mass This is to certify that the thesis entitled Prdbability Learning, Competing Responses, and Interstimulus Interval (181) Effects in Two-Choice Reaction Time presented by Donald Reynolds has been accepted towards fulfillment of the requirements for Ph . D . degree in Psychology " f 4 7 Major professor Date December 1L1961+ 0-169 F LIBRARY Michigan State University ABSTRACT The several experiments reported herein examined the phenomenon known as the Psychological Refractory Period (PRP), earlier assumed.by the writer to be a special case of a "Temporary Inhibition of Response (TIR)" phenomenon (Reynolds, 196M). The behavioral manifestation of double stimulation in close temporal contiguity may, under certain con- ditions, be a delay in response to 222 of them. Exp. I & II explored this possibility under conditions of temporal certainty and event uncer- tainty; no evidence for the PRP was feund. An attempt to train a pre- potent response by means of probability learning was unsuccessful. Exp. III used the double stimulation task with conditions of event certainty and temporal uncertainty) using either a Regular or a Random series of interstimulus intervals (181). The PRP was obtained; this is at some variance with previous work in tracking tasks using bisensory stimulus presentation. Exp. IV explored the role of complete temporal and event certainty in conditions: 1) Where the § made responses to stimuli with only one 181 across all 100 trials; 2) Where the § made re6ponses to stimuli at one 181 fer 20 trials and then went on to respond to another set of sti- muli at a different ISI fer 20 more trials until all ISIs had been used; 3) Where the § was instructed to be equally fast with each hand in a replication of part 2) above. The PRP was obtained at shorter 1815 than in EXP. III, but only in conditions 2) and 3) above. There was no PRP in condition 1) of Exp. IV: All experiments reported herein used a unisensory (visual) task. Results of all four experiments were thought consistent with a com- peting response theory of delays in response, and certain hypotheses partially testing this position were supported. Probability Learning, Competing Responses, and Interstimulus Interval (ISI) Effects in Two-Choice Reaction Time by Donald Reynolds A THESIS Sdbmitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1961» ACKNOWLEDGEMENTS The writer wishes to acknowledge the advice and help of the members of his committee, Drs. M. Ray Denny, Chairman, Donald M. Johnson, Terence M; Allen, and Charles Henley. It must not have been the easiest task in the world for them. For designing and wiring the circuitry involved in these studies, special thanks go to Ronald G. Weisman and Dominic J. Zerbolio, Jr. Many hard hours of verbal and physical work by these two gentlemen resulted in smooth functioning of the equipment at a minimum strain on the E. For figuring out visual angles, and placement of the §_to insure maximum foveal stimulation, the writer is indebted to Richard Haines. TWO sympathetic shoulders to cry on when agonizing over the data and gestating reasonable theory to explain same were provided by Dr. M. Ray Denny and Ronald G. Weismang special mention is hereby made by the writer for patience above and beyond the call of duty. For assistance with the Lindquist Type III Analysis of Variance, thanks go to Tom Teeples and Phil Wynn, Jr. of the Statistical Services Division, HumRRO. My wife Jeanine not only provided moral support, but proved the adage that a good wife is one who knows how to protect her husband at work. This dissertation is for that reason, and a few others, dedicated to her. ii INTRODUCTION . . . . . Definitions . . . TABLE OF CONTENTS SELECTED REVIEW OF THE LITERATURE . . . . Crossmodal Differences . . . . . . . . Event Uncertainty and Time Uncertainty General Discussion Refractoriness and Uncertainty: A Methodological Critique . . . EXPERIMENT I . . . . . EXPERIMENT II . . . . EXPERIMENT III . . . . EXPERIMENT IV . . . . GENERAL DISCUSSION AND REFERENCES . . . . . . CONCLUSIONS . . . . iii Figure l. 2. LIST OF FIGURES RT of lst and 2nd Response as a Function Length. (EXperiment I) . . . . RT of lst and 2nd Response as a Function Length (Experiment II) . . . . . . RT of lst and 2nd Reaponse as a Function Length (Experiment III - Regular) . . . RT of lst and 2nd Response as a Function Length (Experiment III - Random) RT of lst and 2nd ReSponse as a Function Length (EXperiment IV - Part A) . . . RT of lst and 2nd Reaponse as a Function Length (EXperiment IV - Parts B a 0) RT of lst and 2nd Response as a Function of Trials (Experiment IV - Part A) . RT of lst and 2nd.ReSponse as a Function of Trials (Experiment IV - Part B) iv of ISI of 181 of ISI " of ISI of ISI of 181 of Blocks of Blocks Page 22 hh h5 5h 55 61 63 Table 10. LIST OF TABLES Page Mean RTS in msec. for SS by Probability Learning Group . . . . . . . . . . . . . . . . . . 21 Comparison of Mean RTS in msec. of lst and 2nd Responses by 1818 . . . . . . . . . . . . . . . . 2l Lindquist Type III Analysis of Variance of Data From Experiment I (Left Hand vs. Right Hand) . . . 23 Lindquist Type III Analysis of Variance of Data From.Experiment I (First Hand vs. Second Hand) . . 2h Mean RT in msec. for SS According to Prdbability Learning . . . . . . . . . . . . . . . . . . . . . 28 Lindquist Type III Analysis of Variance of Data From Experiment II (Left Hand vs. Right Hand). . . 29 Comparison of Mean RTS in msec. of lst and 2nd ResponsebyISIS................. 3O Lindquist Type III Analysis of Variance of Data From EXperiment II (First Hand vs. Second Hand). . 31 Moan First and Second RTS in msec. in Regular and Random Presentations . . . . . . . . . . . . . h2 Mean First and Second RTS in msec. in Complete Temporal and Event Certainty . . . . . . . . . . . 5l-a INTRODUCTION The general scope of the several studies reported herein deals with a phenomenon earlier labelled as the "Temporary Inhibition of Res- ponse," or the TIR phenomenon (Reynolds, 196A). This phenomenon follows from the presentation of simultaneous, or nearly Simultaneous, stimuli to the subject (S). The behavioral manifestation of such double stimula- tion, when the §_is required to respond to both, is a delay in response to 9&3 of them. The TTR phenomenon is not the result of updating the fact that a complex task requires more time than a simple one, since the delay is posited to occur in only one of the two responses. Although the whole task (making both responses) often takes more time than the two component responses combined additively, the observation that only one of the two responses is delayed.may help to explain.hg! the complex task takes more time. The TIR concept is useful because it clarifies what may be expected to occur when two stimuli follow each other in close temporal contiguity, and.when response to both is required. Further, TIR provides a convenient point of departure for the comparison of different theoreti- cal positions with respect to response output following such stimulation. Explanations of delay in response have been advanced along tradit- ional lines. The stimulus-centered.position maintains that variation in response output is primarily related to variations in stimulus input. Organismpcentered explanations take one of two customary forms: 1) "Expectancy." This position deals with the "set" of the §_and tends to cognitive explanations in describing delays in response, e.g., the S is l. 2 thought to be less than optimally prepared for response when the second stimulus arrives; 2) "Central Limitation." This formulation draws sus- tenance more from the physiological than the cognitive realm, positing a limit on the ability of the §_to respond twice in quick succession. This may be based on the belief that the first response renders the response mechanism insensitive so that it reacts more slowly to the second stimulus (welford, 1952), or that human processing channels are analogous to a ”Y-shaped funnel, " down which simultaneous bits of information (or stimu- lation) must pass successively (Broadbent, 1957, 1958). The third explanation, a response-centered one, calls attention to the competing response tendencies elicited.by'each of the stimuli in the double stimulation paradigm. Berlyne, in a somewhat different context, lays out the discussion as follows: Instead of speaking of a conflict between two partially incompatible responses R1 and R2, we might speak...of a conflict between two completely incompatible responses R2 and 32, the latter being the inhibition of R2 that is tracable to the arousal of R1. The greater the degree of antagonism.between R1 and R2, the great- er the conflict, Since the more nearly equal in strength R2 and R_2 will be (1960, p. 3h). Thus Berlyne views the making of R1 to lead to the arousal of R_2 in certain cases. If R2 must be made anyway, it might well be delayed on the basis of arousal of competing responses. It might further be postulated that the arousal of R_2 might be diminished with practice on the task, such that a wellepracticed § would not have R2 delayed. There are two reasons for this assumption, in increasing order of importance: 1) The belief that there is no reinforcement structure to support R_2, and 2) ‘With repeated elicitation of R2, R_2 may be extinguished, since the two responses are presumed wholly incompatible. The earlier paper (Reynolds, l96h) reviewed findings from.three 3 apparently unrelated areas and attempted to integrate these findings as supportive evidence for the reality of the TIR phenomenon. The areas reviewed were: stereoscopic perception, dichotic listening, and two- choice reaction time. Since a historical Sketch was provided of the latter, no attempt will be made herein at duplication of the same area; rather, selected studies on reaction time (RT) will be presented. Definitions The abbreviation "RT" will be reserved for exclusive use; it shall only refer to the direct time taken (i.e., the latency) in making a single discrete response to a single discrete stimulus. It is a pure speed measure as such, and.refers to what is customarily referred to in the literature as reaction time. Other measures of speed of response have been used by investigators. The most common of these has been called "Response Time," but such a label is confusing. The term "Response Speed" (abbreviated RS) will instead be used to refer to a speed.measure in which the latency includes errors, anticipatory responses, in general any passage of time from the stimulus onset until the "correct" response has been made. SELECTED REVIEW OF THE LITERATURE AS the interval between two stimuli becomes very short the RT to the second stimulus presented.becomes lengthened as the interstimmlus interval (ISI) is shortened and approaches .5 second (Adams, 1961, p. 68). This delay in response to the second.stimmlus at short ISIS is known as the Psychological Refractory Period (PRP). The following studies deal in common ‘with the PRP, herein assumed to be a special