THE ACQUISITION AND EXTINCTION OF INSTRUMENTAL AVOIDANCE AS A FUNCTION OF THE ESCAPE SITUATION Thesis Io:- fIn Dear“ OI pk. D. MICHIGAN STATE UNIVERSITY Robert K. Knapp 1960 (THESIS .‘llll’olrllalll - II I THE ACQUISITIJN AND EXTINCTION UH INSTRULENTAL AVOILAICE AS A FUNCTION UH THE ESCAPE SITUATIUU By Robert R. Knapp AN ABSTRACT Submitted to the School for Advanced Graduate Studies of fiichigan State University in partial fulfillment of the requirements for the degree of DOUToR OF PHlLOooPHY Department of Psychology 1960 792,76 19 Approved / t *g‘A/VZ? J I h ABSTRACT The present experiment was designed to test certain implications of elicitation theory in the analysis of acqui- sition and extinction of instrumental avoidance behavior. According to this VieWpoint an avoidance response is learned because it removes §.to a nonshock area where it learns to relax from an emotional state which was conditioned to shock box cues on early trials. The animal thus learns to approach the nonshock area upon coming in commerce with shock box cues. With successive avoidances of shock (or with omission of shock as during extinction trials) relaxational tendencies generalize and/or chain from the nonshock area back to the shock area according to the degree of similarity between the two areas. When relaxational responses come to predominate in the shock area (where emotional and escape responses pre— viously predominated) the avoidance response no longer occurs. Thus differences in both acquisition and extinction rates are predictable from this framework, since § must learn to relax in the nonshock area and, later, to relax in the shock area as well. In the present experiment it was hypothesized that learning would be facilitated and extinc— tion retarded when the shock and nonshock areas were dissimilar. ii A second set of hypotheses involved nonshock box confinement periods, namely, that variable durations of confinement would retard both acquisition and extinction by reducing Opportunities for relaxation in the nonshock area. Ninety-six male rats were trained to avoid a 1.5 ma. shock by jumping from a shock box With a grid floor to an elevated nonshock box. The shock and nonshock boxes were constructed either of wood or of clear plastic. Four combin- ations of boxes were employed: a wood shock box with a wood nonshock box; a wood shock box with a plastic nonshock box; a plastic shock box with a wood nonshock box; and a plastic shock box with a plastic nonshock box. An equal number of §§_(24) was tested with each of the four arrangements. The period of time s spent in the nonshock box following jumping responses was of fixed duration (90 sec.) for half the rats and of variable duration (5, 25, and ZAO sec., mean-9O sec.) for the other half. After two consecutive avoidances of shock, where the CS was a 5 sec. interval between gig entrance to the shock box and the onset of shock, extinction was begun. During the shock-free extinction trials nonshock box confinement was the same as during acquisition (fixed or variable) for half the gs, and followed the sched- ule opposite that of acquisition for the other gs, Extinction trials were terminated when s failed to jump from the shock box in 180 sec. on each of two consecutive trials. iii Analyses of variance performed on the number of trials to criterion disclosed, as predicted, that §§ trained 'with dissimilar boxes learned more rapidly and extinguished more slowly than animals trained with similar boxes. Although the fixed and variable confinement durations failed to pro- duce significant mean differences in criterion scores, the data suggested that the long-duration confinement (2&0 sec.) in the variable schedule facilitated both learning and extinction. These findings are consistent with elicitation theory in its analysis of instrumental avoidance learning, and tend not to support an interpretation of avoidance learning based upon reduction of an acquired drive such as anxiety. Suggestions for future research stress the need for inves- tigation of the role of the confinement and intertrial interval variables in acquisition and extinction of instru- mental avoidance. iv THE ACQUISITION AND EXTINCTION or INSTRUMENTAL AVoIDANCE AS A FUNCTION or THE ESCAPE SlTUATION By Robert R. Knapp A THESIS Submitted to the School for Advanced Graduate Studies of Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology I 1960 To Barb, Kenny, and Chris, and their grandparents vi ACI’xE‘IO‘.‘ILEDG I‘ll .EI'I TS The writer wishes to express his appreciation to the members of his committee, Drs. h. Ray Denny, Frank Restle, Abram L. Barch, and Charles hanley, for their assistance and guidance in the preparation of this report. It has been especially gratifying to work with Dr. E. Ray Denny, who has always willingly shared both his limited time and his enthusiasm. The author recognizes, further, the value of the constant Support and encouragement offered by his colleagues at the V. A. Hospital, Battle Creek, Michigan; and in this regard Messrs. James Morris and Jay Thomas deserve special thanks. vii TABLE OF CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . l. acquisition of avoidance . . . . . . . . . 2. Extinction of avoidance . . . . . . . . . 3. Elicitation and anxiety reduction theories and the present experiment . . . . 4. Purpose of the present experiment . . . . METHOD . . . . . . . . . . . . . . 1. Subjects . . . . . . . . . . . . 2. Apparatus . . . . . . . . . . . . . 3. Procedure . . . . . . . . . . . . . RESULTS . . . . . . . . . . 1. Acquisition . . . . . . . . 