TH E QUARTE R LY JOU RNAL O F E XPE R IM E NTAL PS YCH OL OG Y, 1996 , 49A (4), 950 ± 971

A Referen tial-codin g E xp lanation for C om patibility E ffects of Physically O rth ogonal Stim ulus and Respon se D im ensio ns Yvonne L ippa M a x-P la nck-I nstitut fuÈr psychologische Forschung, M uÈnchen, Germa ny

T h is stu d y ad d re sses th e d epen d en ce of com pa tibility effects o n resp on d ing h an d w ith ho rizo n tally orie nted stim uli an d ver tically or ien ted resp on ses (H ± V effect) an d w ith vertically orien ted stim uli and h oriz on tally or ien te d resp on ses (V ± H effect) rep orted by B au er an d M iller (1982). E xp erim en t 1 rep licated th e H ± V effect. In E xp erim en t 2, th e su bject w as instru cte d to resp o n d w ith th e ha nd in lin e w ith th e resp on se keys. T h at elim inated the H ± V effect. In E xp erim ent 3, th e resp on se board w as p laced to th e left or righ t sid e of th e su bject, yielding a con sid erably red u ced H ± V effect an d a n ovel com p atibility effect d ep end en t on board loc atio n. In E xp erim en t 4, th e V ± H effect w as p ro d u ced w he n th e su bject w as r eq uired to resp on d w ith th e h an d in lin e w ith th e respo ns e keys. W ith the h an d r otated th ro u gh 90 8 in E xp erim en t 5, th e V ± H effect w as elim inated , an d a m ain effect of m ap p ing w as observ ed . T h e resu lts cha llen ge B au er and M iller’ s m ovem en t-preferen ce hy p othesis an d su pp ort a re feren tial-cod ing h ypo th esis p ro p os ed by th e auth or. T h is assu m es th at resp on se p osition s are cod ed in referen ce to h an d p ostu re, so th at p h ysically or th ogo na l stim u lus an d resp on se dim en sion s can overlap w ith resp ect to their m en tal rep resen tation s. T h e ap plicability of th is hy p oth esis to oth er com p atibility effec ts is d em o nstr ated , and its sign i® can ce for com p atibility th eories is brie¯ y d iscussed .

A ccord ing to F itts an d Seeger (1953), stim ulus± respo nse com patibility (SRC ) m ean s that the quality of a particular resp onse to a particu lar stim ulus, in ter m s of reaction tim e an d er rors, depen ds not on stim ulus or response characteristics alone, but on their interrelation. Fo r instan ce, if subjects respo nd to left an d right stim uli by pressing a left an d a right respo nse key, respectively, faster respo nses an d few er erro rs are observed if the left

Requests for re prints sh ould be se nt to Yvo nne L ip pa, M ax-Plan c k-Institut fuÈ r psychologisc he Forsc hung, Leopoldstr. 24, D-8080 2 M uÈ nche n, Ge rm an y. E m ail: lippa@ mp ip f-muenc hen.m pg.de I wish to thank Bernhard H omm e l, G isa Asc hersleb en, Wolfgan g Prinz, Stephen M on sell, C lair e M ichaels, an d two an onymo us reviewe rs for help ful su ggestion s and comm ents on earlie r versio ns of this article ; Fiorell o Ban ci and K arl- H einz Ho nsbe rg for constructin g the re spo nse devic e, as well as He id i John an d Jo nathan H arrow for checking an d im provin g the En glish. q

199 6 The E xp erim ental Psyc holo g y S ociety

S ± R C O M P A T IB IL IT Y

951

key is assigned to th e left stim ulus an d the right key to the right stim ulus than the co nverse. U sually, SRC effects have been observed under experimental conditions sim ilar to those d escribed, in w hich stim ulus an d respon se d im ensions have physical or sem an tic features in com m o n (e.g. left± right stim ulus position s an d left± right responses). T his sim ilarity of stimulus an d respo nse dim en sions, in so far as it is cognitively coded, is called the dimensiona l overla p of an S± R ensem ble. M ost theories of S± R com patibility rely on th e dimensio nal overlap of S± R sets. Fo r instance, au tom atic activation accounts (D e Jon g, L ian g, & L au ber, 1994; K or nblum , H asbroucq, & O sm an , 1990) assum e th at SRC effects occur b ecau se activated stim ulus cod es au tom atically prime cong ruent (i.e. sp atially or sem an tically sim ilar) response codes, w hich p resup poses som e kind of dimension al overlap. L ikew ise, tran slation accounts argue that the ef® ciency of S± R tran slatio n critically depends on S ± R correspo ndence: Few er an d sim pler tran slatio n rules are need ed in the presence of dim ension al overlap than in its ab sence (H asbroucq, G uiard, & O ttom an i, 1990; U m iltaÁ & N icoletti, 1992; Welford, 1976). T hus, ap art from d ifferences in exp laining the underlying m echan ism s, m ost theories assum e th at overlapping S± R dimensio ns are a necessary precondition for SRC effects to occu r. A gainst this backg round, it is interesting to n ote that S RC effe cts have been ob served w ith p hysically orthogona l stim ulu s an d resp onse dimensio nsÐ for exam ple, w ith top± bottom stim ulus positions an d left± right responses (Bau er & M iller, 1982; C otton, Tz eng, & H ardyck, 1977, 1980; E hren stein, Schroeder-H eister, & H eister, 19 89; L aÁ davas, 1987; L aÁ davas & M oscovitch, 1984; M icha els, 1989; M ichaels & Schild er, 19 91; Weeks & Proctor, 1990; Weeks, Proctor, & B ey ak, 1995). T his raises the follow ing theoretical problem . O n the one han d, o ne m ight assum e that physically orthogonal S ± R dim ensions are cognitively coded orthogonally, too. If so, there w ould be no dimensional overlap between stim ulus an d response codes, an d SRC effects w ith physically orthogonal S± R sets w ould have to be exp lained in som e other w ay. In fact, prior accounts have attributed them to characteristics of the m otor system (Bau er & M iller, 1982), to the cor respondence between the salient polar referents of the S± R sets (Weeks et al., 1995 ), or to cerebral hem isp heric processing (C otton et al., 1977, 1980). O n the other han d, one could assum e that despite physical orthogo nality, S± R dimensio ns can overlap w ith respect to their m ental representations. In this case, the effects could be easily accounted for by existing co m patibility theories. Bu t then the question is how non-orthogonal codes are derived from physically orthogonal dimensio ns. T he purpose of th e p resent study is to give an an swer to the latter question. A referential-coding hypothesis is presented w hich assum es that responses are coded in reference to han d posture an d can therefore b e cognitively represented non -orthogonally w ith respect to the stimulus dimensio n. A ccordin gly, m ost effects w ith physically orthogonal stim ulu s an d response dim ensions require no special explan atio n but can be interpreted w ith in the sam e fram ew ork as the classic SRC effects. In order to test the referential-coding hypothesis, the present stu dy focuses on two SRC effects w ith orthogonal S ± R dim ensions that w ere ® rst reported by Bau er an d M iller (1982) an d ser ved as a starting point for m an y other studies: an SRC effect w ith horizontally oriented stim uli an d vertically oriented resp onses (the H ± V effect) an d an SRC effect w ith vertically oriented stim uli an d horizontally oriented respo nses (the V ± H

952

L IP P A

effect). T he d irectio n of these effects depends o n the han d that is u sed. A ltho ugh b oth effects w ill eventually be reg arded as two variants of the sam e effect, they are ® rst treated separately. T he ® rst three experim ents ad dress the H ± V effect; E xperim ents 4 an d 5 focus on the V ± H effect.

T H E H ± V E FFE C T In Bau er an d M iller’ s (1982) ® rst experiment, subjects respo nded to an ``X ’ ’ presented to the left or right of a central ® xatio n point by p ressing a response key located behin d (upp er key) an d in front of (lower key) a central hom e key w ith their left ind ex ® nger or, in a different block, their right index ® n ger. Apart from the han d, S± R m ap ping w as varied: O ne-half of the subjects respon ded w ith the upper key to left stim uli an d w ith the low er k ey to right stimu li (left± up/right± dow n m ap ping), w hereas the other half w ere given the reversed m ap ping (left± dow n/right± up m ap ping). T he results w ere rather su rprising: W ith left± up/right± d ow n m ap ping, a reaction-tim e advan tage of 34 m sec w as ob served for the left com pa red to th e right han d, w hereas w ith left± dow n/righ t± up m ap ping there w as an advan tage of 84 m sec for the right com p ared to the left h an d . In other w ords, different S RC effects ap peared for left- an d right-h an d resp onses. In order to en sure th at this H ± V effect is reliable an d occurs under the experim ental con ditio n used here, it w as ® rst of all replicated in E xp eriment 1.

