Journal of Experimental Psychology:Human Perception and Performance1985, Vol. 11, No. 2, 221-233

Copyright 1985 by the American Psychological Association, Inc.0096-1523/85/$00.75

A Hand Advantage in Preparation of Simple KeypressResponses: Reply to Reeve and Proctor (1984)

Jeff MillerUniversity of California, San Diego

Miller (1982) introduced an experimental paradigm to discriminate betweendiscrete (e.g., Steinberg, 1969) and continuous (e.g., McClelland, 1979) informationprocessing models. The paradigm required the assumption that preparation oftwo response fingers on the same hand leads to faster responses than doespreparation of two fingers on different hands, at least with the response set used.Data supporting this assumption were obtained using a movement precuingexperiment. However, Reeve and Proctor (1984) challenged the assumption, notinga potential artifact in the design of the experiment supporting it. They repeatedthat experiment, including an additional response set to correct the artifact, andobtained results contrary to the assumption in question. It is argued that the newresponse set used to correct for the artifact was very different from the originalresponse set and that results obtained with this response set cannot be used todraw conclusions about preparation of the original responses. A new experimentis reported that corrects the artifact using a response set similar to the one usedby Miller (1982), and the results support the assumption in question.

Miller (1982) developed an experimentalparadigm to compare discrete stage (e.g.,Sternberg, 1969) and continuous output (e.g.,McClelland, 1979) models of human infor-mation processing. Specifically, the paradigmwas designed to determine whether responsepreparation can begin before stimulus rec-ognition finishes, as assumed by continuousmodels but prohibited by discrete models.The results supported a class of models in-termediate between the discrete and contin-uous extremes, referred to by Miller (1982,1983) as asynchronous discrete coding models.In these models, response preparation canbegin as soon as the recognition process hascompletely identified a distinct stimulus at-tribute or code (Garner, 1974; Posner, 1978),even if other stimulus attributes have not yetbeen fully recognized. Response preparationcannot begin if an attribute has been onlypartly recognized.

I would like to thank Sarah Frampton, Lisa Oakes,and Janice Quackenbush for their assistance in testingsubjects, and William Epstein, Patricia Haden, and DavidRosenbaum for helpful comments on earlier drafts of themanuscript.

Requests for reprints should be sent to Jeff Miller,Department of Psychology, C-009, University of California,San Diego, La Jolla, California 92093.

Reeve and Proctor (1984) disputed an as-sumption critical to the paradigm used byMiller (1982): that a small amount of prep-aration of two response fingers on the samehand leads to faster responses than does asmall amount of preparation of two fingerson different hands, at least with the responsesactually used in the experiments. Miller(1982) obtained data supporting this assump-tion in an experiment using the movementprecuing technique (Rosenbaum, 1980). How-ever, Reeve and Proctor (1984) pointed outa potential artifact in the design of the pre-cuing experiment and repeated the experi-ment using an additional response set tocorrect the artifact. With the new responseset, preparation of two response fingers onthe same hand did not produce faster respon-ses than did preparation of two fingers ondifferent hands, and they concluded that theartifact was probably responsible for thepreparation effect in the original cuing study.These results led them to reject the assump-tion in question and to express serious doubtabout Miller's paradigm and conclusions.

In this article, the paradigm introduced byMiller (1982) is defended against Reeve andProctor's criticisms. It is argued that the newresponse set they used to eliminate the artifactinvolved responses quite different from those

221

222 JEFF MILLER

used in Miller's (1982) original experiments.Preparation effects with these responses mightwell be different from preparation effects withthe original responses, and it is not surprisingthat preparation favored different pairs offingers. Because the original paradigm didnot require the assumption that fingers onthe same hand would benefit most frompreparation for any response set, the resultsof Reeve and Proctor are not damaging to it.A new experiment is reported that correctsthe suspected artifact by using responses sim-ilar to the ones used originally (Miller, 1982),and the results provide further evidence forthe assumption required by Miller's (1982)paradigm.

In defense of Miller's (1982) paradigm, itis also argued that the artifact noted by Reeveand Proctor (1984) could account for at mosta small number of the results reported byMiller (1982) and is therefore not very plau-sible as an alternative hypothesis.

The Hand-Preparation Advantage

To study the interface between perceptualand response processes, Miller (1982) usedchoice reaction-time (RT) tasks with fouralternative keypress responses made by theindex and middle fingers of the two hands.Subjects responded using four keys on thebottom row of a computer keyboard, andthey placed their fingers on keys in the mostnatural fashion (left middle finger on theleftmost key, then left index finger, etc.).Miller's (1982) paradigm was based on theassumption that for this response set responseswould be faster if there were advance prepa-ration of two response fingers on the samehand than if there were advance preparationof two response fingers on different hands, atleast for small amounts of advance prepara-tion (less than 1 s). A movement precuingexperiment (Rosenbaum, 1980, 1983) wasused to establish this hand-preparation ad-vantage. In this experiment a target stimuluswas a plus sign appearing in one of fourpositions in a horizontal row, and targetswere assigned to response keys in the mostcompatible left-to-right order (i.e., leftmostposition to leftmost key, etc.). On each trial,three rows of visual stimuli were presented.First, a warning signal was presented in the

top row. It had four plus signs showing thefour possible stimulus positions, and it re-mained on throughout the trial. Next, themiddle row appeared directly below thewarning signal, and it was a cuing stimulusthat gave partial advance information (Kan-towitz & Sanders, 1972) about the upcomingresponse. This row had plus signs in two ofthe four positions, and subjects could selec-tively prepare the two responses indicated bythe middle row. The cuing stimulus alsoremained on for the remainder of the trial.Finally, the target stimulus was presented inthe bottom row, located directly below theother two rows. This stimulus was a singleplus sign in one of the four positions. Subjectswere to respond by pressing the response keyassociated with this position as quickly aspossible.

