Rate of processing and judgment of response speed: Comparing the effects of alcohol and practice

Transcription

1 Pereption & Psyhophysis 1989, 45 (4), Rate of proessing and judgment of response speed: Comparing the effets of alohol and pratie E. A. MAYLOR, P. M. A. RABBITT, and S. A. V. CONNOLLY University of Manhester, Manhester, England Both alohol and pratie affet hoie reation time. The present study was onduted to investigate the possibility that impairment from alohol and improvement with pratie ould be attributed to hanges in the effiieny of ontrol mehanisms (Rabbitt, 1979a), some of whih depend upon the ability to judge response speed aurately. Twenty subjets partiipated in a four-hoie reation time experiment in whih they reeived no alohol (NA) in Session 1 and either no alohol (10 subjets) or 0.8 ml alohol (A) per kilogram of body weight (10 subjets) in Session 2. The task was to respond as fast and as aurately as possible to eah stimulus. In addition, subjets were required to press a fifth key after any response that they onsidered to be both fast and aurate. Subjets had no diffiulty in performing this task: (1) there was a signifiant differene of 122 mse between the mean response time for orret responses indiated as fast and that for orret responses not indiated as fast, and (2) subjets indiated 1 in 4 orret responses but only 1 in 64 errors. Alohol inreased all response times by approximately 40 mse. In ontrast, pratie dereased response times less for orret responses not indiated as fast than for orret responses indiated as fast. The ability to distinguish between fast and slow responses was thus unaffeted by alohol, but was improved by pratie. Responses indiated as "fast" were signifiantly faster than errors, and appeared to our without warning (unlike errors, whih tended to end a sequene ofinreasingly fast orret responses). The results suggest that alohol and pratie influene hoie reation time in qualitatively different ways: Alohol impairs overall response speed but has no effet on the ability to judge response speed, whereas pratie improves both. Reent models ofhoie reation time have emphasized the importaneoftheability to monitorthe response, both in terms ofspeed and auray. For example, Rabbitt and Vyas (1970) proposed a traking model for performane oftasks in whih the subjet is required to respond as fast and as aurately as possible. This instrution implies that "the subjet knows preisely how fast he an go without running a risk of making errors" (p. 62). Rabbitt and Vyas suggested that the speed-error tradeoff funtion, whih desribes the hange in probability that a response will be an error, ontingent upon the speed with whih it is made, limits the effiieny ofperformane (see, e.g., Shouten & Bekker, 1967). The subjet has to disover the position ofthis funtion to perform optimally. For example, he/she must first make errors by responding too quikly in order to know how fast he/she an respond without making errors. Thus the subjet must "trak" the response rate to stay within the reation time region This researh was supported by Grant G from the Medial Researh Counil ofgreat Britain. The authors are grateful to Hermann Mueller for help with omputer programming. Requests for reprints should be addressed to Elizabeth Maylor at the Age and Cognitive Performane Researh Centre, Oxford Road, Manhester M13 9PL, England. that maximizes speed while maintaining auray (Rabbitt, 1979a, 1980). Rabbitt (1979b) noted that subjets need to know at least three things to ontrol their performane in this way: (1) They must detet errors when they our; (2) they must be able to monitor and vary response speed so as to reognise, and avoid, faster (reation-time) bands at whih errors are intolerably frequent; (3) they must be able adaptively to ontrol response speeds so as to selet optimal (reation time) bands and avoid others whih may be too fast or too slow. (p. 305) It is well established that performane in hoie reation time tasks is impaired by alohol (Jennings, Wood, & Lawrene, 1976) and improved by pratie. Depending upon dose, alohol adversely affets speed (Maylor, Rabbitt, Sahgal, & Wright, 1987), or auray (Wilkinson & Colquhoun, 1968), or both speed and auray (Maylor & Rabbitt, 1987). With pratie, subjets beome faster, partly as a result ofa hange in speed-error tradeoff toward less aurate performane, but also beause of a real inreasein proessing rate (Maylor & Rabbitt, 1989). The impairment with alohol and the improvement with pratie an be onsidered in terms of the three aspets ofontrol desribed by Rabbitt (1979a, 1979b). Preliminary evidene from Maylor and Rabbitt (1987) suggests 431 Copyright 1989 Psyhonomi Soiety, In.

