Online publication date: 01 October 2010

Transcription

1 This article was downloaded by: [BIUS Jussieu/Paris 6] On: 10 June 2011 Access details: Access Details: [subscrition number ] Publisher Psychology Press Informa Ltd Registered in England and Wales Registered Number: Registered office: Mortimer House, Mortimer Street, London W1T 3JH, UK Visual Cognition Publication details, including instructions for authors and subscrition information: htt:// Resonse force in RT tasks: Isolating effects of stimulus robability and resonse robability Stefan Mattes; Rolf Ulrich; Jeff Miller Online ublication date: 01 October 2010 To cite this Article Mattes, Stefan, Ulrich, Rolf and Miller, Jeff(2002) 'Resonse force in RT tasks: Isolating effects of stimulus robability and resonse robability', Visual Cognition, 9: 4, To link to this Article: DOI: / URL: htt://dx.doi.org/ / PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: htt:// This article may be used for research, teaching and rivate study uroses. Any substantial or systematic reroduction, re-distribution, re-selling, loan or sub-licensing, systematic suly or distribution in any form to anyone is exressly forbidden. The ublisher does not give any warranty exress or imlied or make any reresentation that the contents will be comlete or accurate or u to date. The accuracy of any instructions, formulae and drug doses should be indeendently verified with rimary sources. The ublisher shall not be liable for any loss, actions, claims, roceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

2 VISUAL COGNITION, 2002, 9 (4/5), Resonse force in RT tasks: Isolating effects of stimulus robability and resonse robability Stefan Mattes and Rolf Ulrich University of Tübingen, Germany Jeff Miller University of Otago, New Zealand Mattes, UIrich, and Miller (1997) found that as resonse robability decreases in a simle reaction time (RT) task, articiants roduce more forceful resonses as well as longer RTs, suggesting a direct influence of rearatory rocesses on the motor system. In this revious study, however, resonse robability was confounded with stimulus robability, leaving oen the ossibility that resonse force was sensitive to stimulus- rather than resonse-related rearation. The resent study was conducted to unravel the effects of stimulus and resonse robability. Exeriment 1 maniulated stimulus robability and revealed that resonses to a more robable stimulus are less forceful than resonses to a less robable stimulus even when both stimuli require the same resonse. Exeriment 2 demonstrated that this stimulus robability effect does not deend on the overall level of resonse robability. Exeriment 3 showed an analogous effect for resonse robability when stimulus robability is ket constant. The comlete attern of results suggests that both stimulus robability and resonse robability affect the forcefulness of a resonse. It is argued that resonse robability exerts a direct influence on the motor system, whereas stimulus robability influences the motor system indirectly via remotoric adjustments. Prearatory rocesses lay a fundamental role in human erformance, and the study of these covert rocesses is therefore essential for understanding human cognition. Prearatory adjustments seem to oerate at various levels within the Please address all corresondenc e to Rolf Ulrich, Psychologische s Institut, Universität Tübingen, Friedrichstr. 21, D Tübingen, Germany. rolf.ulrich@uni-tuebingen.d e This research was suorted by the Deutsche Forschungsgemeinschaf t (UL 116/3-2). Parts of this aer were resented at the Tagung exerimentel l arbeitende r Psychologen, 1998 (Marburg, Germany). We thank Raymond Klein for helful comments and Frauke Becker, Hiltraut Müller-Gethmann, and Jutta Stahl for assistance in data collection. Ó 2002 Psychology Press Ltd htt:// DOI: /

3 478 MATTES, ULRICH, MILLER CNS. For examle, attention researchers have shown that advance information about stimulus location reares the ercetual system leading to imroved detection and identification of stimuli that aear within the exected location (e.g., Downing, 1988). Psychohysiologica l studies suort the view that selective attention induces rearatory rocesses that have influences at very early ercetual levels (Luck, 1998). At the other end of the information rocessing chain, rearatory rocesses also influence the motoric level (Requin, Brener, & Ring, 1991). For examle, Leuthold, Sommer, and Ulrich (1996) rovided evidence from the analysis of the event-related brain otentials that advance information about a resonse seeds u the motoric ortion in a choice reaction time (RT) task. Reflexogenic and magnetic stimulation studies have shown that rearation even extends to quite distal motor levels (Brunia & Boelhouwer, 1988; Hasbroucq, Kaneko, Akamatsu, & Possamaï, 1997) Because rearatory rocesses have influences at various levels within the CNS, it is often difficult to find out whether a change in overt behaviour is caused by early or late rearatory rocesses (Gehring, Gratton, Coles, & Donchin, 1992). To romote the understanding of human erformance, however, it is theoretically desirable to determine the rearatory roerties of each level and how these roerties are related to certain asects of overt behaviour. The resent study contributes to this research area by investigating the influences of stimulus robability and resonse robability on the motor system. Whereas studies of rearation usually focus on measures of seed and accuracy, the resent research emhasizes a different asect, namely the forcefulness of reared and unreared resonses. Recent studies suggest that articiants exert less resonse force when they are highly reared for making a resonse (Jaśkowski & Verleger, 1993; Mattes et al., 1997; Mattes & Ulrich, 1997; Wascher et al., 1997). For examle, in the study of Mattes et al. (1997) articiants were to ress a force sensitive key with the index finger of their referred hand as soon as a resonse signal aeared. At the beginning of each trial a recue indicated the robability that the resonse signal would occur, with robabilities ranging from.1 to 1.0. As exected, articiants roduced shorter RTs when resonse robability was high than when it was low, indicating that rearation increased with resonse robability. However, rearation influenced not only the seed of the resonse but also how the resonse was erformed. Surrisingly, articiants roduced less forceful resonses when they were highly reared to resond. 1 In line with this finding are the results from recent studies which reorted that resonse force was increased in 1 Interestingly, articiants were not aware of this effect. A ost-exerimenta l interview revealed that most articiants misjudged the relation between resonse robability and force outut. Although these articiant s correctly recognized that rearation shortens RT, they incorrectly believed that rearation would increase resonse force.