case of the TIR ll phenomenon. The PRP has been found to occur at ISIS of 0 milliseconds (Adams, 1962; Creamer, 1963), 50 milliseconds (Davis, 1962), 75 milli- seconds (Marill, 1957), 100 milliseconds (Davis, 1956; Elithorn & Lawrence, 1955), and 500 milliseconds (Telford, 1931; Vince, 19h8). The general trend seems clear: later investigators have succeeded in pushng back the point of maximum refractoriness from 500 milliseconds to at or near 0 milliseconds ISI - Simultaneity of stimulus presentation. The above studies are not strictly comparable due to the different procedures used, some being bisensory and others involving unisensory tasks. In the bisensory tasks the W0 stimuli arrive via different modalities, the two most common being visual and auditory; in the unisensory studies only one modality is used. Crossmodal Differences Is the PRP confined to stimuli from the same modality, or is it general across modalities? Davis (1956) used two visual stimuli and a visual-auditory pair (1957) while varying ISIS from 50 to #00 milliseconds (1956) or 500 millisecond (1957). The pattern of results is the same with unisensory and bisensory presentation: only at the shorter ISIS did delays of response occur to the second stimulus. In a follow-up study Davis (1959) hypothesized that it is the paying of attention to a stimuli, rather than making an overt response to it, that leads to refractoriness. His results were quite conplex, but were taken to show support for a "connnon analyzer or classifying system." Although he did not present statistics to support his conclusion, computation of t-tests from his (presented) data reveals that there is no significant difference between mean reaction times in a unisensory condition in which the _S_ either did 5 or did not respond to the first stimulus (p}D.OS; two sided), but a highly significant difference when a similar comparison is made in a bisensory condition (p<.0001; two-sided). The unisensory vs. bisensory finding is important. If both stimuli come in on the same modality, then whether one makes a response to both or only the second seems to make no difference in the latency of the second response, since the second response in both cases takes about the same time. But note what happens if the second modality stimulated is different from the first: the second reaction may take longer if the S has to respond to 9239 stimuli, but not if he is only to respond to the second. In summary, then: 1) Davis' conclusions are not quite supported by his own data, as there appear to be discriminable differences between the unisensory and bisensory conditions; 2) The making of an overt response is not quite as unimportant as suggested by Davis, given the findings in the bisensory condition, and therefore the evidence for a "common analyzer or classifying system” has not been compellingly set forth; 3) One should be sensitized to the possibility of task-specificity and/or mode-specificity in the double stimulation paradigm. Such Specifi- city has turned up in as unlikely a place as vigilance studies, in which cross-modal vigilance correlations are consistently in the vicinity of + .30 (Buckner a. McGrath, 1963). Event Uncertainty & Time Uncertainty“ Early studies dealing with the PRP confounded uncertainty on the part of the §_as to the event to which he was to respond (event uncertain- ty) with uncertainty as to the time between the two events, both of which required response (time uncertainty). To illustrate: When the S is unsure 6 whether the stimulus-order to which he must respond will be left-to-right or right-to-left, he is said to be in event uncertain conditions. When he knows, say, that the first stimulus will be on the right, but does not know the interval between this event and the next one on the left, he is said to be in a state of time uncertainty. One of the first to inquire into the temporal spacing of stimulation was Klemmer (1956), who studied the effects of time uncertainty on simple (Single-choice) RT. He used five different ISIS of l to 12 seconds in an "a-reaction" (in which the §_makes a single, specified response to a single, specified stimulus) with a variable preparatory interval and found that RT tends to increase with an increase in preparatory interval varia- bility. Since the event was certain, the increase in RT was though due to the temporal uncertainty associated.with variable preparatory inter~ vals. Recent studies have fUrther explored the effects of time- and event-uncertainty in two-choice RT. Using almost exactly the same procedure as Davis (1959), but holding temporal uncertainty to near zero by using fixed ISIs fer each series of stimulus, Borger (1963) presented two stimuli to the S (auditorybvisual or visual-auditory), and varied the ISI from 50 to 500 milliseconds. The second response was found.to be slower when the §_had to make a first res- ponse, but not so when the "second" response was the ggly:response made. This finding held for ISIS of 300 milliseconds or less, and is in close agreement with Davis (1959) with the above noted exceptions. The subjects were very variable and statistics were not presented. Borger felt that a theory of refractoriness based on time uncer- tainty'was not supported by his data, since all 1813 were constant within a series. Since his stimulus-presentation was in part a bisensory one, 7 one should use care in accepting his interpretation. One should ideally inquire if similar results would occur given a unisensory two-choice task with only event uncertainty. Borger brought out a good point: Responses may be stored until after the second stimulus is received, and then made successively or simultaneously. This explanation would seem to indicate that it is the 115232 response, rather than the second, which is delayed. His suggestion that "signals" may also be stored is less convincing, for it implies that the first "signal" remains, in Hartley's terms (1958, p. 99), an "impingement" until the second signal arrives, at which time both become "stimuli." Recast in this terminology the difficulties become apparent. One would have to postulate, among other things, a retroactive effect of the second "signal" on the first - an idea long ago read out of psychological theory. Storage of responses does seem possible, and would in general fit in with the claim made earlier in this paper that when two stimuli are presented in close temporal contiguity and response to both is required that a delay in response will occur to 933 of the stimuli. Adams (1962) varied a time uncertainty condition associated with a 100 millisecond ISI, using an ISI range of O to 800 milliseconds. He found refractoriness at ISIS of less than 200 milliseconds. Errors of anticipation begin to occur at about #00 milliseconds, with little evi- dence of refractoriness. These results were taken to support an "expectancy" position. The task involved can be best described as a tracking task, in which the _S_ maintained a pure tone (auditory task) and proper alignment of lights (visual task). The visual task always preceded the auditory one, so that there was event certainty; time uncertainty was introduced via statistical manipulation of frequencies of each ISI. Adams 8 used as his performance measure what he called "response time," which is what is earlier referred to in this paper as the RS measure (see defini- tions). RS differs from.the "classic" RT in that the former, "...is a function of (a) off-target time between onset of the stimulus and the onset of the response, whether the response is correct or not, (b) number of errors...and, (c) duration of each error before it is corrected (p. 282)." Thus errors of anticipation are permitted, these errors being either "beneficial" or "detrimental" depending on.whether or not they assist the S to make the correct response more or less quickly, respect- ively. USing an experimental set-up quite similar to that of’Adams (1962), Adams & Chambers (1962) held time uncertainty to zero by having all ISIS equal to 0 milliseconds. The stimulus events were either certain or not, with a bisensory and a unisensory tracking task. Findings showed that the bisensory'group had a faster RS when events were certain, and this fact was thought explained by the increased Speed associated with anticipation. These studies seem to Show'that the net effect of time uncertainty is to cause a slowing of the second response only at the shorter ISIS (Adams, 1962), and this defines the PRP phenomenon. The effect of event uncertainty appears to be to generally depress performance, or increase latencies of response in general (Fig. l, p. 201). Methodological diffi- culties are discussed below. Creamer (1963) assigned five groups of S3 to time certain conditions, in which the ISI used.was homogeneous for the group. .A sixth group (time uncertain) had the same five ISIS in a random.order; all groups were pre- sented with two uncertain bisensory events in a tracking task. USing RS, Just as did.Adams (1962), the time uncertain group was found to have 9 slightly slower RS than the time certain group, but this difference was not statistically significant. Upon closer analysis the time uncertain variable was found.to 2321225 the effects of the event uncertain variable, resulting in a delay of RS at longer ISIS fer the time uncertain group. Since the time certain and uncertain conditions were not significantly different from each other on any of the comparisons, Creamer concluded that it is 3322:, not time, uncertainty which is responsible for the PRP. Response delays were ascribed to a "...human limitation for the central decision functions of processing choices among stimuli (p. 19%), "placing him.squarely in the organismecentered, physiologically-oriented theoreti- cal camp. It may be tenable to assume that when events are uncertain (in the double stimulation paradigm) responses to both stimuli must be held in readiness to insure maximum efficiency of response. When either stimulus occurs, there may be competing response tendencies which are associated with the 9223: stimulus which interfere with response to the fi£§§_stimmlus. If so, the £135! response may be delayed without marked delay of the second response. It is usually the case that the second response is delayed at the shorter ISIS, however. Let us consider the implications of assuming equality of response strength: if one of the two responses available to the S is not stronger than the other, then the response elicited by the first stimulus may pgggmg the prepotent response. If this is true we may' expect the second response to be delayed if and only if the second stimulus occurs "too soon," after the first. At greater intervals the effects of response competition may be attenuated. 