P~ychological report^, 1976, 38, 895-911. @ Psychological Reports 1976
EFFECTS OF EXTINCTION PERSISTING THROUGH RECONDITIONING' J O H N L. FLANAGAN A N D DOUGLAS ANGERS
Universiry o f Missouri-Columbia Sr~mmary.-Changes in extinction responses between repeated extinctions were studied under conditions producing large decreases between the first two extinctions (continuous reinforcement, trial procedure). In a base group with 30 conditioning and 350 extinction trials, responses o n the first 50 extinction trials were 49% less on the second extinction than o n the first. Decreasing extinction trials to 50 eliminated the decrease between extinctions. Increasing conditioning trials to 270, or food during extinction (not correlated with pressing) increased responses on each extinction, but without eliminating significant decreases between extinctions. The size of the decrease between extinctions, its dependence o n extinction trials, and independence of both conditioning trials and feeding, indicate the decrease is not due to discrimination, or motivational changes from food withdrawal. Instead the results indicate some residual effect (other than stimulus control by stimuli differing between conditioning and extinction) persists through reconditioning, where it is not visible, and produces a faster second extinction.
Many studies of repeated extinctions have obtained large response decreases between the first and second extinctions. Decreases ranging from 24% to 67% have been found by Bersh, Schoenfeld, and Notterman ( 1950) ; Bullock ( 1960) ; Bullock and Smith (1953) ; Davenport ( 1969) ; Jenkins ( 1961) ; and Perkins and Cacioppo (1950). However, the variables controlling this decrease and its full implications have not been examined. This large change i n extinction responses after only one extinction would seem to make difficult a discrimination explanation of the phenomenon. The discrimination hypothesis proposes that some stimulus cues which are present during conditioning (S+) come to control high responding, while other stimulus cues present during extinction (S-) come to control low responding (Perkins & Cacioppo, 1950; Bullock & Smith, 1953; Wickens & Miles, 1954). However, the usual measure of a discrimination, differential response to two different stimuli, is confounded with other large effects in repeated extinctions. 'This paper is based on a dissertation submitted by the first author in partial fulfillment of the requirements for the Ph.D. degree at the University of Missouri-Columbia. This research was supported by Grant No. GB-27687 to the second author from the National Science Foundation. The authors wish to thank Robert Ayers, Richard Bumb, Thomas Burne, Ray Coffman, Susan Holt, Bruce Horowitz, Ronald Huiatt, Randal Jenkins, Daniel Klaus, Richard Lourie, Patricia Maher, James Nelson, Mark Owens, John Roberds, Mark Schweighoefer, Wendell Stone, and Samuel Zack for their help i n carrying out the experiments. The authors thank Kathleen Anger and John Mueller for helpful suggestions on early versions of this manuscript. 'Requests for reprints should be sent to Douglas Anger, Psychology Department, University of Missouri, Columbia, Missouri 65201.
896
J. L. FLANAGAN & D. ANGER
The high responding during conditioning and the low responding by the end of the first extinction do not necessarily indicate stimulus control or a discrimination, since they may just be due to the presence of reinforcements in one case and their absence in the other. Consequently, other evidence of a discrimination, such as peak shift or negative stimulus control, are needed to demonstrate a discrimination in the repeated extinction situation. Honig, Thomas, and Guttman (1959) did not obtain peak shift or contrast after massed conditioning and extinction, and Weisman and Palmer (1969) failed to find an incremental gradient around a stimulus paired with massed extinction. Admittedly the results are meager and negative, but at present they indicate that one alternation of massed conditioning ( S + ) and extinction (S-) is insufficient to produce a discrimination between exteroceptive stimuli. The objectives of the present study were to determine how the decrease between the first two extinctions was influenced by several variables, chiefly the number of conditioning and extinction trials, since those influences may provide additional evidence concerning the source of the decline. Rats were used in a discrete-trial procedure that permitted only one bar press during each S+ (stimulus during which responses were reinforced) by terminating S+ after the first press. The bar was withdrawn between trials. This procedure was used because it produces a high degree of stimulus control of the response, and seems to minimize influences from the preceding responding (differential reinforcement of different IRTs) over which the experimenter has limited control (Anger, 1956; Ferster & Skinner, 1957; Morse, 1966). Those conditions along with the continuous reinforcement (crf) employed are similar to the conditions of Perkins and Cacioppo (1950) and Jenkins (1961) who reported large decreases between their first two extinctions. After finding a large decrease with 30 crf conditioning trials and 350 extinction trials, changes were made from that base condition in: ( 1 ) number of reinforced trials during conditioning, ( 2 ) number of extinction trials, ( 3 ) feeding during extinction. Concerning the feeding variable, Rescorla and Skucy ( 1969) have shown that low feeding is confounded with low reinforcement in most extinctions. Consequently, in one group of the present study the low-feeding of extinction was reduced by administering food pellets between trials to determine whether the response decline between extinctions persisted when feeding continued during extinction. Those pellets will be called gifts, since that brief term implies little contingency, though not a complete absence of any contingency ( A negative contingency is present, because the gifts never follow closer than 30 sec. after bar-pressing).