2. Extinction . . . . . . . . . . DISCUSSION . . SUMMARY . . . . REFERENCES . APPENDIX . viii TABLE 1. 10. ll. LIST OF TABLES Page Division into 16 groups (N=6) of the 96 Sg and the stimulus conditions prevailing during the acquisition and extinction phases of the experiment . . . . . . . . . . . . . . . . . . . 8 Summary of the analysis of variance for number of trials to the acquisition criterion . . . . . 25 The number of Ss reaching criterion on the Nth trial tabulateHIunder the appropriate apparatus and confinement conditions, related to the variable confinement schedule . . . . . . . . . 27 Summary of the analysis of variance for transformed extinction scores . . . . . . . . . 31 Summary of the analysis of covariance performed on acquisition and extinction scores, and within- and between-groups correlation coefficients . . . . . . . . . . . . . . . . . 33 The number of Ss reaching the extinction criterion within blocks 3? nine trials, tabulated under the appropriate apparatus and confinement . conditions . . . . . . . . . . . . . . . . . . . 35 The number of trials to acquisition for all §§ for all subgroups . . . . . . . . . . . . . . 51 The number of trials to extinction for all Sg for all subgroups (untransformed scores) . . . . 52 Mean and standard deviation for all groups for acquisition and extinction. . . . . . . . . . . .53 Summary of the analysis of variance for untrans— formed extinction scores . . . . . . . . . . . . Supplementary analysis of variance table for achiSj—tion O O O O O O O O O 0 O O O O O O O O 55 Supplementary analysis of variance table for / extinction O O O O O 0 O O O 0 O O O O O O 0 0 0 5:) ix F34 I—I O C 3 9'5 II". I...’ 0 LIST OF FIGURES Page The plastic shock box with grid floor, the plastic chimney with the wood side raised, and the wood nonshock box with 15 a Masonite floor . . . . . . . . . . . . . . The number of Ss reaching the acquisition criterion for all groups pooled (upper curve) and for the separate C and V groups (lower curves) with each acquisition trial. . 23 The cumulative number of Ss reaching the extinction criterion for EIl groups pooled (upper curve) and for the separate C and V groups (lower curves) . . . . . . . . . . . . 29 INTRODUCTION quuisition of avoidance. The ability of animals to acquire an avoidance response which temporally precedes the onset of aversive stimulation has long puzzled learning theo- rists. Explanations of avoidance learning have frequently included the postulating of some mediating process which can elicit the escape response in the absence of the unconditioned stimulus (US). Thus we have behavior mediators posited such as acquired drives (fear, anxiety) and the secondary rein- forcement of the original escape response by cues formerly associated with aversive-stimulus reduction. Mowrer (1940) suggested that one result of aversive stimulation is fear, and that fear functions as an acquired drive such that its reduction reinforces preceding responses. Thus a fear-arousing conditioned stimulus (CS) would evoke a response followed by fear reduction. Since this response temporally follows the CS and precedes the US, it is shock avoiding. Hull (1943) viewed escape learning as behavior motivated by drive reduc- tion. In Hull's system aversive stimulation produces a negative drive state such that responses leading to reduction or termination of the noxious stimulus are reinforced. Avoidance learning is the result of substitution of the CS for the US in evoking the original escape response. The CS-escape response bond is formed because, with contiguous association of the response and the trace of the (neutral) CS, and with reduction of aversive stimulation following the escape reSponse, the previously neutral CS acquires the power to evoke the escape response. In the absence of the US rein- forcement is secondary rather than primary, in that formerly neutral stimuli acquire reinforcing powers through their close association with aversive stimulus reduction. Thus, for Hull, an avoidance response is simply the original escape response evoked anticipatorily by the conditioned stimulus. Howrer (19A6), attempting to show that an avoidance response is more than an anticipatory escape response, trained rats to avoid shock where the avoidance response (e.g. jumping) was required to be different from the original escape response (e.g. running). Although the §§ so trained failed to learn as well as those for which escape and avoid- ance responses were of the same class (e.g. both running responses), the results were interpreted to suggest the Operation of an acquired drive of anxiety and its reduction in motivating new learning. According to howrer (1940) and to Killer (l9h8), S learns to become anxious in the presence of cues associated with aversive stimulation and makes responses which remove it from the anxiety—arousing situation. Removal of these cues results in anxiety reduction which reinforces the tendency for the escape response to occur. The anticipatory occurrence of the escape response (i.e., the transition of the escape response to an avoidance response) follows from the immediate arousal of anxiety by the CS. Recently howrer (1960) has revised his theory of avoidance learning, emphasizing that §_has to learn to be afraid upon presentation of the CS; and in addition §,has to learn what to do about the fear. In accounting for both active avoidance learning (where §,learns to avoid by making some response) and passive avoidance learning (S learns to avoid by failing to make a given response) Mowrer