4 STIMULUS PROBABILITY AND FORCE 479 go/nogo tasks when comared to simle and choice RT tasks (Miller, Franz, & Ulrich, 1999; Ulrich, Mattes, & Miller, 1999). Since the robability to resond was lower in the go/nogo tasks (due to inclusion of nogo-trials) than in the two other tasks, this increase of resonse force in the go/nogo task can be seen as further evidence for the conjecture that resonse force decreases with resonse robability. To account for their findings, Mattes et al. (1997) suggested that the rearatory effects reflected adjustments within the motor system. Secifically, a motor rearation hyothesis based on Näätänen s (1972) motor readiness model was advanced to account for the results. According to this model, the level of motor readiness rior to stimulus onset reflects the amount of rearation. Low resonse robability was assumed to result in oor motor rearation and, consequently, in a low level of motor readiness; conversely, high resonse robability would result in a reared motor system. When the resonse signal is resented, motor activation increases toward a certain threshold value (namely, action limit) to elicit a resonse. According to this modified model, higher resonse force on low robability trials could be due to a greater overshoot of motor activation when the motor system was unreared. In an unreared state the difference between the momentary level of activation (i.e., motor readiness) and the action limit is relatively large. Because a relatively large activation increment is needed, the motor system is unable to roduce a well-calibrated ortion of activation just sufficient to roduce the manual resonse. Instead, activation overshoots the action limit by an unnecessarily large amount, resulting in an esecially forceful resonse. In a reared state, by contrast, motor readiness is already close to the action limit, and the required increment of activation can be calibrated well, roducing only a small activation overshoot and consequently a less forceful resonse. Thus, this elaborated motor-readiness model rovides a ossible exlanation of why resonses are both more forceful and slower when they are unreared. The elaborated motor-readiness model of Mattes et al. (1997) gains some lausibility of its obvious efficiency. At the functional level, the observed relation of resonse robability and force might reflect an economical rincile of the motor system that evolved during hylogenesis. Otimizing a resonse might not only mean executing the resonse as quickly and accurately as ossible but also executing it with the smallest amount of energy. In a recent review of movement economy, Sarrow and Newell (1998) roosed that metabolic energy regulation is a fundamental rincile underlying the learning and control of motor skills. Assuming that resonse rearation is a functionally well-adated rincile in motor behaviour, it is not surrising that wellreared resonses are carried out with less force. The effort to reare a resonse ays out in less effort to execute it. This functional view also suggests that the effect on resonse force should be caused by motor rearation but not by rearation at a remotoric level.

5 480 MATTES, ULRICH, MILLER According to this motor readiness model the robability effect on resonse force originates within the motor system. However, as mentioned reviously, advance information should also cause rearatory adjustments at early levels within the stimulus resonse (S R) rocessing chain. This is because a maniulation of resonse robability in a go/nogo task changes not only the robability of erforming the motor act but also the robability of detecting and encoding the stimulus; thus stimulus-robability and resonse-robability were confounded in our revious study (Mattes et al.,1997). Hence, one might question whether the effect on resonse force is caused secifically by motor rearation or might instead be caused by rearation of remotoric rocesses. For examle, it has been reeatedly shown in RT research that remotoric rocessing links are suscetible to robability maniulations (e.g., Pashler & Baylis, 1991). If similar remotoric effects can be roved for resonse force, this would clearly suggest that rearatory adjustments at remotoric levels can modulate the rocessing stream downwards through the motor system and ultimately modulate the strength of a resonse. To find out whether such downstream effects are ossible, the resent exeriments examined effects of stimulus robability on resonse force. If resonse force deends on stimulus robability with resonse robability held constant, then the motoric interretation of the stimulus/resonse robability effect must be reevaluated. Such a finding would also have imlications for models of the interface between remotoric and motoric rocesses (cf., Ulrich, Rinkenauer, & Miller, 1998). By contrast, if rearatory rocesses at remotoric levels do not affect the force outut of a resonse, this would strengthen the idea of functionally encasulated modules according to which each module oerates in isolation (Fodor, 1983). In such a modular system it seems lausible to assume that rearatory effects oerating at remotoric levels should not influence the strength of a resonse. The general method we chose was to ma two stimuli with unequal robabilities onto the same resonse. This is a common aroach in RT research allowing the maniulation of stimulus robability while keeing resonse robability constant (Bertelson & Tisseyre, 1966; Biederman & Zachary, 1970; Dillon, 1966; LaBerge, Legrand, & Hobbie, 1969; LeBerge & Tweedy, 1964; Rabbitt, 1959; Sector & Lyons, 1976). For examle, LaBerge and Tweedy (1964) assigned two out of three stimuli to one resonse and the other stimulus to a different resonse. Thus, stimulus robability as well as resonse robability could be changed by varying the relative frequencies of the two stimuli that were both assigned to the same resonse. These authors reorted that stimulus robability exerted a larger effect on RT than did resonse robability. This technique has recently been alied to localize sequential effects in choice RT (Cambell & Proctor, 1993; Pashler & Baylis, 1991; Soetens, 1998). In general, the bulk of research clearly indicates that not only the robability of the resonse but also the robability of the stimulus affects the rocessing seed (for a review, see Gehring et al., 1992).

