Back to Realism Applied to Home Page

Pavlovian Journal of Biological Science, Vol 17, No 2, pp.69-79

Direct_Contin80.doc Direct and Continuous Measurement of Relational Learning in Human Pavlovian Conditioning

JOHN J. FUREDY, PH.D., JANE M. ARABIAN, PH.D., EDDA THIELS, B.A., AND LEONARD GEORGE, B.A.

University of Toronto Toronto, Canada Abstract—Three experiments were conducted employing a continuous measure of conditional stimulus/unconditional stimulus (CS/US) contingencies as perceived by the subject (i.e., subjective contingency or SC). It is argued that direct measurement of relational learning, as indexed by SC, can lead to a better understanding of Pavlovian conditioning processes. The first two experiments applied this approach to a methodologic controversy, raising the debate from a procedure-based argument to testing what the subject actually learns about CS/US relationships. While the issue was not resolved, testable hypotheses for future research were generated from the data. The third experiment contrasted the contingency stimulus-stimulus (S-S) account of Pavlovian conditioning with an earlier stimulus-response (S-R) continguity-reinforcement account. In this experiment, both SC and skin resistance were measured. Evidence for the existence of both cognitive-propositional and response-learning processes in conditioning was obtained.

THE RECENT "paradigm shift" (Segal and Lachman tion, according to which "human classical conditioning is 1972) in psychology from stimulus-response (S-R) mediated by an expectancy of, and preparation for, the behaviorism to cognitive approaches has been evident UCS" (Dawson 1973, p. 85). There is, indeed, an implicit in the specific area of human Pavlovian autonomic epiphenomenalism in the recent literature on conditioning. For example, in a symposium on conditioning, which is similar to the earlier behaviorist conditioning and the cognitive processes, held in 1971, epiphenomenalism. According to that earlier tradition, re- the proceedings of which were published two years lational learning or awareness was considered to be a later (Lockhart 1973), there was considerable emphasis mere by-product or epiphenomenon of "true" autonomic on the role of "cognitive factors," "relational learning," conditional responding and was therefore not worthy of and "awareness of the [conditional stimulus- serious consideration. Hence human subjects in a unconditional stimulus] CS-US contingency" in human conditioning experiment tested according to S-R Pavlovian autonomic conditioning. Although some behaviorist traditions were seldom asked about their contributions to the symposium emphasized the limits of cognitions. cognitive control (Furedy 1973), the more predominant The current epiphenomenalism has simply switched approach is represented by statements like the terms, so that now "a naive observer might be excused for conclusion of another contribu- concluding that autonomic researchers now consider cognitive factors to be the only possible factors in conditioned autonomic behavior" (Furedy 1973, p. 108). This paper is based on a paper presented at the meeting The new exclusive emphasis on cognitive factors is also of the Pavlovian Society, Budapest, Hungary, August 1980. exemplified by the current dominance of contingency Address reprint requests to: Dr. J. J. Furedy, Department models of Pavlovian conditioning, as put forward by of Psychology, University of Toronto, Toronto M5S 1A1 Rescorla (1967) and Wagner and Canada. 70 FUREDY ET. AL. Rescorla (1972). Specific responses to the CS, like the and quite sensitive, as in the case of the negative vs. zero autonomic galvanic skin response (GSR) or the skeletal contingency discrimination cited above, there can also be amount-of-lever-press-suppression in the CER cases of nonveridical, yet highly reliable, instances of paradigm are then seen as manifestations or indices of relational learning. Schiffmann and Furedy (1977) used an the underlying cognitive relational learning, arrangement where the USs were delivered at intervals of contingency-analysis process. as short as ten seconds, together with CS-presentations The approach taken in this and previous (e.g., Furedy where the CS- was a signal for a safety period (no US) of and Schiffmann 1973) papers at the Toronto laboratory at least 29 seconds. In terms of objective stimulus has been to try to measure relational learning directly, contingencies, both the US and the CS- are negative i.e., as a phenomenon in its own right, by assessing the (safety) signals, but the CS- is "more negative" than the subjective CS/ US contingency. Assessment by a US since it predicts longer no-US periods. However, the postexperimental rating scale {e.g., Furedy and SC measure showed that, in terms of subjective relational Schiffmann 1971) or by a more sensitive procedure learning, the US was perceived as markedly more negative which seeks to assess continuously changes in than the CS- (Schiffmann and Furedy 1977, Figure 1). subjective contingency (SC) throughout the experiment This paper presents three experiments that continue the (Furedy 1973) has been employed. The latter method approach of directly measuring relational learning. The involves subjects indicating their moment-to-moment first two experiments apply the approach to a changes in belief about the likelihood of US occurrence methodologic controversy which heretofore has been (i.e., SC) by means of a rotary lever, the position of conducted in strictly procedural (objective-contingency) which is recorded continuously just as changes in skin terms. The approach advances and sharpens the enquiry resistance are recorded. beyond strictly procedure-based arguments to ones that One aspect of Pavlovian conditioning that has been deal with what the subject actually learns rather than revealed by this focus on directly and separately what the experimenter has arranged in the Pavlovian measuring relational learning is the presence of certain conditioning experiment. marked instances of fractionation between the process The third experiment was designed