10 General Discussion The area of reaction time studies in which the PRP is examined is one in which there is'a welter of conflicting findings and general incom- patibility across studies. The British studies typically use two to six gs and multiple measurements. Conclusions tend to evolve from.inspection of the results, rather than the testing of hypotheses. In most of these studies it is common practice to omit reports of statistical tests, if in fact they were performed. Generalization from the British to the American studies and vice versa is tenuous, for the fbrmer typically use extremely ‘well-trained Se and the latter typically do not, though tending to use much larger numbers of SS. American studies tend to be both methodologi- cally "cleaner" and more complex. Another difficulty in generalizing is that the British studies are typically reaction time studies in the classic sense of having discrete responses to discrete stimuli, with some attempt made to control or eliminate anticipatory responses. The American studies involve tracking of changing stimuli, and as such provide for many discrete responses per trial, the sum of which is the S5 "RS." In these studies errors of anticipation are permitted. If one is interested in a relatively pure measure of the response of a S'to stimuli which arrive in close temr poral contiguity, tracking studies are somewhat less than optimally effec- tive. It will be recalled that the PRP is assumed to be a special case of the more general TIR phenomenon. In order to explain the more general case a competing response position was advanced (Reynolds, 196%). Other theories, stimulus- and organism-centered, are not incompatible with a competing response theory. There are difficulties with these theories, however. A stimulus-centered explanation, while being squarely in the framework of 11 classic S-R theory, has difficulty in explaining what guides stimulus- selection in, e.g., stereoscopic perception and dichotic listening. If all the information to which the §_must respond is carried in (or by) the stimulus, how can delays in response be handled when the stimuli are iden- tical, differing only in temporal distribution? While these difficulties are not insuperable, they insure that an explanation exclusively in terms of the stimulus is not soon to be expected. The organism-centered explana- tions fare similarly. "Expectancy" as such is more nominal than explana- tory; An explanation of response delays couched in physiological terms holds more promise than the "expectancy" position and seems more amenable to testing once the neural framework has been established. Heretofore these explanations have been "central" in nature, positing a limit on the ability of humans to process information, without specifying in much detail how, much less why, this limit operates. Given the very large number of neural connections available for information processing, it would appear that the burden of proof is on those who claim a "central limitation." Broadbent (1957, 1958) has attempted such an explanation, but his dis- cussion of Filter Theory tends to follow the data and explain specific findings. Its predictive power would seem.almost too good, fer it tends to "explain" differences in opposite directions equally well. The chief advantage in dealing with the TIR phenomenon in general, and the PRP in particular, by means of a competing response theory is that operational referents for the strength of a response tendency may be pro- vided. In an earlier paper Berlyne linked delays in reaction time to competing response tendencies (1957, p. 33%). If we assume that double stimulation elicits competing responses in the S, we may expect the TIR 12 for a short time, after which the prepotent response should emerge (no Hullian endorsement is thereby implied, however). Further, this state of affairs may not be expected when there is a great deal of practice at the task (such that the presently hypothetical R_2 is extinguished). By training responses to stimuli previously unassociated with them we may operationally define a prepotent response. All previous work implicitly assumes equality of response-strengths, or the absence of a prepotent response. As mentioned in the discussion of Creamer's study (1963), in this case the first-elicited response may become the prepotent response, with predictable delay to the second stimulus if it arrives "too soon." What if the first response is relatively weaker in strength than the second? WOuld the first response be delayed and the second facilitated? Would this result in a breakdown of the PRP? 'What is the second stimulus does not arrive "too soon," but at a still relatively short interval after the first? Although previous work indicates that the responses would tend to approach equality of speed (Davis, 1959), would this finding still hold up if the second response was relatively weaker in strength than the first? An interesting finding has come from the probability learning lit- erature. ‘Ss in a probability learning situation typically do not use a "maximum-success strategy," but tend to follow the objective probabilities fairly closely (Restle, 1961). The probability learning task could be used to train a prepotent response and.gng competing response in the double stimulation paradigm. This would permit study of a competing response position.within the TIR framework. If the S is placed in a probability learning situation with, e.g., a probability of .8 associated with the left-hand stimulus, it is posited that the S, in responding to the left- hand stimulus 80% of the time - which he will eventually do - will learn 13 kinesthetic and proprioceptive cues, with knowledge of results as rein- fOrcement. Let us term this "left-hand response" as Ra and the stimulus to which it is made as Sa- The right-hand response and stimulus will be termed Rb and Sb, respectively. Given the above training procedure with the probability of S3 equal to .8 and the probability of Sb equal to .2 the result will be, if the competing response position is correct, that Sa will be a potent elicitor of Ra in subsequent situations. Sb, given this situation, will be a less potent elicitor, both in relation to Rb and to R3 (with which Rb should compete). the that this fermulation does not imply that Ra‘will be made to Sb, but merely that the situation will set up a response competition which should have discriminable effects with respect to the TIR phenomenon, and.which should vary according to whether Sa precedes or follows Sb in the double stimulation paradigm. This procedure has apparently never been used to test the underlying assumptions of a competing response position as it applies to the PRP. Refractoriness & uncertainty: A.Methodological Critique That the PRP is a stable phenomenon is well-known. One may never- theless wish to ask under what conditions may it be attenuated or destroyed, or is any such conditions exist. One set of conditions has been suggested above, i.e., differential training of response probabilities. One may also suggest another set of conditions involving time and event uncertainty. The role of time and event uncertainty in refractoriness, given a unisensory two-choice task in which anticipatory responses are ggt_permitted, is by no means clear. The studies of.Adams (1962), Adams and Chambers (1962), and Creamer (1963) cannot answer this question, and the studies of Borger (1963), lb Davis (1956, 1959), and Klemmer (1956) can be questioned on various grounds. These studies, although using RT as a performance measure, used few §s who were quite variable. Further, Klemmer (1956) used ISIS longer than those used by other workers, and in a single- rather than double-choice task. Creamer's (1963) study comes closest to investigating the role of time and event uncertainty in refractoriness, but uses a tracking task with RS, rather than RT, as a performance measure. Since in the bisensory con- dition the visual task always comes first (except at the O millisecond ISI), the RS score is confounded by the differential reaction times to visual and auditory stimuli in general. The RS measure, as noted above, is not itself an Optimal measure of latency of discrete responses to discrete stimuli. When Creamer's data is examined one sees that the visual (first) response is always made with a shorter latency than the auditory response. This may be due to the instructions to make the visual response first if the S felt a delay (p. 189), and may have little to do with refractoriness as such. Since the RS includes errors and the time needed to correct them, mean latencies of RS may be a poor estimate of actual RT. There are questions left unanswered.by the studies reviewed herein. When the task involves tracking, i.e., the making of many discrete re- sponses and adjustments until the "correct" response terminates the trial, the actual speed of response is confbunded by anticipations and/or errors and a pure speed measure is lacking. In fairness to these investigators it must be said that they deliberately arranged the situation so that the SS could anticipate and not be penalized for it. But, in encouraging anticipatory responses, it is not wholly surprising to find results suppor- tive of an "expectancy" position. The reaction time tasks avoid this difficulty, but there are few unisensory studies, and apparently none 15 expressly manipulating time and event uncertainty. What would happen in unisensory tasks involving a single, discrete response to each of two stimuli? What would curves of response times (RT) to the first and second stimulus look like if errors of anticipation were excluded? Some of these questions have been tentatively dealt with by Davis (1956, 1959). The general finding (with two gas, obtained with time uncertainty only) is that RT to the first stimulus remains constant as ISIs are increased from O to hOO-SOO milliseconds, while the second decreases exponentially for one S and very nearly linearly for the other. The shapes of these curves are not discussed by Davis. At about 325 milli- seconds ISI the response to the second stimulus becomes faster than that to the first. The studies presented herein examined the effects of training pre- potent responses by means of the probability learning task in an attempt to evaluate a competing response theory of the PRP. Further, different manipulations of time and event uncertainty were made using a unisensory two-(choice task in which direct measures of RT were possible. Time and event uncertainty were thus manipulated and their effects on RT were not confounded in the manner of the RS measure. EDCPERIMENT I Exp. I was set up to assess the effects of training prepotent res- ponses by means of the probability learning task. The sequence of events was to give the S single-choice practice trials, then to place him in the probability learning situation and immediately afterward test him on the double-choice reaction task. A11 stimuli to which the S responded were visual. 