In this paradigm, Miller (1982) used cuingstimuli corresponding to four different re-sponse preparation conditions. In the pre-pared: hand condition, the cue indicated thetwo leftmost or rightmost positions, therebyenabling preparation of two fingers on thesame hand, either left or right. In the prepared:finger condition, the cue indicated the twoinside or two outside positions, thereby en-abling preparation of two homologous fingerson different hands, either index or middle.The other two pairs of positions were cuedin the prepared: neither condition, indicatingpreparation of the index finger of one handand the middle finger of the other. Finally,there was an unprepared condition in whichcues appeared in all four positions, so noselective preparation of responses was pos-sible.

Overall, responses were approximately 50ms faster in the prepared: hand conditionthan in any of the others, which were allapproximately the same. Miller (1982) attrib-uted this advantage to more efficient responsepreparation with cuing of two fingers on thesame hand than with cuing of fingers ondifferent hands. It is also relevant that thesize of the hand advantage initially increasedwith the time between cue and test stimulus(preparation interval), leveling off at an in-terval of about 400 ms. As preparation inter-val increased from 125 to 375 ms, the handadvantage more than doubled. Though thesmall hand advantage at the shortest prepa-

HAND PREPARATION 223

ration intervals might result from other fac-tors, the increase in the effect with increasinginterval strongly suggests a preparatory effectof the cuing information.

The hand-preparation advantage is impor-tant for two reasons. First, it can help deter-mine whether certain stimulus attributes en-able response preparation to begin before thestimulus as a whole has been identified, usingthe paradigm developed by Miller (1982). Forexample, Miller (1982, Experiment 2) usedfour consonant-vowel pairs as stimuli (BE,BO, ME, MO) and varied the assignments ofstimuli to responses. When the consonantwas presented before the vowel, responseswere faster if the two responses consistentwith the consonant were on the same handthan if they were not, and this advantageincreased with the time between the conso-nant and the vowel. This finding suggests thatthe consonant can be used to prepare respon-ses on the indicated hand before the vowelhas been recognized. In another experiment(Miller, 1982, Experiment 3), the stimuliwere four letters varying in name and size—s, S, t, T, with the name discrimination mucheasier than the size discrimination. With thesestimuli, responses were faster if the two re-sponses consistent with each letter name wereon the same hand than if they were not,suggesting that letter name could be used tostart preparing responses before size had beendetermined. Thus, the paradigm indicateswhether or not an attribute allows responsepreparation by comparing conditions thatdiffer only in the benefit that would be re-ceived from preparation. Such conditions willdiffer only if preparation does take place.

Second, the hand-preparation advantagehas implications about the structure of themotor system components responsible forcontrol of finger movements, componentsthat are particularly interesting because ofthe sophistication of finger control, the prac-tical implications for tasks like typing andpiano playing (Shaffer, 1975, 1981), and themethodological implications for studies usingmanual RT. It is generally assumed that prop-erties of the motor system are revealed instudies of how various pairs of movementsare selected and prepared together (e.g., Ro-senbaum, 1980, 1983; Rosenbaum & Korn-blum, 1982). Specifically, if a pair of responses

is especially easy to prepare together, theyare thought to share common attributes at arelatively high level within the motor system.Given this assumption, the hand-preparationadvantage suggests that the attribute of handoften takes precedence within the system formanual control. Data in support of this con-clusion have also been obtained in studies ofserial finger movements, both typing andpiano playing (Shaffer, 1975, 1981). One ob-vious possibility, consistent with much neu-rological evidence (e.g., Brinkman & Kuypers,1973; Gazzaniga, 1970; Wiley, 1975), is thatthe system is structured hierarchically withhand at the highest level, but other possibilitieshave also been suggested (e.g., Goodman &Kelso, 1980). Though the mechanisms ofmotor preparation are interesting and impor-tant in their own right, it should be notedthese mechanisms need not be specified beforeone can use the hand advantage to determinewhich stimulus attributes allow responsepreparation to occur.

Reeve and Proctor's (1984) Objections tothe Hand-Preparation Advantage

Reeve and Proctor (1984) questioned thehand-preparation advantage because the pre-pared: finger and prepared: neither conditionswere no faster than the unprepared condition.The lack of preparation in these two condi-tions suggests that response hand must bespecified before any other attribute of themovement can be prepared. Most studieswith other cued movements have found thatany type of preparation is better than nopreparation at all (Rosenbaum, 1983; but seeZelaznik, 1981), suggesting that movementattributes can be prepared in any order.Therefore, Reeve and Proctor conducted threeadditional experiments to see if the hand-preparation advantage observed by Millerwas actually caused by response preparationor if it was caused instead by some nonmo-toric factors) in the experimental setup. Theirfirst two experiments were not directly in-compatible with the hand-preparation advan-tage assumed by Miller (1982) and so will bediscussed only briefly in this article. Thethird presented a major challenge to thecritical assumption, however, so it will beconsidered in detail.

224 JEFF MILLER

In Experiment 1, Reeve and Proctor (1984)replicated the hand-preparation advantage forpreparation intervals of less than 1 s butfound that the advantage disappears whenthe preparation interval is extended to 3 s.This finding is consistent with other evidencesuggesting that hand preparation would leadto the slowest responses if even more prepa-ration time were provided (Kornblum, 1965;Rosenbaum & Kornblum, 1982). As theyacknowledged, one explanation of these re-sults is that hand preparation is the best forbrief preparation intervals, even it if is notthe best for longer preparation intervals. Be-cause all the experiments of Miller (1982)involved brief preparation intervals, this ex-planation is completely consistent with theassumption required for his paradigm.