to empirically measured by the SC index and the autonomic process contrast the contingency S-S account of Pavlovian measured by such indices as the GSR. Whether SC was conditioning (e.g., Rescorla and Wagner 1972), with an measured after (e.g., Schiffmann and Furedy 1972) or earlier S-R contiguity-reinforcement one (e.g., Jones during (e.g., Schiffmann and Furedy 1977) the ex- 1961). The accounts lead to opposite predictions rather periment, fractionation was manifested by: (a) the than the weaker pattern of predictions, more common in finding that, whereas the cognitive SC measure psychologic research, where one account predicts a differentiated between zero and negative CS/US difference and the other does not. contingency, i.e., between what Rescorla (1967) has called, respectively, the "truly random" and "explicitly Actual Relational Learning About Purportedly unpaired" control CS, the autonomic conditional Random Control CSs: Experiments I and II responses (CRs) did not differentiate between the two contingencies (cf. Furedy and Schiffmann, who The above-mentioned finding of a failure in the employed the GSR; Schiffmann and Furedy 1972, who autonomic system to differentiate between zero and examined peripheral vasomotor responding; and Szalai negative CS/US contingency is contradicted by only one and Furedy 1978, who measured heart-rate human autonomic study in which there was some deceleration); (b) the finding that the extent of SC evidence for this sort of differentiation (Prokasy et al. discrimination between the CS associated with the US 1973). One feature which was unique to that experiment (CS+) and the control CS (CS-) was not correlated was that all three CSs were delivered to each subject in with the extent of autonomic CR discrimination (e.g., a totally within-subject design. Specifically, all subjects Furedy and Schiffmann 1974). received a CS +, a CS that was negatively correlated The more sensitive, within-experiment SC measure with the US (i.e., CS-), and a CS that was said to be has also made it possible to distinguish between random with respect to the US (i.e., RS). Within-subject objective (procedural, experimenter-arranged) CS/US designs seem ideal for assessing the relationships between contingency and subjective (subject-perceived) CS/US RS, CS-, and CS + in autonomic conditioning because of contingency. For example, although the SC measure the increased sensitivity and avoidance generally shows the relational learning of subjects to be veridical RELATIONAL LEARNING 71

FIG. 1. Relational learning (vertical axis: mean subjective contingency, percent full scale) in a within-subject design as a function of whether the purportedly random stimulus ("RS") is generated by the "independence" (Experiment I) or the "overlap" (Experiment II) method. The horizontal axis represents seconds after CS onset. of such problems as the law of initial values. empirical significance, if only because without its However, Furedy, Poulos, and Schiffmann resolution it is impossible to empirically evaluate the (1975a) questioned the validity of Prokasy et al.'s influential position advanced by Rescorla (1967) and (1975) results on the grounds that their RS was accepted by many Pavlovian conditioners that the actually excitatory, i.e., like a CS +. On the other "proper" control for Pavlovian conditioning is the "truly hand, Prokasy et al. (1973, p. 152) had questioned random" (RS) rather than the "explicitly unpaired" the validity of such earlier studies from the Toronto (CS-) arrangement. laboratory (e.g., that of Furedy and Schiffmann However, it is at least arguable from a psychology-of- 1971), on the grounds that the nominal RS in those learning point of view that the more relevant issue is not studies was actually inhibitory, i.e., like a CS-. the procedural one of which arrangement leads to an A summary of the procedural debate is that it objectively random stimulus but rather the question of turns on the question of whether an RS should be how the subject perceives the two sorts of nominally generated by equating occasions when the RS does random stimuli. The continuous within-subject SC mea- and when the RS does not overlap with the US sure was employed in the present two experiments to (Prokasy et al/s "overlap" method) or by ensuring answer this question. The RSs were generated either by independence between RS and US onsets (Furedy et use of the overlap (Experiment I) or independence al's "independence" method). Beyond this summary, (Experiment II) method. however, no further advance has been made in resolving the problem. The last paper from the Materials and Methods Toronto laboratory asserted that the overlap method of generating RSs led to "logical problems" (Furedy Subjects. Twenty-four and ten college-aged subjects et al. 1975b), such as being forced to misclassify a were used for Experiments I and II, respectively. CS+ in a trace and delay-without-overlap Subjects received course credit for writing a brief conditioning paradigm (Kimble 1961) as a CS-. On account of the experiment after it was over. the other hand, the last paper from the Utah laboratory asserted that the independence method led to an Apparatus. The indicator for continuously inhibitory CS, as shown by "actuarial data" (Prokasy measuring SC (cf. Furedy 1973, pp. 110-111) was a 1975b). It bears emphasis that, although technical, rotary lever mounted on a scale positioned in front of the debate has considerable the seated subject. The horizontal 72 FUREDY ET. AL. left, vertical, and horizontal right positions were 1973). All subjects then received the above-mentioned marked, respectively, -100, 0, and +100. The subject sequence of stimuli. During the sequence of stimuli, the was told that these positions represented, experimenter was in an adjacent room, and the subject was respectively, certainty that the noise would occur, continuously visible through a one-way mirror. complete uncertainty, and certainty that the noise In Experiment II, the aim was to adapt the would not occur. The output from this dial was "independence" method of