16 SubJects The 2% Se were volunteers recruited from the introductory psycho- logy course at Michigan State University; All were male, right-handed, and had at a minimum.20-2O vision, corrected. Ss were randomly assigned to experimental treatments. Apparatus The apparatus was a specially constructed stimulus-presentation unit. Facing the S, at a distance of approximately 86 inches from.his frontal plane, was a panel 18 by 2h inches. Three holes l%~inches square were cut such that a horizontal line through their centers would be 12 inches above the table top on which the panel rested. The centers of the holes were 7 inches apart. The center hole was covered from.the rear with a square of white frosted glass on which.was affixed a small red fixation square, %-inch square. The two other holes were covered from.the rear by the stimulus projection units of Lehigh valley Electronics (LVE) Multi- stimulus Panels, Model #l3h6-h62. The whole panel was painted flat black. Each visual target subtended an approximate visual angle of four degrees, and by placing the S'at least 80 inches from the targets, this provides a visual fixation angle of 10 degrees. Given a stable fixation with "normal" wander, this insures that images will fall on the foveal region of the retina, giving maximum acuity. Response keys provided for the S were Lafayette Radio Telegraph Keys,.Model #MS-h28. The keys were mounted on a heavy fiberglass board 1 inch thick by'6 inches deep by 2h inches wide. Keys were spaced 9% inches apart at their centers. The distance between contacts on each key was calibrated and maintained at .01 inch. A plywood board % inch thick by 17 ll 3/h inches high by ll%-inches deep was placed exactly between the two keys on the fiberglass board and painted flat black. The supporting hardware consisted of a LVE Probability Randomizer, Model #1h85; a LVE Tape Programmer, Model #1319FC; a LVE Digital (Sodeco) Counter bank, Model #lu25; Two LVE Bicircuit Pulse Fbrmers, Model #1537; a LVE Power Supply (5 amp.), Medel #lh2h; TWO Standard Electric Milli- second Clocks, Mbdel #MS-lOO; Three Hunter Timers, Model lOO-B, Series D. All supporting hardware, with the exception of the clocks, were mounted on a desk-type relay rack. The clocks were independently mounted on a separate panel, out of sight of the S.. Clocks were read visually by the S and times recorded on a specially devised form. Procedure All Ss were run individually. Each S was given 25 practice trials on single-choice responses with each hand, and latencies greater than A50 milliseconds were discarded. This served to establish a baseline control for each hand separately as well as to allow the S_to practice making the key-pressing response. Woodworth & Schlosberg (l95h) point out that Ss are near their asymptote on practiced RT with 50-100 trials; further practice adds little proficiency. Instructions to S for this part were: Take a comfortable position in your chair. The keys in front of you are to be used for your reaction to the stimulus lights. I will present you with a signal from one of the two darker screens, left or right, one at a time. I want you to react to the onset of a light by pressing down on the key as fast as you can. First we 'will use the left (right) light. Each time the light goes on, press down on the left (right) key with the first finger of your left (right) hand. Always use the first finger of each hand for your response. I will let you know when we switch to the other side. Any noises which you may hear from the equipment are to be ignored. Remember, you will have to be on your toes, as there will 18 be no warning signal. Is that clear? Finger on the left (right) key now, please. . There was a 20 second rest after each block of 10 trials. After the single-choice practice, Ss went directly into the pro- bability learning (PL) situation. Instructions to S were: I want you to guess which light will go on by pressing down on the key which corresponds to your guess. If you think the left light will go on, press down on the left key. If you think the right light will go on, press down on the right key. As before, use only the first finger of each hand for your guess. As soon as you press the key, one of the two lights, left or right, will go on telling you if your guess was correct. .As soon as the light goes out, this is your signal to make another guess. Try to work at a reasonable pace, that is, do not spend too much time in making your guess. Try to get as many right as you can. Is that clear? Fingers on keys now, please. Depression of either key by the S then turned on a light according to a random schedule under the control of the probability randomizer. The bank of Sodeco counters registered for each stimulus (left or right) two events: 1) The number of times the S actually pressed that key (left or right), and 2) The number of times each light was illuminated. The background of the left hand light was always red and of the right hand light always green; each had three white dots superimposed (use of studs numbered h & 10 and 5 & 11, respectively, on the Multistimulus Panel achieved this result). The dots were horizontally aligned on the left hand light, and vertically on the right, yielding two discriminative cues for the S (color and alignment of dots). This situation is analogous to the free-operant situation in that the S generates his own sequence of events by pressing the keys to signify his guesses. Each _s received 600 trials in the PL task. The vertical panel between keys insured that S used different hands for each response instead of crossing over and making both responses with the same 19 hand. 12 Ss received 80% of the stimuli on the left, and 12 received 80% on the right. Immediately after completion of the PL task, instructions for the double-choice task were read to the S as follows: I want you to look at the little red square on the panel and place the first finger of each hand on the prOper key. I will be showing you some lights in pairs, and I want you to respond to each. You are to respond to the left light with the left key, and the right light with the right key. You must be very careful not to jump the gun, that is, to press down on a key before the light goes on.. Remember, each light has its own response. Sometimes there will be only one light, and not two. If you press the second key before the second light goes on, we will have to throw out that trial and the five just before it, so stay on your toes. There will be no warning signal befbre the lights go on. Sometimes the first light will be on the right and sometimes on the left. Respond only when the light goes on. Is that clear? Fingers on keys now, please. ITIs (times between presentation of pairs of stimuli) were pro- grammed on a VI 10 schedule, with a range of 5 to 20 seconds. This means that pairs of stimuli were randomly presented 5 to 20 seconds apart, with a mean of 10 seconds. (The actual mean ITI was 10-18 seconds.) Each stimulus was on the screen for 95 milliseconds, the same length of time as in the PL situation. There were 50 test trials for each S. The design was a 2 x 6, with 2 Ss in each cell. As a control against anticipatory responses and to introduce event uncertainty, the stimuli were presented with the right stimulus first in half the trials in a random order; the left stimulus was first in the other half, such that the probability of a right-to-left order equalled that of a left-to-right order = .5. Each S always received the same ISI; intervals of O, 100, 200, 300, #00, and 500 milliseconds were used with four Ss at each ISI. Two of each four Ss were those trained in the PL task with 80% of the signals on the right-hand side; the other two had 20 been trained with 80% of the signals on the left. Note that this design provides event uncertainty with complete temporal certainty for all Ss (except those in the O millisecond group, who had both complete event and temporal certainty). Latencies of response greater than 800 milliseconds for either hand were discarded, and the trial re-run. This was in an attempt to control "grouping" of responses (in which the S may wait until both stimuli have occurred, and then respond with a joint, much delayed, response to both), and to eliminate extremely atypical RTS. Previous pilot work showed that a latency of 800 milliseconds in this task were rare. After trials #13, 25, and 37, the S received a 20 second rest. flypotheses It was predicted that Ss in the PL condition where 80% of the responses were made on the left-hand.key would have faster RTs in the two-choice RT task with the left-hand than Ss who had 80% of their PL choices made on the right-hand key. This hypothesis grew out of a belief that 3,, (the left-hand light) would become a potent elicitor of Ra (the left-hand key press) in subsequent conditions. Ra should therefore be made faster to Sa than Rb should be made to Sb. This hypothesis holds, of course, only for those Ss who made 80% of their responses in PL on the left-hand key. The reverse was hypothesized for those Ss who made 80% of their responses on the right—hand key. Results Mean RTS in milliseconds (msec.) for the two groups of Ss according to their PL treatment are presented below in Table l. 21 Table 1 Mean RTS (in msec.) for Subjects bprrobability Learning Group, 80% Probability of 80% Probability of Left-Hand Right-Hand if Left-Hand 32' Right-Hand 3E Left-Hand i Right-Hand 35h.h0 368.20 351.68 382.24 i lst Response* 2 2nd Response* i lst Response* E 2nd Response 369.03 352.62 382.68 351.2h *N.B. These means exclude data from the O msec. Ss. AS can be seen from Table 1, mean latencies for both PL groups were higher for right-hand responses than for left-hand responses, and so no further analyses were made. Averaged across five ISIS, with the 0 msec. groups excluded from the analysis, the mean first response took slightly longer than the second, but this difference is not statistically significant (P) .05, two-Sided, Sign test). This analysis confounds ISIS, and so Table 2 presents a comparison of mean RTS for groups of SS at each ISI, regardless of the PL group from which the SS came. The 0 msec. groups are omitted. Table 2 Comparison of Mean RTS (in msec.) of lst and 2nd Responses by ISIS n ISI in i’Pirst Response 2 Second Response Msec. h 500 338.78 291.80 u too 395.2h 368.33 A 300 378.30 335.5u h 200 372.3u 382.3u h 100 310.78 313-h8 IT in msec. _475 — 550“ !?§"’ !QO 275:? “o n l . ~ 10; m Figure l. .———o In Response I- .- .2nd Response l l _2oo_ 300 W. 93:22.93! .1400 Reaction Time oft-lst and 2nd Response as a Function of ISI Length (Experiment I). 