Reeve and Proctor (1984) also emphasizedthe fact that they had obtained a hand ad-vantage with simultaneous presentation ofprecue and target (Experiment 1). They ar-gued that this effect "suggests that responsepreparation is not the cause of the advantage,because the time for response preparationshould be minimal, at most, when the precuedoes not precede the target" (p. 546). It ispossible, however, that subjects adopt a strat-egy of using cues for some preparation evenwith simultaneous presentation. That is, sub-jects may partially process the cue beforeattending to the target even when both comeon at the same time. Furthermore, even ifresponse preparation is not the cause of thehand advantage with simultaneous presenta-tion, the increase in this advantage withincreasing preparation interval still suggestsa preparation advantage for responses on thesame hand.

In Experiment 2, Reeve and Proctor (1984)varied the percentage of trials with simulta-neous presentation within a block (either20% or 80%), using a 3-s preparation intervalfor the remaining trials. In previous experi-ments (Miller, 1982; Reeve & Proctor, 1984,Experiment 1) the hand advantage with si-multaneous presentation was primarily aninterference effect: The prepared: finger andprepared: neither conditions were slower thanthe unprepared condition, and the prepared:hand condition was only slightly faster thanunprepared. In Experiment 2, this interferencewas again found in blocks with simultaneous

presentation on only 20% of trials, but it wasnot observed in blocks where 80% of thetrials had simultaneous presentation. Thisfinding suggests that the interference resultsfrom active strategies of cue processing (Reeve& Proctor, 1984). This conclusion is notincompatible with the idea that the handadvantage is due to response preparation,however, because response preparation isknown to be influenced by strategies (Requin,1980; Sanders, 1983). In any case, the prep-aration intervals used in this experiment (0 sand 3 s) preclude any unambiguous conclu-sions about the hand advantage observed inthe first few hundred milliseconds of prepa-ration, and it is the latter advantage that isrequired for the paradigm of Miller (1982).

The third and most critical experiment ofReeve and Proctor (1984) addressed a poten-tial cuing artifact in the experiment used toestablish the hand-preparation advantage.Specifically, Miller (1982) used different pairsof stimulus locations as cues in the prepared:hand, prepared: finger, and prepared: neithercuing conditions. Reeve and Proctor (1984)pointed out that responses might have beenfastest in the prepared: hand condition becausethe cues were easiest to process in that con-dition. Response preparation might have beenidentical for the three prepared conditionsbut simply initiated sooner in the prepared:hand condition because of faster cue process-ing. Thus, the hand-preparation advantageobserved by Miller (1982) may not havearisen in the process of response preparation.If not, Miller (1982) erred in using the ad-vantage to infer that response preparationwas caused by preliminary information abouta stimulus attribute. The effects attributed topreparation of responses using preliminaryinformation might actually have arisen fromsome other mechanism.

To test this alternative explanation, Reeveand Proctor's (1984) third experiment wasdesigned to show that the hand-preparationadvantage disappears when the confoundingbetween cue positions and type of responsepreparation is removed. They used the samestimuli, cues, and responses as Miller (1982)but varied the assignment of fingers to re-sponse keys. For half of the subjects, fingerswere assigned to response keys in the com-patible left-to-right order (adjacent hands

HAND PREPARATION 225

condition) used by Miller (1982). For theother half, the hands were overlapped, so thatthe four response keys, from left to right,were pressed by the right index, left middle,right middle, and left index fingers (overlappedhands condition). Positional and responsepreparation effects of the cues could be sep-arated by including both adjacent and over-lapped hand positions, because the cues thatindicated the prepared: hand condition forone group indicated the prepared: neithercondition for the other group, and vice versa.

As predicted by their hypothesis of cuingstimulus effects, Reeve and Proctor (1984)found faster responses following cuing of thetwo leftmost or two rightmost response keys(a side advantage), even with overlappedhands. Averaging across hand positions toremove effects of cues, there was no overalladvantage for cuing two fingers on the samehand versus cuing two fingers on differenthands. Unfortunately, it is difficult to interpretthis average, because the overall RTs weremuch larger with overlapped than adjacenthands (751 ms vs. 529 ms). With two un-equally difficult response sets, it may be in-appropriate to measure a "true" effect ofhand preparation with an average across them.The effects with overlapped hands may havebeen exaggerated by whatever factor(s) in-creased overall RT (cf. Biederman & Tsao,1979), and a true average hand advantagemay have been concealed by an exaggeratedcounteracting effect in the overlapped condi-tion.

A Reinterpretation of Reeveand Proctor's Results

Despite Reeve and Proctor's (1984) discus-sion, the results of their third experiment donot directly contradict the assumption onwhich the paradigm of Miller (1982) is based.The crux of Reeve and Proctor's (1984) ar-gument is that the preparation effects resultedfrom differential processing of cuing stimuli(e.g., recognition, S-R translation) rather thanfrom differential response preparation. How-ever, the fact that the same cuing stimuliproduced the fastest responses for both handpositions does not prove that the cuing stimuliwere responsible for the speed of those re-sponses,1 nor does it rule out the possibility

that those responses benefited from differen-tial response preparation, consistent with theassumption of Miller (1982).

It is easy to explain all of the obtainedcuing effects (Miller, 1982; Reeve & Proctor,1984) in terms of response preparation, be-cause response preparation could favor dif-ferent pairs of fingers in different conditions.This would allow response preparation toproduce a hand advantage with adjacenthands and a side advantage with overlappedhands. One plausible explanation2 is thatresponses are coded with respect to spatialposition (e.g., left or right of body midline)as well as hand (Nicoletti, Anzola, Luppino,Rizzolatti, & Umilta, 1982; Wallace, 1971)and that preparatory cues are especially usefulif they completely specify a response code.In the condition with overlapped hands, theunnatural arrangement of the fingers mayhave caused subjects to code responses interms of spatial position rather than hand, inwhich case a stimulus indicating responseside would produce a larger preparation effectthan would a stimulus indicating responsehand. With adjacent hands, of course, handand spatial codes are completely confounded,so side cuing would produce a big responsepreparation effect no matter which responsecoding scheme the subjects used. Indeed, thehand-preparation advantage observed withadjacent hands may have resulted as muchfrom cuing of spatial response codes as fromcuing of hands.