states that whenever a stimulus-signal (either response-produced or environmental) precedes marked drive increment, as the onset of aversive stimulation, fear becomes conditioned to that stimulus-signal. If the signal is response-produced, conditioned fear can produce response inhibition as in punishment. If the signal is an environmental stimulus and not response-produced, active avoidance learning follows from the conditioned fear. In either case, behavior which eliminates the signal or the stimulus constellation producing the fear will be rewarding and will reinforce the activity (or inactivity) involved. Reinforcement for Nowrer occurs according to the principle of "Type 1" secondary reinforcement, defined as the reward experienced when a danger signal terminates. howrer's remarks on the extinction of avoidance will be examined later. An analysis of instrumental avoidance behavior not based upon acquired drive and its reduction or upon secondary reinforcement, but upon the kinds of responses S makes in the "danger" and "safe” stimulus situations is offered by 4 elicitation theory, as formulated by Denny and Adelman (1955, 1956). This viewpoint posits that both emotional- and escape- type responses are directly elicited by aversive stimulation such as electric shock. The emotional reSponses such as freezing, urinating, and biting, thus become conditioned to cues associated with shock as, e.g., in a shuttlebox with a low central barrier and with discriminable ends. 0n the other hand the escape-type responses, e.g. running and jump- ing, are incompatible with each other (cannot occur simultaneously), have relatively equal initial strength, and initially may be inappropriate in removing S to the nonshock area. This situation prevails at the start of training because only one sequence of responses, running to the barrier and jumping to the nonshock area, will terminate the shock. Variation from this sequence, e.g. §_jumps before running to the barrier, can result in receipt of shock unless the correct sequence is performed within the CS-US interval. once the correct sequence is performed and a jump leads S to the nonshock area, absence of shock results in the occurrence of 1T relaxational-approach reSponses. hesc responses are con— ditioned to cues of the nonshocx side of the apparatus and 1The principle of secondary elicitation (Denny and Adelman, 1956) holds that omission of a consistent elicitor (in this case shock) from an established behavior sequence elicits a characteristic class of responses (in this case relaxational-approach responses) and mediates the acquisition of a new response tendency (in this case the tendency to approach cues of the nonshock side of the apparatus). on subsequent trials are conditioned to proprioceptive cues attending the jump response, to cues dominant just prior to a jump, and finally, to visual stimuli of the barrier region. From this point the entire behavior sequence follows a discrimination learning paradigm, where inappropriate and out-of-sequence escape responses in the shock side decrease in probability, and relaxational-approach responses to the nonshock side increase in probability and decrease in latency to the point where they become shock-antedating. Throughout the acquisition process shoCk side cues continue to elic1t emotional responses in 3. It is presumably from this emotional state that S relaxes in the nonshock side. In short, elici- tation theory takes the position that in instrumental avoidance ,S learns to approach nonshock area cues as well as to avoid shock area cues. ”his position is taken in the present paper as well. It can be seen then, that elicitation theory explains avoidance learning in terms of the relaxational-approach responses conditioned to the nonshock area rather than in terms of any reduction of anxiety that might occur there, or instead of in terms of the secondary reinforcement of the jumping response by cues associated with the termination of shock. At least, the explanation emphasizes what the animal gggg in the nonshock stimulus situation, rather than what stimuli occur there. However, the elicitation analysis of avoidance learning does recognize the importance of nonshock area stimuli, as will be discussed later. Extinction of avoidance. The acquired drive reduction viewpoint has occasionally generated.research in which the instrumental avoidance response became highly resistant to extinction (Solomon, Kamin, and Wynne, 1953) or in which the acquired drive (emotionality) through its reduction has been employed to mediate the acquisition of novel responses (Miller, 1948). It is perhaps for this reason that anxiety reduction theorists have had difficulty in explaining the extinction of instrumental avoidance. For example, Solomon and wynne (1954) have suggested the principle of "anxiety conservation" to explain the durability of avoidance behaviors. This notion holds that the typically short latency of learned avoidance at asymptote serves to minimize §L§ anxiety because the animal responds to the CS so rapidly that anxiety is only moderately aroused. Linimization of anxiety, however, results in increasing latency of response until anxiety is again fully aroused by the CS plus a longer exposure to the anxiety- arousing situation. The latency of response again becomes minimal, and the entire cycle is repeated. In this manner anxiety is presumed to be "conserved" at the limit of learn- ing. More recently howrer (1960) has defined extinction as the deveIOpment of a resting response incompatible with the avoidance