22 9 23 Figure 1 presents a graph of mean RTS, and reflects the data in Table 2. (See page 22) Table 3 summarizes a Lindquist (1953) Type III Analysis of Variance on the data of Exp. I, where Left Hand vs. Right Hand corres- ponds to Lindquist's "A" classification, and the Probability Levels and ISIS correspond to "B" and "C" respectively. Table 3 Lindquist Type III Analysis of Variance of Data From Experiment I (Left Hand vs. Right Hand) Source of Variance df Mean Square F Between Subj ects Probability Levels (PL) 1 93,060.00 (1.00 Interstimulus Inter- vals (ISI) 5 7,965,186.80 < 1.00 PL x ISI 5 2h2,h29.h0 .05, two-sided). A similar test for the mean second response was also not Significant (U = 7; p>.05, two-Sided). A Mann-Whitney U-test comparing the individual Ss at the 100 msec. ISI for the mean second response between experiments was also made; this particular test compared the points of largest differ- ences between SS in both studies. The obtained U’of 5 was non-significant (p:>.05, two-sided). These comparisons Show in general that the data obtained in Exp. I & II are roughly comparable in terms of mean first and second responses across ISIS. General Discussion, Exp. I & II The results of Exp. I & II, in which a PL task was used to attempt to train a prepotent response, were unifbrmly negative with respect to effects of the PL training upon subsequent two-choice RT;* This may be due to the dissimilarity of the double-choice RT’task and the PL task. In the latter, the S presses a key prior to the onset of the stimulus, and with a single, rather than double, response. In the RT task the response is made after the stimulus occurs, and two responses are made in quick succession. If competing responses trained via PL are present the dissimilarity between the two tasks may override them. Furthermore, training via the PL method may be a fragile phenomenon; to overcome whatever existing habit-strengths the SS bring into the experimental situation may require much more than the few minutes the PL task takes. Finally, the effects of complete time. * The research hypothesis dealing with the effects of PL on two- choice RT is tested by the PL x H interaction terms in Tables 3 and 6. Neither are significant. 31+ certainty in these two experiments may also provide for the attenuation of any response strengths presumed trained in the PL Situation. One must conclude that whatever merit the PL task holds for the training of response strengths has yet to be demonstrated. The Lindquist Type III Analyses of Variance (Tables 3 and 6, h and 8) Show generally similar findings with respect to Exp. I and II, and so will be discussed in common. The finding of a significant within subjects effect of handedness (H) in both Tables 3 and 6 seems artifactual. The reason is simply that all subjects were right-handed, and it makes no psychological sense to expect that conditions of the experiment EEE.§E. caused the left hand to respond with a shorter latency than the right (as Subsequent investigation revealed). A more likely explanation is a constant error in the calibration of the equipment, probably in the tension on the spring of the telegraph key, which led to this effect. It is important to note that a constant error of this type would p23 affect the interaction term of PL x H, but merely reduce the latency contribution of the left hand to the interaction term. Analogously, the finding of a significant within subjects effect of order of hand responding (O), i.e., first vs. second hand as revealed in Tables h and 8 may be considered. In this case, any equipment (or constant) error is evenly distributed across trials (and the hand responding) and the classification (0) is not thereby affected. Figures 1 and 2 show that the mean second response was made with a.much shorter latency than the mean first response, predominantly at the longer ISIS. The interaction of ISI x O is not Significant in either study. The conclusion seems tenable that there is no pronounced refractoriness at the short ISIS, while at the longer ISIS the occurrence of the first stimuli serves to act as a warning 35 signal for the second. The finding of no overall difference between the two sets of curves of Exp. I as compared to EXp. II was unexpected. Since Exp. II had a somewhat greater degree of event uncertainty, one would expect greater RTS at each ISI compared to Exp. I. That this was not the case poses a theoretical problem. Creamer (1963) suggested that the PRP was caused mainly by event uncertainty. If this is the case we should expect a greater difference between the two sets of the curves at or near the point of simultaneous stimulation, and the second response should Show this effect. .Although this trend occurred, comparison of these points showed that the difference was not statistically Significant (Sign test between mean first and second response across SS at 100 msec. ISI; p>.05, one-sided for each Exp.). Furthermore, an almost equally great difference was obtained at a 200 msec. ISI for the mean §i£S§_response. There seems too little evidence from the first two experiments to give unqualified support to Creamer's position. The discrepancy may be resolved by looking at each experiment separately. In Exp. I, the shapes of the curves for mean first and second res- ,ponse are curvilinear and not very widely separated. Note that a cross over of the two curves occurs, with the second response having a shorter latency at the longer ISIS. In Exp. II the shape of each curve is irregular, IJut the same effect in general holds: at the longer ISIS the second Iwasponse has a Shorter latency. It will also be noted that the mean first alld.second response made in Exp. I was shorter than.the mean first and SENcond response in Exp. II by 39.13 and 32.97 msec., respectively. Since thee two sets of curves (Exp. I vs. II) are not significantly different, these differences cannot be, but they are in the "right" direction. 36 Note that the points of mean first and second response at 100 msec. ISI were not significantly different from each other in Exp. I or II. One may tentatively conclude that the effects of event uncertainty on refractori- ness can be attenuated or destroyed by having complete temporal certainty. Any refractoriness in these conditions seems at an absolute minimum, and poses a prdblem for those who claim that the PRP is due to a "central limitation" of the nervous system to process information or stimulation. This finding also poses a prdblem for the expectancy theorists, for they would expect greater delays under the conditions of Exp. II as compared to Exp. I. These delays would be posited to occur in the second response at the short ISIS, for any "expectancy" which the S could build up is at a much lower level in Exp. II than in Exp. I. Statistical and visual inspection of the two sets of curves Shows that this was not the case. EXPERIMENT III Exp. III explored the temporal uncertainty variable while holding events certain. Since the PL task had not provided evidence for the train- ing of response-strengths, no further work with it was undertaken. The general manipulation within Exp. III was the presentation of "Random" or "Regular" series of ISIS, while specifying on which side the first stimulus would occur. Subjects The 28 SS were a fresh sample from the same pOpulation as in the fixrst two experiments. All were male, right-handed, and had at a minimum 20-20 vision, corrected. SS were randomly assigned to treatments. 37 Apparatus The same apparatus as in Exp. I & II was used. The probability randomizer was disconnected from the circuit. Procedure All SS were run individually. The order of tasks was single-choice RT, then double-choice RT. Each S received #0 trials on the single-choice task (20 for each hand, alternately, with an approximate 30 second rest after the 20th trial), and then went into the double-choice task. Single choice RTS with latencies greater than #50 msec. were discarded. There were two types of stimulus-presentation in the two-choice task: 1) "Random" presentation of ISIS, in which ISIS were randomly presented to the S with the restrictions that no ISI could follow itself, and that adjacent ISIS had to be at least 200 msec. apart, and 2) "Regular" presentation, in which a block of four identical ISIS (either 0 or #00 msec. in length) were presented to the S, There followed a block of four more homogeneous ISIS, either 100 or 300 msec. in length, respectively. This was continued until the last block of four homogeneous ISIS were either #00 or O msec. in length, respectively. The series of trials starting at O and ending at #00 msec. is termed an "ascending" sequence, while the series beginning at #00 and ending at 0 msec. is termed a "descending" sequence. At no time was the S_informed that the interval was to be changed in the Regular series. All SS proceeded through both the Regular and Random series; latencies greater than 800 msec. were discarded and the trial re-run. Half the SS (N = 1A) got the Random series first, followed by the Regular series, and half the reverse. There were a total of #0 trials in each series; a 20 second rest separated each block of 10 trials. There was 38 approximately a 30 second pause between the Random and Regular series. Half of the _S_s (N = 11+) received the ascending series first; half the descending series first. Half the SS (N = l#) received the first order of stimulus-presentation left-to-right, and half the reverse. The same order (left-to-right or the reverse) was kept for the first 20 trials in the Regular and Random series; after the 20th trial the S was told that the order would be reversed. Instructions for the single-choice task were identical to those in Exp. II. When the order of series for the S was Regular, then Random, the following instructions were given: I want you to look at the little red square on the panel and place the first finger of each hand on the prOper key. I will be Showing you some lights in pairs, and you are to respond to each. USe the left key to respond to the left light whenever it goes on, and the right key to respond to the right light whenever it goes on. As before, use only the first finger of each hand for your response. Each light has its own response, and you are to respond to each light only when it goes on. The time between lights will be very short, so be sure to catch both lights and make both responses. As before there will be no warning Signal, and you will have to be on your toes. The first light will be on the left (right) followed by one on the right (left) very shortly after. Respond to each light only when it goes on, and as quickly as possible. Is that clear? After completion of the Regular series, the E said: Now we’ll continue the same way, only this time instead of a regular time between lights there will be different times between lights in a random, mixed-up order. Sometimes the lights will come on together, sometimes with a relatively long or short interval between them. Respond to each light only when it goes on. The first light will be on the left (right) followed by one on the right (left) very Shortly after. IS that clear? When the order of series for the S was Random, then Regular, the E read the instructions for the Regular series (first paragraph above) and then 39 continued reading the second.paragraph starting with the words, "...there will be different times between lights...." After completion of the Random series, the E merely said, "Now let's continue with the first light on the left (right)." Note that the Random group was advised