2 432 MAYLOR, RABBITT, AND CONNOLLY that alohol does not influene the ability to detet an error and make an immediate harateristi adjustment to the response riterion (see Rabbitt, 1966). There is less evidene relating to the ability to distinguish between a fast and a slow response with and without alohol and pratie. As Rabbitt (1979b) noted, very fine degrees of temporal disrimination are required in hoie reation time tasks sine the range over whih speed-error tradeoff operates is only approximately 200 mse, for mean reation times of mse with standard deviations of mse (see Maylor & Rabbitt, 1987). There is evidene from Sanford (1970) that for auditory reation time, subjets an rate their responses with regard to speed in a meaningful way. An attempt was made to investigate this with alohol by interrupting subjets in a hoie reation time task and asking them to rate the speed of the last response made with respet to the average speed of the previous blok oftrials (Mayloret al., 1987). Thesubjets estimated the speed of their responses with only hane auray, whether they had onsumed alohol or not. Also, the information transmitted, as a proportion of the information presented, was low and unaffeted by alohol. However, the number of trials in a session was large ompared with the number of interruptions requesting ratings, so that it was suggested that the judgment task may have appeared relatively unimportant. Also the requests for ratings were distrating and the times taken to give ratings were long, with several subjets ommenting that, beause of the interruption, they had forgotten the speed of their last response. The present study was an attempt to provide more aurate ratings by requiring the subjets, in effet, to rate every response. They were asked to press a separate key immediately after making any response that they onsidered to be partiularly fast. We adopted an experimental design in whih eah oftwo groups partiipated in two sessions. Both groups reeived no alohol in Session 1; in Session 2, one group again reeived no alohol (NA/NA) and the othergroup was given alohol (NAJA). The advantages of this design inlude the following: (1) the omparisons of interest are essentially withinsubjets, (2) pratie effets are taken into aount, and (3) only halfthe subjets are required to onsume alohol. MEmOD Subjets Twenty-two subjets (mostly students and tehnial staff) volunteered in response to advertisements at the University ofmanhester. The data from 2 of these subjets were subsequently deleted from the analysis for reasons given in the Proedure. Details of the remaining 20 subjets are given in Table I. Eah subjet partiipated in two I-h sessions at the same time on onseutive days. Payment on ompletion of the experiment was 10. Apparatus and Stimuli The subjets were given alohol in the form of vodka (37.5% alohol by volume) mixed with pure orange juie and water (see Proedure for exat quantities). A Lion Alolmeter S-D2 Table 1 Subjet Details Group Age (in years) Weight (in kilograms) Time of Testing (in hours:minutes) 13:06 1:48 13:30 2:12 Note-NA = no alohol. A = alohol. NAmA (N = 10) NAtA (N = 10) M SD M SD (Breathalyser) obtained from Lion Laboratories Ltd. (Barry, UK) was used to estimate blood alohol onentration to the nearest 5 mg alohollloo rnl blood. An LSI omputer with a Cambridge Eletronis Design 502 laboratory interfae was used to ontrol the experiment and reord responses with milliseond auray. The stimuli appeared (white on blak) in the enter of a Hewlett-Pakard 1362 XY display monitor with P-31 phosphor. The stimuli were the haraters?, A, B, C, and D, eah measuring approximately 7 x 10 rom. The display program used was the EMDISP system (Shepherd, 1984). The response keys formed part ofa keypad, whih was plaed on the table in front of the subjet. There were two keys (unused) on the top row; four keys on the middle row labeled 1,2,3, and 4 orresponding to the letters A, B, C, and D (operated by the seond and first fingers ofthe left hand and the first and seond fingers of the right hand, respetively); and four keys on the bottom row labeled 5,6,7, and 8 (key 6 operated by the left thumb). The subjet sat approximately 600 rom from the sreen. Design and Proedure Eah subjet was tested individually and asked not to eat for at least 2 h before eah session and to drink no alohol on the previous evening. A questionnaire had previously established that there were no exessive drinkers or teetotalers in the sample. In addition, the subjet was asked to give details of any urrent mediation or serious illnesses, and to sign a onsent form agreeing to take part in the experiment. The subjet was warned not to drive, yle, or handle dangerous mahinery for 24 h and was instruted to inform the experimenter ofany side effets experiened as a result of the alohol. Session 1 began with the subjet's drinking 500 rnl in exatly 5 min (a stopwath was provided). The drink ontained 250 rnl pure orange juie, 250 rnl water, plus a few drops of vodka that were arefully floated on the top of the drink. The subjet was given no information onerning the ontents of the drink. After a rest of 5 min, the experiment began. Breath readings