response in a situation involving unrewarded responding or, in other words, where the emotional response (fear) is repeatedly unconfirmed. The fact that extinction of avoidance does occur, and fairly rapidly when shock and nonshock areas are similar as in Denny, (cons, and Mason (1959), does not support the principle of anxiety conservation. Extinction of avoidance response can be explained in elicitation theory terms, how- ever. Extinction for Denny and Adelman (1956) does not involve the unhooking of responses, weakening of the instrumental response tendency, or the building up of an inhibitory drive state. Rather, extinction is held to involve simply additional learning of new responses in the presence of shock box cues. These responses are relaxational—approach responses, and as such are incompatible with the emotional responses previously elicited in the shock box. According to the principle of secondary elicitation (outlined above) the omission of shock results in the eliciting of a class of responses different from those directly elicited by shock. That is, relaxational- approach responses are elicited in the shock box in the absence of shock rather than emotional and escape responses. This situation prevails when the level of shock has been less than traumatizing. Thus it would seem that extinction (the acquisition of relaxation-approach in the shock area where previously only emotional and escape responses were elicited) begins the first time S avoids the shock. However, the elicit- ing of relaxation-approach in the shock box following omission of shock is inhibited by the consistent eliciting of emotional responses by shock box cues. As shock—free extinction trials progress and as relaxational-approach responses continue to be elicited in the nonshock area the relaxational pattern eventually generalizes back to the shock area via a chaining process. Occasional emotional upsurges (due to the presence of emotional response—eliciting cues) and resulting jumps to the nonshock area notwithstanding, relaxational-approach tendencies finally gain greater relative strength than escape responses in the shock area. The escape response, jumping, no longer occurs. Elicitation and anxiety reduction theories and the present experiment. Superficially the anxiety reduction point of view and elicitation theory do not appear markedly divergent in their analyses of instrumental avoidance learn- ing. Both frameworks posit that emotional and escape responses accompany receipt of strong aversive stimulation, and that the emotional responses become conditioned to the CS or other signal-stimuli. However, the role that each of these theo- retical treatments seems to assign the nonshock stimulus situations in mediating the acquisition of avoidance points up a major difference between the theories. Another major difference is apparent when one considers the nature of predictions implied in each viewpoint as to the variables affecting the speed of acquisition and extinction of avoid— ance. An explanation of avoidance learning based upon reduction of an acquired drive of anxiety implies that any nonshock area that S enters would serve to reduce the anxiety aroused in the shock area. That is, the appearance of the nonshock area, whether similar to the shock area or dissimi- lar, should not alter the effectiveness of the nonshock area in reducing anxiety. A further implication of the anxiety reduction viewpoint is that reduction of anxiety by cues of a nonshock area is immediate and complete. That is, as soon as,§ enters a nonshock area its anxiety is reduced to recur only when the CS is again presented. Elicitation theory, in emphasizing the role of relaxational-approach responses which come to occur in a nonshock area does not imply that S relaxes as soon as it enters the nonShock area. Rather, the tendency for relaxation- approach to predominate in the nonshock area must be acquired during either a long exposure of,S to the nonshock area or in the course of several short visits to that area. Since the acquisition of relaxational-approach responses proceeds, presumably, as does the learning of other instrumental acts, it is possible from this framework to predict differences in speed of acquisition as well as in speed of extinction of avoidance. That is, the extent to which shock and nonshock areas are discriminable may determine speed of acquisition. If the shock and nonshock areas are closely similar the proprioceptive stimuli present before and after the initial escape responses are the only cues available for the discrim- ination. If the two areas are dissimilar, however, many cues facilitate the discrim nation. During extinction relaxation-approach which occurs in the nonshock area can, if this area is similar to the shock 10 area, generalize to the shock area and thereby facilitate extinction of avoidance; To the extent the two areas are dissimilar, generalization is minimized and relaxation-approach must chain back to the shock area; thus extinction may be retarded. Results supporting this interpretation were obtained by Denny 33 a1 (1959): rats were permitted to jump from a wood shock box to any of four identical elevated nonshock boxes, or to an elevated table top. For Sg jumping to the boxes, a stimulus complex similar to the shock box, extinction took place more rapidly than for those jumping to the table top. Another variable which may affect the acquisition of relaxation-approach tendencies (and thus govern the speed Of both acquisition and extinction of avoidance) is the period Of time §_is confined in the nonshock area prior to being placed in the shock area for the next trial. Simply stated, if §.is to relax, it must be given time in which to do so. Presumably, consistent periods of confinement would provide greater Opportunities for relaxing than inconsistent periods, and long periods would be more conducive to relaxation than short periods. Dinsmoor and Hughes (1956) have reported results relevant to the latter point: increasing the shock- free intertrial intervals from five to to sec. improved the acquisition of a shock-terminating bar pressing response (where the response being learned is not one directly elicited by shock). 