that ISIS would be different on successive trials. This was an attempt to eliminate some of the "uncertainty" which the expectancy theorists maintain accrues in situations in which the S does not expect a short ISI and so is "less than Optimally prepared to respond." As a further safeguard against objections which expectancy theorists might raise, the statistical distribution of frequencies of ISIS was the same for all SS across the Regular and Random series. Each series contained eight ISIS of 0, 100, 200, 300, and #00 msec.; four were in the order left-to-right, and four in the reverse order. Within the Random series each block of ten trials had two of each ISI. ITIS were identical to those used in Exp. I. Since SS always knew on which side the first stimulus would occur, this experiment compares two-choice RTS in two conditions of temporal uncertainty while holding event uncertainty to zero. The direct comparison, on the Spp§_SS (a "within SS" comparison), between the Random and the Regular series permits an assessment of the effect of increasing temporal uncertainty while holding event uncertainty to zero. Hypotheses Based on the conclusions tentatively arrived at in the discussion of the work of Adams (1962) and.Adams & Chambers (1962), as well as in the discussion of Exp. I & II, it was hypothesized that there would be a delay #0 of response to the second stimulus at "Short" ISIS, i.e., the PRP would be obtained under the conditions of Exp. III. The basis for this was that event uncertainty per se is thought to result in longer latencies across ISIS, and not necessarily in the PRP. (Creamer, 1963, would not agree.) This is because each response, when separately held in readiness, may compete with the other and result in generally longer latencies. When events are certain, the only "problem" with which the S_must deal, is ppgp to get the second response out with minimal delay. (This of course assumes a motivated S.) At longer ISIS the first stimulus can possibly serve as a warning Signal for the second, this is less true at Shorter ISIS, for the first response may already be underway when the second stimulus arrives. If the competing response theory of delays in response is correct, one would look for maximum delays in response to any stimulus if that stimulus arrives while another response is ongoing. This is intended as a generali- zation; specific exceptions may be found. The generalization was first developed in the discussion of a paper by Helson & Steger (1962) by Reynolds (196#) as follows: Helson & Steger (1962) presented two "stimuli" to the S, only the first of which required response, at ISIS of O to 180 msec. The "first" (only) response was found to be delayed when the ISI was 10 to 170 msec., with maximum effect in the range #O-l#0 msec. If‘S is not required to respond to both stimuli, how could his only response be delayed by a presumably irrelevant signal? Here one sees the inadequacy of a stimulus- centered or organism-centered explanation; instead one might wish to consider what a theory of response competition has to contribute. #1 Let us assume a mean reaction time close to 200 msec. for each S, and let us further assume that a response deveIOpS or "unfolds" over time. One may then conceptualize a graph of response-segments plotted against unit time (on the abcissa). If we assume the simplest (but not necessarily correct) function, that of a straight line, then each increment of 20 msec. results in roughly 10% of the response moving toward completion. If the maximum effect of the application of a second signal iS in the range #O-l#0 msec., then we can see that the response is somewhere between 20% and 70% complete when the second (irrelevant) signal is applied. It seems likely that a positively accelerated curve more accurately depicts response- segments as a function of unit time. If so, the above discussion must be modified. The curve of reSponse—segments over time will be displaced to the right of the straight-line function, indicating that relatively'pppg time must pass before a response is, say, 50% complete. Thus the effect of the formation of competing responses may be magnified, and the more rapid the acceleration of the curve, the greater the effect. The empirical questions remaining unanswered are these: I) Will competing responses form to a signal which is not to be responded to? 2) Does enough time elapse during the elaboration of a response (such as made in RT) for a competing reSponse to form? The effects of increasing temporal uncertainty were hypothesized to be greater in the Random than the Regular series; a greater delay in the second response was thus expected in the Random series. Results Table 9 is an analogue to Table 7, and presents mean RTS in msec. for the first and second response of all 28 SS in both the Regular and #2 Random series. Table 9 Mean First 8c Second RTS (in msec.) in Regular 8: Random ISI Presentations Regular . Random ISI E First T'Second I'First E'Second too 273.86 2A2.A7 300.67 273.20 300 269.86 252.u1 292.67 277.05 200 266.89 270.93 280.33 295.02 100 262.91 272.63 273.58 3A2.17 0 258.13 3A3.A7 267.33 Ah3.86 Note that the Random series has a longer latency at every ISI for both the mean first and second response. A Wilcoxon matchedapairs, signed-ranks test shows this discrepancy to be significant (p (.01, Egg- sided). Latencies in the Random series are thus seen to be longer than those in the Regular series, and this is in accord with the prediction ‘based on the greater temporal uncertainty in the Random series. It will also be noted that the ordering of mean responses is the same for both the Random and Regular series. .AS the ISI decreases from #00 to 0 msec., the latency of the first response decreases, while the latency of the second response increases. For both series the second response becomes faster than the first at between 200 and 300 msec. A Friedman two-way analysis of variance by ranks was performed on each set of data on which the four columns of Table 9 are based. Only the ranks of the first column (Regular series, mean first response) was statistically non-significant (p >.05, two-sided). Each of the remaining #3 three sets of ranks was highly significant (p<.001, two-Sided). The interpretation of this finding is that when mean RT latencies are converted to ranks one cannot maintain that there is no difference in latencies across ISIS. Stated in another way, it appears that two-choice RT (given conditions of Exp. III) is systematically affected by increasing or decreasing ISIS. In order to graphically demonstrate this, the results of Table 9 were plotted; this is presented as Figures 3 and #. Figure 3 presents the Regular series; Figure # presents the Random series. (See pages ## and #5) Note that in both Figures 3 and # the curve for the mean first response rises slowly and linearly from O to #00 msec. (statistically significant in both by means of a sign test, p<.03 and p<.OO#, two-Sided, respectively). The drop in mean second response from O_to #00 msec. ISI is also Significant by means of a sign test on each curve at well beyond p<.OOOl, two-sided. Sign tests were performed on points between mean first and second response on each of the figures at 200 and at 100 msec. The difference between mean first and second response at 200 msec. on each figure is not significant (p§>.05, two-sided), but the same comparison at 100 msec. is significant (p4{.OO8, two-sided on Figure 3; p4{.OOOl, two-sided on Figure #). Since the O msec. ISI points for mean first and second response on each figure are more widely separated, they are at least as significantly different as the 100 msec. points. Thus the PRP is found in Exp. III at an ISI of at (or less than) 100 msec. This.is in accord with the prediction that the PRP will occur at the "short" ISIS. In an attempt to determine if the ascending or descending series within the Regular series were differentially affecting the overall shapes (RT in _msec. :7 1‘, .. 4&3" .590"- §Z§'-' 32° ‘- 925 — \ poo - 1“275 "— \_ ___.. w ‘ — — ‘ .__. _ ___..——— \ \ ‘0‘ 250»—- -; _11 ‘ i ‘225 — 0" l . l l _ l J ._..‘ .0“, ‘00 i 200 “39,; 13 400 1511.93! Figure 3. Reaction Time of lst and 2nd Response as a Function of ISI Length (Experiment III-Regular). 81:31} .mse'c. \ 450'— e..- 425— \ .__..;__.g \ ‘ .— _s fig; 409 e- \ . . \' 12.5 "' , \ 'u is l f ‘u n o. I / f2§o— 325— A! -o 1 1 1 1 1 —— p _1_o_9_ 390 300 400 ””151 “(.135 Figure #. Reaction Time of lst and 2nd Response as a Function of ISI Length (Experiment III-Random). Q #5 #6 of the Figure 3 curves, a Wilcoxon matched-pairs, signed-ranks test on the mean latencies of first and second response at each ISI was computed between the ascending and descending series. The difference between the ascending and descending series was not Significant (p).05, two-sided), justifying the combination of the two sequences into one Regular series for the analysis. Discussion The results of Exp. III demonstrate that the PRP ppp be obtained under conditions of temporal uncertainty, even with complete event certainty. This is in Sharp variance with Creamer's (1963) discussion, although it is in accord with Klemmer's (1956) findings. Since neither of these two studies used a discrete, two-choice RT task, generalization across studies is tenuous at best. It will be recalled that Adams (1962) found refractoriness at "short" ISIS using a tracking task under conditions analogous to those of Exp. III. R8 was the speed measure, however. Taken together with the findings of Adams 3. Chambers (1962) discussed above, the conclusion that the effect of event uncertainty is to elevate latencies of‘pppp_responses, while the effect of temporal uncertainty is to elevate latencies of the second response only at "short" ISIS, seems supported by the results of Exp. III. Even though the results of Exp. III are in general accord with those of Adams (1962) and Adams & Chambers (1962), they are not necessarily best explained by the "expectancy" position. Since Exp. III did not permit anticipatory responses, one may reasonably minimize the role of "expectancy" in accounting for the data. The response competition outlined in the Hypotheses section of Exp. III offers an at least equally plausible account. A A7 The difference in the Random and Regular sets of curves was Obtained from the same SS in a "within Ss" design. (Each point on each curve reflects mean data of pl;_Ss.) The effect of manipulating ISI randomness (increasing temporal uncertainty) seems to lead to a greater change in the shape of the mean second response curve than the first. The curve for mean first response rises somewhat more steeply as ISI length increases in the Random series compared to the Regular series. The curve for mean second response changes from a near-linear or perhaps slightly exponential decreasing function to a sharply delineated expon- ential decreasing function. It is tempting to Speculate that the effects of increasing temporal uncertainty results