E X PE R IM E N T 1 M eth o d Subjects E leven fem ale an d ® ve m ale su bjects aged b etw een 20 an d 39 w ere p aid to p ar ticipate in th e experim en t. A ll claim ed to be righ t-h an d ed an d all h ad no r m al or corrected -to-n or m al vision .

Stim uli and App aratus Stim ulu s p resentation an d d ata acqu isition w ere co ntrolled by a Rh otron V M E sy stem . T h e stim uli, stand ard u p p er-case ``X ’ ’ s (0.3 3 0.4 cm ) in black o n a w h ite backg ro u n d, w ere presen ted on an A tari S M 124 m o nitor. T h ey ap p eared w ith th eir centres 1.5 cm to th e left or righ t of a cen tr al ® xation asterisk. T h e view ing distan ce w as ap p roxim ately 60 cm . Respo n se s w ere m ad e on a h orizon tal 31 3 31-cm r esp on se board interfaced w ith the com p u ter. T h ree squ are sen so r keys w ere m oun ted on th e board : a cen tral h om e key (2.5 3 2.5 cm ) an d tw o respo ns e keys (3 3 3 cm ) located 2.7 cm (e dg e-to-edg e) in fro nt of an d beh ind the h om e key relative to th e s ub ject’s body. T h ese are refer red to, from th e su bject’ s p ersp ective, as th e ``low er’ ’ an d ``u p p er’ ’ key, respectively.

S ± R C O M P A T IB IL IT Y

953

Procedure Su bjects w ere seated at th e table o n w hich th e m on itor an d th e resp on se board w ere located an d w ere requ ired to keep th eir sa gittal m idlin e in lin e w ith the resp o n se keys an d the ® xation po int (see F igu re 1). T h e resp on se board w as 3 cm abov e table heigh t an d close to the ed ge of th e table. T h eir task w as to m ove th e ind ex ® ng er from th e h om e key to tou ch the u pp er or low er key in respo ns e to left or righ t stimu li. H alf th e su bjects resp on d ed to a left stim ulu s w ith th e up p er key an d to a righ t stim u lus w ith the low er key, w h ereas th e oth er h alf u sed the rever se m ap p ing. T he sessio n w as d ivided into tw o blocks of 90 trials con sistin g of 10 p ractice trials an d 80 test trials (40 for each stim ulu s location ). In on e block of tr ials su bjects used th e left ind ex ® n ger, an d in th e oth er block th e righ t index ® n ger on ly. T he ord er of h an d s w as balan ced acro ss su bjects. E ach trial beg an w ith a d isp lay of the En glish com m an d ``H O M E K E Y !’ ’ in the cen tre of th e screen, u ntil the su bject tou ch ed th e h om e key w ith th e ind ex ® n ger of th e res po nd ing h an d . A fter 100 m sec, a ® xatio n asterisk w as presen ted for 1 sec, followed by the stim ulus d isplayed r an d om ly to the left or righ t of th e ® xation asterisk. T h e stim ulu s w as visible u n til th e su bject resp on d ed or 1 sec h ad pa ssed (m issin g trial). If su bjects left the h om e key before stim ulu s p re sen tatio n , th e trial w as restarted w ith the ``H O M E K E Y ’ ’ disp lay. Au d itory feedback (a beep) w as given o n erro r an d m issin g trials (0.6% ), w h ich w ere th en rep eated at r an d om p os ition s in th e block.

R esu lts M ean reaction tim e (RT ), that is, the tim e n eeded to leave the hom e key, m ean m ovem ent tim e (M T ), that is, th e tim e from m ovem ent o nset to eng agem ent o f the respo nse key, an d percentages of er ror were calculated as a function of stim ulus an d respo nse location an d responding han d (see Table 1). Trials w ith RTs of less than 120 m sec (0.3% ) were excluded

FIG . 1. Left pan el: The hand posture of the le ft an d the rig ht han d, de® ning upper and lowe r re sp onse key positio ns as left and rig ht. Right pan el: The H ± V effec t in E xperim ent 1: total tim e an d erro r rate as a functio n of m apping an d respo ndin g hand.

954

L IP P A TABLE 1 a a b c M e an R T , M T , T T , a n d E rro r D ata for E xp erim e n t 1 as a F u n c tio n o f R e s p o n d in g H a n d an d S tim u lu s a n d R es p o n s e L o c a tio n

Left H a nd

Right H a nd

S timulus Loca tion

S timulus Loca tion

R esponse Loca tion

Left

Left

Up

RT MT TT Errors

286 131 417 1.8

331 136 467 10.6

33 2 13 0 46 2 7.3

303 127 430 2.1

D ow n

RT MT TT Errors

314 132 446 7.6

288 111 399 1.5

29 5 11 5 41 0 1.4

350 110 460 7.5

a

b

Right

R ight

c

In m sec; tota l time in m sec; in % .

from calculation . A s th e experim ental situation allow ed subjects to lift th eir ® n gers b efore the decision p rocess or m ovem ent prog ram m ing w as com plete, the relevan t SRC effects m ay b e d istrib uted over RT an d M T d ata. T herefore, total tim e (T T )Ð that is, the sum of RT an d M T Ð w as cho sen as th e primary m easure, an d separate an alyses of RT an d M T are reported only w hen their outcom es reveal relevan t deviatio ns from the m ain an alysis. A nalyses of variance (A N OVA) w ere condu cted on the d ata, using the between-subject variable S± R m ap ping (left± up/righ t± dow n vs. left± dow n/rig ht± up) an d the w ithin-subject variables response han d (left vs. right) an d response location (upward vs. d ow nw ard). Tota l Times. T he m ain effect of response location reached signi® can ce, F(1, 14) = 4.9 2, p < .05, indicating that dow nw ard responses w ere faster (42 9 m sec) th an upward responses (444 m sec). In ad dition, the relevan t M ap p ing 3 H an d interactio n w as reliable, F(1, 14) = 17.95, p < .01. A s Figure 1 show s, w ith the left han d left± up/right± dow n m ap ping w as easier (408 m sec) than left± dow n/right± up m ap ping (457 m sec), w hereas the p reference w as reversed w ith th e right hand (461 vs. 420 m sec). E rror Da ta . O nly the M ap ping 3 H an d interactio n reach ed signi® can ce, F(1, 14) = 60.72, p < .001 (see Figure 1). L eft-han d respo nses w ere less error-prone w ith left± up / right± dow n m ap ping (1.7% ) than w ith left± dow n/right± up m ap ping (9.1% ); for righthan d responses, the opp osite patter n occur red (7.4 vs. 1.8% ).

D iscussio n In sum , the H ± V effect is replicable in total tim es as w ell as in error d ata. U nlike Bau er an d M iller’ s study (1982, E xp. 1), the experim ent yielded a cross-over in teraction. T hat is, in ad dition to a difference in perfor m an ce for the two han ds across m ap pings, there w as also a difference in perfor m an ce for the two m ap pings across han ds. T his difference

S ± R C O M P A T IB IL IT Y

955

is due to a no n-replicable overall reactio n tim e advan tage of 57 m sec, found by B au er an d M iller fo r the left± up/ right± dow n m ap ping. But there is no obviou s explanation for it. Studies that report a m ain effect of orthogonal S± R m ap pings (e.g. Weeks & Proctor, 1990 ) found only a superiority of up± right/dow n± left m ap ping. T he follow ing sectio ns present two different explanations for the H ± V effect: the m ovem ent-preference hypothesis proposed by B au er an d M iller (1982) an d the referential-coding hypothesis proposed by the p resent au thor.