0n the basis of the present interpretation it would appear that the greater the opportunities for relaxing 11 during the intertrial interval (or in a nonshock area), the more rapidly will the approach response be acquired. Purpose of the present experiment. The purpose of the present experiment, in general, was to test the elicitation analysis of instrumental avoidance. Specifically, certain hypotheses pertaining to the speed of acquisition and extinc- tion of avoidance behavior were tested; these hypotheses stem from elicitation postulates but would not seem to follow from an anxiety reduction point of view. The study to be reported investigated the effect of two variables on the acquisition and extinction of an avoid- ance response in a two-chambered avoidance-learning apparatus: (1) similarity vs. dissimilarity of shock and nonshock boxes and (2) consistent vs. inconsistent durations of confinement in a nonshock box following a jump response from a shock box. Rats were shocked in a box of one type of construction and were permitted to jump to an elevated nonshock box which was as similar to the shock box as possible for half the §§ and quite different for the other half. Periods of confinement in the nonshock box following jump responses were constant (90 sec.) for half the rats and variable (5, 25, and 240 sec.) for the other half. The experimental design counterbalanced the nonshock box confinenent schedules and the shock box- nonshoek box combinations. The latter provision controlled for possible visual preferences in the 35, a variable which was not controlled in Denny EE.§£ (1959). The number of jump 12 responses during the acquisition and extinction phases of the experiment constituted the dependent variable for the study. The following hypotheses were tested with regard to acquisition. If initial acquisition is sufficiently slow, then: (1) variable periods of confinement in the nonshock box following a jump response will retard acqui- sition of the tendency to approach the nonshock box (2) more crucially, learning to approach a nonshock box dissimilar to the shock box Will be more rapid than learning to approach a nonshock box which is similar. The hypotheses tested with regard to extinction were that: (l) primarily when shock and nonshock boxes are similar, a variable nonshock box confinement schedule will retard extinction (2) more crucially, independent of nonshock box confinement, extinction will be more rapid when the shock and nonshock boxes are alike than when they are unlike 13 I-‘ETI-iul) Subjects. The gsnwere 107 naive male rats of mixed strains from the colony maintained by the Department of Psychology of Michigan State University. At the start of the experiment the mean age of the §§_was 112 days, and the age range, 7A to 220 days. Of this number two were discarded following the occurrence of extreme emotionality rendering them impossible to handle; three were discarded after com- pletion of the experiment because of gig adopting of a more rigorous extinction criterion; and six were discarded from one group (N=l2) by means of a table of random numbers in order that all groups would contain an equal number of gs, Thus the final number of animals was 96, with each of the 16 groups containing six rats. Fifty-eight of the 96 §§_were albino rats, 18 were hooded, and 20 were of the grey-hooded strain. The typical group consisted of four albino, one hooded, and one grey- hooded. The g; were on ad lib feeding in the home cages throughout the course of the experiment, and the weights for all rats were recorded prior to running. The group mean weights were very similar, ranging from 297 to 357 gms. 14 Apparatus. The apparatus consisted of a shock box with an electrifiable grid floor, and a nonshock box situated above and to the side of the shock box (see Fig. 1). The non- shock box.was the compartment into which §.jumped in escaping and subsequently avoiding shock. The shock and nonshock boxes could be either opaque wood or transparent plastic. All boxes were approximately 12 in. by 12 in. at the floor and top (outside measurements) and were 11 in. high. The wood boxes were of 1/2 in. natural plywood diagonally striped on the inside surfaces with 3/4 in. black plastic tape spaced approximately 1 in. apart, except for the side under the nonshock box which remained plain. The plastic boxes were of 1/8 in. clear Plexiglas with three of their four sides bent inward 2 in. at the midline. The side of the plastic shock box directly under the nonshock box was flat in order to provide, as with the plain side in the wooden box, additional cues for direction of jumping. The plastic and wood boxes, then, were discriminable on the basis of shape, construction materials, and to the extent to which external stimuli were perceptible from within the boxes. The shock and nonshock boxes were so arranged that the top edge of the shock box was level with the floor of the nonshock box. Entrance to the wood shock box was by means of a A in. by 4 in. overhead-hinged door cut into the box at floor level on the side 90 degrees clockwise from the direc- tion of 5's jumps. The entrance door of the plastic shock box was a removable plastic panel covering an Opening in the Fig.1. The lastic shock box (1) with grid floor, the plastic chimney ()2) with the wood side raised, and the wood nonshock box (3) with a Itiasonite floor. In other apparatus arrangements the shock and nonshock boxes were positioned in the same manner. 