Another factor may also have contributedto the tendency for response preparation tofavor different pairs of fingers in differentconditions: Preparation effects may dependon the movements being prepared (cf. Stern-berg, Monsell, Knoll, & Wright, 1978). Itshould be emphasized that overlapping thehands not only reassigns movements to re-

1 Note that it could easily have been just chance thatthe same cues signaled optimal preparation in bothconditions. Because only three different classes of cuingstimuli were compared, this would occur randomly 33%of the time—hardly a remote possibility. Certainly, theco-occurrence cannot be taken as conclusive evidencethat cue processing speed is an important source ofvariation among preparation conditions, let alone theonly important source.

21 am indebted to David Rosenbaum for major con-tributions to this explanation.

226 JEFF MILLER

sponse keys, but also alters the set of responsemovements. With hands overlapped, the fin-gers impede each others' movements, so theact of pressing a single key requires a morecomplicated motion than is needed with ad-jacent hands. Indeed, the data reported byReeve and Proctor show quite clearly thatoverlapping the hands increases response dif-ficulty. Though stimulus conditions wereidentical, RTs averaged 529 ms with adjacenthands versus 751 ms with overlapped hands,a highly significant difference, F(\, 30) =31.37. p < .001). Introspection also suggeststhat different movements are required in thetwo conditions, as the reader can easily verifyby typing the four letters VBNM with thehands in the adjacent and overlapped posi-tions. If overlapping the hands introducednew movements, albeit with the same fingers,it could certainly have introduced new typesof preparation as well, and these new typesof preparation may have favored differentpairs of fingers.

Because response preparation effects caneasily account for the results with both ad-jacent and overlapped hands, the results ofReeve and Proctor's (1984) third experimentmay simply indicate that preparation of twofingers on the same hand does not alwayslead to faster responses than does preparationof two fingers on different hands: specifically,not in the condition with overlapped hands.In Miller's experiments, the hands were alwaysadjacent, however, and the paradigm requiresonly that there be a hand-preparation advan-tage with hands in this position. The paradigmuses the hand advantage with this responseset to determine whether or not various stim-ulus attributes cause preparation, and it doesnot require specific assumptions about themechanisms of preparation responsible forthe advantage. Because the paradigm doesnot depend on the mechanism of preparation,it does not require a hand advantage for allpossible sets of finger responses (i.e., all pos-sible positions and movements), but onlywith the particular responses actually used.Admittedly, the label hand-preparation ad-vantage is poor terminology to the extentthat it suggests the advantage should be ob-tained for all hand positions.

The side advantage with overlapped handsmay have important implications concerning

the motor system even if it does not invalidatethe paradigm of Miller (1982). It stronglysuggests that the hand advantage with adjacenthands is not a consequence of a motor systemrequirement to specify hand first (Reeve &Proctor, 1984; Rosenbaum, 1980, 1983). Thismeans that the hand-preparation advantagedepends on the particular movements ratherthan being universal—a conclusion that alsohas other support. For example, Heuer (1982)showed that preparation of similar movementsis more efficient than preparation of dissimilarmovements, other things being equal. Thus,a normal hand advantage might disappearwith overlapped hands if movements by fin-gers on the same hands became highly dis-similar in this condition. It is important toemphasize, however, that a dependence of thehand advantage on the response set does notcontradict the assumption necessary for theparadigm introduced by Miller (1982), be-cause that paradigm requires only that therebe a hand-preparation advantage with adja-cent hands.

Though response preparation provides areasonable account of the results of Reeveand Proctor (1984) as well as those of Miller(1982), it is also instructive to consider Reeveand Proctor's alternative idea that the cuingstimuli cause the advantages attributed toresponse preparation. One shortcoming ofthis alternative explanation is that it does notaddress the hand advantage observed withother stimulus sets (e.g., BE, BO, ME, MO;s, S, T, and 7"), in which cuing stimuli andpreparation conditions could be counterbal-anced to eliminate cuing stimulus artifacts.Such effects would be expected if the handadvantage results from response preparation,and they are strong evidence that the handadvantage in the cuing paradigm was not anartifact of the cuing stimuli (Rosenbaum,1983). Indeed, no explicit cues were evenpresented in Miller's (1982) later studies.Second, the explanation does not account forthe lack of preparation effects in prepared:finger and prepared: neither conditions, nei-ther of which was faster than the unpreparedcondition. If preparation in these conditionssimply began later due to difficulty in pro-cessing the cues, there should still have beensome preparation with a full second betweencue and stimulus. Third, the cuing explana-

HAND PREPARATION 227

tion also seems incompatible with the timecourse of the hand advantage. If the advantageresults from faster recognition of certain cues,it should develop across preparation intervalslong enough to allow recognition of thosecues but too short to allow recognition of theother cues (assuming preparation effects arenegatively accelerated, as they seem to be inmost cuing studies). However, the hand ad-vantage continues to increase over severalhundred milliseconds, and it seems unlikelythat recognition could be that much fasterfor the side cues than for the other cues. Infact, given the stimulus conditions, it seemsunlikely that recognition was much faster forthe side cues than for the others. The warningsignal, which remained on throughout thetrial, provided reference points that shouldhave made identification of all pairs of cuedpositions relatively easy.