Th e M o vem en t-preferen ce Hy po thesis Bau er an d M iller (1982) presum ed that the structure o f the m otor system is responsible for th e H ± V effect. S peci® cally, they postulated that each m ovem ent in the H ± V design is not a sim ple ver tical m ovem ent, but actu ally a com b ination of two m ovem ents: T he ® rst (im p licit) m ovem ent aim s at the relative positio n of the stim ulus (left vs. rig ht), in the sense of responding tow ards the source of stim ulatio n (Simo n, 1969); the second (explicit) m ovem ent is the o ne requ ired by instructio n (upw ard vs. dow nw ard). C om bining these two m ovem ents resu lts either in a joint clockw ise (e.g. ® rst left, then upw ard) or in a joint co un terclockw ise (e.g. ® rst left, then dow nw ard) m ovem ent. T he H ± V effect is explained by the assum ptio n that clockw ise m ovem ents, h ere cau sed by left± up /right± dow n m ap ping, are preferable for the left rather than the righ t han d, w hereas counterclockw ise m ovem ents, cau sed by left± dow n /right± up m ap ping, are preferable for the right rather than the left han d. Bau er an d M iller em ph asized that these m ovem ent p references are n ot ® xed ch aracteristics of each han d that occur independently of the experim ental situatio n. In several pilot studies, they stud ied the speed of simple rotational m ovem ents for each h an d, b ut they found no interactio ns between response han d and direction of rotation. T herefore, they co ncluded that ``these preferences are m an ifest through com binatio ns of implicit m ovem ents tow ards the stim ulus an d explicit m ovem ents tow ards the respo nse key’ ’ (p. 367). T h us, w hen exam ining this hypothesis, it is im portant to m aintain an experimental situation that evokes an im plicit an d an ex plicit m ovem ent. A s long as this is present, the m ovem ent preferences should b e effective, an d a H ± V effect com parable to that ob served in E xp eriment 1 should occur.

Th e R eferentia l-co ding H yp oth esis A ccord ing to the referential-coding hypothesis proposed here, th e H ± V effect is a simple sp atial SRC effect. Speci® cally, it is assum ed that physically ``upw ard’ ’ and ``dow nw ard’ ’ responses are coded as ``left’ ’ an d ``right’ ’ responses an d thus show dimensional overlap w ith the horizontally oriented stim ulus set. To understand th is assum ption, one has to re-exam ine th e experim ental situatio n that generates the H ± V effect. Subjects are in structed to respo nd w ith the index ® nger of one han d. A s no special instructio n is given about han d posture, su bjects probably ho ld th eir han d at a com fortable an gle (about 45± 90 8 ) to the line of the response keys (see F igure 1). If so, the han d (i.e. the intrinsic ® ngertip-to-w rist axis) m ight function as a reference

956

L IP P A 1

point d e® n ing the position of the response keys . T hat is, everything located to the left of the han d m ay be cod ed as LE FT an d everyth ing located to the right m ay b e coded as R IG H T. In left-h an d resp onding, the upper key w ould be coded as L E FT an d the low er key as R IG H T. C onversely, w ith right-han d respond ing, th e upp er key would b e coded as R IG HT an d the lower key as L E FT. If this is the case, the occurrence of the H ± V effect is no longer surprising. In left-han d responding, left± up/ right± dow n m ap ping is com patible, because stim ulus an d coded response position s cor resp ond, w h ereas the left± dow n/right± up m ap ping is incom patible, as stim ulus an d coded respon se position s do not cor respond. W ith right-h an d responses, the opp osite coding is expected, so that left± dow n/ right± up m ap ping is com patible an d left± up/righ t± dow n m ap ping is incom patible. In sum , the referential-coding hypothesis assum es that th e han d functions as a left± right reference, so that upw ard an d dow nw ard respon ses are coded as left an d righ t, thus sh ow ing d im ensional overlap w ith the stim ulus set. G iven standard han d postures, it follow s that the m ental representatio n of respo nse d irectio n depends on w hether it is the left or the right han d that respo nds. In the case of left-han d responses, the upper key is coded as L E F T an d the low er key as RIG HT, w hereas the op posite coding is valid for righthan d respo nses. T his explains w hy the SRC effect depends on the responding han d. T he follow ing two experim ents w ere carried out to test the referen tial-coding hypothesis ag ainst the m ovem ent-preference hypothesis.

E X PE R IM E N T 2 T he referen tial-coding hypothesis exp lains the H ± V effect by proposing th at th e positioning of the han d at a 45± 9 0 8 an gle to the line of the response keys cau ses the left± right co ding of the respon se set. T herefore, it should be possible to elim inate the H ± V effect, or at least to reduce it con siderably, by m an ipulating han d posture. T his prediction w as tested by instructing su bjects to hold their respo nding hand an d ® nger in line w ith the response keys, so that the keys were above an d b elow the han d an d, thus, sh ould be coded as T O P an d BO TT O M (see F igure 2). Reg ardless of th e responding han d, left± up/righ t± dow n m ap ping should now lead to sim ilar perfo r m an ce to left± dow n /right± up m ap ping. In contrast, the m ovem ent-pre ference hypothesis predicts an H ± V effect com p arable to the on e obtained in E xp erim ent 1, because the experim ental situation still evokes the co m bination of an implicit an d explicit m ovem ent, w h ich is n ecessary for the m ovem ent preferences to be effective. So, w ith clo ckw ise m ovem ents, better p erform an ce for the left han d is expected, an d w ith cou nterclockw ise m ovem ents, better perfor m an ce for the right han d.

1

E videnc e for a refe renc e function of the han d w as found by C arpenter an d Eisenb erg (1978 ) in a haptic le tter-judgem ent task . They sh owe d that blindfolded su bjects coded the orientatio n of a le tter w ith respec t to hand positio n. A letter pre sented at 0 8 was id enti® ed fastest whe n the su bject’s hand and ar m were perpe ndic ular to the sub ject’ s fronta l plane, w hereas letters presented at 300 8 were id enti® ed fastes t with the forear m positio ned at a 300 8 an gle to the su bjec t’s frontal plane.

S ± R C O M P A T IB IL IT Y

957

FIG . 2 . L eft panel: The hand postur e req uired in Expe rim e nt 2 for le ft- and right-han ded respon ses. R ight pane l: T he abse nce of the H ± V effe ct in Experim ent 2: total tim e an d er ror rate as a function of m ap ping an d respo nding hand.

M eth o d Subjects Seven fem ale an d ® ve m ale su bjects (11 righ t-h an d ed an d 1 am b idextrou s by self-rep ort) aged betw een 16 an d 38 w ere p aid to p ar ticipate in th e exp erim en t. T h ey h ad n or m al or cor rected -ton or m al visio n.

App aratus and Procedure T h ese w ere th e sam e as in E xp erim en t 1, w ith th e follow ing e xcep tio ns: ® r st, in add ition to er ro r an d m issing trials (0.5% ), trials w ith RTs less th an 120 m sec (0.2% ) w ere re peated at r an d om p osition s in th e bloc k. S eco n d, su bjects w ere re quired to h old th eir resp on d ing h an d in lin e w ith the respo n se keys. To avoid un com fortable h an d po stu res, su bjects w ere allow ed to sit ap p roxim ately 1 m aw ay from the screen . T he su bjec t’ s sagittal m id-lin e an d th e h an d were then in lin e w ith th e respo ns e ke ys an d th e ® xation asterisk.

R esu lts an d D iscu ssion RT, M T, an d error data were treated an d an alysed as in Experim ent 1, an d their m ean s are presented in Table 2. Tota l Times. O n ly the m ain effect of response location w as reliable, F(1, 10) = 6.23, p < .05, indicating that d ow nw ard resp onses were faster (446 m sec) than upward responses (466 m sec). T h e relevan t M ap ping 3 H an d interaction w as far from being signi® can t, F < 1 (see F igure 2).

958

L IP P A TABLE 2 a a b c M e an R T , M T , T T , a n d E rro r D ata for E xp erim e n t 2 as a F u n c tio n o f R e s p o n d in g H a n d an d S tim u lu s a n d R es p o n s e L o c a tio n

Left H a nd

Right H a nd

S timulus Loca tion

S timulus Loca tion

R esponse Loca tion

Left

Left

Up

RT MT TT Errors

345 118 463 5.3

341 150 491 12.1

34 1 11 3 45 4 5.0

325 132 457 6.5

D ow n

RT MT TT Errors

340 129 469 4.4

338 89 427 4.2

34 4 11 7 46 1 2.4

343 82 425 2.0

a

b

Right

R ight

c

In m sec; tota l time in m sec; in % .

E rror Da ta . O nly the m ain effect of respo nse locatio n ap proached reliability, F(1, 10) = 3.9 0, p = .077, in dicating that dow nw ard resp onses w ere less error-pro ne (3.3% ) than were upward respon ses (7.2% ). A g ain, the relevan t M ap ping 3 H an d interactio n w as far from being sig ni® can t, F < 1 (see Figure 2). In sum , E xperim ent 2 revealed no M ap ping 3 H an d interactio n. T his m eans that the H ± V effect is absent w hen the reference point provided by th e responding han d is such as to encourage an upw ard ± dow nw ard coding of the response direction that does not correspond to the left± righ t stim ulus dim ension. T his supp or ts th e proposal that it is the reference provided by the responding han d that is critical for the H ± V effect an d not the m ovem ent p reference of each han d .