METHOD Subjects The subjects were 70 male Sprague-Dawley retired-breeder rats from Car-
EXTINCTION PERSISTING THROUGH RECONDITIONING
897
worth Farms, about 15 mo. old at the experiment start. They were housed in individual cages and their feeding adjusted to maintain a body weight of 75% of their free-feeding weight, with only slight occasional feeding adjustments during the experiment. The animals were fed 3 hr. after the end of the experimental session. Water was available at all times except during the experimental session. Four Skinner boxes of the same design were used. The experimental chamber was 31.8 cm long, 20.3 cm wide and 21.6 cm high. The ceiling, the two long side walls, and the floor were 0.24-cm stainless-steel rods centered 1.3 cm apart; the front and back walls were sheet stainless steel. Eight sections of the rods that formed the floor (two rows of four sections) were supported separately, and were so hinged and counterweighted that when more than 43 gm. of the rat's weight was on a section, it moved down about 0.16 cm and operated a microswitch. These switches recorded the activity and location of the rat. The bar was made from thin sheet stainless-steel, and was 1.11 cm thick, and 5.24 cm wide, but tapered to 3.97 cm wide at the rounded front edge. The front edge could be pressed down 0.64 cm, and a microswitch operated near the middle of this excursion. A magnet and the weight of the bar were so adjusted that 15 gm. weight on the front edge moved the bar but after 0.16-cm movement, then 8 gm. weight completed the depression. The top of the bar was 10.2 cm above the floor and its center was 4.4 cm from rhe left wall. A motor moved the bar either out of reach of the rat or to a position where the bar projected 1.3 cm into the cage. A 5.1-un square hole to the right of the bar was covered by a very light stainless-steel door. The door was hinged at its upper edge, and rats easily pushed it open to reach a small food chamber into which all pellets (P. J. Noyes 0.045 gm.) were dropped. Door openings were recorded by a switch. The S+ was a 4000-8000-Hz band of white noise at 73 db above reference level produced by a speaker on the right side of the cage. Each apparatus was enclosed in a sound attenuating box, and two such units were enclosed in a sound attenuating booth. A blower introduced air into the boxes and provided a masking noise. Cages were illuminated through a diffuser in the box roof by a 60-wart bulb in series with 60 ohms. The temperature was 75' F in the colony room and booths. Electromechanical control circuits were in a separate room and were inaudible in the booth room. O n the first day of magazine training the food chamber contained two Noyes pellets when the subject was placed in the experimental cage. Afcer the rat had eaten those pellets, 40 pellets were delivered about 1 min. apart, and
898
J. L. FLANAGAN
&
D. ANGER
40 more pellets were given on the following day. After the first few pellets, they were not delivered while the rat's head was in the food chamber. If, by the end of the second day, the rat did not immediately go to the food chamber when the magazine operated, then a third day of magazine training was given. Only three subjects needed a third day. Conditioning occurred on the next day. Three additional magazine trials trial the white noise were given first, then 50 sec. after the third magazine stimulus (S+) came on, and the bar came into the cage. The bar stayed in and the S+ stayed on until a bar press occurred. When a press occurred, the S+ terminated immediately, and the bar was withdrawn 6 sec. later. The next S+ began 60 sec. after the last Sf onset when the response latency to the last Sf was less than 10 sec., but when the response latency to the last S+ exceeded 10 sec., then the next S+ began 50 sec. after the last response. These conditions lasted for 30 trials (30 bar introductions). Beginning on the 31st trial, if a response did not occur within 10 sec. of the introduction of the Sf and bar, then the S+ was terminated and the bar was withdrawn. If a response did occur within 10 sec., then the S+ went off immediately, and the bar retracted 2 sec. later. These conditions were in effect for the remainder of the first conditioning and during all extinction trials. The two conditionings were identical; the whole sequence just described was repeated on the second conditioning. On a reinforced trial one pellet was delivered after the first response of that trial. Only the first response during a trial was ever reinforced, and since the S+ terminated at the time of the first response, only one response during the Sf was possible. Since the bar stayed in for 2 sec. following the response, it was possible for the rat to make more than one response per trial. However, extra responses were never reinforced, and their frequency was quite low in nearly all