Experiment

For the paradigm of Miller (1982), themost important point in dispute is that, withthe adjacent-hands response set, preparationof two fingers on the same hand leads tofaster responses than does preparation of twofingers on different hands. This experimentwas designed to replicate the hand advantageobtained by Miller (1982) in a situationsatisfying three constraints. First, the experi-ment had to allow counterbalancing of cuesand response preparation conditions, so thatdifferences among preparation conditionscould not be attributed to a cuing stimulusartifact. Second, the mapping of stimuli toresponses had to be at least reasonably com-patible, because cuing effects may changewith incompatible mappings (Goodman &Kelso, 1980). Third, the response set had tobe as similar as possible to that used byMiller (1982), because small differences inthe response set are unlikely to alter therelative effects of preparing different pairs ofresponses.

In this experiment, stimuli were asterisksthat appeared in one of four positions directlyabove, below, to the left of, or to the right ofa fixation point, as shown in the top panel ofFigure 1. As in the earlier experiments, partialadvance information was presented by dis-playing plus signs in two of the four possible

stimulus positions. In this experiment thetwo cued positions were always along one ofthe diagonal sides of the diamond, as shownin the lower panels of Figure 1.

Responses were keypresses with one of theindex or middle fingers. Four response keyson a computer keyboard were chosen tomatch the diamond configuration of the stim-uli and to allow finger positions comparableto those of the conditions with adjacent hands.The top stimulus in the diamond was assignedto a response key in the top row of letters onthe keyboard, such as (Y), the bottom stim-ulus to a key in the bottom row, such as (B),and the left and right stimuli to left and rightkeys in the middle row, such as (G) and (H).Fingers were assigned to response keys in oneof two orientations that were identical exceptfor reversed rotations of the arms and wrists.In the positive hand-diagonal group, the leftmiddle and index fingers were assigned to thetop and left response keys, respectively, andthe right middle and index fingers were as-signed to the right and bottom response keys.In the negative-hand diagonal group, the leftmiddle and index fingers were assigned to theleft and bottom response keys, and the rightmiddle and index fingers were assigned to thetop and right response keys. Thus, two fingersof the same hand were on either the twopositive or the two negative diagonals of thefour response keys arranged in a diamond.

Though these hand positions are not iden-tical to the adjacent position used by Miller(1982), it is unlikely that they are differentenough to alter substantially the relative effectsof preparing different pairs of fingers. Thepositions differ only in the angles of the armsand wrists, and these are unlikely to havemuch influence on the particular fingermovements necessary to press response keys.Relative to the adjacent-hands condition, thesepositions maintain the relations of the fingersto each other, to the palms of the hands, andto the keys. Thus, it is likely that the fingermovements used to produce responses inthese positions are very similar to those withadjacent hands. Note that movements in thecondition with overlapped hands would bemuch different from both of these, becausein that condition the fingers (and hands) actas obstacles to each others' movements.

Across the two hand-diagonal groups, cuing

228 JEFF MILLER

TIME >

Figure 1. Top center frame shows the potential positions of the test stimulus, indicated with asterisks, inrelation to the fixation point (:); next three rows show examples of the sequence of displays on trials withthe lower positive diagonal cue, the upper negative diagonal cue, and the cue for the unprepared condition.(In all three examples, the correct response to the test stimulus is to press the rightmost response key.)

stimulus and response preparation conditionwere completely counterbalanced. When thehands were on the positive diagonal responsekeys, positive diagonal cues enabled prepa-ration of two fingers on the same hand, andnegative diagonal cues enabled preparationof homologous fingers on different hands.When hands were on the negative diagonalresponse keys, the roles of the cuing stimuliwere reversed.

An additional factor in this experimentwas the distance between the hands. In thenear-keys condition, the four keys in thediamond configuration were located togetherat the center of the keyboard. In the far-keyscondition, they were located at the left andright edges of the keyboard. In the far con-dition the configuration of keys within eachsame-hand diagonal was maintained eventhough the keys making up these two diago-nals were much farther apart. It was antici-

pated that this manipulation would influencethe response coding system used by the sub-jects. Coding of responses in terms of spatialposition should have been much more salientin the far condition than in the near condition.Thus, same-hand preparation could benefitfrom advance specification of both positionand hand codes in ihefar condition, but onlyhand codes in the near condition. It was notanticipated that response movements wouldbe changed significantly by this manipulation,because it involved changes only in wrist andarm positions.

Method

Apparatus and stimuli. An Apple 11+ computercontrolled stimulus presentation and recorded responsesand response latencies. Stimuli were plus signs, about 3mm square, presented as light figures on the dark back-ground of the video monitor. The four stimulus positions

HAND PREPARATION 229

were about 12 mm above, below, to the left of, and tothe right of a fixation point in the center of the monitor.Subjects viewed the displays from a distance of about60 cm.

Responses were made by pressing one of four keys onthe Apple keyboard. Four groups of subjects used differentkeys and/or assignments of fingers to keys, as shown inTable 1. The four groups made up a factorial of twohand positions (positive vs. negative diagonal) and twodistances (near keys versus far keys).

Subjects and procedure. Subjects were 36 undergrad-uate students at the University of California, San Diego,who served in the experiment as a requirement of anintroductory psychology class. Each subject served in asingle session of about 45-min duration, and subjectswere tested individually in small rooms with normaloffice illumination.

Each subject was tested in two blocks of trials, eachstarting with 10 warm-up trials that were not recorded.There were 36 different kinds of trials, defined by afactorial of four target stimulus positions (points of thediamond), three cue types (unprepared, prepared: hand,and prepared: finger), and three preparation intervals(400, 1,200, and 3,000 ms). Each type of trial was testedseven times in each block, and errors and trials with RTsof less than 200 ms or more than 2 s were discarded andrerun later in the block. A new random order of trialswas generated for every block.