were taken as near as possible to 20,35, and 50 min after the subjet finished the drink. The third reading almost always oinided with the ompletion of the last blok of trials; that is, eah session lasted just under 1 h. The proedure for Session 2 was very similar. Again, the subjet was given no information about the drink. For 12 of the subjets, the drink was idential to that in Session 1, whereas for the remaining 10, the drink ontained 250 rnl pure orangejuie, 2.13 rnl vodka (Le., 0.8 rnl alohol) per kilogram body weight, and water (total volume of500 rnl). The extra 2 subjets in the NAtNA group were inluded to replae 1 highly inaurate subjet (average error rate in the first session over 15 %) and 1 exeptionally slow subjet (mean response time in Session 1 almost three standard deviations above the mean). The task was idential in the two sessions. There were 12 bloks oftrials with breaks of approximately 30 se between bloks. There were 302 trials in Blok 1, and 202 trials in eah ofbloks 2 through 12. The beginning of a blok of trials was signaled to the subjet by the appearane of "?" in the enter of the sreen. Bloks 1 to

3 ALCOHOL, PRACTICE, AND JUDGMENT OF RESPONSE SPEED 433 II provided extensive pratie at the hoie reation time task; the subjet lightly plaed the first two fingers of eah hand on the keys labeled I, 2, 3, and 4. When ready to begin, the subjet was required to press anyone of the keys on the keypad. The sreen then went blank for I se before the presentation of the first stimulus (randomly hosen from A, B, C, or D). The subjet was required to press the orresponding key as quikly as possible. This immediately removed the stimulus from the sreen unless the response was made within the first 100 mse of stimulus presentation. The omputer then waite4 until a "legal" response was made (i.e., more than 100 mse after stimulus onset) before removing the stimulus. Following a delay (the response-stimulus interval, RSI), the next stimulus appeared, whih again was randomly hosen but with the restrition that eah stimulus was different from the previous one. For Bloks I-II the instrutions were merely to respond to eah stimulus as quikly as possible with a single keypress. The subjets were also requested to avoid making more than 10% errors, that is, a maximum of 30 in Blok I and 20 in eah of Bloks 2 through 12. As indiated earlier, all but 1 of the 22 subjets were able to omply with these instrutions. The RSI for the first blok in eah session was 300 mse. For Bloks 2 to 11, the RSIs were 50, 100,200,400, and 800 mse (two bloks at eah RSI), and the presentation order for these was ompletely randomized for eah subjet and for eah session. The inlusion of the shorter RSIs ensured that prior to the ritial task (Blok 12), the subjets reeived over 2,000 trials of the hoie response task in only 40 min. There were 202 trials in Blok 12, with an RSI of 1,000 mse. The first 2 trials were for pratie only and were therefore not inluded in the data analysis. In addition to the hoie response task, the subjets were instruted to press the key numbered''6" (under the left thumb) after any response that they onsidered to be partiularly fast and aurate. No indiation was given as to how often they should use this extra key. The subjets were requested to indiate "fast" responses during the RSI (although if key "6" was pressed during the first 100 mse ofthe next stimulus presentation, it was aepted as a rather slow indiation that the previous response was "fast"). RESULTS AND DISCUSSION As indiated above, the data from 2 subjets were not inluded in the analysis. Table 1 and the following results and disussion are therefore based on 20 subjets, 10 in eah of the two groups (NAJNA and NAJA). Blood Alohol Conentration The mean levels of alohol in milligrams per 100 ml of blood estimated from the Breathalyser readings taken 20,35, and 50 min after ompletion of the drink in Session 1 were all zero. They were also zero for the NAJNA subjets in Session 2. The means (and standard deviations) for the NAIA subjets in Session 2 were 63.0 (17.0), 70.5 (21.1), and 69.0 (19.8) after 20,35, and 50 min, respetive1y. The third reading was always taken either immediately before or after the present task. (The legal limit for driving in England is 80 mg per 100 ml blood. This was exeeded by only 2 of the 10 subjets.) The results reported in the remaining setions are from Blok 12 only (i.e., with the extra key to signal fast and aurate responses). Overall Error Rates The errors were divided into two types: (1) errors aused by pressing the "fast" key in response to the stimulus (i.e., more than 100 mse after its appearane), and (2) errors aused by making an inorret hoie from the four response keys (A, B, C, and D). The numbers of errors were then subjeted to an analysis of variane with group (two levels: NAJNA and NAJA) as a betweensubjets fator, and session (two levels: first and seond) and error type (two levels as desribed above) as withinsubjets fators. There were no signifiant effets ofgroup [F(l,18) = 2.82, P >.1] or session [F(l,18) = 2.09, P >.1], but there was a highly signifiant effet of error type [F(1,18) = 33.70, P <.01]. With