E X PE R IM E N T 3 E xperiment 3 w as carried ou t to test the referential-coding explanatio n of w hy the H ± V effect depen ds o n the respo nding han d. It w as assum ed that in E xperim ent 1, becau se of the m irrored posture of the two han d s, the upper key w as coded as LE F T an d the low er key as R IG H T w hen the left han d w as used, w hereas the opposite coding ap plied w hen the right han d w as u sed. A ccordingly, it shou ld also be possible to elim inate the H ± V effect by elim in ating this difference between th e coding for the two han ds. T he simplest experim ental design to test this assum ption is one in w hich the po stures of the left an d the right han dÐ relative to the response keysÐ are identical rather than m irrored. To ach ieve this, the response b oard w as placed either far to the left or far to the right of the su bject’s sagittal m id-line, so th at the posture of either h an d relative to the keys w as sim ilar an d responding reaso nably com fortable (see Figure 3). W ith the respo nse board positio ned o n

S ± R C O M P A T IB IL IT Y

959

FIG . 3. Left panel: The hand postur e required in E xp erim ent 3 for left- and rig ht-han ded respon ses on a leftside and a right-s id e-locat ed response board. Right panel: The conside rably reduced H± V effect in E xp erim e nt 3: total tim e an d erro r rate as a function of m ap ping and re sp onding han d.

the left, the upp er key sh ould now b e coded as R IG H T an d the lower key as L E F T for either han d, w h ereas w ith the respon se b oard position ed on the right, the up per key should be co ded as L E F T an d the lower key as R IG H T. T he referential-codin g hypothesis predicts two results. F irst, th e task shou ld yield a SRC effect, becau se the posture of the han d relative to the keys causes a left± right coding of the resp onses, w h ich overlaps w ith the coding of the stim ulus dim en sion. Speci® cally, perfor m an ce w ith the respo nse board o n the left should be better w ith left± dow n/righ t± up m ap ping than w ith left± up/right± dow n m ap ping, w hereas p erfor m an ce w ith th e response b oard o n the righ t should be b etter w ith left± up/right± dow n m ap ping than w ith left± dow n/right± up m ap ping. Second ly, no H ± V effect, or at least a con siderably reduced one (i.e. a depen dence of th e SRC effect o n respo nding han d), is predicted, because both the left and th e righ t han d provide the sam e left± right reference. T he cur rent version of th e m ovem ent-preference hypothesis does not provide for an in¯ uence of response locatio n, an d the presen t experim en tal situation m ain tains th e co nditions u nder w hich m ovem en t preferences should be effective. T hus, a H ± V effect co m parable to that in E xperiment 1 is predicted.

M eth o d Subjects Seven fem ale an d n ine m ale su bjects (15 righ t-han ded an d 1 am bidextrou s by self-rep ort) aged betw een 21 an d 42 w ere p aid to p ar ticipate in th e exp erim en t. T h ey h ad n or m al or cor rected -ton or m al visio n.

960

L IP P A

App aratus and Procedure T h ese w ere identical to E xp erim en t 1, excep t for th e variation of respo nse board location . T h e session w as d ivided into 4 blocks of 90 trials con sistin g of 10 pr actice trials and 80 test trials (40 in each stim ulus location ). In tw o blocks, th e resp on se board w as located 37 cm to o ne sid e, an d in th e oth er tw o blocks, 37 cm to th e oth er side of th e su bject’ s sagittal m id-lin e. For both resp on se board p osition s, su bjects u sed th e left ind ex ® n ger for on e block and th e righ t ind ex ® n ger for th e oth er. T h e ord er of h an d s an d of resp on se board location w as balan ced acros s sub jects. S u bjects w ere instru cted to h old th eir respo n d ing h an d at a 45± 90 8 an gle to th e lin e of th e resp on se keys, p articu larly in th e less com fortable con d ition , w h ere th e resp on se h an d an d th e board w ere o n th e sam e side. T h e su bject’ s sagittal m id-line w as in line w ith th e ® xatio n asterisk. E r ror and m issin g trials (1.5% ) w ere rep eated at a ra nd om p osition in th e block.

R esu lts M ean RT, M T, an d error data w ere calculated as a functio n of stim ulus an d response location, respo nding han d an d resp onse b oard location (see Table 3). Trials w ith RTs less than 120 m sec (0.4% ) were exclud ed from calculation. A N OVAs w ere co nducted on the data, w ith the between-sub ject variable S ± R m ap ping (left± up/righ t± dow n vs. left± dow n /right± up) an d the w ithin-sub ject variables respo nse b oard location (left vs. right), response h an d (left vs. right), an d respo nse location (upw ard vs. dow nw ard). Tota l Times. T he m ain effect of response location reached signi® can ce, F(1, 14) = 16.98, p < .01, in dicating that dow nw ard responses w ere faster (420 m sec) than upward responses (437 m sec). T he M ap ping 3 Response B oard L ocation interaction w as highly signi® can t, F(1, 14) = 134.96, p < .001. Perfo rm an ce w ith the respo nse board on the left w as better w ith left± dow n/right± u p m ap ping (379 m sec) th an w ith left± up/right± dow n TABLE 3 a a b c M e a n R T , M T , T T , an d E rr o r D a ta fo r E x p e rim en t 3 a s a Fu n ctio n o f R es p o n s e B o a rd L o c a tio n , R es p o n d in g Ha n d , a n d S tim u lu s an d R e s p o n s e Lo ca tio n

R esponse B oa rd on Left

R esponse B oa rd on R ight

Left H a nd

R ight Ha nd

Left H a nd

R ight H a nd

S timulus Loca tion

S timulus Loca tion

S timulus Loca tion

S timulus Loca tion

R esponse Loca tion

Left

R ight

Up

RT MT TT Errors

319 158 477 3.7

291 120 411 0.6

315 135 450 8.3

271 94 365 0.6

273 119 392 0.3

335 126 461 7.6

301 136 437 0.3

365 140 505 8.9

Dow n

RT MT TT Errors

266 112 378 0.3

327 128 455 2.1

271 89 360 0.6

332 128 460 5.7

339 124 463 8.8

269 113 382 0.6

344 110 454 6.3

289 118 407 0.6

a

b

c

In msec; total time in msec; in % .

Left

R ight

Left

R ight

Left

Right

S ± R C O M P A T IB IL IT Y

961

m ap ping (461 m sec), w h ereas w ith the resp onse board on the right th e opposite data patter n occu rred (471 vs. 405 m sec). In addition , a M ap ping 3 H an d interaction w as found, F(1, 14) = 7.75, p < .05. A s Figure 3 sh ow s, right-h an d respo nding w as faster w ith left± dow n/right± up m ap pin g (421 m sec) than w ith left± up/right± dow n m ap ping (439 m sec), w hereas for left-han d respon ding no difference occurred (428 vs. 427 m sec). M oreover, there w as a m ain effect of response b oard locatio n, F(1, 14) = 7.99, p < .05, w hich w as m odi® ed by a higher-o rder Respo nse B oard L ocatio n 3 H an d interactio n, F(1, 14) = 9.42, p < .01, an d, ad ditionally, by a high er-order Respon se Board L ocation 3 H an d 3 Respo nse interactio n, F(1, 14) = 59.90, p < .001. Fo r upw ard responses, b oth the left an d th e right han d show ed a reaction tim e disadvantage for responses o n the response board located on the sam e side (444 an d 471 m sec for left an d right han d, respectively) an d an advan tage for responses on the response b oard located on the op po site side (427 an d 408 m sec). For dow nw ard responses, this in¯ uence of a com fortable an d less com fortable response b oard location w as o nly partially effective. Right-h an d perfor m an ce w as better w ith the response board on the left than on the right (410 vs. 431 m sec), but left-han d perfor m an ce w as scarcely affected by board locatio n (417 vs. 423 m sec). A s an add ition al check for the signi® cance of the p resent M ap ping 3 H an d interaction, an A N OVA w as conducted on the data of E xperim ent 1 an d E xperim ent 3, w ith ex periment as an ad ditio nal variable. A s expected, a signi® can t E xperim ent 3 M ap ping 3 H an d interaction w as revealed, F(1, 2 8) = 10.00, p < .01, indicatin g that the siz e of the present H ± V effect (10 m sec) w as considerably sm aller than that of th e H ± V effect obtained in E xperim en t 1 (45 m sec). RT a nd M T Da ta . T he M ap ping 3 Respo nse Board L ocation inter action w as reliable in both RT, F(1, 14) = 53.51, p < .001, an d M T data, F(1, 14) = 16.58, p < .01 . In contrast, the M ap ping 3 H an d interactio n failed to reach signi® cance in RT data, F(1, 14) = 1.36, p = .262, an d occu rred only in M T data, F(1, 14) = 8.74 , p < .05, m odi® ed by a higher-order M ap ping 3 H an d 3 Response interactio n, F(1, 14) = 4.64, p < .05; the H ± V effect w as present w ith dow nw ard , bu t not w ith upw ard m ovem ents. E rror Da ta . T he m ain effect of m ap ping, F(1, 14) = 5.97, p < .05, of respo nse b oard location, F(1, 14) = 5.84, p < .05, an d the interaction between these factors, F(1, 14) = 97.15, p < .001, reached signi® can ce. W ith the response board o n the left, few er errors were m ad e w ith the left± dow n/right± up m ap ping (0.5% ) than w ith the left± up/righ t± dow n m ap ping (5.0% ), w hereas resp onding w ith the response board on the right reversed the effects (7.9 vs. 0.5% ). In add itio n, the M ap ping 3 H and interaction ap proached reliability, F(1, 14) = 4.40, p = .055 (see F igure 3). L eft-han d perform an ce w as less er rorprone w ith the left± up/right± dow n m ap ping (1.7% ) th an w ith the left± dow n/right± u p m ap ping (4.3% ), but right-han d perfor m an ce rem ained un affected (3.7 vs. 4.1% ). Finally, a Respo nse Board L ocation 3 H an d in teraction reached signi® can ce, F(1, 14) = 4.7 1, p < .05. For left-han d resp onses, there were fewer errors w ith the response board o n the left than o n the right (1 .7 v s. 4.3% ), w hereas for right-han d responses, the sam e num ber of er rors w ere m ad e (3.8 vs. 4.0% ).