Each trial began with the appearance of a fixationpoint that remained on throughout the trial. Two positionswere cued 500 ms after the onset of the fixation point.After the appropriate preparation interval, the cues werereplaced by the target stimulus, which remained until aresponse was made. The next trial began about 2.5 safter the response. Subjects were instructed to respondas quickly as possible without making too many errorsand to try to prepare as much as possible by using theinformation provided by the cues.

Results

For each subject, average correct RT andpercentage of correct response (PC) werecomputed for each block, preparation con-dition, and stimulus-onset asynchrony (SOA).Figures 2 and 3 show average RT as a functionof preparation condition and SOA for subjectsin the near-keys and far-keys conditions, re-spectively. Table 2 shows the correspondingPCs. Overall, responses in the prepared: handcondition were 19 ms faster and 1% moreaccurate than those in the prepared: fingercondition.

ANOVAS were computed to compare theprepared: hand and prepared: finger conditionswith respect to both RT and PC. The twobetween-subjects factors, key distance andhand diagonal, were included in the analysis

along with the within-subjects factors of block,preparation type, and preparation interval.In the analysis of RT, significant main effectswere obtained for preparation type (hand vs.finger), F(l, 32) = 14.3, MSe = 2,760, p <.01, and preparation interval, F(l, 32) =14.4, MSe = 1,809, p < .01. The interactionof these two factors was also significant, F( 1,32) = 4.29, MSe = 958, p < .02, reflecting adecrease in the hand preparation advantagewith increasing preparation interval. Asshown in Figures 2 and 3, different patternsof results were obtained in the near-keys andfar-keys conditions. Increases in preparationinterval reduced RT more with far keys thanwith near keys, F(2, 64) = 7.04, MSC = 1,809,p < .01. There was also a significant three-way interaction of preparation interval, keydistance, and hand diagonal, F(2, 64) = 3.77,MSe = 1,809, p < .05. The beneficial effectof increased preparation interval was greaterfor the positive-hand diagonal with far keysand for the negative-hand diagonal with nearkeys. If this significant interaction is not aType I error, it is difficult to explain.

In a comparable analysis of PCs, significanteffects were also obtained for preparationtype, F(l, 32) = 9.1, MSe = 13.2, p < .01;preparation interval, F(2, 64) = 27.9, MSe =6.3, p < .01; and their interaction, F(2, 64) =4.97, MSe = 5.0, p < .02. In addition, re-sponses were significantly more accurate withfar keys than with near keys (98.6% vs.97.2%), F(l, 32) = 6.45, MSe = 31, p < .02.

Discussion

For the paradigm of Miller (1982), thecritical assumption was that preparation oftwo fingers on the same hand produces fasterresponses than does preparation of two fingerson different hands, at least with the responseset used in his experiments. Miller (1982)used the hand-preparation advantage to in-dicate which stimulus attributes enable pre-liminary response preparation, and, for thatpurpose, it is critical that the hand advantageindicate response preparation with the re-sponse set he used.

The present experiment provides furtherevidence for the assumed hand-preparation

230 JEFF MILLER

Table 1Response Key and Response Finger as a Function of Group and Stimulus Position

Group Top

Stimulus position

Bottom Left Right

Positive farKeyFinger

Positive nearKeyFinger

Negative farKeyFinger

Negative nearKeyFinger

(W)Left middle

(Y)Left middle

(P)Right middle

(Y)Right middle

(.)Right index

(B)Right index

(Z)Left index

(B)Left index

(A)Left index

(G)Left index

(A)Left middle

(G)Left middle

(;)Right middle

(H)Right middle

(;)Right index

(H)Right index

advantage with simple keypress responsessimilar to those used by Miller (1982), thussupporting the assumption needed for theearlier paradigm. Subjects were both fasterand more accurate in the prepared: handcondition than in the prepared: finger condi-tion, and cuing stimuli were counterbalancedacross preparation conditions to eliminatethe possibility of an artifact involving thecues. From the overall RTs, it appears thatthe response sets used in this experimentallowed unconfounding of cuing stimulusfrom preparation condition with movementsthat were similar not only to each other butalso to the movements required with adjacenthands (Miller, 1982). Average RTs for thefour placement groups (Key Distance X HandDiagonal) ranged from 506 ms to 542 ms,

and neither of the grouping factors nor theirinteraction approached significance (all Fs <1). These times are similar to those obtainedwith adjacent hands (Miller, 1982; Reeve &Proctor, 1984), suggesting that movementswith this response set are comparable tothose with adjacent hands.

It is clear that the hand-preparation advan-tage is not a universal effect across all responsesets, but rather depends on the exact responseset used. The strongest hand-preparation ad-vantages were obtained in the adjacent-handscondition (Miller, 1982; Reeve & Proctor,1984) and the far condition of the presentexperiment. It is not surprising that thesetwo conditions gave very similar preparationeffects, because they are very similar response

soo

550

500

NEftR KEYS

'UNPREPftRED•PREPORED: FINGER

*—* PREPARED: HOND

HOC 1200 3000PREPARATION TIME IN MSEC

Figure 2. Reaction time (RT) for subjects in near-keyscondition as a function of preparation condition andpreparation interval.

600

550

500

FOR KEYS

UNPREPflREDPREPORED: FINGERPREPARED: HOND

HOO 1200 3000PREPARATION TIME IN MSEC

Figure 3. Reaction time (RT) for subjects in far-keyscondition as a function of preparation condition andpreparation interval.