6 STIMULUS PROBABILITY AND FORCE 481 In summary, then, the following exeriments were designed to assess whether stimulus robability exerts effects on resonse force. If changes in stimulus robability do affect resonse force, this would argue against Mattes et al. s (1997) motor readiness exlanation of the resonse robability effect and strengthen the view that force is artially or comletely determined by the robability of the stimulus rather than the robability of the resonse. In contrast, if resonse force is not affected by stimulus robability, then it becomes more likely that the rearatory motor adjustment model is an accurate descrition of the mechanisms resonsible for the effect of robability on resonse force. In Exeriment 1 three letters were emloyed as stimuli. The articiants were instructed to ress the same key when either of two letters aeared but to withhold the resonse when the third letter was resented. Table 1 shows the stimulus robabilities for the three stimuli. The letters A, B, and C exemlify the stimulus letters. The two stimuli that required a resonse (A or B) had different robabilities (.64 vs.16). If the dynamics of the resonses do not deend on stimulus robability, then resonse force should not differ between the two stimuli. Exeriment 2 in addition varied resonse robability, which was announced by a recue as in the study of Mattes et al. (1997) (Table 2). TABLE 1 Stimulus robabilities emloyed in Exeriment 1 Probability RT (ms) Force (cn) Go a go A B C a Resonse: Keyress with index finger to any go-stimulus. TABLE 2 Stimulus robabilities emloyed in Exeriment 2 Go go A B C Probability: Precue 20% RT (ms) Force (cn) Precue 80% RT (ms) Force (cn)

7 482 MATTES, ULRICH, MILLER TABLE 3 Stimulus robabilities emloyed in Exeriment 3 Go go A B C D Probability: Precue 20% RT (ms) Force (cn) Precue 80% RT (ms) Force (cn) Exeriment 3 was designed to investigate the influence of resonse robability when stimulus robability is held constant (Table 3). EXPERIMENT 1 This exeriment tested if mere stimulus robability affects the force outut of a resonse. Three different letters, one of which was resented in each trial, served as stimuli. The articiants were instructed to resond to two letters (resonse letters) with the same keyress, but to withhold the resonse to the third letter (no-resonse letter). Most imortant, the two resonse letters aeared with unequal frequencies. One was resented four times as often as the other, resulting in stimulus robabilities of.16 and.64 for the two resonse letters, while the sum of both stimulus robabilities constitutes the resonse robability.8. If these two resonse letters roduce different resonse force, we can conclude that resonse force is sensitive to stimulus robability. After the exeriment a questionnaire was given to the articiants as in our revious study (see Aendix). This questionnaire assessed whether articiants were aware of the effects of robability on RT and resonse force. In our revious study articiants correctly redicted that resonse robability shortens RT; however, interestingly enough, articiants were unaware of the influence of resonse robability on resonse force. Secifically, articiants incorrectly believed that resonse force would increase with resonse robability. Therefore, it seems interesting to know whether the same misconcetion alies to stimulus robability. Method Particiants. Nineteen male and 17 female students (mean age: 25.7 years) took art in a single session. Two of them claimed to be left handed. They were aid for their co-oeration and were naive about the urose of the exeriment.

8 STIMULUS PROBABILITY AND FORCE 483 Aaratus and stimuli. The exeriment was carried out in a dimly illuminated room. A microcomuter controlled signal resentation and recorded resonse force. All signals were resented in the middle of a comuter screen. A grey cross (1.1 ) served as warning signal and fixation cross. The stimuli were the caital letters X, S, and G. They were light grey, subtended a visual angle of 0.6 horizontally and 1 vertically and their intensity was 41 cd/m 2 against a background intensity of about 0.25 cd/m 2. Resonses were measured means of a force sensitive key, which consisted of a leaf sring ( mm) with strain gauges attached on it. The leaf sring was fixed at one end and any force alied to the free end was registered with a rate of 500 Hz throughout the whole trial. The articiant ressed this key with the index finger of his or her referred hand while the forearm rested comfortably on the table. The average resonse force of 695 cn that was registered in this exeriment bent the sring aroximately 0.7 mm. The resolution of this device was about 2 cn. RT was measured as the interval between the onset of the resonse signal and the oint in time when resonse force attained the criterion force of 50 cn. A chin rest was used to maintain a constant osture and viewing distance of 60 cm. Procedure. A session lasted about 50 min and consisted of 550 trials. The first 50 trials were considered ractice and excluded from data analysis. The whole set of 500 exerimental trials was randomized and artitioned into blocks of 50 trials each. At the end of each block articiants received feedback about mean RT and resonse errors for their erformance in this block. Particiants were encouraged to resond quickly but to avoid errors. They initiated each block with a keyress on the comuter keyboard with their nonresonding hand. Particiants resonded to two letters with their dominant hand but had to withhold the resonse when the no-resonse letter was resented. They were told which of the two resonse letters would be resented four times as often as the other resonse letter. The assignment of the three letters as frequent resonse letter, infrequent resonse letter, and no-resonse letter was balanced over articiants. A trial started with the resentation of the warning signal (fixation cross) for 300 ms. The interval between the offset of the warning signal and the onset of the letter was 1750 ms. The letter was resented for 150 ms. The comuter dislayed an error message if the articiant failed to resond to a resonse letter within 1000 ms or if he or she resonded to the no-resonse letter. Other aroriate error messages, for examle if articiants ressed the key before stimulus onset, were given as well. The next trial started 2 s later. Design. The session was slit into two halves to assess otential ractice effects. Thus, there were two within-subjects factors stimulus robability (.64 vs.16) and ractice (first vs second half of session). The deendent variables