rats. In the following results and discussion, all references to responses or responses per trial refer only to the first responses. If the response latency of a rat exceeded 700 sec. on any of the first five trials of the first conditioning then the rat was shaped to bar-press. Shaping was only done on the first five trials and only on trials where the latency exceeded 700 sec. About a third of the rats had shaping, usually only for one trial. Four rats that exceeded a 700-sec. latency between Trials 5 and 30 were dropped from the experiment. Two rats were discarded because they failed to respond on over five trials after the Sf was limited to a 10-sec, duration during the reinforcement phase on the first day of conditioning. The 70 rats were divided into five groups of 12 and two groups of 5 . The different groups will be referred to by a pair of numbers, the first designating the number of conditioning trials (crf), and the second designating the number of extinction trials. Thus 30-350 refers to the group with 30 conditioning trials on both conditionings and 350 extinction trials on the first extinction. The three groups with different conditioning trials were 30-350 (base group),
EXTINCTION PERSISTING THROUGH RECONDITIONING
899
90-350, and 270-350. The three groups with different extinction trials were 30-50, 30-170, and 30-350 (same base group). The two groups with five rats were the group with gifts (30-35OG) and a control group, 30-50R, that received the same conditionings and extinctions as 30-50 but were not tested for five days ("rest") in order to provide the same time between the two conditioning periods as occurred in the 30-350 group. Group 30-350G received the same conditioning and extinction treatment as Group 30-350. However, during extinction (but not conditioning) gifts (single pellets) were delivered while the bar was retracted. Delivery of these gifts was independent of the occurrence of bar presses on the preceding trial. Gifts were only delivered at one particular time in the cycle, the gift-time, which was 40 sec. after the last bar introduction (therefore all gifts occurred at least 30 sec. after the bar had been withdrawn). Twenty-four gifts were given per hour (per 60 trials) at variable intervals. The delivery of gifts was independent of all responses of the rat, except that delivery of a scheduled gift was delayed until the next gift-time if the food-chamber door was opened during the 15 sec. preceding the gift-time. Such a delay prevented the door-open time from increasing to high levels. Such increases have sometimes been observed without such a delay. For each group the last 30 trials of conditioning and the first 50 trials of extinction were given on the same day. On this day the experimental session was 80 trials long. On all other conditioning and extinction days the experimental session was 60 trials ( 1 hr.) long. A nonresponse during extinction was recorded as a 10-sec. latency for that trial. All transitions from the end of the first extinction to the beginning of the second conditioning occurred between days. After the second conditioning only 50 extinction trials were given.
RESULTS Since one group had only 50 extinction trials, to enable comparison of all groups, the measures employed throughout will be the number of responses and average latency during the first 50 extinction trials. Fig. 1 shows the average number of extinction responses for each group during the first 50 extinction trials of both extinctions. The solid line and circles labeled E l show the first extinction results; while the dashed line and crosses labeled E2 show the second extinction results. The left panel of this figure shows the effect of changing extinction trials while holding conditioning trials constant. The middle panel shows the effect of changing conditioning trials with extinction trials constant. The right panel shows the effect of adding gifts with conditioning and extinction trials constant. The base group (30-350, large circles and crosses) is repeated in each panel for comparison. Separate analyses of variance using both extinction responses and mean latencies as the dependent measures were carried out for each of these t h e e comparisons.
J. L. FLANAGAN & D. ANGER
I C30 E 50
G -
30 170
-
C - CONDITIONING
30 350
-
TRllLS
I
. 30 350
-
90 350
-
270 350
. 30
30
350 350 +
-
0. G I F T S
E~EXTINCTION TRIALS
FIG. 1. Number of trials with a response during the first 50 trials of the first extinction ( E l ) and second extinction (E2). Groups differing in number of extinction trials are shown in the left panel; groups differing in conditioning trials are in the middle panel; groups differing in gifts are in the right panel. The base group of 30 conditioning trials and 350 extinction trials is repeated in each panel for comparison, and is distinguished by a larger circle and cross.