generating RS employed in two- recorded on an E & M Physiograph in an adjoining group (e.g., Furedy 1971) and three-group (e.g., room; the chart paper ran at a continuous speed of 1 Schiffmann and Furedy 1977) designs to the single-group, mm/second. The USs and CSs were the same as within-subject design used by Prokasy et al. (1973). The those used by Prokasy et al. (1973): a 115-dB sound- CS and US durations were the same as in Experiment I pressure-level (spl) noise source from a standard (five and 0.2 seconds, respectively); a randomly ordered white-noise generator delivered through headphones sequence of 15 CS + , 15 CS-, and 15 US-alone trials was for the US and three white lights arranged in a created. In this arrangement, the CS+ was always followed column on a box located adjacent to the SC dial for the by a US occurring at CS+ offset (i.e., a five-second CS-US CS + , CS-, and RS. The sequence of lights and noises interval), and the intertrial intervals were randomly was controlled by a combination of paper tape and assigned duration of 30, 40, or 50 seconds (so that a CS- relays located in another room. was never followed by a US for at least 29 seconds). Procedure. In Experiment I, the purpose was to Then, the experimental session was divided into 191 ten- duplicate the computer-generated arrangements as second intervals, and 15 five-second RSs were randomly described by Prokasy et al. (1973, pp. 147-8) and as assigned to these periods. The exact placement of each further detailed in a protocol sent by Prokasy (1976). RS within a given ten-second interval was determined by First, the 46-minute session was divided in 540 5.1- dividing the interval into 20 0.5-second subintervals and second periods; the five-second CS occurrence was distributing RS onset randomly among these.* In other re- randomly assigned to these periods with a spects, the procedure for these subjects was identical to probability of .25. The 134 CS events were then that used in Experiment I. randomly allocated to the CS + , CS- or RS categories. Results. The index of relational learning, SC, was Next, the session was again divided into 2700 1.02- measured second by second from 0 to five seconds second periods, and US occurrence was randomly as- following stimulus (CS + , CS-, and RS) onset. The results signed to these with a probability of .033, there being are summarized in the left (Experiment I) and right 95 such 0.2-second US (noise) occurrences. Finally, (Experiment II) panels of Figure 1, averaged over trials the two adjustments made by Prokasy et al. (1973) and subjects. The SC response topography indicated that were made, namely: (a) any US falling within 13 SC level at second 5 (the point of noise onset for CS + ) seconds of CS- onset was eliminated,* and (b) any US could be used as a measure of relational learning. In terms falling within 13 seconds of CS+ onset was moved to of that index, statistical analyses confirmed the main CS-\- offset, i.e., a CS-US interval of five seconds. trends suggested by Figure 1, namely that: (a) in The sequence containing the period with the first 50 Experiment I (left panel), both CS+ and RS responding US presentations was transcribed to paper tape exceeded that to CS- (/ (23) = 9.1 and 7.3, respectively, because Prokasy et al. (1973) reported data for only P < .001), but that between CS + and RS did not differ this part of their experiment. (/ (23) = 1.8); (b) in Experiment II Before the start of the experiment, subjects were informed that there would be lights and noises and were instructed to use the SC dial to indicate "moment- * This procedure resulted in an independence or lack of to-moment" changes in belief about the likelihood correlation between RS and US occurrence so that "the that the noise would or would not occur, as in probability of a US is the same given either the presence or absence of the_CS" (Rescorla 1969, p. 25), i.e., p(US/RS) = previous studies using this SC measure {e.g., p(US/RS). However, as has been detailed elsewhere (cf. Furedy and Schiffman Prokasy 1975a, especially Table 1), the procedure results in far fewer physical RS/US overlaps than instances wherein the * It is this adjustment that, according to critics of the RS and US do not occur together. On these grounds, it has overlap method of generating RSs, results in the RS been argued that the independence method of generating RSs becoming excitatory (cf. Furedy et al. 1975a, p. 100). results in the nominal RS becoming inhibitory (cf. Prokasy On the other hand, as indicated by Prokasy (1975a), the 1975a, Prokasy et al. 1973). method does result in an approximately equal number of occasions on which the RS and the US do and do not physically overlap with each other, i.e., an equal overlap-nonoverlap ratio for the RS. RELATIONAL LEARNING 73 (right panel), CS+ responding exceeded that to RS and < .001), RS, and CS- (t (9) = 35.3 and 35.7, respectively, P < .001), 6.7, respectively, for CS-{P < . whereas that to RS and CS- did not differ (t = .31). 01)). These statistical outcomes clearly indicate that the Discussion nominal RS in Experiment I generated by the overlap The results of central concern are those that relate to procedure, was perceived as being as excitatory as the the relational learning induced by the two nominal RSs. CS + (as argued by Furedy 1975a, b) and that the In this respect, the outcomes are clear. Both the criticisms nominal RS in Experiment II, generated by the indepen- by the Toronto laboratory {e.g., Furedy et al. 1975a) of the dence procedure, was perceived as being as inhibitory Utah laboratory overlap method of generating RSs (cf. as the CS- (as argued by Prokasy et al. 1975a, b). results of Experiment I) and those by the Utah laboratory In addition to these main outcomes, statistical {e.g., Prokasy 1975a) of the Toronto laboratory's analyses were performed to determine the reliability of independence method of generating RSs {cf. results of other trends suggested by Figure 1. The first of the Experiment II) appear to be strongly confirmed by the questions raised by inspection of the