9 484 MATTES, ULRICH, MILLER were mean RT, mean resonse force, and error rates. As in revious studies (e.g., Mattes et al., 1997) the maximum force level (eak force) of each trial was determined as an index of resonse strength. Questionnaire. The articiants comleted a questionnaire when they had finished the exeriment. The questions assessed whether the articiants could redict the results of the exeriment. The comlete questionnaire along with the results is rovided in the Aendix. Results A searate ANOVA was conducted for each deendent variable. Table 1 shows data for stimulus A and B. Resonse errors. Anticiations (RT < 100 ms) and misses (RT > 1000 ms) occurred in 0.3% and 0.5% of the trials, resectively. These figures were too low to ermit a meaningful statistical analysis. There were 2.4% false alarms, that is, resonses to the no-resonse letter. Only trials without errors were included in the following analyses. Reaction time. Particiants resonded faster to the more frequent stimulus (327 ms) than to the less frequent stimulus (369 ms), F(1, 35) = 61.9, <.001. Practice yielded a main effect, with mean RTs being 358 ms in the first half of the exeriment and 338 ms in the second half, F(1, 35) = 24.7, <.001. However, ractice did not modulate the effect of stimulus robability on RT, F(1, 35) = 1.6, >.2. Resonse force. More interestingly for the uroses of this aer, resonse force was influenced by stimulus robability, F(1, 35) = 7.0, =.012. The mean resonse force was 704 cn for the infrequent letter and 686 cn for the frequent one. Although resonse force decreased from the first to the second half of the exeriment from 750 cn to 640 cn, F(1, 35) = 7.2, =.011, ractice did not modulate the effect of stimulus robability (F < 1). Questionnaire. The Aendix rovides the results of the questionnaire. Particiants correctly judged how stimulus robability would influence RT and resonse force, although there was less agreement on the effect of resonse force than on RT. Discussion This exeriment showed that with a high robability stimulus RTs become faster and resonses are erformed with less force. The first finding was of course exected on the basis of earlier studies that varied stimulus robability

10 STIMULUS PROBABILITY AND FORCE 485 (e.g., Bertelson & Tisseyre, 1966; LaBerge & Tweedy, 1964), but the evidence concerning resonse force is new. In articular, the finding that resonse force deends on stimulus robability shows that the motor-readiness model of Mattes et al. (1997) outlined in the introduction is at best an incomlete account of the effect of stimulus resonse robability on resonse force. Before considering alternative exlanations for the effects of stimulus and resonse robability on force, we first sought to examine the combined effects of stimulus and resonse robability in Exeriment 2. EXPERIMENT 2 Exeriment 2 was designed to test whether the stimulus robability effect found in Exeriment 1 is modulated by resonse robability. To this end, the aradigm of Exeriment 1 was sulemented by a condition with low resonse robability. In each trial a recue announced the robability to resond (.2 or.8). The stimulus set and the resonse instructions were identical to Exeriment 1, that is, one resonse letter was resented four times as often as the other resonse letter, resulting in conditional stimulus robabilities of.2 and.8 for the infrequent and frequent resonse letter, resectively. Method The method was identical to Exeriment 1 excet some modifications as indicated in the following. Particiants. Thirty-six students (18 male, 18 female; mean age: 26.2 years) took art in a single session. Three of them claimed to be left handed. They were aid for their co-oeration and were naive about the urose of the exeriment. ne of them had articiated in Exeriment 1. Procedure. As in our revious study (Mattes et al., 1997), resonse robability was announced at the beginning of each trial. One of two ercentages (20% or 80%, visual angle of 4.5 horizontally and 1.6 vertically) was resented in red or green, resectively. In addition to Exeriment 1, the articiants were told that the ercentage number at the beginning of each trial reresented the robability that they would have to resond at all. Particiants were again told which of the two resonse letters would be resented four times as often as the other resonse letter. Table 2 rovides the stimulus robabilities for both resonse robability conditions. A trial started with the resentation of the resonse robability cue for a duration of 500 ms. After a blank eriod of 800 ms a warning signal (fixation cross) of 300 ms duration followed. The interval between the offset of the