16 side of the box the same size and in the same relative location as that of the wood shock box. Above either the wood or the plastic shock box was situated a "chimney" of 1/8 in. clear rlexiglas stock. This device, 12 in. square and 12 in. high, was actually an exten- sion of the shock box walls to a total height of 23 in. The side of the chimney enclosing the nonshock box was raised to permit entrance to the nonshock box; and could be replaced with a plywood panel whenever the nonshock box was plywood. The grids were of 1/8 in. steel welding rod spaced 5/8 in. apart, and so wired that conduction across any two adjacent rods completed a circuit with the shock source. Current directly to the grids was 1.5 ma., and was supplied by an Applegate hodel 228 Stimulator Operated through a self-returning hand switch. In order to provide a "footing" for §1§_jumps, the bars of the shock box were arranged parallel with the entrance to the nonshock box. The bars of the nonshock box grids, which were covered with 1/8 in. Lasonite panels whenever the shock and nonshock boxes were of different materials, were parallel to those of the shock box. Illumination of the apparatus was not Specifically controlled since the general light level of the laboratory room was adequate owing to overhead lights and to the presence of windows on three sides of the room. The timing of all intervals, as between §;§_introduction to the shock box and the onset of shock, the latencies of 5's jumps, and periods 17 of confinement in the nonshock box, was accomplished with a two-handed stepwatch; one hand of which could be stopped and restarted independently. Procedure. The 16 groups, as they represent the shock and nonshock box arrangements, and the nonshock box confine- ment conditions for the acquisition and extinction phases of the experiment, are all presented in Table 1. In summary, for half the §§_shock and nonshock boxes were similar, and for the other half they were different. Half the groups were trained in a plastic shock box, and half in a wooden box. The nonshock box confinement periods were constant for half the §§_and variable for the other half. Confinement during extinction was the same as during training for half the animals, while the remaining 92 found confinement conditions during extinction trials different from those prevailing on acquisition trials. Thus the present experiment represents a counterbalanced 2 by 2 by 2 by 2 experimental design. All g§_were placed in the shock box for 60 sec. prior to the start of training in order to determine whether initial jumping tendencies existed. None of the rats jumped to the nonshock box during this period. During acquisition the onset of shock occurred 5 see. after g was placed on the grid. If no jump to the nonshock box occurred within this 5 sec. period (CS-US interval) shock was turned on and continued for a maximum of 115 sec. Throughout this period the door to the nonshock box was open, permitting entrance by g, If the animal failed to jump 18 TABLE 1 DIVISIOH Ir-.TO 16 GROUPS (N:6) OF THE 96 Ss ahD TIIE bTIMULUS Cu-DILIOHQ PRMVAILIHG DURING THE ACE UlblTIUN AND LillbelUN PHths Ob THL EXPL HILLNT Shock Box Nonshock Box Nonshock Box Nonshock Box Construction Confinement Construction Confinement (Acquisition) (Extinction) Group -.,.. ___I __,._...-—--C l. .‘Fz'flloc VJ:::lI-~‘-N V 2.WWCV 'Constant (C) ”‘““*--P"""“"”'”'_C 3.WPCC ““‘“‘*‘-~v A.NPCV I) ‘ —#flflfl_flg,..—-C 5.WWVC \\\\\\\\\ ,,,7’-’”’”"'V\ \V 6.wwvv Wood (1 Variable \flf’“,,,lflaw-U 7.dPVC \P\ V 8.WPVV _ ’flflflfifflllflflec 9.PWCC w ///,,,/»””” “~““‘“*-~v 10.chv /////////’ ““‘-~11‘\\\‘ __flfl,,_,....-u ll.PPCC LI \ V 12.PPCV Plastic (P) ’#”',,,,,.—c 13.Pwvc M-N V2. \\\ll\\\\\\‘\ ’flflflfl’,,,.,—c 15,ppvc P ‘““““-~—-v 16.PPVV #*-*- "-"« 19 following 115 sec. of shock, shock was terminated for that trial and g was permitted to remain in the shock box an additional 120 sec. At the end of this period, 240 sec. in all, §_was lifted by g from the shock box directly into the nonshock box. ' The confinement period in the nonshock box when con- stant was 90 sec., and when variable was 5, 25, and 2A0 sec. (mean=90 sec.). The three values of variable confinement were presented in the following order for all gg'which were to have this schedule: 25-2AO-5-25-5-240-5-25-2h0-2h0—25-5-5-240—25-240-5-25 This schedule was repeated every 18 trialsuntil §_had reached either the acquisition or the extinction criterion, and was presented from the beginning when confinement was to be vari- able during extinction. The schedule satisfied the require- ments that (1) each value occurred once in each block of three trials, and (2) at no point in the variable schedule (or as a result of its repetition) was one value followed immediately by itself more than once. Training was terminated when g had made two consecutive jumps to the nonshock box within the 5 sec. period between introduction to the shock box and the onset of shock (CS-US interval). The extinction