figure is whether data: the RS generated by the overlap method (Figure 1, the intertrial interval SC levels, as defined by the 0- left panel) was perceived to be as excitatory as the CS + , second (stimulus-onset) values, were the same prior to and the RS generated by the independence method (Figure each stimulus. These 0-second values did not differ be- 1, right panel) was perceived to be as inhibitory as the tween the CS + , RS, and CS- conditions in either CS-. experiment, F < 1, suggesting that subjects were unable On the face of it, then, it appears that the resolution of to predict the nature of the succeeding stimulus. the overlap-independence procedural controversy about These results, therefore, also indicate that the 0- what is an RS is that both methods are equally second SC values can be validly compared with the erroneous, the errors being in the opposite direction and five-second SC values to assess the significance of the as specified by the respective critics. However, in the effect of each stimulus on relational learning. These case of the independence-generated RS of Experiment tests, on the SC difference between seconds 0 and 5, II, there is other SC-based evidence that is contrary to showed that: (a) in Experiment I, CS + and RS were this symmetric resolution of the issue. This evidence, both perceived as significantly positive (t (23) = 6.0 and referred to above, is that, in all previous studies where 6.4, respectively; P < .001), but the perceived negativity this sort of RS was measured in terms of SC taken either of CS- failed to reach significance (t (23) = 1.0); (b) in after the experiment (Furedy and Schiffmann 1971, Experiment II, the perceived positivity of CS + and the Schiffmann and Furedy 1972) or during the experiment negativity of RS and of CS- were significant in all (Furedy and Schiffmann 1973, Schiffmann and Furedy three cases (/ (9) = 27.0, 10.5, and 10.2, respectively;/5 1977), the RS was differentiated from, and perceived as < .001). more positive than, the CS-. Moreover, use of the latter Finally, the development of these individual-stimulus within-experimental SC measure also indicated that the SC effects over the session was studied by dividing the RS was perceived as neutral or random with respect to the trials in the experiments into three blocks. In US, there being no difference in SC levels at 0 and four to Experiment I, the number of trials was not divisible by five seconds following stimulus onset for the RS in the three, so that the first, second, and third blocks Schiffmann and Furedy (1977, Figure 1) study. contained, respectively, seven, seven, and eight trials One feature that differentiates the present in- with each of the three stimuli (CS + , RS, and CS-). dependence-generated RS of Experiment II and those of Oneway analyses of variance, with blocks as the factor the previous studies is that only in this study was the and the algebraic SC difference between 0-second and experiment completely a within-subject design, with all five-second values as the dependent variable, did not three CSs being presented to all subjects. Moreover, in show any significant development over blocks for CS + , Experiment II, the schedule of reinforcement for CS+ RS, or CS- (F (2.60) = 3.0, 2.3, and 1.4). In Experiment was maximal at 100% in contrast with the low schedule of II, the trials were blocked into three 15-trial blocks of under 40% for Experiment I. The alternative, testable five of each of the three stimuli. Analyses of variance interpretation of the RS results of Experiment II, then, is showed that both the perceived positivity (in the case of that the three-stimulus, within-subject design, when used CS + ) and the perceived negativity (in the case of RS with a high-reinforcement schedule CS + , produces a and CS-) significantly increased over blocks of trials categorization error with subjects who, when {F (2.27) - 10.1, 6.2, and 74 FUREDY ET. AL. faced with the strong contrast between the strongly tioning and learning up to the mid-1960s (e.g., Kimble positive CS (CS + ) and other two CSs (RS and CS-), 1961), as indicated by the fact that the critical temporal adopt a dichotomous view of the CSs (CS+ vs. the interval for conditioning was stated to be that between rest) rather than being sensitive to the finer the two stimulus elements, the CS and the US. contingency differences that exist between RS and One indication of the lack of a maximal difference CS-. This categorization-error interpretation, then, between the contingency and pairings S-S positions is asserts that it is not the independence method that the fact that the differential predictions of the two results in RS being perceived as negative (and like positions have typically been ones whereby one the CS-), but the combination of the within-subject account predicts a difference and the other a null (no design with the high schedule of reinforcement for difference) result. Thus, whereas the pairings S-S view CS + . The interpretation has at least two testable predicted differences as a function of the CS-US consequences: (a) reducing the CS+ reinforcement interval (e.g., Kimble 1961), Kamin (1967), in one of schedule should produce a differentiation between the early versions of the cognitive-propositional RS and CS- as measured by SC, and the former contingency S-S position, presented extensive data stimulus should be perceived as neutral (as in based on CER studies that the CS-US interval made no Schiffmann and Furedy 1977, Figure 1), and (b) difference. Conversely, in the matter discussed above, rerunning the arrangements of Experiment II but of whether so-called "explicitly unpaired" and "truly- omitting the CS+ and CS- trials (thereby transforming random" control CSs (Rescorla 1967) produce the experiment into a between-subjects design) should differential performance, the pairings S-S position produce a perception of RS as a neutral stimulus, predicts that they do not, whereas the contingency S-S whereas carrying out the same elimination for position predicts that they do. Experiment I should leave that