properly called mo- for or c&ctor specific, however, that might still be multi- level. Even a stroke in letter production involves coordi- nating muscles, and contribution of different muscles may vary as the stroke takes place. That is, even compo- nents of movement more elementary than the stroke i t - self may become organized into a stroke, and those levels of organization may be dependent on particular effectors. Likewise in typing, it is possible that with ex-

28 Journal of Motor Behavior

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

Sequence Modularity

treme practice, the expert may come to group some key- strokes together into some more efficient or unified pat- tern that is idiosyncratic to the effector mode. That is, practice may enlarge the temporal domain over which sequential representation is motor, but a stage will be reached in which higher levels of representation remain effector independent. This is likely to be the case even with highly practiced skills and with units, such as words, that are highly practiced and familiar.

Thus, although it may be too extreme to conclude that all knowledge that organizes the most molecular ele- ments of movement into larger groupings is effector inde- pendent, we conclude that at least relatively high levels of sequential representation-perhaps levels above that of key-press strokes or possibly some stroke combina- tions, or above handwriting strokes-learned in the con- text of a specific effector system are independent of the effector system itself. In our particular experimental par- adigm, sequential knowledge may reflect primarily where successive stimuli occur. The same sequential representa- tion can be accessed by different response systems, a con- clusion in accord with modular conceptions. Some se- quential representation may also reside in an intermediate stage of response prior to effector specifi- cation, that can be shared again by different modes of executing the response.

ACKNOWLEDGMENT

We are grateful for an Oflice of Naval Research grant (Con- tract #N00014-87-K-0279) and a grant to the Center for Cogni- tive Neuroscience of Attention at the University of Oregon from the Pew Memorial Trust and McDonnell Foundation, which provided financial support for this research. Richard Ivry and Tim Curran provided appreciated advice throughout the project. Reviewers of the manuscript suggested notable im- provements.

REFERENCES

Abrams, R. A., & Landgraf, J. 2. (1990). Differential use of dis- tance and location information for spatial location. Percep- tion and Psychophysics, 47, 349-359.

Abrams, R. A., Meyer, D. E., & Kornblum, S. (1990). Eye-hand coordination: Oculomotor control in rapid aimed limb move- ments. Journal of Experimental Psychology: Human Per- ception and Performance, 16. 248-267.

Berkinblit, M. B., & Feldman, A. G. (1988). Some problems of motor control. Journal of Motor Behavior, 20, 369-373.

Bernstein, N. A. (1947). On the construction of movement. Moscow: Medgiz (in Russian).

Cohen, A., Ivry, R. I., & Keele, S. W. (1990). Attention and structure in sequence learning. Journal of Experimental Psy- chology: Learning, Memory, and Cognition, 16, 17-30.

Curran, T., & Keele, S. W. (1993). Attentional and nonatten- tional forms of sequence learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 189-202.

Fendrich, D. W., Healy, A. E, & Bourne, L. E., Jr. (1991). Long-term repetition effects for motoric and perceptual pro- cedures. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17, 137-1 5 1.

Fitts, P. M., & Posner, M. I. (1967). Human performance. Bel- mont, CA: Brooks-Cole.

Fodor, J. (1983). The modularity of mind. Cambridge, MA: M.I.T. Press.

Gonzalez. L. J.. & Heilman, K. M. (1985). Ideomotor apraxia: Gestural discrimination, comprehension and memory. In E. A. Roy (Ed.), Neuropsychological studies of apraxia (pp. 65-74). New York: North-Holland.

Heilman, K. M., Rothi, L. J., & Valenstein, E. (1982). Two forms of ideomotor apraxia. Neurology, 32, 342-346.

Hillis, A. E., & Caramazza, A. (1988). The graphemic buffer and attentional mechanisms. Reports of the Cognitive Neu- ropsychology Laboratory (Rep. No. 30). Baltimore, MD: Johns Hopkins University.

Ivry, R. I., Keele, S. W., & Diener, H. (1988). Differential con- tributions of the lateral and medial cerebellum to timing and to movement execution. Experimental Brain Research, 73,

Jordan, M. I. (in press-a). Serial order: A parallel distributed processing approach. In J. L. Elman & D. E. Rumelhart (Eds.), Advances in connectionist theory: Speech. Hillsdale, N J Erlbaum.

Jordan, M. 1. (in press-b). The organization of action se- quences: Evidence from a relearning task. Journal of Mo- tor Behavior.

Keele, S. W. (1986). Motor control. In L. Kaufman, J. Thomas, & K. Boff (Eds.), Handbook ofperception andper- formance. New York: Wiley.