in some logarithmic increase of RT to the second stimulus presented, whereas the effect on RT to the first stimulus presented would appear linear and additive. EXPERIMENT IV Borger (1963; discussed on pp. 6 & 7) could not support a theory of refractoriness based on temporal uncertainty from his data. He combined temporal certainty with event uncertainty in a bisensory RT task. He found the PRP at ISIS of 300 msec. or less. Exp. IV explored the possibility of extending and generalizing his findings, and to further investigate possible experimental manipulations affecting the shapes Of the curves for the mean first and second response. Exp. IV has three parts: Part A is a "between Ss" design in which each S received one and only one ISI. Part B is a "within Ss" design in which each S received different ISIS in blocks Of 10 homogeneous ISIS. Part C is a replication of Part B, with somewhat different instructions to #8 the S, In each part the S always knew to which side the first response would have to be made, hence there was complete event certainty. There was complete temporal certainty as well; Exp. IV thus explores the role of both event and temporal certainty in a "between SS" and in a "within SS" design, using the two-choice RT task. subjects 'Ss (N = 36) were a fresh sample from the same population as in Exp. I. All were male, right-handed, and had at a minimum 20-20 vision, corrected. There were 12 different SS in each part, assigned randomly to experimental treatments A and B, and run consecutively in Part C. Within Part C, SS were randomly assigned to ISI sequences. Apparatus The apparatus used was identical to that used in Exp. III. Procedure All SS were run individually. The procedure for Part A was as follows: SS received 20 single-choice practice RT trials (10 with each hand alternating; latencies greater than 275 msec. were discarded, and the S told he was "too slow"). The following instructions were then read: Now I will present you with some lights in pairs, and you are to respond to each, using the left key for the left light and the right key for the right light. The time between lights will be very short, and as before there will be no warning signal, so you will have to be on your toes. The time between the two lights will always be the same, and you.must respond as quickly as you possibly can to each of them. (The first light will be on the left (right) followed by one on the right (left) Shortly thereafter.) I am.now going to Show you a pair of lights to acquaint you with the interval between them, but do not respond to this first pair (Demonstration). All right, fingers on keys now, 1+9 please. (The first light will be on the left (right) just as I have shown you.) Respond to each light only when it goes on. Try not to "jump the gun." IS that clear? The sentences in parentheses were omitted for the SS whose only ISI was 0 msec. Ss were then given 50 two-choice RT trials in the Specified order at 9g; of six ISIs; 0, 50, 100, 200, 300, and #00 msec. ISIS were used in all three parts Of Exp. IV. There was a 15 second break between blocks Of 10 trials. After the 50th trial, SS received a two to three minute break, and then were informed that the experiment would continue, but with the order of stimuli reversed. One trial to which the S did not respond was again given him to acquaint him with the new stimulus-order. The S was then given 50 more trials in blocks Of 10 at the same ISI, just as above. Only one randomly assigned ISI was used per S; there were two SS at each ISI. Part B differed from Part A in that each S_after completing the 20 practice trials, was read the same instructions as for Part.A, and was then given a block Of 10 trials in which the ISIS were identical. The S then received a 15 second rest, and was then informed that the E was going to change the interval while keeping the previously specified order the same. One trial to which the S did not respond was given to acquaint him with the new interval. This was repeated after each‘block of 10 homogeneous trials, each with a different ISI. ISIS were selected randomly for the sequence with the restriction that each had to appear equally Often as the first, second, third, fourth, fifth, and sixth in order. After 60 trials, exhausting the possible ISI orders for a specified sequence (either right-tO-left or the reverse), the S was given a two to three minute break. 50 He was then told that the experiment would continue with the reverse stimulus order, and was given one demonstration trial to acquaint him with both the new interval and order. Latencies greater than 800 msec. were discarded, and the trial re—run. Part C, in an attempt to explain the rise in the mean first response latency as the ISI increases from O to #00 msec., was run after Parts A & B had been completed; Part C duplicated the procedures of Part B, but tried via instructions to eliminate the rise in mean latency of the first response. Instructions to the S were identical to those for Part B, and the following was added at the end of the instructions: When you respond to each light be sure that you are as quick in your response to the first as to the second light, that is, each light is to be responded to equally quickly, and each accounts for 50% of your "speed score." It's just like a two-item test with each item weighted equally. Is that clear? In addition, before each new block of 10 trials, the S was reminded as follows: Remember that each response has Squal weight in your speed score. Respond to each light only when it goes on, and be equally fast with each hand. ITIS throughout Exp. IV were identical to those in Exp. I. Hypotheses Fbr the data of Part A it was postulated that no delay of response to either stimulus would be found. With completely predictable events (foreknowledge Of‘Wthh response to make first) and completely predictable ISIS (fOreknowledge of when to make the second response), there may be in fact only one stimulus to which the S responds - the first. If this is true then there is but one response, a composite of two motor reactions, and no competing responses may be fermed. 51 Part B may be meaningfully compared with the Regular series of Exp. III, for the only differences are slight: 1) The former uses 10, rather than four, trials at each ISI, and 2) In the former (Part B, Exp. IV) the S.is told when the ISI is changed, but not in the latter experiment. Part B thus has more event certainty than the Regular series Of Exp. III. Accordingly, it was expected that any refractoriness Obtained would be less in magnitude than obtained in Exp. III, Regular series. Instructions in Part C were aimed at flattening out the curve for the mean first response, and possibly also affecting the mean second response as well. In Exp. III it was noted that the curve for the mean first response rises as the ISI increases from O to 500 msec. By instructing the §.t0 be equally fast with each hand, it was posited that the curve for the first response would flatten. It was hypothesized that the points on the mean first response curve for O to #00 msec. would not be different. Results Table 10 is an analogue to Table 7 and 9, and presents mean first and second responses at each ISI for Parts A, B, and C. A Sign test performed on the data Of Part A for all ISIS except the O msec. one compared the mean first with the mean second re- sponse for each block of 10 trials per S. The difference between responses was highly significant (p‘:.OOOl, two-sided), indicating that the mean Sl-a Table 10 Mean First and Second RTs (in msec.) in Complete Temporal & Event Certainty Part A Part B Part C ISI i1 72 71 t a 7. hoo 250.16 172.22 286.61 224.10 270.90 228.52 300 268.h9 180.55 277.8h 223.19 277.77 231.59 200 2h1.20 192.9% 267.99 250.h9 260.h2 2h3.16 100 200.92 178.h2 25h.80 261.1h 25h.90 260.78 50 220.29 186.60 25h.5h 270.5h 251.51 280.18 0 190.51* 2h5.6h 289.22 2h1.80 272.0h l90.Sl* *N.B. No "first" response. Figures represent extrapolations from.means of each hand, and are presented fbr comparison only. 52 second response was made with a shorter latency than the first in the vast majority of blocks of 10 trials each. A Friedman two-way analysis of variance by ranks was computed for the mean first and second response in Part B. Both the ranks of mean first and mean second responses were significantly different across 1815 (p < .001 and p<.Ol, respectively, two-sided). The interpretation possible from this test is that when mean RT is converted to ranks there appears to be differences across ISIs. When one looks at the magnitude and direction of the differences across ISIS, one finds that the mean first and second RTS are inversely ordered by ranks. A Friedman two-way analysis of variance by ranks was computed in an analogous fashion for Part C. The ranks of mean first and second RTS are also significant across ISIs (both at p< .001, two-sided). The results of Parts B and C indicate that as the ISI lengthens the mean first response tends to increase in latency while the mean second response tends to decrease in latency. When Part A is compared with Part B by means of a Mann-Whitney U-test on mean responses at each ISI (excluding the O msec. ISI), Part A is significantly different from Part B with respect to both the mean first and second response (U = 3, p = .028, one-sided and U a 0, p < .00LL, one-sided, respectively). When Part A is similarly compared with Part C, the identical result occurs, with the same levels of significance. A comparison of Parts B and C on the mean first response by means of a Mann— Whitney U~test was not significant (U = 16; p>.OS, two-sided). A 53 Mann-Whitney U-test between mean second responses between Parts B & C was also not significant (U = 17; p>.05, two-sided). These tests indicate the comparability of Parts B and C, and show that the results of Part A is quite different from Parts B & C. Figure 5 presents a graph of the data in Table 10, Part A. (See p. 5%) The mean first response curve rises irregularly but sharply as the ISI lengthens. A sign test comparing the SO and 500 msec. point is some- what on the conservative side, since the lowest and highest points are at 100 and 300 msec., respectively. Nevertheless, the Sign test was highly significant (p(.OOl, two-sided), indicating that the rise in the curve H 1! cannot be due to chance factors. The generally flat appearance of the curve for the mean second response was tested by means of a sign test for the points at 200 and #00 msec. ISI, and this difference was found to be highly significant (p(.OO6, two-sided). As the ISIS increase beyond 200 msec. there seems to be some room for decrease in latency of the second response. These results partly confirm the first hypothesis. Although there is no evidence of refractoriness in the second response, the mean first response does increase in latency as ISI length increases. In order to confirm the first hypothesis, both curves should be flat, with the mean second below the mean first response curve. Figure 6 presents a graph of the data in Table 10, Parts B & C. (See p. 55) In Figure 6 the curve for the mean first response is gain seen to rise sharply, and again the rise is significant by means of a sign test (p‘<.OO6, two-sided). Similarly the drOp in the mean second response ‘37 firms“: 270 -260 7250 240 230 320 210 'I'l'l'l'l"lr|‘| I . fil'l‘l' Figure 5. "is: 1.3.». Reaction Time of lst and 2nd Response as a Function of ISI Length _(Experiment IV-Part A). \ 51+ 290 [RT .ipnumsoc. Part B M ‘3'] P — q.2nd. Pan C Ono-Inoouuuooo I“ 0 - - — 0 20d 1 1 g 1 o 1 1_ 1 ‘ _Q 55). 