RS subjectively The problem with these differential-null patterns of excitatory. predictions is that, whenever the null alternative is Accordingly, although the present two experiments obtained, the outcome is considered to be less than do not resolve the methodologic controversy about completely convincing for theory testing; the the generation of RSs, the measuring of subjective outcome is classed as a "negative result" and hence contingency does take the enquiry beyond purely intrinsically indecisive (cf. Furedy 1978). For this procedure-based arguments to ones concerned with reason alone, arrangements where both competing the psychologic issue of what the subject actually accounts predict differential results in the opposite learns during Pavlovian conditioning procedures. direction are preferable. That issue, like all psychologic ones, is likely to have It is possible to generate such an arrangement if both a complex resolution, but, as indicated above, the is- relational learning is measured in the way proposed here sue can be tackled in terms of testable hypotheses and the account that is contrasted with the cognitive- by using sensitive, within-experimental measures of propositional S-S contingency account is not the relational learning. pairings S-S account but a more thoroughgoing In the next experiment, both relational learning and response-learning account: the S-R contiguity- autonomic conditioning were conjointly measured in reinforcement theory of Pavlovian conditioning put an arrangement where cognitive-propositional and forward by Jones (1961) and by Champion and Jones response-learning principles were contrasted. (1962). According to contiguity-reinforcement theory, the Contingency S-S vs. Contiguity-Reinforcement critical interval is not that between CS and US, but that S-R Accounts of Pavlovian Conditioning: between CS and UR, so that a US-CS arrangement Experiment III will produce conditioning as long as the UR follows CS As noted in the introduction, the contingency onset. Such is the case whenever the response in stimulus-stimulus (S-S) account of Pavlovian question is of long enough onset latency to exceed the conditioning is a cognitive-propositional one ac- US-CS interval. In the line with this view, there is some cording to which the relevant process is the learning evidence of conditioning with the long-latency GSR of the predictive relationship between the CS and the and US-CS intervals of less than one second US. Although the account that has been most (Champion and Jones 1962, Furedy 1967, Trapold, commonly contrasted with the S-S contingency Homzie, and Rutledge 1964), although failure to obtain position is the so-called "pairings" (Rescorla 1967) conditioning has also been reported (Furedy 1965, view, this latter account is still an S-S one. Pavlov Zimny, Stern, and Fjeld, 1966). In any event, this S-R (1927) held such an S-S view, as did most account of Pav- textbooks on condi- RELATIONAL LEARNING 75 lovian conditioning predicts that a CS paired with the Beckman electrode paste) for skin-resistance (GSR) US in a short-backward (bCS) arrangement will be measurement were on the palmar surface of the ring and excitatory and will therefore elicit a larger GSR than index finger of the left hand. The signal was amplified by that elicited by a CS that is (explicitly) unpaired (euCS). a standard low-level D.C. Amplifier (5P) and recorded According to the S-S contingency account, on the on a Grass Model 5 Polygraph at a chart speed of 2.5 mm/ other hand, the bCS is even more inhibitory than the second. euCS. The bCS and euCS are inhibitory because they Procedure. In order to generate the explicitly unpaired both signal US-free ("safety") periods. The former (euCS) paradigm, the 20-minute session was first divided CS, because it always closely follows the US, is never into 1200 seconds; 0.5 second noise-USs were randomly closely followed by the US; the latter CS, because it is allocated across the session with the restriction that the separated from any consequent US by a period of time, minimum and maximum US-US intervals were 31 and 130 also is never closely followed by the US. In the seconds, respectively. Fifteen five-second light-CSs, one conditional probability terms often used to state the per inter-US interval, were then placed randomly CS/US contingencies, it is possible to write for both CSs anywhere within this interval with the restriction that the that, p(US/CS) < p(US/CS), i.e., the probability of US minimum CS-US interval must exceed 29 seconds. All occurrence in (or closely following) their presence (CS) randomizations were computer-generated and resulted in is less than that in (or closely following) their absence a stimulus sequence in which the average CS-US interval (CS). Conditional probability terms may also be used to varied around a mean of 54 seconds, and the average indicate in what way the bCS is more inhibitory than the US-CS interval around a mean of 26 seconds. euCS. With respect to conditional US probability, given In order to obtain similar CS-US intervals for the CS presence, the two CSs are equal, i.e., p{US/bCS) = backward (bCS) paradigm, the 15 USs of the backward p(US/ euCS). However, the conditional US probability paradigm were allocated such that they resulted in an given bCS absence exceeds that given euCS absence, average US-US interval of 55 seconds with a range of i.e., p(US/bCS)> p(US/euCS).