the exeption of an interation between group and error type that approahed signifiane [F(l, 18) = 3.42, P <.08], there were no other effets (all remaining Fs < 1). The results are summarized in Table 2. First, it is lear that most errors are of the seond type, that is, an inorret hoie between A, B, C, and D. Seond, the group X error type interation an be seen as a tendeny for the NAJNA group to make more errors of the seond type than the NAJA group. However, sine there were no interations involving session, the overall onlusion is that alohol had no signifiant effet on error rate. Numbers of "Fast" Responses One subjet in the NAJA group failed to indiate any "fast" responses whatsoever in Session 1. He was the slowest of the 20 subjets in terms of his mean orret response time in Session 1, and ommented afterward that he felt he required longer than the response-stimulus interval (l,000 mse) to indiate a "fast" response. In Session 2 he indiated six "fast" responses, with a mean time to indiate of 913 mse. No other subjet had any diffiulty in arrying out the task, although several reported that they had erroneously pressed the "fast" key in response to the stimulus. There were 7,490 orret responses and 510 errors altogether. Of these, 1,947 and 8 responses, respetively, were indiated as "fast." Sine the instrutions were to indiate fast and aurate responses, these results suggest that subjets were highly effiient at least in terms of the Table 2 Mean Nwnbers oferrors (out of200) Caused by Pressing the "Fast" Key, and Pressing the Inorret A, B, C, or D Key Group Session I Session 2 NA/NA NAJA Pressing the "Fast" Key Pressing the Inorret A, B, C, or D Key NA/NA NAJA Note-NA = no alohol. A = alohol

4 434 MAYLOR, RABBITT, AND CONNOLLY latter requirement sine they seleted a muh higher proportion of orret responses (approximately I in 4) than error responses (approximately 1 in 64). This is, of ourse, onsistent with previous work demonstrating that in suh reation time experiments, subjets generally know whether or not a response is orret (Rabbitt, 1968). An analysis of variane was arried out on the numbers of "fast" responses indiated, with group and session as between- and within-subjets fators, respetively. There was no effet of group (F < 1) or of session [F(l,I8) = 2.65, P >.1], and there was no interation (F < 1). Clearly neither alohol nor pratie had any influene on the proportion ofresponses indiated as "fast." The overall mean number was 49 (out of200 responses), but there were large individual differenes, the numbers ranging from 0 to 176, with a standard deviation of U 450 Ql. '" 400 ~ II: Session 1 CF' CF E RESPONSE NAJNA F&iI NAJA CF+ Overall Mean Response Times The mean response times are summarized in Figure 1. (There were insuffiient numbers of errors aused by pressing the "fast" key in response to the stimulus, and errors indiated as "fast," to inlude in the response time analysis.) Two analyses of variane were initially onduted on the data. The first ompared the three types of responses made to the stimulus, namely, orret responses not indiated as "fast" (CF'); orret responses indiated as "fast" (CF); and errors (E) for whih the response was A, B, C, or D. Group was the betweensubjets fator, and session and response (three levels: CF', CF, and E) were within-subjets fators. There were no effets of group or session (both Fs < 1), but there was a highly signifiant effet of response [F(2,34) = 29.90, P <.01]. The only signifiant interation was that between group and session [F(l,I7) = 6.39, p <.05]. Sheffe tests revealed that CF' responses were signifiantly slower than E responses (F = 16.78, P <.01), whih in tum were slower than CF responses (F = 13.27, P <.01). The improvement from Session 1 to Session 2 was greater for the NAtNA group (25 mse) than for the NAtA group (-11 mse), leading to the group X session interation and indiating an overall alohol effet of 36 mse (see Figure 2). The seond analysis of variane ompared the time to make a "fast" orret response (CF) with the time to indiate it (CF+). Again, there were three fators: group (between-subjets), session (within-subjets), and response (within-subjets, two levels: CF and CF+). There was no effet of group (F < 1), the effet of session approahed signifiane [F(l,17) = 3.80,p <.07], and the differene between CF and CF+ was signifiant [F(I,I7) = 7.71, P <.05]. There was also an interation between group and session [F(l,17) = 7.46, P <.05]. Although it appears from Figure 2 that the effet of alohol was larger for CF+ (90 mse) than for CF (33 mse), the three-way interation was not signifiant [F(l,I7) = 1.45, P >.1]. There are several points to emphasize from these analyses ofoverall mean response times. First, the influene U 450 Ql. '" 400 ~ II: Session 2 CF' CF E RESPONSE NAJNA F&iI NAJA CF+ Figure 1. Mean response times for Session 1 and Session 2. CF' = orret respolllies not indiated as "fast." CF = orret respolllies indiated as "fast." E = errors for whib tbe response was A, B, C, or D. CF+ = the time to indiate a "fast" orret response. of alohol is lear: responses were slowed by approximately 40 mse. Furthermore, the effet was similar