962

L IP P A

D iscussio n T here are two m ain resu lts: First of all, a M ap ping 3 Respo nse Board L ocation interactio n w as found . Respond in g w ith the response board on the left w as faster an d less er ror-pro ne w ith the left± dow n/right± up m ap ping than w ith the left± up/right± dow n m ap ping, w hereas the opposite data patter n occurred w ith the respo nse b oard on the right. T he secon d m ain result concer ns the M ap p ing 3 H an d interaction obtained. A ctually, the data pattern cor respond s to the H ± V effect, but its size show s a statistically signi® can t decrease from 45 m sec in E xperiment 1 to 10 m sec. M oreover, there is reaso n to query the reliability of this H ± V-like patter n: It w as obtained only in M T da ta an d w ith dow n w ard m ovem ents only. H ence, it sh ould b e cau tiously inter preted. T hus, E xperim en t 3 yielded a S RC effect dependent on respon se b oard locatio n an d a co nsiderably red uced H ± V effect. B oth results ag ain supp ort th e proposal that it is the reference provided by th e responding han d an d n ot the m ovem ent preference of each han d that is critical for the H ± V effect. U p to now, the referential-coding idea has cor rectly predicted the co nditio ns for the presence an d the absence of the H ± V effect as w ell as the dependence of the SRC effect o n response board location . In contrast, the m ovem en t-preferen ce hypothesis cannot read ily account for these results, as, in its cur rent versio n, it is unclear w hy eccentricity of response location or han d postu re should m odulate the H ± V an d SRC effects. H owever, Bau er an d M iller (1982, E xp. 3) offered further evidence for their m ovem ent-preference idea. T h ey predicted an other S RC effect dependent on the response hand by exam ining vertically oriented stim uli an d h orizontally oriented respo nses. T his V ± H effect is discu ssed in the next section .

T H E V ± H E FFE C T In B au er an d M iller’ s third experim ent, subjects had to press a left or righ t key w ith one index ® nger in respo nse to a stim ulus ap pearing above or below a ® xation point. A g ain, they p redicted that left-han d perfo r m an ce w ould be better w ith clockw ise m ovem ents an d that right-han d perfo rm an ce w ould be better w ith counterclockw ise m ovem ents. N ote that in this design clockw ise m ovem ents are cau sed by up± righ t/d ow n± left m ap ping, w hereas co unterclockw ise m ovem ents are caused by up± left/dow n± right m ap ping. Bau er an d M iller indeed yielded a M ap ping 3 H an d interactio n, b ut, again, there w as no crossover interactio n becau se, this tim e, right-han ded responding w as faster overall. T hat is, a V ± H effect occurred in that the advan tage of up± right/ dow n± left m ap ping w as g reater for the left (65 m sec) th an for the right (20 m sec) han d. T here have been several attem pts to replicate the V ± H effect, bu t its reliab ility seem s to be question able: Som e studies have obtained a M ap ping 3 H an d interaction (M ichaels, 1989 , E xp. 1; M ichaels & S ch ilder, 1991, E xp. 3, pro ne position), but others have not (M ichaels & S childer, 19 91, E xp. 1, m idline positio n; Weeks & Proctor, 199 0, Ex p. 1). T he referential-coding idea casts light on this rather co nfu sing data pattern because, dependin g on the particular han d posture, b oth the presence an d th e absence of the V ± H effect can be p redicted. T he V ± H effect should be present if the responding h an d is in line with the respo nse keys, as show n in F igure 4. U n der these conditions, resp onses should be coded not as ``left’ ’ an d ``right’ ’ , but as ``upw ard’ ’ an d ``dow nw ard’ ’ . A s th e vertical

S ± R C O M P A T IB IL IT Y

963

FIG . 4 . L eft panel: The hand postur e req uired in Expe rim e nt 4 for le ft- and right-han ded respon ses. R ight pane l: the V± H effect in Experim ent 4: total tim e and erro r rate as a function of mappin g an d respon ding hand.

response dimension th en overlaps w ith the vertical stim ulu s d im ension an d th e coding of a physically left or right key as T O P or BO T T O M dep end s on the respo nding han d, a V ± H effect is expected in this situation. In co ntrast, the referential-coding hypothesis pred icts the a bsence of or at least a considerably reduced V ± H effect if the respo nding han d is positioned at a 90 8 a ngle to the line of the respo nse keys (as in F igu re 5). In th is case, the left key is to the left an d the right key to the right side of the respo nding han d, so that the response positions sh ould be coded orthogo nally to those of the vertical stim ulus set. In this situatio n, S RC effects depen dent o n han d should be absent or sm all. In order to test the referential-c oding exp lanations for the presence an d absence of the V ± H effect, E xperiments 4 an d 5 w ere car ried ou t.

E X PE R IM E N T 4 E xperiment 4 w as d esigned to create a situation in w hich the referential-coding hypothesis predicts a V ± H effect. T he sub jects’ task w as to respond to top an d bottom stim uli by tou ching a left or right response key u nim anually. In addition, they w ere explicitly instructed to keep their respon din g han d in line w ith th e keys (see Figure 4). T his situation should cau se the physically left an d right response keys to be coded as T O P an d BO T T O M , depending on the respo nd in g han d. T h erefore, th e referential-codin g hypothesis predicts, for the left han d, b etter p erform an ce w ith up± right/dow n± left m ap ping than w ith up± left/dow n± right m ap ping, w hereas for the right han d the opposite data patter n is predicted.

964

L IP P A

M eth o d Subjects Ten fem ale an d tw o m ale su bjects aged betw een 19 an d 33 w ere p aid for th eir p ar ticipation . A ll claim ed to be righ t-h an d ed , an d all h ad no r m al or cor rected -to-n or m al visio n .

App aratus and Procedure T h ese w ere as in E xp erim en t 1, w ith the follow ing exce ption s. First, th e exp erim en t w as c on trolled by a M eg a/S T E com pu ter. S eco nd , in ad ditio n to er ro r and m issin g trials (0.7% ), trials w ith RTs less than 120 m sec (0.2% ) w ere rep eated at a ran d om po sition in th e block. T h ird , th e reve rsed S ± R set w as u sed : Stimu li ap pear ed 1.6 cm above or be low th e ® xatio n aster isk, an d th e respo ns e board w as tu r n ed th ro ug h 90 8 so th at th e resp on se keys w er e oriented h oriz on tally. O n e ha lf of th e su bjects tou ched th e left an d th e righ t key in r esp o nse to top or bottom stim uli, resp ectively, w h ereas the oth er ha lf u sed the rever sed m ap p ing. In ad d ition, th ey w ere requ ired to h old their r esp o nd ing h an d in lin e w ith th e resp on se keys th ro ug h ou t. T h e su bje ct’ s sa gittal m id-lin e w as in lin e w ith th e h om e key an d the ® xation asterisk.