10‘ __L 300 L490 ”ISI “'6ch 1, Figure 6. Reaction Time of lst and 2nd Response as a (75. Function of ISI Length ,(Experiment IV-Parts B 8c C). \ 55 56 latency as the ISI lengthens is also significant (p<.006, two-sided). A Sign test between mean first and second response at the 50 msec. ISI is not significant (p >.05, two-sided), nor is a Wilcoxon signed-ranks test. (Siegel, 1956, p.78 suggests that a sign test is not very powerful as a test of this type, and so the Wilcoxon was run as a double-check). A sign test at the 0 msec. point is highly significant, however (p4f.006, two-sided). Thus refractoriness, in conditions of complete event and temporan certainty, is found only at the O msec. ISI. Thus the second hypothesis is given some support as the point of refractoriness of the second response includes only the 0 msec. ISI in Part B of Exp. IV, whereas the Regular series of Exp. III shows the PRP at 100 msec. as well as at 0 msec. The curves for mean first and second response for Part C are quite similar to those for Part B. The rise in mean first response and the drop in mean second response latency as ISI length increases is significant (p<.006 and p = .006, respectively, two-sided). A sign test comparing mean first and second response curves at 50 msec. was not significant (p >.05, two-sided), but a Wilcoxon test was (.02.05, pile-sided). A Sign test on the differ- ence between Blocks 1 and 5 on the mean second response curve was highly significant (p = .011, gpgfsided). The prediction made above is thus borne out, and support is given for the "extinction of competing responses" explana- tion. Returning to the discussion of hypotheses made for Part A of Exp. IV, one can see that the original formulation was in error. The double-choice task elicits Ego responses, not one composed of two part-reactions. Com- peting responses d2 appear to form to the second response, and these competing responses are apparently extinguished over time. Part A 329"' ~ ' ao— " ‘2 "' ’———-—"" 35.0"" ' . 2 film-5. '. ' ,_ . ,230'-' .220— U _- -0 .3010— \ .s - \ a: 39°: \ lzo— \ 8180"" F '- —- 4‘ 179— ‘l-— _ __ _I 160— -- F‘ ~l5_0 . .1 . 1, J 1' p 1 —~ I 2 : "i 2 4 Li; Blocks of 20 Trials-.7 Figure 7. Reaction Time of lst and 2nd Response as a Function of Blocks of' Trials (rs-periment IV—Part a). \ 61 62 Figure 8 (see p. 63) presents an analogous graph for the data of Part B. Note that the curves for both the mean first and second response lgige slightly over blocks of 20 trials. Any "practice" effect associated with the second response is certainly at a minimum. This is additional evidence that in Part B a different process (or processes) is involved. In Part B there appears to be no extinction of the competing response R_2, and consequently, the PRP is found at the 0 msec. ISI, and the "crossover" effect of the curves for mean first and second response is Obtained. When data is likewise plotted for the analogous comparison in Part C, identical results are obtained. Both the mean first and second responses have longer latencies after the sixth block of 20 trials than after the first, providing fUrther substantiation of the belief that the major factor influencing the relative shapes of the curves in Part.A vs. those of Parts B and C is the presence or absence of R_2 in competition with R2. 4 RT in msec. -290 3.39 _210 1'.ch ‘0: 0 Part B", I— —- —-l 2nd ,_. (I 1 1 1 1 l 1 1L -13 ' '_ 3 '1; 5 A. _leclu of 20 Tria_|_s__ Figure 8. Reaction Time of lst and 2nd Response as a Function of Blocks of Trials (Experiment IV—Part B). 63‘ 6h GENERAL DISCUSSION AND CONCLUSIONS The effect of event uncertainty in combination with temporal certainty was seen to result in generally longer latencies of response to both stimuli in the two-choice task, using unisensory stimuli. An attempt to train a prepotent response by means of probability learning was unsuccessful (Exp. I & II). The effect of temporal uncertainty in combi- nation with event certainty was seen to result in longer latencies of the second response at the short ISIS - the Psychological Refractory Period. In general, latencies were shorter for both responses in this latter con- dition (Exp. III). The effect of complete temporal and event certainty was examined in conditions analogous to having "well-practiced" vs. not so "well-practiced" §Seompared on latencies of response. The "well- practiced" SS had Significantly shorter latencies than did the "less well- practiced" group (Exp. IV). The results were generally in line with a theoretical position which explains delays in responding as due to the arousal of competing responses. Two types of delay were seen to have different antecedents: 1) Delay of both responses, due to the experimental combination of event uncertainty and temporal certainty; 2) Delay of the second response at short ISIS, due to the experimental combination of event certainty and temporal uncertainty. The second response was also found delayed, to a lesser degree, in conditions where complete event and temporal certainty were combined with relatively few trials. The general explanation of these delays follows Berlyne's (1960) discussion: Making the first response (Ri) leads to the arousal of a response which competes with the second response; this response is labeled R_2, and the second reSponse is labeled R2. Given event uncertainty and 65 temporal certainty (Exp. I & II), the S must keep both responses (right- hand and left-hand) maximally prepared if he is to respond quickly to the onset of the stimuli. This accounts for the increase in latency to both stimuli, for the occurrence of either stimulus arouses responses associated with 2225 stimulus. These responses are presumed to indirectly (via arousal of R_2) compete with each other. Since the leftéhand and right-hand stimulus-orders were randomly presented in Exp. I & II, there is no possibility for R_2 to be extinguished by the repeated elicitation of R2, Since R2 randomly changes over trials from a left-hand to a right- hand response. Thus R2 is delayed across ISIS due to the inhibitory effects of R_2. R1 was somewhat longer than R2 in latency in both Exp. I & II, but this difference was not significant. Where events are certain (Exp III & IV), the first response is not delayed, and latencies decrease when compared to uncertain event conditions (Exp. I & II). The second response is delayed at Short ISIS due to the arousal of R_2, and there is a clear cross over of mean first and second response curves: R1 increases and R2 decreases in latency as the ISI lengthens. The increase in R1 was thought due to either temporal summation of stimuli or brightness enhancement, or possibly increased physiological arousal at the short ISIS - at or below 100 msec. The decrease in R2 latency was thought adequately explained by the extinction of R_2 through the repeated elicitation of R2. r/ 00 REFERENCES Adams, J.A. Human tracking behavior. Psychol. Bull., 1961, 58, 55-79. Adams, J.A. Test of the hypothesis of psychological refractory period. J. exp. Psychol., 1962, 6h, 280-87. Adams, J.A. & Chambers, R.W. ReSponse to simultaneous stimulation of the two sense modalities. J. exp. Psychol., 1962, 63, 198-206. Bartley, S.H. The psychophysiology of vision. In Stevens, S.S. (Ed.) Handbook of experimental psychology. N.Y.: Wiley, 1951. Bartley, S.H. Principles of perception. N.Y.: Harper, 1958. Berlyne, D.E. Uncertainty and conflict: a point of contact between information theory and behavior theory concepts. Psychol. Rev., 1957, 611. ‘ 329-39. Berlyne, D.E. Conflict, arousaILiand curiousity. N.Y.: McGraw-Hill, 1960. Borger, R. The refractory period and serial choice-reactions. Quart. J. exp. Psychol., 1963, 15, 1-12. Broadbent, D.E. A mechanical model for'human attention and immediate memory. Psyphol. Rev., 1957, 6h, 205-15. Broadbent, D.E. Perception and communication. N.Y.: Pergammon, 1958. Buckner, D.N., & MCGrath, J.J. Vigilance: a symposium. N.Y.:‘McGraw-Hill, 1963. Clark, W.C. Relations between thresholds for Single and multiple light pulses. Unpubl. Ph.D. dissert.,'U. of Michigan, 1958. Creamer, L.R. Event uncertainty, psychological refractory period, and human data processing. J. exp. Psychol., 1963, 66, 187-9h. Davis, R. The limits of the "psydhological refractory period." Quart. J. exp. Psychol., 1956, 8, 2h-38. "“" Davis, R. The human operator as a single channel information system. Quart. J. exp. Psychol., 1957, 9, 119-29. Davis, R. The role of "attention" in the psychological refractory period. Quart. Jo em. PSyChOlo, 1959, 11, 211-200 Davis, R. Choice reaction time and the theory of intermittency in human performance. Quart. J. exp. Psychol., 1962, 1h, 157-66. Dember, W.N. The psychology of perception. N.Y.: Holt, Rinehart, 8c Winston, 1960. 67 Elithorn, A. & Lawrence, C. Central inhibition: some refractory observa- tions. Quart. J. exp. Psychol., 1955, 7, 116—27. Helson, H. Current trends and issues in adaptation level theory. Amer. Paypholog., 196A, 19, 26-38. Helson, H. & Steger, J.A. 0n the inhibitory effects of a second stimulus following the primary stimulus to react. J. exp. Psychol., 1962, 6h, 201-05. Klemmer, E.T. Time uncertainty in simple reaction time. J. exp. Psychol., 1956, 51, 179-85. Lindsley, D.B. Psych0physiology and perception. In Current trends in description and analysis of behavior. (No author cited.) PittSburgh: U. of Pittsburgh Press, 1958. Lindquist, E.F. Design and analysis of experiments in psychology and education. Boston: Houghton Mifflin, 1953. Marill, T. The psychological refractory phase. Brit. J. Psychol., 1957, #8, 93-97- Restle, F. Psychology of judgment and choice. N.Y.: Wiley, 1961. Reynolds, D. Effects of double stimulation: temporary inhibition of response. Psychol. Bull., 196M, 61. Siegel, s. Noneparametric statistics. N.Y.: McGraw-Hill, 1956. Telford, C.W. The refractory phase of voluntary responses. J. exp. Psychol., 1931, 19, 1—35. Vince, Margaret A. The intermittency of control movements and the psychological refractory period. Brit. J. Psychol., 19h8, 38, lh9~ 57- Walker, Helen M. & Lev, J. Statistical inference. N.Y.: Holt, Rinehart & Winston, 1953. Welford, A.T. The psychological refractory period and the timing of high Speed performance - a review and a theory. Brit. J. Psychol., 1952, 43, 2-19. Woodworth, R.S. & Schlosberg, H. Experimental psychology. N.Y.: Holt, 195A (Rev. ed.). “1111111111111111111111111111“