for all types of responses, both to the stimulus (CF', CF, and E) and to the response (CF+). Seond, the task was performed well both with and without alohol in that CF' responses were signifiantly slower than CF responses (mean differene of 122 mse). It is perhaps even more surprising that CF responses were even faster (by 57 mse) than E responses, whih generally our in suh speeded hoie tasks before suffiient information has been aquired about the stimulus (see, e.g., Maylor & Rabbitt, 1987). Third, the time taken to indiate a "fast" response was remarkably short (approximately 300 mse). Subjets were not instruted to indiate as quikly as possible, but were merely asked to press the "fast" key before the appearane of the next stimulus (I,OOO-mse response-stimulus interval). This absene of a speed instrution was probably responsible for the large between-

5 ALCOHOL, PRACTICE, AND JUDGMENT OF RESPONSE SPEED Ql '" 20 E.... a: CF' NA/NA mi NA/A a= E CF+ RESPONSE Figure 2. Improvements in mean response times from Session 1 to Session 2. CF' = orret responses not indiated as '1"ast." CF = orret responses indiated as '1"ast." E = errors for whih the response was A, B, C, ord. CF+ = the time to indiate a '1"ast" orret response. subjets variability for CF+ response times ompared with those for CF', CF, and E. By inluding E responses in the first analysis of variane, it seemed that two features ofthe CF'/CF data were obsured (the relevant interations failing to reah signifiane). The first was that the differene between CF' and CF response times (a measure of how well the subjets were performing the task ofdistinguishing between fast and slow responses) was larger for the NAtA group than for the NA/NA group (150 and 94 mse, respetively). This less-than-perfet mathing of groups is of less interest than the seond feature, namely that the CF'tCF differene inreased from 106 mse in Session 1 to 138 mse in Session 2. Thus there appears to be evidene that pratie atually improves the ability to reognize fast responses. A third analysis ofvariane was therefore performed on the mean response times for CF' and CF only. The main results were as before: a signifiant effet of response and an interation between group and session. However, there was also a signifiant interation between session and response [F(1, 17) = 7.67, P <.05], and the interation between groupand response approahed signifiane [F(1, 17) = 3.71, P <.07]. Thus there is statistial support for the seond of the two suggestions; that is, improvement aross sessions was smaller for CF' responses (-8 mse) than for CF responses (24 mse) (see Figure 2). What Happens Before and After an Error or a "Fast" Response? Several studies have demonstrated that subjets gradually inrease their speed in hoie response tasks until an error is made; an immediate adjustment then ours so that the next response is abnormally slow and aurate (Laming, 1968; Maylor & Rabbitt, 1987; Rabbitt, 1966, 1979a). The question of interest here is whether or not the same pattern applies to orret responses indiated as fast. In other words, do subjets gradually speed up until a partiularly fast orret response is made, then immediately beome autious again sine to ontinue at suh a speed would involve a high risk of making an error? Also, how are these ontrol mehanisms affeted by alohol and pratie? The first analysis onerned CCECC sequenes (C = orret and E = error, where the response was A, B, C, or D) in whih eah response was not followed by an extra response (e.g., an error orretion-see Rabbitt, 1968-or the "fast" key) either during the responsestimulus interval or within 100 mse ofthe next stimulus presentation. A three-way analysis of variane was onduted on the mean response times with group (betweensubjets), session, and trial type (5 levels: C, C, E, C, and C) as fators. There were no CCECC sequenes for I subjet in one session, so the results are based on 19 subjets. There was no effet ofgroup [F(1, 17) = 1.18, P >.1] or session (F < I), but there was a highly signifiant effet of trial type [F(4,68) = 14.29, P <.01]. The only interation was between group and session [F(1,17) = 10.38, p <.01], with all other Fs < l. Again, the overall improvement from Session I to Session 2 was greater for the NAtNA group (34 mse) than for the NAtA group (-40 mse), indiating an alohol effet of 74 mse. The trial-type effet is shown in Figure 3. The general pattern was as expeted: a gradual inrease in speed until an error ourred (although on Sheffe tests, the differene between the first three trial types did not reah signifiane), followed by a very slow response. Although there were overall effets on response times of alohol (an inrease from Session 1 to Session 2 for the NAtA group) and of pratie (a derease from Session 1 to Session 2 for the NAtNA group), they did not interat with the CCECC response time sequene. In otherwords, there was no influene of either alohol or pratie on the ontrol mehanism of deteting and responding to errors , ' , ' E C TRIAL TYPE Figure 3. Mean response times for CCECC sequenes (where the responses were not followed by a orretion response or the '1"ast" key). C = orret response. E = error.