R esu lts an d D iscu ssion RT, M T, Table 4. m ap ping response

an d er ror data w ere treated as in E xperim ent 1, an d their m eans are presen ted in A N OVAs w ere conducted o n the d ata, w ith the between-subject v ariable S± R (up± left/d ow n± right vs. up± right/dow n± left) an d the w ithin-su bject variables h an d (left vs. right) an d respo nse locatio n (left vs. right).

Tota l Times. A M ap ping 3 H an d interaction did occur, F(1, 10) = 11.02, p < .01. A s Figure 4 sh ow s, left-han d responding w as faster w ith up± right/d ow n± left m ap ping (452 m sec) than w ith u p± left/dow n± righ t m ap pin g (527 m sec), w h ereas right-h an d TABLE 4 a

a

b

c

M e an R T , M T , T T , a n d E rro r D ata for E xp erim e n t 4 as a F u n c tio n o f R e s p o n d in g H a n d an d S tim u lu s a n d R es p o n s e L o c a tio n

Left H a nd

Right H a nd

S timulus Loca tion

S timulus Loca tion

R esponse Loca tion

Up

Up

L eft

RT MT TT Errors

399 111 510 6.1

349 85 434 1.6

33 9 13 8 47 7 3.0

384 127 511 4.5

Righ t

RT MT TT Errors

345 125 470 1.6

403 141 544 6.5

38 1 80 46 1 1.2

333 107 440 0.4

a

b

c

In m sec; tota l time in m sec; in % .

Down

Down

S ± R C O M P A T IB IL IT Y

965

responding w as faster w ith u p± left/d ow n± right m ap ping (459 m sec) than w ith up± right/dow n± left m ap p ing (486 m sec). In addition, th e H an d 3 Respo nse L ocatio n interaction reached signi® can ce, F(1, 10) = 41.84, p < .00 1, indicatin g that a cor respo nden ce between han d an d respon se location, such as the left han d m ak ing a leftw ard response, yield ed better perfor m an ce than non-correspondence, such as the left han d m aking a righ tward response. E rror Da ta . T he m ain effect of han d ap proached reliability, F(1, 10) = 4.83, p = .053, an d w as m odi® ed by the higher-order M ap ping 3 H an d interaction, F(1, 10) = 14.99, p < .01 (see F ig ure 4). A test of sim ple m ain effects show s that left-han d perfor m an ce w as less er ror-pro ne w ith up± right/dow n± left m ap ping (1.6% ) than w ith up ± left/dow n± right m ap ping (6.3% ), F(1 , 10) = 4.83, p = .053, w hereas right-han d perfor m an ce w as not signi® can tly affected (2.9 vs. 1.7% ), F < 1. In sum , a V ± H effect occurred w h en the respond ing han d w as situated in line w ith the response keys. T his is consistent w ith the assum ptio n that SRC effects dependent o n han d arise w hen han d posture cau ses a response coding overlapping w ith the stim ulus dimensio n. In ad dition, it sug gests that som e previous stu dies m ay have failed to ® nd a V ± H effect becau se han d posture w as not p roperly co ntrolled.

E X PE R IM E N T 5 E xperiment 5 exam ined a situatio n in w hich refe rential-coding predicts a sm all or absent V ± H effect. It w as identical to E xperim ent 4, w ith the exception th at subjects w ere ex plicitly instructed to hold th eir respon ding han d at a right an gle to the line of the response keys throughout (see Figure 5). Re gardless of th e resp onse han d, this situ atio n sh ou ld cause a left± righ t coding of the respo nse set that is orthogonal to the top± bottom co ding of the stim ulus set, an d, thu s, a V ± H effect is not expected.

M eth o d Subjects, A pparatus, and Procedure E ight fem ale an d eigh t m ale su bjects aged be tw een 19 an d 39 w ere p aid for th eir p ar ticipation . A ll claim ed to be righ t-h an d ed an d all had n or m al or c or rected-to-n or m al vision . T h e app ar atu s an d th e p roc edu re were th e sam e as in E xp erim en t 4, bu t th e su bjects w ere requ ired to keep th eir r esp o nd ing h an d at a righ t an gle to th e lin e of th e re sp on se keys. To avoid u n com fortable h an d p ostu res, su bjects w ere allowed to sit ap p roxim ately 1 m aw ay from the screen . T h e su bject’ s sa gittal m id-lin e an d th e h an d w ere th en in lin e w ith th e h om e ke y an d ® xation asterisk. E r ro r an d m issin g trials (0.4% ) as w ell as trials w ith RTs less th an 120 m sec (0.0% ) w ere repe ated at a ran d om p osition in the bloc k.

R esu lts RT, M T, an d error data were treated an d an alysed as in Experim ent 4, an d their m ean s are presented in Table 5. Tota l Times. T he m ain effect of m ap ping reached sig ni® can ce, F(1, 14) = 9.88, p < .01. W ith u p± right/dow n± left m ap ping, perfor m an ce w as better (386 m sec) than w ith up± left/ dow n ± right m ap ping (440 m sec). In ad dition, th e H an d 3 Respo nse L ocation interactio n

966

L IP P A

FIG . 5 . L eft panel: The hand postur e req uired in Expe rim e nt 5 for le ft- and right-han ded respon ses. R ight pane l: The abse nce of the V± H effec t in E xpe rim ent 5: total time and er ror rate as a function of m apping an d respo nding hand.

w as sign i® can t, F(1, 14) = 10.27, p < .01, indicating that right-han d respo nses w ere faster for rightward than for leftw ard responses, w hereas left-han d responses, did not differ. T he relevan t M ap ping 3 H an d inter action w as far from being signi® can t, F < 1 (see Figure 5). A s an ad ditio nal check for the disapp earan ce of th e M ap pin g 3 H an d interactio n, an A N OVA w as con ducted on the data of E xperiment 4 an d E xperim en t 5 w ith experim ent as an ad ditio nal variable. A s expected, a signi® cant E xperiment 3 M ap ping 3 H an d interaction w as revealed, F(1, 24) = 1 2.59, p < .01, ind icating that the V ± H effect w as present in E xperim ent 4 but not in E xperiment 5.

RT a nd M T Da ta . W hereas in RT d ata no M ap ping 3 H an d interaction occurred, F < 1, M T data revealed a M ap p ing 3 H an d 3 Response L ocation interaction, F(1, 14) = 7.1 5, p < .05, indicating a reversed V ± H effect for leftw ard responses on ly. L eft-han ded responses to the left had shorter m ovem ent tim es in response to top than to bottom stim uli (114 vs. 128 m sec), w hereas right-han ded respo nses to the left w ere scarcely affected by the stim ulus positio n (121 vs. 1 15 m sec).

E rror Da ta . A s F igure 5 show s, error data suggested an advantage of the up± right/ dow n ± left m ap ping for the left h an d, but this M ap ping 3 H an d interaction w as only m argin ally signi® can t, F(1, 14) = 3.15, p = .097.

S ± R C O M P A T IB IL IT Y

967

TABLE 5 a a b c M ea n R T , M T , T T , a n d E rro r D a ta for E x p e rim e n t 5 as a F u n c tio n o f R e sp o n d in g H a n d an d S tim u lu s a n d R es p o n s e Lo c atio n

Left H a nd

R ight Ha nd

S timulus Loca tion

S timulus Loca tion

R esponse Loca tion

Up

Up

L eft

RT MT TT E rrors

325 114 439 2.1

265 128 393 0.6

323 121 444 2.4

27 3 11 5 38 8 1.9

R igh t

RT MT TT E rrors

268 125 393 0.4

324 127 451 3.2

266 105 371 1.6

31 8 10 9 42 7 1.8

a

b

Down

Down

c

In m sec; Tota l time in msec ; in % .

D iscussio n A s predicted by the referential-coding hypoth esis, th e V ± H effect disap peared w hen the responding han d w as situated at a right an gle to the line of th e respo nse keys, so that the response po sitions w ould be coded orthogonally to the stim ulus position s. T his con ditio n w as even better m et by M ichaels an d Schilder’ s ® rst experim ent (1991, m idline position). T hey used a response ap p aratus that locked the respo nding han d in front of the b ody, so that o nly left± right m ovem ents perpendicular to the ® ngertip-to-w rist axis could be perfor m ed. U nder this experim en tal condition, the V ± H effect disapp eared com p letely. T herefore, it can be co ncluded that the ab sence of the V ± H effect resu lts from the response han d b eing situated at a right an gle to the lin e of the response keys or, m ore generally, from the left± right codin g of the resp onse set. T here is no ready explanation for w hy M T data revealed a reversed V ± H effect for leftw ard responses. N either the m ovem ent-preference hypothesis, w hich would have predicted a positive V ± H effect, nor the referential-codin g hypothesis, w hich predicted no SRC effects at all, com e up w ith a reasonable account for this particular ® nding. In conditions that elim inated the V ± H effect, an other com patibility effect occurred, nam ely a signi® can t overall superiority of the up± right/dow n± left m ap ping to the up± left/dow n± right m ap ping. Such an effect w as ® rst rep orted by Weeks an d Proctor (1 990) an d is beyon d the scope of the referential-coding hypothesis, b ecau se, for the V ± H design, this p redicts either the p resence of a V ± H effect or no SRC effect at all. I w ill refer to this point later o n.