11 486 MATTES, ULRICH, MILLER warning signal and the onset of the letter was 1750 ms. The letter was resented for 150 ms. Each resonse robability was emloyed in half of the trials in random order. Design. Crossing of the within-subject factors of conditional stimulus robability 2 and resonse robability (.2 vs.8, due to the advance cue) resulted in four exerimental conditions. As in Exeriment 1, ractice (first vs second half of exeriment) was included as an additional factor to assess whether otential robability effects are modulated by ractice. Questionnaire. Again a questionnaire was given at the end of the session to find out whether the articiants could redict the results of the exeriment (see Aendix). In comarison to Exeriment 1 this questionnaire contained additional questions concerning the resonse robability recue. Results A reeated measures ANOVA was conducted for each deendent variable. Table 2 shows data for stimulus A and B for both levels of resonse robability. Resonse errors. The ercentages of anticiations (RT < 100 ms) and misses (RT > 1000 ms) were 0.01 and 0.6%, resectively. These trials were excluded from further data analysis. As one might exect, resonses to the noresonse letter were more frequent in trials with high resonse robability (5.0%) than in trials with low resonse robability (2.8%), F(1, 35) = 6.3, <.05. Reaction time. Mean RT was 380 and 370 ms for resonse robabilities.2 and.8, resectively, F(1, 35) = 8.8, <.01. Stimulus robability also revealed a highly significant main effect, F(1, 35) = 61.1, <.001, with mean RTs of 357 ms for the frequent resonse letter and 393 ms for the infrequent one. The interaction of the two factors did not reach statistical significance, F(1, 35) = 1.8, <.2, indicating that the effect of stimulus robability on RT does not deend on the absolute level of resonse robability. The ractice factor revealed the exected learning effect with mean RTs declining from 383 ms to 367 ms between the first and the second half of the exeriment, F(1, 35) = 18.9, <.001, but showed no interaction with either robability factor. The threeway interaction was also not significant (F < 1). 2 In trials with resonse robability.2 the unconditione d stimulus robabilities were now.16 and.04, and in trials with resonse robability.8 these figures changed to.64 and.16 for the frequent and the infrequent resonse letter, resectively.

12 STIMULUS PROBABILITY AND FORCE 487 Resonse force. As resonse robability increased, resonse force decreased from 647 to 628 cn, F(1, 35) = 10.9, <.01. In agreement with Exeriment 1, resonse force was also strongly influenced by stimulus robability. Mean force was 653 cn for the infrequent letter and 623 cn for the frequent letter, F(1, 35) = 11.7, <.01. The interaction between the two robability factors was not significant, F(1, 35) = 2.4, <.2. As in Exeriment 1, resonses were more forceful in the first half of the exeriment (704 cn) than in the second half (571 cn), F(1, 35) =17.8, <.001. However, ractice did not modulate either the effects of stimulus or resonse robability (Fs < 1) or the interaction between them (F = 1.1). Questionnaire. The results of the questionnaire are rovided in the Aendix. Generally, the articiants were quite accurate in their redictions for RTs but not for resonse force. As in our revious study (Mattes et al., 1997) articiants believed incorrectly that an increase of resonse robability would increase resonse force. However, there was less agreement on how stimulus robability would affect resonse force. Discussion This exeriment relicates the findings of Exeriment 1 that stimulus robability affects both RT and resonse force. In addition, the size of the effect on resonse force was not modulated by resonse robability. The articiants again resonded faster and less forcefully to the more frequent stimulus as comared to the less frequent one. Surrisingly, the effect of stimulus robability on resonse force was even larger than that of resonse robability. Unfortunately, we cannot derive accurate measures for the relative contributions of stimulus and resonse robability from these data. The factor of resonse robability, desite its interretative labelling, is still confounded with stimulus robability. This is because the advance robability cue changes not only the robability of resonding but of course also the robability of each stimulus. As can be seen from Table 2, stimulus robabilities of.16 and.04 for the 20% recue change to.64 and.16 when the 80% recue announces a higher resonse robability. Thus, in view of the effects of stimulus robability on resonse force and the confounding of resonse robability with stimulus robability, it remains unclear whether resonse robability actually lays any role at all in the resent designs. This question of a ossible effect of resonse robability on force is of considerable imortance since the exlanation in terms of the motor readiness model outlined in the introduction was entirely based on the assumtion that different levels of resonse robability result in corresonding adjustments of the motor system. Although Exeriment 1 clearly shows that stimulus

13 488 MATTES, ULRICH, MILLER robability affects resonse force, it is still ossible that resonse robability also does so. For examle, it is conceivable that a maniulation of resonse robability exerts a secific effect on motor rearation (e.g., as assumed by the motor-readiness model), whereas a maniulation of stimulus robability exerts an indirect effect on resonse force that is mediated by non-motoric rocesses. Hence, if both factors roduce relatively searable effects on resonse force, a significant main effect of resonse robability would still suort the motor readiness model as a artial exlanation of the S R robability effect. Although it aeared that resonse robability layed a minor role in Exeriment 2, there is nevertheless some evidence that it did have some effect. False alarm resonses to the no-go letters occurred more frequently when resonse robability was high, in accordance with the idea that motor readiness was closer to the motor action limit when resonse robability was high. Thus, this asect of the results is consistent with Mattes et al. s (1997) motor-readiness model. This evidence is indirect, however, and it would be more convincing to demonstrate an effect of resonse robability in a design where resonse robability was not confounded with stimulus robability. Exeriment 3 was designed to tackle this matter. EXPERIMENT 3 Exeriments 1 and 2 rovided evidence that stimulus robability exerts an effect on resonse force. The question arises whether resonse robability has any influence at all on resonse force, because resonse robability was confounded with stimulus robability both in our earlier study (Mattes et al., 1997) and in Exeriment 2 of the resent study. Exeriment 3 was designed to test the effects of resonse robability on RT and resonse force for constant levels of stimulus robability. To achieve this, the design of Exeriment 2 was slightly changed. This exeriment emloyed four stimuli, three of which were resonse letters and the fourth of which was the no-resonse letter (see Table 3). One resonse letter could be resented in both resonse robability conditions, and the other two resonse letters were each assigned to just one of the two resonse robability conditions. For examle, assume that the resonse letters were A, B, and C and the noresonse letter was D. As in Exeriment 2 each trial was receded by a recue 20% or 80% that announced the resonse robability for that trial. Then the recue 20% would be followed either by the resonse letter A with =.04, by B with =.16, or by the no-resonse letter D with =.80. The recue 80% would either be followed by the resonse letter A with =.64, by the third resonse letter C with =.16, or by the no-resonse letter D with =.20. Only the resonses to B and C are of theoretical interest and thus will be analysed