6 436 MAYLOR, RABBITT, AND CONNOLLY The seond analysis was onduted on CCFCC sequenes (C = orret and F = orret and indiated as "fast"). The data from 5 subjets ould not be inluded in the analysis as there were no CCFCC sequenes in at least one ofthe two sessions. One subjet has already been mentioned: in Session 1 he was unable to press the "fast" key in the time allowed. The remaining 4 subjets (2 in eah group) indiated as "fast" over 100 trials (out of 2(0) in eah session, and as a result produed no CCFCC sequenes in at least one of the two sessions. Again a three-way analysis of variane was performed with group, session, and trial type (five levels: C, C, F, C, and C) as fators. There was no effet ofgroup [F(l,13) = 2.65, p>.1] or session [F(l,13) = 1.18,p >.1], but there was a highly signifiant effet of trial type [F(4,52) = 37.47, P <.01]. There were interations between group and session [F(I,13)' = 7.44, P <.05] and between group and trial type [F(4,52) = 3.22, p <.05], and the interation between session and trial type approahed signifiane [F(4,52) -= 2.29, P =.07]. The three-way interation was not signifiant (F < 1). Figure 4 illustrates the overall trial type effet. Sheffe tests revealed that Trial Type 3 (F) was signifiantly different from all other trial types (p <.01 in every ase) and that Trial Type 4 was signifiantly different from both Trial Types 1 and 5 (ps <:.05). The group X session interation again reflets a greater improvement aross sessions for the NAINA group (21 mse) than for the NA/A group (-49 mse), giving a very similar alohol effet of 70 mse. The group x trial type interation (whih was marginally signifiant in the analysis of all the CF' and CF response times) is simply further evidene that the two groups were not perfetly balaned in terms ofthe ability to indiate "fast" responses. Thus the differene between the mean ofthe four C trial types and the F trial type was 116 mse for the NA/NA group but 207 mse for the NA/A group. Although the interation between session and trial type did not quite reah signifiane in the present analysis of CCFCC sequenes, , 550 '0 500 (l) III.s ~ l...--,---r---, ' F TRIAL TYPE Figure 4. Mean response times for CCFCC sequenes. C = orret. F = orret response indiated as "fast." the trend is similar to the signifiant effet found in the analysis ofcf' and CF response times. Thus, the differene between the mean of the four C trial types and the F trial type inreased from Session 1 (137 mse) to Session 2 (186 mse). (In onsidering these last two interations, it should be remembered that the analysis of the CCFCC sequenes was based on fewer subjets and fewer responses per subjet than the previous CF'/CF analysis. We should therefore plae greater emphasis on the results from the analysis of CF' and CF response times.) To summarize what happens before and after a "fast" response, first, it is lear from Figure 4 that subjets do not gradually inrease their speed until a partiularly fast response is made and then indiate it by pressing the "fast" key. Instead, a "fast" response appears to ome out of the blue, being on average 158 mse faster than the previous response. The next orret response is rather slower than normal, although it is diffiult to know whether this is simply a result ofhaving to make an extra response (the "fast" key) during the RSI, or an atual adjustment in response riterion, similar to that following an error (see Figure 3). Seond, there is a tendeny for the differene between the F trial type and the four C trial types immediately surrounding it to inrease from Session 1 to Session 2. (Remember also that there was a signifiant interation between session and response in the CF'/CF analysis.) We would argue from this that performane of the task of seleting "fast" responses improves with pratie, sine there was no differene between the two sessions in the proportion of responses indiated as fast. We know from previous studies (e.g., Rabbitt & Vyas, 1970) that pratie dereases mean orret response time primarily beause it redues the size of the slow tail of the response time distribution. Consider the data from the NA/NA group: Beause subjets performed the task after having reeived 2,324 hoie response trials in the same session (see Proedure), they were already wellpratied and thus the mean response time for all orret responses dereased by only 20 mse from Session 1 to Session 2. We an safely assume that the derease at the fast end of the response time distribution was even less than this. However, the derease for CF responses aross sessions for the NA/NA group was 41 mse. [It is worth noting that the results ofan analysis of variane omparing all orret responses with CF responses, produed the same effets as that for CF' and CF responses, inluding the highly signifiant interation between session and response: F(I,17) = 9.78, p <.01.] It is therefore extremely unlikely that the signifiant inrease in the differene between CF' and CF response times with pratie (or the tendeny for the differene between Cs and the F in the CCFCC sequenes to inrease with pratie) an be attributed to an effet of pratie on the number of fast responses available. Sine the proportion of responses indiated as fast did not hange with pratie, and the response time distribution hanged very little, it is proposed that the observed effets an be explained as