* 31-80 seconds. Materials and Methods Since the US-CS interval was fixed at 0.7 second, none Subjects. There were 24 subjects from the University of the 15 CSs preceded a US within 20 seconds, and the of Toronto who had volunteered to participate. The data average CS-US interval varied around a mean of 54 from another eight were not used because of seconds, as in the case of the euCS paradigm. apparatus failure (four), early termination by the To increase sensitivity, a within-subject design was subject because of the aversiveness of the experiment employed in which subjects received the two paradigms (three), and misunderstanding of the instructions (one). in two phases. The phase factor (first vs. second), Apparatus. The SC apparatus was identical to that introduced by this within-subject design, was used in Experiments I and II. The silver/ silver- counterbalanced over paradigms (euCS vs. bCS) by chloride finger electrodes (used with presenting half of the subjects with the euCS paradigm in the first phase and the remainder with the bCS paradigm in the first phase. Preliminary trials preceding the first * This last inequality may also be derived by considering phase included a (to-be) euCS, a US, a (to-be) bCS, a US, the amount of excitation elicited by the background stimuli because, on the contingency account, excitation to the and a (to-be) euCS, separated by intervals of 35, 40, or 45 background determines inhibition to the CS (Wagner and seconds. In addition, following each paradigm (of 15 CS Rescorla 1972, p. 315). Under the euCS condition, there are and 15 US presentations as detailed above) there were two types of background periods. One type is followed by three CS-alone test trials separated by the same a US and the other by a CS. Under the bCS condition, there (unsystematically arranged) intervals. It will be noted that is only one substantial type of background period, that such trials are necessary for testing GSR conditioning in followed by a US. The other type of background lasts only the short-backward paradigm because the latency of onset 0.2 second (i.e., the interval between US offset and CS of the GSR exceeds the US-CS interval. onset) and may therefore be neglected. Hence, the The instructions to subjects concerning use of the SC background under the bCS condition is more excitatory dial and the stimuli to be presented were identical to those (being always followed by a US) than that under the euCS used in Experiments I and II except that after the first condition (being sometimes followed by a CS and sometimes by a US), and thus the bCS is more inhibitory phase had been completed (euCS or bCS paradigm plus than the euCS (Wagner and Rescorla, 1972). the three 76 FUREDY ET. AL. CS-alone test trials) subjects were informed over the of subjects in the paradigm x phase matrix does not allow a intercom that the second part of the experiment would split-plot ANOVA model to be applied. Such a model is be presented. The total experiment lasted about one applicable, however, to an "interactive" factorial, in hour. which the phase factor is reclassified so that it is Results connected to the paradigm factor. It has been shown that The results of the second-by-second SC measure are the statistical results from this interactive factorial are summarized in Figure 2. The SC response topography logically equivalent to effects describable in the simple shows a marked difference between the euCS and bCS factorial terms of paradigm, phase, and trials (Furedy conditions with respect to initial (0-second) level and 1967). In these simple factorial terms, the an approximate asymptoting of the stimulus-elicited counterbalancing for phase controls only for any main SC change by seconds 4 and 5. Therefore, in this effect due to that factor and not for any interaction experiment, relational learning to the two CSs was involving phase and paradigm as terms (cf. Furedy 1967). defined as the algebraic difference between the SC However, in the present experiment, no such phase x par- level at second 0 and the mean of the SC levels at adigm interactions emerged either in the case of SC or of seconds 4 and 5, and it was this index that was used in the GSR, so that the counterbalancing for the phase factor the statistical analyses to be reported. appears to have been successful. The first analysis was conducted to test the Of the significant effects to emerge from the paradigm x prediction drawn from cognitive S-S theory (i.e., phase x trials ANOVA, the first was the paradigm effect that the bCS would be more inhibitory than the euCS). already reported in the previous paragraph. The main This prediction was confirmed, the difference between effect of phase was also significant, F (1,22) = 15.15, P < . the mean SC change to bCS (-102.96) and that to euCS 005; as suggested by the trends in Figure 2, the nature of (3.72) being significant (/ (22) = 11.79, P < .001, one this effect was that CSs presented in the second phase tailed). were perceived as more inhibitory (X = -67.24) than in A further analysis examined the main and interactive the first phase (X = -16.00). In addition to this evidence effects of the following factors: paradigm (euCS vs. for between-phase learning of the inhibitory nature of both bCS), phase (first vs. second), and trial blocks (three CSs, there was evidence for within-phase learning as blocks of five trials each). In this regard it will be noted manifested by a significant trial-blocks effect, F (2,44) = that the counterbalanced phase factor makes it difficult 6.05, P < .01; this ef- to statistically assess the data because the distribution RELATIONAL LEARNING 77 feet arose from an increased perceived negativity of 1961, Furedy 1967 and 1970, Trapold et al. 1964). In CSs, especially from the first to the second block of S-R terms, this paradigm is not really "backward" trials (X = -37.20, -55.80, and -55.84 for the three because the CS is followed by the (GSR) UR though trial blocks, respectively). not by the US. However, failures to obtain GSR However, the within-phase learning was restricted conditioning under these arrangements have been to the bCS paradigm. Thus there was a significant reported (Furedy 1967, p. 162, Zimny, Stern, and paradigm x trials interaction, F (2.44) = 7.88, P < . Fjeld 1966). Moreover, the use of CS-alone test trials 005, with the increased perceived negativity of the to assess conditioning raises the possibility that any CS over trial blocks occurring in the bCS paradigm difference may be totally due to the reinstatement of (X = -77.52, -112.20, and -119.20 for the three trial the orienting reaction rather than to conditioning (cf., blocks, respectively, but not in the euCS paradigm e.g., Badia and Defran 1970). On the other hand, it has (X = 3.12, 0.60, and 7.48 for the three trial blocks, been argued more recently (e.g., Furedy and Arabian respectively). 