14 STIMULUS PROBABILITY AND FORCE 489 further. These stimuli have identical conditional occurrence robabilities (.16) as well as identical overall resentation frequencies (8% of all trials each). However, the resonse to the letter B has an a riori robability of.2, whereas resonse robability for C is.8. If resonse robability has the exected effect on resonse force, then resonse force should be higher for stimulus B than for stimulus C. In contrast, if resonse robability does not affect resonse force, then resonse force should be identical for these two letters because they have identical stimulus robabilities. Method Particiants. Thirty-six students (sixteen male, twenty female; mean age: 26.7 years) took art in a single session. Three of them claimed to be left handed. They were aid for their co-oeration and were naive about the urose of the exeriment. ne of them had articiated in Exeriment 1 or 2. Aaratus and stimuli. This was identical to the revious exeriment with the following excetions. The letters G, H, M, and S served as stimuli. Assignment of stimuli to conditions was balanced over articiants. A frame centred around the osition of stimulus resentation was visible throughout the trial. The frame had the same colour as the recue to enhance its efficiency. The outer border of the frame was 9.9 wide, 7.2 high, and the line had a thickness of 0.7. The colours of the recues were changed to brown and blue, because the colours of the revious exeriment roduced undesirable after-effects in the area of the frame when they were changed at the beginning of a trial. Procedure. The course of a trial was identical to Exeriment 2, excet for the coloured frame, which was visible during the whole trial. It changed its colour when the resonse robability changed. Particiants were told that they should resond to three letters but withhold the resonse when the fourth letter was resented. They were further told that one resonse letter would be resented more often than the others and that the other two resonse letters would be resented with equal frequency. The resulting robabilities are rovided in Table 3. Design. Only the factor resonse robability (.2 vs.8) was of interest in this exeriment. As in the revious exeriments a ractice factor (first vs second half of exeriment) was included to take ractice effects into account. Questionnaire. The questionnaire of Exeriment 2 was only slightly changed in wording to account for the frame that accomanied the recue (see Aendix). All other questions were identical.

15 490 MATTES, ULRICH, MILLER Results Searate reeated measures ANOVAs were conducted for each deendent variable. The analysis only included the resonses to the stimuli B and C because this comarison is critical for isolating the effect of resonse robability on resonse force. Table 3 shows data for stimulus B and C and, for comleteness, also the data for stimulus A. Resonse errors. There were 0.2% anticiations (RT < 100 ms) and 0.3% misses (RT > 1000 ms). These trials were excluded from further analysis. Resonses to the no-resonse letter were more frequent in trials with high resonse robability (18.1%) than in trials with low resonse robability (4.1%), F(1, 35) = 33.1, <.001. Reaction time. RT analyses revealed a significant main effect of resonse robability, F(1, 35) = 10.9, =.002. Mean RT was 406 ms for resonse robability.2 and 387 ms for resonse robability.8. RT decreased with ractice from 402 ms in the first half of the session to 391 ms in the second half, F(1, 35) = 5.0, =.031. The difference between the two resonse robability conditions was 13 ms in the first half and 24 ms in the second half, but this interaction was not significant, F(1, 35) = 2.3, =.139. Resonse force. Most imortant, there was a highly significant main effect of resonse robability on force, F(1, 35) = 21.8, <.001. Mean eak force was 724 cn when resonse robability was.2 and 685 cn when resonse robability was.8. Peak force decreased from the first to the second half of the exeriment but this ractice effect was not significant, F(1, 35) = 1.9, =.17. The interaction of both factors was also not significant (F < 1). Questionnaire. The results of the questionnaire are rovided in the Aendix. te that the articiants redicted fast but weak reactions for high stimulus robability, whereas they exected fast and strong reactions for high resonse robability. This suggests that the redictions on force are not simly guided by the beliefs about RT but are rather generated indeendently. In sum, the redictions about RT were again correct, whereas the redictions about resonse force were correct for stimulus robability but incorrect for resonse robability. Discussion In this exeriment different levels of resonse robability were announced by a recue. Two stimuli of equal frequency were assigned to the two resonse robability conditions to avoid a confound with stimulus robability. The