7 ALCOHOL, PRACTICE, AND JUDGMENT OF RESPONSE SPEED 437 0~ 50 ~ () + U. () if () Blok Blok ;; <:> <:> <:> C> C> C> C> C> C> C> C> (") '" '" '" '" '" <D V 1\ ":' ""t ""t ":' ":' ~ ~ ;; ;;; ;; ;;; ;; ;;; ;; (") (Y) '<t '<t RT (mse) Figure 5. Results ofadditional study (N = 36) showing the effet of pratie on the slope ofthe funtion relating the number oforret responses indiated as "fast" (CF) as a perentage of the total number of orret responses (CF' + CF) for eah of nine response time intervals: 400 trials per subjet (200 in eah blok). follows. In Session 1, subjets indiated some fast and some not-so-fast responses, whereas in Session 2 they indiated fast responses only. In other words, they "sharpened up" the distribution ofresponses indiated as "fast." Beause ofomputer problems, we were unable to plot for the present data the number of responses indiated as "fast" as a perentage of the total number of orret responses for several response time intervals, to examine whether or not theslope of the funtion beame steeper with pratie, as the above would suggest. Therefore, 38 new subjets were eah given 100 pratie trials of the four-hoie reation time task alone, prior to two bloks of 200 trials in whih they were required to indiate fast and aurate responses as desribed earlier (see Proedure). Figure 5 summarizes the results. (The data from 2 subjets, who learly misunderstood the instrutions, were not inluded: One indiated no "fast" responses whatsoever, and the other indiated almost all responses as "fast.") First, the subjets' ability to perform this task is learly shown by the monotoni derease in the probability of Table 3 Further Analysis of Results from Additional Study: Numbers of CF' and CF Responses for Bloks 1 and 2 Less Than and More Than 450 mse Blok CF' CF I 2 I 2 Responses < 450 mse Responses > 450 mse '" '" <D indiatinga response as fast with inreasing response time. Seond, in agreement with the present onlusions, it does appear that the effet of pratie is to redue the probability that slow responses will be indiated as fast while not affeting the probability that fast responses will be indiated as fast. This was onfirmed statistially by two hi-square tests on the numbers ofcf' and CF responses for Bloks I and 2 (see Table 3), first for responses less than 450 mse Ix 2 (1) =.33, p >.05], and seond for responses more than 450 mse [x 2 (1) = 25.86, P <.001]. To summarize, pratie affets the slope and this moves the interept further to the left along the response time axis (see Figure 5). Returning to the present study, there is no evidene that alohol would have any adverse influene on the slope (although, as indiated above, we were unable to plot the omplete funtions). The presene of an overall alohol effet and the absene of an interation between alohol and response (Figure 1) or trial type (Figure 4) together suggest that with alohol, subjets are as effiient in distinguishing between fast and slow responses within a session as without alohol. However, sine all responses (orret, error, and "fast") are slower with alohol, the interept along the response time axis would be further to the right than without alohol, both for the speed-error tradeoff funtion (see Maylor & Rabbitt, 1987) and for the funtion desribed above. CONCLUSIONS First, it an be onluded that the two fators ofinterest, namely pratie and alohol, had signifiant effets on the overall rate of proessing. Alohol inreased, and pratie dereased, mean response times for all types of trial. Moreover, the effets did not interat with response type for CCECC sequenes, whih represent the limits of performane in terms of the traking model outlined in the introdution (see also Maylor & Rabbitt, 1989). Neither pratie nor alohol influened the overall error rate. This is onsistent with the results of Maylor et al. (1987), who argued that for tasks suh as hoie reation time, in whih subjets are aware ofmost oftheir errors, alohol adversely affets speed but not auray (unless the dose is very high; see Maylor & Rabbitt, 1987). Seond, it is lear that subjets were suessful at the task of indiating fast and aurate responses. They were able to make a deision regarding eah response within a third of a seond. Approximately 1 in 4 orret responses were seleted, ompared with only 1 in 64 errors, demonstrating the ability to distinguish between aurate and inaurate responses. The analysis ofresponse times revealed that subjets were also highly effiient at distinguishing between fast and slow responses. Furthermore, responses indiated as "fast" were signifiantly faster than errors, whih is perhaps surprising sine errors in hoie response tasks have been attributed to premature deisions based on insuffiient evidene. Also,