HAND PREPARATION 231

Table 2Percentage Correct as a Function of Key Distance, Preparation Condition,and Preparation Interval (in ms)

Interval

Preparationcondition

UnpreparedPrepared: HandPrepared: Finger

400

98.798.097.0

Near keys

1,200

98.399.499.2

3,000

97.999.098.9

400

96.897.294.3

Far keys

1,200

97.499.197.4

3,000

96.797.997.4

sets, differing only in slightly different orien-tations of the wrists and arms. However, asReeve and Proctor (1984) have shown, over-lapping the hands reverses the advantage forhand preparation, possibly because it makesresponses much more difficult or changes theway they are coded. The present results in-dicate that even the spatial separation of thehands can modify the hand-preparation ad-vantage, because it was reduced in the nearcondition. This finding supports the hypoth-esis that spatial coding of responses has animportant influence on the relative efficiencyof different kinds of preparation, possiblyeven more important than hand coding ofresponses. Thus, all of the preparation effectsobtained in the present study and two pre-vious ones can be explained by assumingthat response preparation depends primarilyon the match between the cue and a spatialcode for a subset of the responses. In theprecuing studies, spatial response codes werespecified by presenting two stimulus alterna-tives, and in the other studies (Miller, 1982,Experiments 2-8) spatial response codes werespecified by an easily recognized attribute ofthe stimulus. The assumption that preparationeffects depend on cuing a spatial responsecode is completely consistent with the para-digm of Miller (1982), because that paradigmdid not assume any particular mechanismfor producing the hand-preparation advan-tage.

It should be emphasized that the lack of auniversal hand advantage does not contradictthe assumption needed for the paradigm usedby Miller (1982), because that paradigm re-quires only that there be a hand advantagewith the response set used. The reversal ofthe hand advantage with overlapped hands

does support an interesting claim about motorpreparation, however. It appears that fingerresponses can sometimes be prepared even ifthe responding hand has not yet been speci-fied, as noted above.

An important difference between this studyand the previous ones is that there was alarger overall preparation effect (i.e., benefitfor the prepared conditions relative to theunprepared.) One possible explanation is thatthe cues produced more perceptual facilitationin the present experiment. Alternative stim-ulus locations were much farther from thefixation point than in the previous studies,and cues might have allowed the perceptualsystem to prepare for certain of these locations(cf. Posner, 1980; Posner, Snyder, & Davidson,1980).

The overall advantage for preparation ofhand as opposed to preparation of finger wassomewhat smaller in this study than in thatreported by Miller (1982), but this differencemay be a simple function of procedural dif-ferences. One procedural difference is thatMiller (1982) used only preparation intervalsof 1 s or less, the times producing the greatesthand-preparation advantage in the presentexperiment. It is also possible that the slightdifferences in arm and wrist placement causedthe slight difference in preparation effects.Finally, the cuing artifact noted by Reeve andProctor (1984) may have inflated the hand-preparation advantage in the previous study.

One of the most interesting results of thisexperiment was that there was more prepa-ration in the far-keys condition than in thenear-keys condition. This greater preparationwas reflected in a greater hand-preparationadvantage, a greater effect of preparationinterval, and a greater difference between

232 JEFF MILLER

unprepared and prepared conditions. Becausethe most obvious correlate of the distancebetween hands is the extent to which theresponse fingers are perceived as two separategroups,3 this result suggests that responsepreparation depends on a match betweencuing information and a spatial responsecode. However, spatial separation may alsoincrease subjects' awareness of the subgroupsof responses, thereby encouraging them toadopt strategies that channel more effort intoresponse preparation. Others have also sug-gested strategic components to various typesof preparation. Requin (1980) and Sanders(1983) have emphasized the effortful natureof preparing to respond, arguing that it isvoluntary, time consuming, briefly main-tained, and difficult to combine with otherprocessing. To the extent that response prep-aration is effortful, of course, it is also subjectto strategic influences. Kantowitz and Sanders(1972) pointed out that in many experiments,like the present one, subjects can respondcorrectly without processing the partial ad-vance information at all. Their results indicatethat subjects prepare for specific stimulusattributes only when the task makes it con-venient to do so. Similarly, the results ofReeve and Proctor's (1984) second experimentsupport the notion that preparation has animportant strategic component.

In conclusion, the results of this experimentprovide further evidence that the hand-prep-aration advantage assumed by Miller (1982)is real with the response set he used. Themost direct implication of this advantage isthat the inferences of Miller (1982) were valid,or, at least, not invalid because of erroneousassumptions about preparation. Thus, the re-sults provide reassurance concerning Miller'sconclusions about discrete and continuousinformation processing models.

3 Distance could also influence the response movementsthemselves, but this seems unlikely because overall RTdid not depend on distance.

References

Biederman, I., & Tsao, Y. (1979). On processing Chineseideographs and English words: Some implications from

Stroop-test results. Cognitive Psychology, 11, 125-132.Brinkman, J., & Kuypers, H. (1973). Cerebral control of

contralateral and ipsilateral arm, hand, and fingermovements in the split-brain rhesus monkey. Brain,96, 653-674.

Garner, W. R. (1974). The processing of information andstructure. Potomac, MD: Erlbaum.

Gazzaniga, M. S. (1970). The bisected brain. New York:Appleton-Century-Crofts.

Goodman, D., & Kelso, J. A. S. (1980). Are movementsprepared in parts? Not under compatible (naturalized)conditions. Journal of Experimental Psychology: Gen-eral, 109, 475-495.

Heuer, H. (1982). Binary choice reaction time as acriterion of motor equivalence. Acta Psychologica, 50,35-47.

Kantowitz, B. H., & Sanders, M. S. (1972). Partialadvance information and stimulus dimensionality.Journal of Experimental Psychology, 92, 412-418.

Kornblum, S. (1965). Response competition and/or in-hibition in two-choice reaction time. PsychonomicScience, 2, 55-56.