G EN E R A L D IS C U S S IO N T he present study h as ad dressed the d ependence of SRC effects o n respond ing han d w ith horizo ntally oriented stim uli an d vertically oriented resp onse s (H ± V effect) an d w ith vertically oriented stim uli an d horizo ntally oriente d resp onses (V ± H effect) reported by

968

L IP P A

Bau er and M iller (19 82). T he ® rst th ree experim ents focused on the H ± V effect. E xperim ent 1 replicated th e H ± V effect. In E xperiment 2, the basic H ± V d esign w as m aintained, but subjects w ere instructed to respond w ith their h an ds in lin e w ith th e response keys. T his elim inated the H ± V effect. In E xperiment 3, the resp onse board w as placed to the left or right side of the subject. T his yielded a co nsiderably red uced H ± V effect as w ell as a novel SRC effect dependent o n respon se board location. E xp eriment 4 an d 5 focused o n the V ± H effect. In E xperim ent 4, this effect w as prod uced w hen subjects were required to position the respon ding h an d in line w ith the resp onse keys. W ith th e han d rotated through 90 8 in E xperim en t 5, th e V ± H effect w as elim inated an d a signi® can t superiority of up ± right/dow n± left m ap ping com pared to up± left/dow n± righ t m ap ping w as ob served. T he present evidence p rovides am ple support for the referential-coding hypothesis, favouring it over Bau er an d M iller’ s m ovem ent-preference hypothesis. F irst of all, the ex planatory scope of the referen tial-coding hypothesis is b road er than that of the m ovem ent-preference hypothesis. W hereas the latter accounts only fo r the presence of the H ± V an d V ± H effect, th e referential-coding hypothesis also predicts the con ditio ns th at elim in ate b oth effects an d those that produ ce the SRC effect dependent on response board location. O f course, it m ight be possible to com e up w ith a m o di® ed m ovem entpreference hyp othesis that accounted fo r the entire ran ge o f the present results by introducing assum p tions about posture-d ependent m ovem ent preferences. But these would b e u nsatisfactorily post hoc an d certainly less pa rsimo niou s than the referentialco ding hypothesis, w hich gets by o n a single assum ptio n. In fact, the p resent experiments have exam ined on ly som e of the possible ap plications of the referential-coding log ic: If han d posture cau ses a respon se coding that overlaps w ith the stim ulus dimension, SRC effects dependent on han d should occur; an d if han d posture cau ses a respon se coding that does not overlap w ith the stim ulus set, no SRC effects depen den t on han d should be found. W ithin the parad igm of physically orthog onal S ± R dimensions, the present study is the ® rst to test the referential-coding account system atically. H ow ever, other au thors have previou sly co nsidered the idea to explain their effects, but rejected it in favour of m ore co m plex explanatio ns (Co tto n et al., 1977; E hren stein et al., 1989). To w hat extent these are needed or how far the referential-coding idea is ap p licable to other SRC effects of physically orthogo nal S± R sets is discussed in the next sectio n.

Th e E xp lan atory S co p e of th e R e fe ren tial-cod in g H ypo thesis First of all, let u s focus o n som e ® ndings from H ± V-like designs. To the best of m y know ledge, there are o nly three relevan t experim ents, an d they all prod uced results co nsistent w ith the referen tial-coding hypothesis. M ichaels (1989 , Exp. 2) asked her su bjects to de¯ ect a toggle switch, located on a horizontal table, upw ards (i.e. aw ay from the sub ject’s b ody) or dow nw ards in respo nse to left an d right stim uli. In a m idline co ndition, the toggle sw itch w as located in front of the su bjects; in eccentric conditions, it w as located 60 cm to the left or the right of the subject’ s sagittal m idline. In the m idline co ndition, sub jects w ere instructed to hold the sw itch w ith the index ® nger on the top an d the thum b o n the bottom ; in the eccentric co nditio ns, the thum b w as o n the top an d the index ® nger on the bottom . In the m idline condition a H ± V effect w as observed, but in the

S ± R C O M P A T IB IL IT Y

969

eccentric conditions there w as a reversed H ± V effectÐ that is, a left-han d advan tage w ith left± dow n/right± up m ap ping an d a rig ht-han d advan tag e w ith left± u p/right± dow n m ap ping. In prin ciple, this experim ental design com bin es the co nditions of E xperiments 1 an d 3 of the present study. T h erefore, the referential-coding explanation is the sam e: G iven the instructed han d postures, it can be assum ed that in both m idline an d eccen tric co nditions, th e up- an d dow nw ard de¯ ectio ns were coded as left an d right w ith respect to the ® ngertip-to-w rist axis. A s the reference provided by the left an d right han d in the eccentric conditions w as opposite to that in the m id line co nditions, th e H ± V effect at m idline po sition sh ould be reversed at eccentric p ositions. E hrenstein et al. (1989) investigated SRC effects o f physically orthogonal S ± R dimension s w ith uniman ual two-® nger choice reactions in a Simo n task. Two co nditions of th eir ex periment w ere sim ilar to the experimental design used here. S ubjects respon ded to red an d g reen lights ap p earing in the left or righ t visual ® eld by pressing vertically oriented response keys w ith the index an d m iddle ® nger of the left or right han d. In o ne co nditio n, the subjects’ palm s faced dow n; in the other, the palm s faced u p. E hrenstein et al. ex pected that SRC effects w ould depend o n the an atom ical left- an d rightness of the index an d m iddle ® nger, so that w ith palm s up SRC effects opp osite to those w ith palm s dow n were predicted . Inste ad , for both con ditio ns, S RC effe cts referring to the response position w ere obtained: L eft-h an d perfor m an ce w as b est for left± up/righ t± dow n trials, w hereas right-han d perfor m an ce w as best for left± dow n/ right± up trials. A ccording to the referential-coding hypothesis, this data patter n is expected if it is assum ed th at the han d 2 reference axis reverses w ith respon ding hand, but not w ith supin e or pro ne position . C otton et al. (1977, E xp. 2; 1980, E xp. 1) used the sam e design as in the present E xperiment 1, excep t for the stim ulus ar ray. T he stimu li presen ted varied not o nly on the horizo ntal but also on the vertical dimension, so that left an d right stim uli ap p eared either above or below a central ® xatio n po int. Subjects w ere instru cted to respo nd to top stim uli w ith the u pper key an d to bo ttom stim uli w ith the low er key, independent of w hether they were left or right of the ® xation point. T he results revealed that left-han d respon din g w as faster w h en top stim uli ap peared on the left an d bottom stim uli o n the right sid e, w hereas right-han d respond ing w as faster w hen top stim uli ap p eared on the right an d bottom stim uli on the left side. In the light of the referential-coding hypothesis, the in ¯ uence of the irrelevant horizo ntal stim ulus dimensio n is at least reason able. T hough C otto n et al. do not report han d posture in detail, on e m ay speculate that the respo nd ing han ds w ere situated at a 45± 90 8 an gle to the line of the response keys, so th at the upp er an d lower response keys w ere coded as left and righ t, too. T herefore, both stim uli an d respo nses varied horizo ntally an d vertically, w hich per m its a H ± V effect for b oth dim ensions. Studies using V ± H -like designs have largely investig ated the in¯ uence of response eccentricity an d revealed th at the V ± H effect obtained in m idline position reverses w ith 2

For left-ha nd respon ding M ic haels and Sc hilder (1991 , E xp. 3) found an advantage of up± right /down± left m apping with the palm facing dow n an d an advantage of up± left/dow n± righ t m appin g with the palm facing up, hence differe nt SRC effects for pro ne an d su pine posture s. U nfortunately, M ichaels an d Sc hilder do not repor t wh ether the prone vs. supin e postur e m anipulatio n als o produced a change of the intrins ic ® ngertip ± to± w rist axis relativ e to the respo nse positio n (le ft± righ t de¯ ections of a toggle sw itch) and wh ether the hand axis w as contro lle d at all. So, their failure to replicate E hrenste in et al.’s (1989 ) results m ight be due to in tra- an d/or in terindividually varying hand postures.