phase of the experiment, during which no shock was given, was then begun. Extinction trials (AC on day l, 60 on day 2, and 100 on day 3) were continued until the jumping latency exceeded 180 sec. on two consecutive trials. Iflg failed to jump from 20 the shock box in 190 sec. on an extinction trial it was lifted to the nonshock box for confinement and, after the appropriate period of time, was returned to the shock box. Iflg again remained in the shock box 180 see. it was lifted to the nonshock box for confinement and, following removal from the latter box, returned to the home cage. All §§_were tested for Spontaneous recovery approximately 24 hrs. after the last extinction trial, and carried to the same criterion. The possible correlating of seasonal factors (weather, temperature, etc.) and apparatus combinations was minimized in the following manner: on a given day a block of six or eight §§ was spread across the CC, CV, VC, and VV conditions under the same given arrangement of shock and nonshock boxes. After the §§_were run the apparatus combination was changed, and another sample of animals was run in like manner. Thus one—fourth to one-third of the §§ to be run with an apparatus combination was run at any one time. The same apparatus arrancement occurred approximately one month later. All §§ L: were run during daylight hours. In order to define more clearly what is meant here by a nonshock box confinement period, i.e., to control for inter- trial interval, 19 control §§_were run in the same apparatus by another E where intertrial interval was a fixed 90 see. but the amount of time 9 was confined in the nonshock box was variable.2 2The writer wishes to thank 1r. Neal Finley for running the control gg, 21 Ten §§.of group PNVV and nine §§_of NWVV were confined 5, 25, and 90 sec. in the nonshock box on both acquisition and extinction trials. The remainder of the 90 sec. inter- trial interval (ITI) when confinement was 5 or 25 see. was spent by the pg on a wooden stool several feet from the apparatus. No significant mean differences in either acquisition or extinction trials to criterion occurred when these groups were compared with groups PWVV and WWVV from the present study. 22 RESULTS Acquisition. The cumulative number of Sg'reaching the acquisition criterion (two consecutive jumps to the nonshock box with latencies of less than five see.) with each acqui- sition trial for all groups pooled is presented in the upper curve of Fig. 2. The two lower curves of Fig. 2 represent the separate "C" and "V" distributions. Considering the upper curve, it can be seen that half the animals learned by three trials, and that the curve quickly levels off after five trials. A holmogorov-Smirnov test was employed to determine whether the separate "0" and "V" distributions are from the same population. By trial 2, 17 "C" 22 and 8 "V" Sg had reached criterion. This difference, 9, yielded D=.l9. The value of D required for significance at the .05 level is D=.27. Thus the "C" and "V" distributions would appear to be from the same population. The number of trials to criterion ranged from l—lh; this range is more than three times that reported by Denny §E_§l (1959). The mean trials to criterion for all groups, n.12, is nearly twice that reported by those eXperimenters (2.2). Thus learning was slower in the present study, where the §§_were permitted to jump in only one direction rather than four, and the nonshock box confinement period was, on the average, shorter in duration than was the case in Denny et a1. 23 10° 1 C+V(——-) , N=% 90 4 so - 70 ~. 60 -( 50 — he 4 30- 20-1 CUMULATIVE FREQENCY 104 '6 i 7211;; 6 7 S910Ef213f1+ AQUISITlON TRIALS Fig. 2. The number of fig reaching the acquisition criterion (the two criterion responses are excluded) for all groups pooled (upper curve) and for the separate C and V groups (lower curves) with each acquisition trial. 21+ Bartlett's test for homogeneity of variances, performed on the acquisition data, indicated that the variances were homogeneous (x2=1a..97, d.f.=l5, P<.Ao). This finding, in conjunction with the experimental design, made it appropriate to perform an analysis of variance on the acquisition data. Since the confinement schedule for extinction was not a variable during acquisition, groups CC and CV and groups VV and V0 were pooled for each of the four apparatus conditions. Thus the analysis (summarized in Table 2) involved eight groups of 12 i3 each. The analysis yielded only one significant main effect, the shock box--nonshock box similarity——dissimi- larity variable (F-lO.6l, d.f.=l, 87, P<:.Ol). The construc- tion materials (wood or plastic to control for possible visual preferences) and the nonshock box confinement schedules (constant 90 sec. or variable, mean=90.sec.) did not produce significant mean differences. Nor did interactions of either first or second order approach significance. The significant shock box—-nonshock box similarity—- dissimilarity variable refers to the fact that groups with the shock and nonshock boxes dissimilar (mean=3.36) learned faster than the groups with similar boxes (mean=4.88). Thus, the hypothesis that learning Would be faster when shock and nonshock boxes are different than when they are similar is supported. Although the constant and variable nonshock box confinement schedules failed to have a significant effect in acquisition and thus the hypothesis that variable confinement TABLE 2 SUMMARY OF ANALYSIb 0F VARIANCE FDR NULBEd OF TRIALS T0 ACquSlTION CRITERION 25 Source of Variation sum of Squares d.f. Mean