1979, Furedy and Poulos 1977), that the evidence, The GSR was defined as the drop in skin resistance when critically examined, does not support this sort of initiated between one and five seconds after stimulus confounding through the reinstated orienting reaction. onset and was transformed into conductance change Nevertheless, neither the present results nor the scores in micromhos. The optimal point at which to pattern of results in the literature suggest that the test the S-R contiguity-reinforcement theory's superiority of the bCS over the euCS arrangement in prediction that the bCS would produce greater terms of GSR conditioning is a particularly robust responding (i.e., be more excitatory) than the euCS effect. What is clear, however, is that there is no was the first CS-alone test trial following the 15 CS evidence at all for any reversed relationship of bCS < trials of each paradigm. Later CS-alone trials were euCS, as predicted by an S-S contingency account. less desirable because the first CS-alone trial is also In contrast, the continuous SC measure of relational an extinction trial. learning yielded very marked differences and hence The mean GSR in micromho units, elicited by the merits more detailed discussion. Because of the bCS and euCS first test trials, was, respectively, 2.93 marked difference between bCS and euCS in the and 1.56; this predicted difference was significant, / predicted direction, cognitive SC can be said to (22) = 1.99, P < .05, one-tailed. It will be noted that accurately reflect objective CS/US contingencies and to on an S-R contiguity-reinforcement account the euCS confirm the notion that in Pavlovian conditioning paradigm is an unpaired control condition which is the "cognitive system" does act as a "contingency expected to elicit less responding than a paired, analyzer." However, there are also aspects of the SC excitatory condition like the short-interval bCS data that strongly contradict this notion of a paradigm. correspondence between subjective and objective Discussion contingencies: relational learning, that is, may be The directionally opposed predictions derived from cognitive, but it is not always veridical. the S-S and S-R accounts were each confirmed for The nonveridical aspect of the present SC data is the cognitive SC and autonomic GSR measures, that the true inhibitory nature of the euCS had not, respectively. Thus both the cognitive-propositional apparently, been learned. Although there was an overall and response-learning processes reflected by each learning of the inhibitory nature (i.e., negative objective measure appear to occur in Pavlovian conditioning. CS/US contingency) of both CSs, as evidenced by the These results, therefore, constitute further evidence between-phases effect, within-phase (trials effect) against both the behaviorist and the more recent learning occurred only for the bCS and not for the euCS cognitivist forms of epiphenomenalism referred to at paradigm. Also, as is evidenced both from Figure 2 and the outset: can be reduced to the other, both from the algebraic values of the mean SC change scores processes occurring and following different laws during associated with the euCS, this CS was not perceived as Pavlovian conditioning. clearly inhibitory but as slightly excitatory. In addition, The GSR results themselves, though significant, are an aspect of Figure 2 that is of relevance to this point is not particularly strong and, therefore, merit only brief represented by the 0-second values which reflect the comment. The superiority of the bCS over the euCS perceived US expectancies during the intertrial-interval paradigm is a form of "backward" conditioning or background-stimulus periods. These values are which has been reported previously to occur when markedly positive for the bCS paradigm, in line with the CS-US interval is exceeded by the (relatively the long) latency of the (GSR) UR (Champion and Jones 78 FUREDY ET. AL. notion that if a CS is perceived as inhibitory, then the process-reflecting measures like the autonomic GSR, background is perceived as (relatively) excitatory vasomotor response, and heart-rate deceleration, it (Wagner and Rescorla 1972). However, the 0-second SC appears that these measures are completely insensitive to values for the euCS paradigm are markedly negative, negative contingency per se, a finding to which we which is another indication that the real inhibitory referred in the introduction. The conditions, then, under nature of this CS was not learned by the subjects. which negative contingency is or is not registered is a The failure of relational inhibitory learning to occur to matter for empirical investigation which is likely to the euCS in this experiment contrasts strongly with the increase our understanding of the processes that go on occurence of this learning in similar euCS paradigms in during various Pavlovian conditioning paradigms; both within- (Furedy and Schiffmann 1973) and between- continuous measurement of relational learning is a (Schiffmann and Furedy 1977) subject designs which procedure that is likely to contribute to that employed the continuous SC measure used here. In understanding. those experiments, by the last third of a 15-trial series, References the euCS was producing negative SC changes, and the 0- second SC values representing the contingency Badia, P., and Defran, R. H. Orienting responses and GSR perception of the background stimuli had become conditioning: A dilemma. Psychological Review, 1970, 77, positive. The most obvious procedural difference 171-181. Champion, R. A., and Jones, J. E. Drive level (D) and extinction between those studies and the present one is that there in classical aversive conditioning. The Journal of General the sequence of CS and US presentations was unsys- Psychology, 1962,67, 61-67. tematic, whereas here it was a systematically alternating Dawson, M. E. Can classical conditioning occur without one. The systematic