16 STIMULUS PROBABILITY AND FORCE 491 results demonstrate that resonse robability affects RT and resonse force even when stimulus robability is held constant. 3 One might object that stimulus exectancy for stimuli B and C differs desite the identical stimulus robability. When the recue had announced low resonse robability, stimulus B was the more exected stimulus of the two ossible stimuli (B vs A). In contrast, when the recue had announced high resonse robability, stimulus C was the less exected one (C vs A). Such a otential influence of stimulus exectancy, however, is not crucial for the main conclusion of this exeriment. First, a comarison of stimulus A and B reveals only a small and insignificant RT effect, t(35) = 0.77, >.4, suggesting that stimulus exectancy has little effect in this situation. Second, and more imortant, a otential stimulus exectancy effect would tend to counteract the resonse robability effect, because articiants tend to resond more quickly and less forcefully to exected stimuli than to unexected ones. Therefore, at the worst, the effect of resonse robability could be somewhat underestimated in the resent exeriment and the conclusion that resonse robability exerts an effect on RT and resonse force would still be warranted. GENERAL DISCUSSION Human erformance benefits greatly from rearation. Covert rearation is assumed to increase both the seed and the accuracy of various information rocessing mechanisms, as demonstrated by the measurement of RT and resonse errors. Recent research has revealed that rearation even affects the force outut involved in erforming a resonseresonses in a simle RT task to stimuli with low occurrence robability are relatively forceful (Jaśkowski & Verleger, 1993; Mattes et al., 1997). As mentioned in the introduction this robability effect may reflect resonse rearation at a motoric level roducing not only short RTs but also a moderate force outut that is aroriate to erform the task. The rimary question in the resent study was whether the effects of robability on resonse force generally reflect resonse rearation at a motoric level or might even be caused by rearation at receding levels. To answer this question we maniulated stimulus robability while keeing resonse robability constant, and vice versa. As argued in the introduction, the effects of robability on force in designs confounding stimulus robability with resonse robability may actually have been due to stimulus robability rather than resonse robability. 3 te that the effect of resonse robability in Exeriment 3 was even somewhat larger than in Exeriment 2 (39 cn vs 19 cn) desite the confound with stimulus robability in Exeriment 2, which should have exaggerated the effect.

17 492 MATTES, ULRICH, MILLER Exeriment 1 and 2 evaluated the effects of stimulus robability while keeing resonse robability constant. Particiants ressed a key when one of two resonse letters aeared but refrained from resonding when the noresonse letter was resented. One of the resonse letters aeared more often than the other. Particiants resonded faster and less forcefully to the more robable of the two resonse letters. Exeriment 2 showed that this stimulus robability effect is obtained whether the robability of resonding is low or high. Exeriment 3 made it lain that a maniulation of resonse robability affects resonse force. Although stimulus robability was constant, resonses were faster and less forceful when resonse robability was high than when it was low. Thus, the conclusion that emerges from the results of all three exeriments is that both stimulus robability and resonse robability affect resonse force in the absence of variations in the other factor. A noteworthy difference between the resent study and our earlier one is that in the earlier one resonse robability was given by the frequency of stimulus trials among blank trials, whereas in the resent study a stimulus was resented in every trial and the articiants had to decide whether or not the stimulus was a resonse letter. Thus, the resonse robability effect could be demonstrated for a task when stimulus detection is sufficient to trigger a resonse and also for a task that requires identification of the stimulus. A ost-exerimental questionnaire revealed that the articiants were not fully aware of the directions of the effect of stimulus and resonse robability on force. Although most articiants correctly recognized that faster RTs are associated with higher stimulus and resonse robabilities, there was somewhat less agreement on how the two tyes of robabilities affect resonse force. Generally, articiants tend to resume correctly a decrease of resonse force when stimulus robability increases. By contrast, they incorrectly believe that increasing resonse robability yielded more forceful resonses. This differential suggests that the articiants judgements about resonse force are not merely the counterart of their beliefs about RT. The main conclusion from the resent study is that an interretation of robability-related changes in motor erformance requires consideration of remotoric stages of stimulus rocessing. The increase of resonse force relatively unlikely stimuli resonse airs cannot solely be attributed to rocesses that oerate at a muscular level, because stimulus secific rearation also affects resonse force. More generally, this finding sheds some light on the structure of cognitive rocesses. The demonstration that stimulus robability affects resonse force aears to be in conflict with the notion of functionally encasulated modules. In such a modular system it is difficult to see how rearatory effects oerating at a remotoric level can influence the strength of a resonse. However, it is not entirely clear whether this rearatory effect reflects structural rocessing features or stems from non-secific activation rocesses that

18 STIMULUS PROBABILITY AND FORCE 493 byass the comutational rocessing route. These and further interretations are discussed next. Motor-readiness model An extended version of Näätänen s (1971) motor-readiness model can account for resonse robability effects on resonse force, although it cannot account in its resent formulation for the stimulus robability effects on resonse force. As we outlined in the introduction, this model assumes that oor motor rearation results in a high activation overshoot. Because low resonse robability is thought to result in oor motor rearation, it follows that manual resonses should be executed with more force when resonse robability was low. This extended motor readiness model can exlain the effects of resonse robability in the resent study and our revious study (Mattes et al., 1997). According to the model, robability effects are due to motor adjustments that take lace rior to the onset of the resonse signal. However, in the case of stimulus robability effects, the level of readiness at the moment of stimulus resentation should always be the same, because resonse robability was controlled. Therefore, the model in its resent formulation cannot exlain the observed force effects associated with the maniulation of stimulus robability. One might seculate whether a rincile similar to the one assumed in the motor-readiness model might oerate at an earlier level. For examle, the robability effect on resonse force could arise in the resonse selection stage where stimuli are assigned to their associated resonses. When the link between a stimulus and its associated resonse is weak more activation might be necessary to trigger the resonse. For examle, more activation could be needed to overcome cometition with more robable links. As in the original formulation of the model, this would imly a resonse roduction mechanism which inherently roduces more motoric activation whenever a weak stimulus resonse link must be rocessed. In the remainder of this discussion, we will consider some further ossible interretations of the resent results. The next two subsections consider alternative accounts for the stimulus robability effect and comare these to the modified motor-readiness model. The third section considers recent neurohysiological findings that might be imortant for localizing the effects on force within the CNS. Comensation hyothesis One alternative exlanation for the force effects obtained in this study comes from a comensation hyothesis that was suggested by Jaśkowski and Verleger (1993). According to this hyothesis, articiants want to comensate oor rearation at the moment of stimulus detection by activating more motor