Keele, S. W., & Ivry, R. (1990). Does the cerebellum provide a common computation for diverse tasks: A timing hypothesis. In A. Diamond (Ed.), The development and neural bases of higher cognitive function (Vol. 608, pp. 179-21 1). New York: Annals of the New York Academy of Sciences.

Keele, S. W., Jennings, P., Jones, S., & Cohen, A. (1992). On the modularity of sequence representation (Tech. Rep. 92-4). Eugene, OR: Institute of Cognitive and Decision Sciences, University of Oregon.

Lewicki, P., Hill, T., & Bizot, E. (1988). Acquisition of proce- dural knowledge about a pattern of stimuli that cannot be articulated. Cognitive Psychology, 20, 24-37.

MacKay, D. G. (1982). The problem of flexibility and fluency in skilled behavior. Psychological Review, 89, 483-506.

Margolin, D. I. (1984). The neuropsychology of writing and spelling: Semantic, phonological, motor, and perceptual pro- cesses. Quarterly Journal of Experimental Psychology, 36A, 459-489.

Nissen, M. J., & Bullemer, P. (1987). Attentional requirements of learning: Evidence from performance measures. Cognitive

Nissen, M. J., Knopman, D. S., & Schacter, D. L. (1987). Neu- rochemical dissociation of memory systems. Neurology, 37,

Raibert, M. H. (1977). Motor control and learning by the state space model (Tech. Rep. No. AI-M-351). Cambridge, MA: M.I.T.

Rozin, P. (1976). The evolution of intelligence and access to the cognitive unconscious. In J. M. Sprague & A. N. Epstein (Eds.), Progress in psychobiology and physiological psychol- ogy (pp. 245-280). New York: Academic Press.

Schacter, D. L. (1987). Implicit memory: History and current status. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13. 501-518.

Stadler, M. A. (1989). On the learning of complex procedural knowledge. Journal of E-xperimental Psychology: Learning, Memory, and Cognition, 15. 106 1-1 069.

Willingham, D. B., Nissen, M. J.. & Bullemer, P. (1989). On the development of procedural knowledge. Journal of Experi- mental Psychology: Learning, Memory, and Cognition, 1.5.

Wright, C. E. (1990). Generalized motor programs: Reexamin-

167-1 80.

Psychology, 19. 1-32.

789-794.

1047-1060.

March 1995, Vol. 27, No. 1 29

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013

S. W. Keele, P. Jennings, S. Jones, D. Caulton, & A . Cohen

ing claims of effector independence in writing. In M. Jeannerod (Ed.), Attention and performance XI11 (pp. 294- 320). Hillsdale, NJ: Erlbaum.

Wright. C. E., & Lindemann, P. (1993, November). Effector in- dependence in hierarchically structured motor programs for

handwriting. Paper presented at the meeting of the Psy- chonomics Society, Washington. DC.

Submitted March 9, I992 Revised October 28. I993

Devoted to an understanding of motor behavior as it is most broadly defined, this

various perspectives and dif- fering levels of analysis. It encompasses all areas of mo- tor behavior including psy-

m m m m m m m m m m m m m m m m m m m m m

ORDER FORM 0 YES! I would like to order a one-year subscription to Journal of Motor Behavior, published quarterly. I understand payment can be made to Heldref Publications or charged to my VISA/MasterCard (circle one). 0$45.00 individuals 0 $84.00 institutions

: journal presents articles from

A C C O U N T # E X P I R A T I O N D A T E

S I G N A T U R E m

N A M E l l N S T l T U T l O N

A D D R E S S

C l T Y l S T A T E l Z l P

C O U N T R Y

ADD $10.00 FOR POSTAGE OUSIDE M E U.S. ALLOW 6 WEEKS FOR DELIVERY OF FIRST ISSUE.

SEND ORDER FORM AND PAYMENT TO

HELDREF PUBLICATIONS, JOURNAL OF MOTOR BEHAVIOR

1319 EIGHTEENTH STREET, NW, WASHINGTON, DC XXl3&1802

PHONE (202) 296-6267 FAX (202) 296-5149

SUBSCRIPTION ORDERS 1 (800) 365-9753

chology, kinesiology, neuro- physiology, and biomechan-

the processes underlying mo- tor control, learning, and de- velopment. In addition to original research, the journal publishes theoretical and re- view articles, notes and com- ments, and invited articles by recognized authorities.

m ics. Of special interest are

m

30 Journal of Motor Behavior

Dow

nloa

ded

by [

Uni

vers

ity o

f Su

ssex

Lib

rary

] at

09:

42 0

9 Se

ptem

ber

2013