8 438 MAYLOR, RABBITT, AND CONNOLLY in ontrast to errors, there appeared to be no warning (in the form ofa gradual inrease in speed) prior to a "fast" response. The main question of interest was how this ability to judge response speed was affeted by fators that affet rate of proessing, namely pratie and alohol. There was a lear improvement with pratie suh that subjets initially indiated some fast and some not-so-fast responses, but then subsequently tightened up the response time distribution for "fast" responses by exluding slower responses. In ontrast, alohol did not impair the ability to judge response speed: All response times were signifiantly inreased by the same amount (inluding the time to indiate a "fast" response, whih was nonspeeded). In terms ofthe ontrol model desribed in the introdution, it is lear that the present dose of alohol had no effet on subjets' ability to monitor the auray or speed of the response: subjets deteted their errors both with and without alohol and made an immediate adjustment to their response riterion following an error. They were also unimpaired by alohol in their ability to monitor response speed, that is, to distinguish between fast and slow responses. However, the results do suggest that this latter aspet of ontrol was improved by pratie. Rabbitt (1979b) ompared hoie response times from young and elderly subjets and similarly disussed the results in terms ofthe ontrol proess model. One possibility was that the elderly "simply annot respond at the same rate as the young. That is, that all their responses are lagged by a onstant relative to those ofthe young. ', This was rejeted.,sine for partiular ategories of very fast responses (errors...) both groups an respond in the same [response time] range" (po 310; see also Rabbitt & Goward, 1986). Although this simple lagging explanation fails to aount for the effet ofage, the present study suggests that it annot be ruled out with respet to the effet of alohol in speeded hoie response tasks. These results, together with those from Maylor and Rabbitt (1989), provide an interesting omparison between the effets ofpratie and alohol on hoie reation time. Thus pratie improves at least three things: (I) judgment of response speed (subjets learn to reognize slow responses), (2) rate of proessing (subjets beome faster at responding for the same number oferrors), and (3) rate ofpreparation (subjets ahieve optimal preparationin less time) (see Maylor & Rabbitt, 1989). In ontrast, alohol impairs performane by slowing the rate of proessing alone; that is, it has no adverse effet on either judgment ofresponse speed or rate ofpreparation. At the very least, these onlusions illustrate the inadequay ofusing mean reation time and error rate only for desribing performane in hoie response tasks. For example, 2 subjets may have idential mean reation times and error rates but differ in terms of rate of preparation and judgment ofresponse speed, sine one is sober but unpratied and the other is drunk but pratied. Finally, it is interesting to speulate on the existene ofsome fator (a drug, stress, age, personality?) that would adversely affet all three aspets of performane disussed above, and then to test whether or not pratie ould ompletely ompensate for the impairments observed. REFERENCES JENNINGS, J. R., WOOD, C. C., & LAWRENCE, B. E. (1976). Effets ofgraded doses ofalohol on speed-auray tradeoff in hoie reation time. Pereption & Psyhophysis, 19, LAMING, D. (1968). Information theory ofhoie reation times. London: Aademi Press. MAYLOR, E. A., & RABBITT, P. M. A. (1987). Effets of alohol and pratie on hoie reation time. Pereption & Psyhophysis, 42, MAYWR, E. A., & RABBITT, P. M. A. (1989). Relationship between rate of preparation for, and proessing of, an event requiring a hoie response. Quarterly Journal ofexperimental Psyhology, 41A, MAYWR, E. A., RABBITT, P. M. A., SAHGAL, A., & WRIGHT, C. (1987). Effets of alohol on speed and auray in hoie reation time and visual searh. Ata Psyhologia, 65, RABBITT, P. M. A. (1966). Errors and error orretion in hoie-response tasks. Journal ofexperimental Psyhology, 71, RABBITT, P. M. A. (1968). Three kinds of error-signalling responses in a serial hoie task. Quarterly Journal ofexperimental Psyhology,20, RABBITT, P. M. A. (1979a). Current paradigms and models in human information proessing. In V. H. Hamilton & D. M. Warburton (&Is.), Human stress and ognition: An information proessing approah (pp ). New York: Wiley. RABBITT, P. M. A. (l979b). How old and young subjets monitor and ontrol responses for auray and speed. British Journal ofpsyhology, 70, RABBITT, P. M. A. (1980). A fresh look at hanges in reation times in old age. In E. Stein (Ed.), The psyhobiology ofaging: Problems and perspetives (pp ). Amsterdam: Elsevier North-Holland. RABBITT, P. M. A., & GOWARD, L. (1986). Effets ofage and raw IQ test sores on mean orret and mean error reation times in serial hoie tasks: A reply to Smith and Brewer. British Journal ofpsyhology, 77, RABBITT, P. M. A., & VYAS, S. M. (1970). An elementary preliminary taxonomy for some errors in laboratory hoie RT tasks. Ata Psyhologia, 33, SANFORD, A. J. (1970). Rating the speed ofa simple reation. Psyhonomi Siene, 21, SCHOUTEN, J. F., & BEKKER, J. A. M. (1967). Reation time and auray. Ata Psyhologia, 27, SHEPHERD, M. (1984). EMDISP: A visual display system with digital and analogue sampling. Behavior Researh Methods, Instruments; & Computers, 16, WILKINSON, R. T., & COLQUHOUN, W. P. (1968). Interation of alohol with inentive and with sleep deprivation. Journal ofexperimental Psyhology, 76, (Manusript reeived June 29, 1988; revision aepted for publiation Otober 17, 1988.)