McClelland, J. L. (1979). On the time relations of mentalprocesses: A framework for analyzing processes incascade. Psychological Review, 86, 287-330.

Miller, J. O. (1982). Discrete versus continuous stagemodels of human information processing: In searchof partial output. Journal of Experimental Psychology:Human Perception and Performance, 8, 273-296.

Miller, J. O. (1983). Can response preparation beginbefore stimulus recognition finishes? Journal of Exper-imental Psychology: Human Perception and Perfor-mance, 9, 161-182.

Nicoletti, R., Anzola, G., Luppino, G., Rizzolatti, G., &Umilta, C. (1982). Spatial compatibility effects on thesame side of the body midline. Journal of ExperimentalPsychology: Human Perception and Performance, 8,664-673.

Posner, M. I. (1978). Chronometric explorations of mind.Hillsdale, NJ: Erlbaum.

Posner, M. I. (1980). Orienting of attention. QuarterlyJournal of Experimental Psychology, 32, 3-25.

Posner, M. I., Snyder, C., & Davidson, B. (1980). Attentionand the detection of signals. Journal of ExperimentalPsychology: General, 109, 160-174.

Reeve, T. G., & Proctor, R. W. (1984). On the advancepreparation of discrete finger responses. Journal ofExperimental Psychology: Human Perception and Per-formance, 10, 541-553.

Requin, J. (1980). Toward a psychobiology of preparationfor action. In G. E. Stelmach & J. Requin (Eds.),Tutorials in motor behavior (pp. 373-398). Amsterdam:North-Holland.

Rosenbaum, D. A. (1980). Human movement initiation:Specification of arm, direction, and extent. Journal ofExperimental Psychology: General, 109, 444-474.

Rosenbaum, D. A. (1983). The movement precuingtechnique: Assumptions, applications, and extensions.In R. A. Magill (Ed.), Memory and control of action(pp. 231-274). Amsterdam: North-Holland.

HAND PREPARATION 233

Rosenbaum, D. A., & Kornblum, S. (1982). A priming control and learning (pp. 117-152). New York: Aca-method for investigating the selection of motor respon- demic Press.ses. Acta Psychologica, 51, 223-243. Wallace, R. (1971). S-R compatibility and the idea of a

Sanders, A. F. (1983). Towards a model of stress and response code. Journal of Experimental Psychology,human performance. Acta Psychologica, 53, 61-97. 88, 354-360.

Shaffer, L. H. (1975). Control processes in typing. Quar- wi,ey> j (1975) Hemispherectomy and consequent motorterly Journal of Experimental Psychology, 27, 419- behavior. In W. Spirduso & J. King (Eds.), Proceedings:432- Motor control symposium. Austin, TX: University of

Shaffer, L. H. (1981). Performances of Chopin, Bach, Texas Pressand Bartok: Studies in motor programming. Cognitive ^^ R (19gl) Jhe ̂ ^ force and

SteST0969). Memory scanning: Mental processesrevealed by reaction-time experiments. American Sci-enlist, 57, 421-457. '

Steinberg, S., Monsell, S., Knoll, R. L., & Wright, C. E.(1978). The latency and duration of rapid movementsequences: Comparisons of speech and typewriting. In Received June 21, 1984G. E. Stelmach (Ed.), Information processing in motor Revision received October 31, 1984

Instructions to Authors

Authors should prepare manuscripts according to the Publication Manual of the AmericanPsychological Association (3rd ed.). All manuscripts must include an abstract of 100-150 wordstyped on a separate sheet of paper. Typing instructions (all copy must be double-spaced) andinstructions on preparing tables, figures, references, metrics, and abstracts appear in the Manual.Also, all manuscripts are subject to editing for sexist language. For further information oncontent, authors should refer to the editorial in the February 1982 issue of the Journal (Vol.8, No. 1, pp. 171-172). For information on the other three JEP journals, authors should referto editorials in those journals.

APA policy prohibits an author from submitting the same manuscript for concurrent con-sideration by two or more journals. APA policy also prohibits duplicate publication, that is,publication of a manuscript that has already been published in whole or in substantial part inanother publication. Authors of manuscripts submitted to APA journals are expected to haveavailable their raw data throughout the editorial review process and for at least 5 years afterthe date of publication.

Anonymous reviews are optional, and authors who wish anonymous reviews must specificallyrequest them when submitting their manuscripts. Each copy of a manuscript to be anonymouslyreviewed should include a separate title page' with authors' names and affiliations, and theseshould not appear anywhere else on the manuscript. Footnotes that identify the authors shouldbe typed on a separate page. Authors should make every effort to see that the manuscript itselfcontains no clues to their identities.

Manuscripts should be submitted in quintuplicate (the original and four photocopies), andall copies should be clear, readable, and on paper of good quality. Authors should keep a copyof the manuscript to guard against loss. Mail manuscripts to the Editor, William Epstein,Department of Psychology, University of Wisconsin, W. J. Brogden Psychology Building, 1202West Johnson Street, Madison, Wisconsin 53706.

For the other JEP journals, authors should submit manuscripts (in quadruplicate) to one ofthe editors at the following addresses: Journal of Experimental Psychology: General, SamGlucksberg, Department of Psychology, Princeton University, Princeton, New Jersey 08544;Journal of Experimental Psychology: Learning, Memory, and Cognition, Henry L. Roediger III,Department of Psychology, Purdue University, West Lafayette, Indiana 47907; and Journal ofExperimental Psychology: Animal Behavior Processes, Michael Domjan, Department of Psy-chology, University of Texas, Austin, Texas 78712. When one of the editors believes a manuscriptis clearly more appropriate for an alternative journal of the American Psychological Association,the editor may redirect the manuscript with the approval of the author.