19 494 MATTES, ULRICH, MILLER neurons. These authors observed that articiants roduce more forceful resonses under seed stress and thus concluded that articiants increase the force outut to generate shorter RTs (Jaśkowski, Verleger, & Wascher, 1994). Although the authors did not further secify this hyothesized comensation rocess, there are at least two versions conceivable. First, one may assume an automatic mechanism that times the resonse and if the redicted RT turns out to be slow, this mechanism increases the force level of the resonse outut. This version suggests a ositive correlation between RT and force; slow resonses should be generally associated with higher resonse force due to the comensatory mechanism. Several studies, however, reorted correlations for RT and resonse force that were calculated across trials within each exerimental condition, so that the correlation was not influenced by the exerimental maniulation (Giray, 1990; Giray & Ulrich, 1993; Mattes et al., 1997; Miller et al., 1999; Mordkoff, Miller, & Roch, 1996; Ulrich & Mattes, 1996; Ulrich et al., 1998). The finding that the average correlation has always been close to zero is difficult to exlain within the latter version of the comensation hyothesis. The second version resumes that the comensation tendency is strategically emloyed rather than an automatic rocess. Accordingly, the comensation rocess is not based on online information about the redicted slowness of the forthcoming resonse. Instead, the articiants decide in advance to resond with more force to a stimulus which they susect roduces slower reactions, e.g., one with low robability. However, the results of the questionnaire rovide evidence against this version of the comensation hyothesis. If the articiants emloyed such a comensation strategy, they should have been able to redict correctly how robability influences resonse force, just as they were for RT. However, they tended to exect increased resonse force only for low stimulus robability, and not for low resonse robability. Finally, and robably more crucial for the comensation hyothesis, both versions imly that every factor which shortens RT should also roduce less forceful resonses. Therefore, it is difficult to see how the comensation hyothesis can exlain why RT decreases yet resonse force increases with stimulus intensity (Angel, 1973; Jaśkowski, Rybarczyk, Jaroszyk, & Lemanski, 1995; Miller et al., 1999), stimulus duration (Ulrich et al., 1998), the number of stimuli (Giray & Ulrich, 1993; Mordkoff et al., 1996), word frequency in a lexical decision task (Abrams & Balota, 1991; Balota & Abrams, 1995), and with the set size in a memory scanning task (Abrams & Balota, 1991). Arousal effects A third ossible exlanation of the stimulus robability effect roceeds from the well-established fact that unlikely events roduce a non-secific arousal effect. For examle, infrequent events are known to elicit certain neuronal

20 STIMULUS PROBABILITY AND FORCE 495 activity atterns, as documented for an event-related otential comonent known as the mismatch negativity (Näätänen, 1995). Another comonent that is also sensitive to the robability of stimulus events is the P300. The amlitude of this ositive deflection increases as the robability of a rare event decreases. For examle, Gehring et al. (1992) found a larger P300 for target letters that were not redicted. These findings demonstrate that deviant or rare events seem to elicit certain activity atterns in the brain. If an infrequent stimulus generally elicits some non-secific arousal effects, this activation could result in a higher resonse force (Miller et al., 1999; Ulrich & Mattes, 1996). This activation might even exert an indirect influence on the motor system, which is mediated by a route that byasses the information rocessing channel (Giray, 1990; Kramer & Sinks, 1991; Miller et al., 1999; Öhman, 1987). At resent, the arousal exlanation of stimulus robability effects on force must be regarded as at least as lausible as the modified motor-readiness model, because of the clear-cut evidence that low robability stimuli increase arousal (Sokolov, 1963) and that increases in arousal tend to increase resonse force (Jaśkowski, Wroblewski, & Hojan-Jezierska, 1994; Miller et al., 1999; Ulrich & Mattes, 1996). Resonse force and cortico-sinal excitability Finally, reflexogenic studies and studies that have emloyed transcranial magnetic stimulation rovide some suggestive hints to localize force effects within the CNS. These studies robe sinal and cortico-sinal excitability, resectively. Reflexogenic studies (Brunia & Boelhouwer, 1998; Requin, Bonnet, & Semjen, 1977) examined sinal reflex athway activity during the foreeriod between a warning signal and the resonse signal. Monosynatic reflexes were evoked during this interval, and the sizes of the evoked reflexes were unrelated to resonse robability. This result suggests that the robability effect on resonse force does not reflect changes in sinal excitability and is thus not a eriheral motor henomenon. However, as Hasbroucq et al. (1997) ointed out, the use of reflexogenic techniques is restricted to lower limbs, and functional differences of the uer and lower limbs in humans might render a generalization of reflexogenic studies unwarranted. Therefore, these authors robed the motor excitability of the uer limbs during the foreeriod in a RT task by emloying transcranial magnetic stimulation. Their study revealed a decrement of cortico-sinal excitabilityresumably due to cortical modulations (Hasbroucq, Kaneko, Akamatsu, & Possamaï, 1999a)when the foreeriod condition facilitated resonse rearation. Hasbroucq et al. suggested that this decrement reflects an adative mechanism increasing the sensitivity of the motor structures to the forthcoming voluntary command. They further seculated that such a mechanism could consist of the active filtering of task-unrelated afferents to