970

L IP P A

increasing eccentricity (M ichaels, 1989, Exp. 1; M ichaels & Schilder, 1991, E xp. 1 ; Weeks, Proctor & Beyak, 1995). T hat is, w ith a resp onse pan el located on the subject’s left or right sid e, left-han d perfor m an ce w as better w ith up± left/dow n ± right m ap ping an d righthan d perfor m an ce w as better w ith up± right/dow n± left m ap ping. W ith respect to th e ex periments carried out w ith toggle switch es, one could argue that in creasing eccentricity ch an ged han d posture so that it w as not left± right but upward an d dow nw ard de¯ ections that w ere perfor m ed. If so, th ere w as once ag ain a dim ensional overlap w ith the stim ulus set, w hich p er m itted SRC effects. T hus, it w as not the eccen tricity of respo nse locatio n per se, but the resulting han d posture that w as respon sible for the reversal. A lthough this ex planation is probably valid for som e of the results, it is not com p lete. M ichaels an d Schild er (1991, E xp. 1) sh ow ed that the reversal of the V ± H effect at eccen tric positions also occurs if the han d is locked such th at only left± right de¯ ectio ns are per m itted. A s the han d-related coding of the response set should b e or tho go nal to the top± bottom stim ulus dimensio n, th e refere ntial± coding hypothesis w ould predict no S RC effects at all. A ctu ally, the explanatory scope of the present referential-cod ing hypothesis ends here. A s it postulates an in¯ uence of han d posture o nly, its predictive pow er is lim ited. N evertheless, the basic idea b eh ind the referen tial-coding hypothesisÐ nam ely, that actions are co ded w ith respect to fram es of referen ceÐ is m ore generally ap plicable. For instance, it is possible that in the tasks described above, responses w ere coded w ith respect to th e body. If so, left± right de¯ ection s, as responses aw ay from an d tow ard s the body, m ay have been co ded as upw ards an d dow nw ards, w hich w ould ag ain result in a dim ensional overlap w ith the stim ulus dimensio n. To sum up, th e present version of the referential-coding hypothesis accounts for a ran ge of ® n din gs obtained w ithin H ± V-lik e an d V ± H -like d esigns. M oreover, generalizing from the present referential-coding hypoth esis to som e other fram es of reference m ight help to explain other know n effects an d probably also to predict novel S RC effects.

Th e S ig ni® can ce of the R eferen tial-co ding Id ea for Th eo ries of S ± R C om p atibility In gener al, the referential-coding hypothesis can be reg arded as a m issing link that bridges the g ap between S RC theories that rely on the dimensional overlap assum ptio n an d com patibility effects observed w ith S± R sets that do not seem to ful® l th is criterio n. W ith the referential-coding hypothesis, m ost of the effects of physically orthogonal S± R dim ensions can be show n to depend on coding in a spatial d im ension com m o n to stim ulus an d response representations an d are thus open to explanation by conventional SRC th eories. H owever, there are still som e SRC effects th at can not be accom m odated to conventional SRC theories by the referential-coding hypothesisÐ for exam ple, the overall superiority of up± righ t/ dow n± left m ap ping to up± left/dow n± right m ap ping obtained in E xperim ent 5. A ccord ing to the referen tial-coding hypo thesis, this m ap ping effect occurs w ith orthogonally coded S± R dimensio ns, as the h an d-related coding of the respon se set did not overlap w ith the cod ed stim ulu s set. T his effect requires an other explanatio n such as the salientfeatures coding p rinciple proposed by Weeks an d colleagues (Weeks & Proctor, 1990; Weeks et al., 1995). T his ``attrib utes the preference fo r the up± right/dow n± left m ap ping to th e correspo ndence between the salien t polar referents of the stimu lus an d response

S ± R C O M P A T IB IL IT Y

971

dimensio ns (up an d right, respectiv ely) that it instantiates’ ’ (Weeks et al., 1995, p. 380). T hat is, up ± right/dow n± left m ap ping is prefer red, becau se the salient features of the S± R set m atch. H ence, like the referential coding hypothesis, the salient features codin g princip le also tries to solve the p roblem of orthogonal stim ulus an d respo nse dimensio ns by prop osing a for m of cognitive overlap between them . By co m bining these hypotheses, it m ay b e possible to inter pret all effects of com patibility between physically orth ogo nal stim ulus an d respo nse d im ension s w ithin the sam e fram ew ork as the classic SRC effects.

R E FE R E N C E S Bauer, D.W., & M iller, J. (1982) . Stimulus± respon se com patibility and th e m oto r system . The Qua rterly J ourna l of Experimenta l P sychology, 34A , 367 ± 380. Car penter, P.A., & Eisenberg, P. (1978) . M ental rotatio n an d th e fram e of reference in blind and sighted individuals. P erception & P sychophysics, 23, 117± 124. Cotto n, B., Tz eng, O.J.L ., & Hardyck, C. (1977) . A respo nse instruction by visual-® eld interaction : S± R com patibility effect or? B ulletin of the P sychonomic S ociety, 10, 475± 477. Cotto n, B., Tz eng, O.J.L ., & H ardyck , C. (19 80) . Role of cerebral hemispheric processing in th e visual half-® eld stimulus± response com patibility effect. J ourna l of E xperimenta l P sychology : Huma n P erception a nd P erforma nce, 6 , 13± 23. De Jo ng, R., L iang, Ch.Ch., & L aub er, E. (19 94) . Con ditional and unconditional au tomaticity : A dualprocess m odel of effects of spatia l stimulus± respo nse correspon dence. J ourna l of E xperimenta l P sychology : H uma n P erception a nd P erforma nce, 20, 731 ± 750. Ehrenstein, W.H., Schroeder-H eister, P., & Heister, G. (1989) . S patial S± R com patibilit y with or th ogona l stimulus± respon se relationship. P erception & P sychophysics, 45, 215 ± 220. Fitts, P.M ., & S eeger, C.M . (19 53) . S± R compatibility: Sp atia l characte ristics of stimulus an d respo nse cod es. J ourna l of Experimenta l P sychology, 46, 199± 210. Hasbroucq, T h., Guiard, Y., & O tto man i, L . (1990) . P rinciples of respo nse deter minatio n: T he list-rule mod el of S R compatibility. B ulletin of the P sychonomic S ociety, 28, 327± 330 . Kornblum , S., Hasbro ucq, T., & O sman , A. (1990) . D im ensional overlap: Cognitive basis for stimulus± respon se compatibilityÐ A m odel an d tax o nomy. P sychologica l R eview, 97, 25 3± 270. L aÁ davas, E. (1987) . In¯ uence of han dedness on spatial compatibility effects with perpendicular ar rangement of stimuli an d responses. Acta P sychologica , 64, 13± 23. L aÁ davas, E., & M os covitc h, M . (1984) . M ust egoce ntric an d environm ental fram es of reference be aligned to produce spatial S± R com patibility effects? J ourna l of Experimenta l P sychology: H uma n P erception a nd P erforma nce, 10 , 205± 215. M ichae ls, C.F. (1989) . S ± R com patibilities depend on eccentricity of respon ding han d. The Qua rterly J ourna l of Experimenta l P sychology, 41A , 263 ± 272. M ichaels, C.F., & S childer, S. (1991) . S timulus± respon se compatibilities between vertically oriented stimuli an d horizo ntally oriente d respon ses: Th e effects of han d position an d posture. P erception & P sychophysics, 49, 342± 348. Sim on , J.R. (19 69) . Reactio ns toward th e source of stimulatio n. J ourna l of Experimenta l P sychology, 81, 174± 176. Um iltaÁ , C., & N icoletti, R. (1992) . An integ rated mod el of th e Sim on effect. In J. Alegria, D. H olender, J. Junca de M or ais, & M . Rad eau (E ds.), Ana lytic a pproa ches to huma n cognition (p p. 331 ± 35 0). Am sterdam : Elsevier. Weeks, D.J., & P roctor, R.W. (1990) . Salient-featu res coding in th e tran slatio n between or th ogo nal stimulus and respon se dim ensions. J ourna l of Experimenta l P sychology : Genera l, 119 , 355± 366. Weeks, D.J., P roctor, R.W., & Beyak , B . (1995) . Stimulus± respon se comp atibilit y for ve rtically oriented stimuli an d horizon tally oriented respon ses: Evid ence for spatial coding. The Qua rterly J ourna l of Experimenta l P sychology, 48A , 367± 383. Welford, A.T. (1976) . S killed performa nce: P erceptua l a nd motor skills. Glenview, IL : Scott, Foresm an . Origina l ma nuscript receiv ed 13 J une 199 4 Accepted revision receiv ed 11 October 199 5