Square E A. Similarity-- 55.51 1 10.61* dissimilarity of boxes (nonshock) B. Construction 0.01 l <:l.OO material (shock box) (wood vs. plastic) C. Confinement 6.51 1 1.2h schedule (acquisition) Interactions: A x B 1.76 l <:l.00 .. x C 0.26 1 <1.00 B x C 11.34 1 2.16 1 x B x C 1.77 l <:l.00 within groups h60.58 88 5.23 Total 537.7h 95 *Significant beyond the .01 level 26 would retard learning is not confirmed, a more detailed analy- sis of the data suggests that length of the confinement period rather than variability of confinement is an important variable in acquisition of avoidance response. Tablel3 presents the number of trials that each animal took before reaching criterion. Opposite each criterion trial (the trial on which g made the first of two consecutive avoidance responses) is presented the number of gs requiring this many trials to reach criterion. Animals confined in the nonshock box for constant or variable periods are tabulated in the various "C" and "V" columns. These columns contain the pooled data of WW and PP apparatus combinations under the heading "Boxes similar", and the pooled data of WP and PW apparatus combin- ations under the heading "Boxes dissimilar". The "All C" and "All V" columns consist of the sums of "C" and "V”, respec- tively, for similar and dissimilar boxes. The final column presents the schedule for variable confinement. Trial 1 for "V" §§ concluded with a 25 sec. confinement; trial 2, with a 2hO sec. confinement; etc. Trials 2 and 6, both concluding with a 2h0 sec. confinement, are underlined to draw the reader's attention to the long confinement period which, the data suggest, may have a Special effect. Inspection of the "All C" and "All V" columns reveals that no §,made an avoidance response on the first acquisition trial, and that only one §f (a "C" animal) avoided shock on the second trial. On the third trial_(which followed a 240 sec. confinement for "V"' §§) 16 "C" gs made the first criterion response, and eight TABLE 3 THE NULBLR OF §§ REACHING CdITLHION 0? THE ch TRIAL TABULATED UNDmd THE AFrRUhRIATE APPARATUS AND CUNFINE— LENT CONDITIONS, RELATED TO THE VARIABLE CUNEIWBLBHT SCHEDULE N Trials to Boxes Boxes Variable Criterion* Similar Dissimilar Schedule C V C V All C All V l O 0 O O O 0 25 sec. 2 O O 1 C l O 2h0 " 3 5 2 ll 6 16 8 5 " h 6 4 4 9 10 13 25 ” 5 5 h 4 3 9 7 5 " 6 3 4 3 3 6 7 ZAO " 7 l 5 o l 1 6 5 n 8 O 2 l 0 1 2 25 h 9 l l 0 2 1 3 21,0 " 10 l 2 O O 1 2 2hu " ll 0 O O O O O 25 " 12 O O O O O O 5 " 13 l O O O l O 5 '" 1h 0 O O O O O 240 " 15 l O O O l O 25 " Sum 24 2h 24 24 A8 48 * The trial on which § made the first of two consecutive avoidance responses. 28 "V" §§ first avoided shock. On the fourth, fifth, and sixth trials 25 "C" §§ and 27 "V" §§ reached criterion although, considering cumulative frequencies, the "V" animals were deficient in learning after six trials. However, on the seventh trial, which for "V" SE followed another long con- finement period, six "V" gs made the first criterion response whereas only one "C" §_met the criterion on that trial. On trials 8, 9, and 10 (the latter two involving long confinements for "V" SE) the remainder of the "V" Sg met the criterion. Thus all "V" as learned within ten trials; i.e., after receiving three 240 sec. confinements. Two "0" animals required 13 and 15 trials to make the first criterion response, which attenuated any mean difference in learning brought about by an initial faster learning in "C" fig. To the extent these data can be considered reliable they suggest that the short intervals of the variable nonshock box confinement schedule retarded acquisition, but that this deficit was quickly overcome when a long confinement period was given. This analysis, as well as a similar one for the extinction data, is presented more to emphasize the need for research on the confinement variable than to present anything like conclusive results. Extinction. In Fig. 3 the uppermost curve presents cumulatively the number of Si reaching the extinction criter- ion (no jumps to the nonshock box in 180 sec. on two consecutive trials) on each extinction trial for all groups pooled. The two lower curves in Fig. 3 cumulatively present the same data 29 x" V= 30 x 9}D: 11/48 C37 23 C(-~-) AND V(--—)) N=48 1 xi— -wnm-me—w1‘.ru fi---- ' -'--.- on 1 O 5 10 15 20 25 30 35 #0 and lOO‘L 90- j»- (_) 80- Z LlJ [:3 70~ 0 us (Z 60-4 UL llJ 5o. 2 F- < 40‘) ...J :3 E 30~ :3 LJ 20d 10« 01 Fig. extinction for all gro C and V gro above EXTINCTION TRlALS 3. The cumulative number of Ss reaching the criterion (the two criterion trials are excluded) ups pooled (upper curve) and for the separate ups (lower curves). 30 for "C" and "V" separately.3 Because the range of criterion trials (the two jump-free trials are excluded) is large, 1-91, the data is plotted for every fifth trial from O-AO. It can be seen that approximately half the leextinguished by 20 trials, and that the separate "C" and "V" curves intersect in a crossover effect, and run essentially parallel after 15 trials. A Kolmogorov-Smirnov test was employed to determine whether the "C" and "V" distributions are from.the same pop- ulation. By trial 20, 30 "V" and 19 "C" §§.had reached criterion. This difference, 11, yielded D=.23. The value of D required for significance at the .05 level is D=.27. Thus the "C" and "V" distributions would appear to be from the same pepulation. When a Bartlett's test was performed on trials to criterion for all §§ for all subgroups (Table 8, Appendix) it indicated that the variances were heterogeneous (X2-29.8O, d.f.=15, P