sequence was adopted in the contingency learning? A review and evaluation of the present experiment in order to equate the euCS to the evidence. Psychophysiology, 1973, 10, 82-6. bCS condition as closely as possible; the latter condition Furedy, J. J. Reinforcement through UCS offset in classical is by definition a systematically alternating one (i.e., aversive conditioning. Australian Journal of Psychology, US-CS, US-CS, US-CS, etc.). If time is ignored and 1965, 17, 205-212. the subject does not register the presence of safety Furedy, J. J. Explicitly-unpaired and truly-random CS- periods following both (objectively inhibitory) CSs, then controls in human classical differential autonomic conditioning. Psychophysiology, 1971, 8, 497-503. it is possible to conceive of both CSs as excitatory in the Furedy, J. J. Some limits on the cognitive control of conditioned sense that following each CS there will always be a US. autonomic behavior. Psychophysiology, 1973,10, 108-111. Of course, the instructions to indicate "moment to mo- Furedy, J. J. "Negative results": Abolish the name but honor ment" changes in US expectancy are contrary to the the same. In J. P. Sutcliffe (Ed.), Conceptual Analysis and ignoring of the time variable. The results of the bCS Method in Psychology. Studies in Honor of W. M. O'Neil. paradigm here and of the euCS paradigms in other studies Sydney: University of Sydney Press, 1978. (Furedy and Schiffmann 1973, Schiffmann and Furedy Furedy, J. J., and Arabian, J. M. A Pavlovian 1977) indicate that, under some conditions, the psychophysiological perspective on the OR: The facts of the relational learning process is sensitive to such time vari- matter. In H. D. Kimmel, E. H. van 01st, and J. F. ables. In Experiment III, however, subjects were not Orlebeke (Eds.), The Orienting Reflex in Humans. New able to learn the negative contingency property of the Jersey: Lawrence Erlebaum Associated, 1979. Furedy, J. J., and Poulos, C. X. Short-interval classical SCR euCS. conditioning and the stimulus-sequence-change-elicited OR: This result, we would suggest, should not be seen as The case of the empirical red herring. Psychophysiology, some sort of failure of the SC measure but rather as 1977,14, 351-359. telling us something more about the process of Furedy, J. J., and Schiffmann, K. Test of the propriety of the relational learning in Pavlovian conditioning. Even from traditional discriminative control procedures in Pavlovian the previous studies with the continuous SC measure, it electrodermal and plethysmographic conditioning. Journal of is apparent that negative contingencies are more difficult Experimental Psychology, 1971,91, 161-164. to learn than positive ones: the excitatory nature of a Furedy, J. J., and Schiffmann, K. Concurrent measurement of CS paired with the US (CS+) is registered earlier and autonomic and cognitive processes in a test of the more markedly than the inhibitory nature of an euCS traditional discriminative control procedure for Pavlovian electrodermal conditioning. Journal of Experimental (Furedy and Schiffmann 1973, Figure 1, Schiffmann and Psychology, 1973, 100 210-217. Furedy 1977, Figure 1). Indeed, to return for a moment to response- RELATIONAL LEARNING 79

Furedy, J. J., and Schiffmann, K. Interrelationships between havior Research Methods & Instrumentation, 1975b, 7, human classical differential electrodermal conditioning, 524-526. orienting reaction, responsivity and awareness of stimulus Prokasy, W. F., Williams, W. C, Kumpfer, K. I., Lee, W. contingencies. Psychophysiology, 1974, 11, 58-67. Y., and Jensen, W. R. Differential SCR conditioning with Furedy, J. J., Poulos, C. X., and Schiffmann, K. Contingency two control baselines: Random signal and signal absent. theory and classical autonomic excitatory and inhibitory Psychophysiology, 1973, 10, 145-153. conditioning; Some problems of assessment and Prokasy, W. F. Personal communication, 1976. interpretation. Psychophysiology, 1975a, 12, 98-195. Rescorla, R. A. Pavlovian conditioning and its proper control Furedy, J. J., Poulos, C. X., and Schiffmann, K. Logical procedures. Psychological Review, 1967, 74, 71-80. problems with Prokasy's assessment of contingency Schiffmann, K., and Furedy, J. J. Failures of contingency and relations in classical skin conductance conditioning. cognitive factors to affect long-interval differential Behavior Research Methods & Instrumentation, Pavlovian autonomic conditioning. Journal of 1975b, 7, 521-523. Experimental Psychology, 1972,96, 215-218. Jones, J. E. Contiguity and reinforcement in relation to Schiffmann, K., and Furedy, J. J. The effect of CS-US CS-UCS intervals in classical aversive conditioning. contingency variation of GSR and on subjective CS-US Psychological Review, 1962,69, 176-186. relational awareness. Memory & Cognition. 1977,5, Kamin, L. J. Temporal and intensity characteristics of the 273-277. conditioned stimulus. In W. F. Prokasy (Ed.), Classical Segal, E. M., and Lachman, R. Complex behavior or higher Conditioning. New York: Appleton-Century Crofts, mental process: Is there a paradigm shift? American 1967. Psychologist, 1972, 27, 46-55. Kimble, G. A. Hilgard and Marquis' Conditioning and Szalai, J. P., and Furedy, J. J. Is the effective tilt US onset Learning (2nd edition). New York: Appleton-Century merely coy and elusive or should we welcome back Crofts, 1961. backward conditioning—Pavlov's prodigal son? Lockhart, R. A. Cognitive processes and the multiple Psychophysiology, 1978,75, 272 (Abstract). response phenomenon. Psychophysiology, 1973,10, 112- Trapold, M. A., Homzie, M., and Rutledge, E. Backward 118. conditioning and UCR latency. Journal of Experimental Prokasy, W. F. Random control procedures in classical Psychology, 1964,67, 387-391. skin conductance conditioning. Behavior Research Zimny, G. H., Stern, J. A., and Fjeld, S. P. Effects of CS Methods & Instrumentation, 1975a, 7, 516-520. and UCS relationships on electrodermal response and Prokasy, W. F. Random controls: A rejoinder. Be- heart rate. Journal of Experimental Psychology, 1966,72, 177-181.