Change in Number of Extinction Trials The effect of number of extinction trials was not significant ( F = 1.47, > .2). However, the difference between successive extinctions was significant ( F = 26.43, df = 1/33, j~ < .0001), as was the interaction of number of extinction trials with repetition of extinction (F = 5.04, d f = 2/33, p < .01). Between-group comparisons were made for the three groups at each level of extinction (horizontal comparisons in left panel of Fig. 1 ) . There were no significant differences between any of the groups in che number
df = 2/33, p
-
-
E- E X T I N C T I O N
IRIaLS
0 . GIFTS
F I G . 2. Mean latency/response in sec. during the first 50 trials of the first extinction ( E l ) and second extinction ( E 2 ) . Groups differing in number of extinction trials are shown in the left panel; groups differing in conditioning trials are in the middle panel; and groups differing in gifts are in the right panel. The base group of 30 conditioning trials and 350 extinction trials is repeated in each panel for comparison, and is distinguished by larger symbols. The ordinate scale is inverted to simplify a comparison with Fig. 1.
EXTINCTION PERSISTING THROUGH RECONDITIONING
90 1
of extinction responses in the first extinction. In the second extinction, Group 30-350 had fewer second extinction responses than Group 30-50 ( q = 4.54, d f = 66, p .01; Tukey's ratio, Kirk, 1968, p. 268). However, Group 30-170 did not differ significantly from either Group 30-50 or Group 30-350. Withingroup comparisons between the first and second extinctions (vertical comparisons in Fig. 1 ) showed that both Group 30-170 and Group 30-350 ( q = 5.02, df = 11, p < .01; q = 6.86, df = 11, p < .01 respectively) had fewer responses on the second extinction than on the first extinction but there was no difference between extinctions for Group 30-50 (q < 1.00). The average latency is shown in Fig. 2 where latency was used as the dependent measure. The ordinate scale is inverted to simplify comparison with Fig. 1. Not only are the two figures quite similar but the significant differences were essentially the same. The mean latency/response over five trial blocks during the initial and repeated conditioning-extinction series is shown in Fig. 3 for the three groups that differed in the number of extinction trials. Fig. 3A shows the group receiving 50 extinction trials (Group 30-50), Fig. 3B shows Group 30-170, and Fig. 3C shows Group 30-350. In Groups 30-170 and 30-350 shown in Figs. 3B and 3C it can be seen that the first and second extinction curves separated after the first 10 extinction trials.
EFFECTS OF EXTINCTION PERSISTING THROUGH RECONDITIONING' J O H N L. FLANAGAN A N D DOUGLAS ANGERS
Universiry o f Missouri-Columbia Sr~mmary.-Changes in extinction responses between repeated extinctions were studied under conditions producing large decreases between the first two extinctions (continuous reinforcement, trial procedure). In a base group with 30 conditioning and 350 extinction trials, responses o n the first 50 extinction trials were 49% less on the second extinction than o n the first. Decreasing extinction trials to 50 eliminated the decrease between extinctions. Increasing conditioning trials to 270, or food during extinction (not correlated with pressing) increased responses on each extinction, but without eliminating significant decreases between extinctions. The size of the decrease between extinctions, its dependence o n extinction trials, and independence of both conditioning trials and feeding, indicate the decrease is not due to discrimination, or motivational changes from food withdrawal. Instead the results indicate some residual effect (other than stimulus control by stimuli differing between conditioning and extinction) persists through reconditioning, where it is not visible, and produces a faster second extinction.
Many studies of repeated extinctions have obtained large response decreases between the first and second extinctions. Decreases ranging from 24% to 67% have been found by Bersh, Schoenfeld, and Notterman ( 1950) ; Bullock ( 1960) ; Bullock and Smith (1953) ; Davenport ( 1969) ; Jenkins ( 1961) ; and Perkins and Cacioppo (1950). However, the variables controlling this decrease and its full implications have not been examined. This large change i n extinction responses after only one extinction would seem to make difficult a discrimination explanation of the phenomenon. The discrimination hypothesis proposes that some stimulus cues which are present during conditioning (S+) come to control high responding, while other stimulus cues present during extinction (S-) come to control low responding (Perkins & Cacioppo, 1950; Bullock & Smith, 1953; Wickens & Miles, 1954). However, the usual measure of a discrimination, differential response to two different stimuli, is confounded with other large effects in repeated extinctions. 'This paper is based on a dissertation submitted by the first author in partial fulfillment of the requirements for the Ph.D. degree at the University of Missouri-Columbia. This research was supported by Grant No. GB-27687 to the second author from the National Science Foundation. The authors wish to thank Robert Ayers, Richard Bumb, Thomas Burne, Ray Coffman, Susan Holt, Bruce Horowitz, Ronald Huiatt, Randal Jenkins, Daniel Klaus, Richard Lourie, Patricia Maher, James Nelson, Mark Owens, John Roberds, Mark Schweighoefer, Wendell Stone, and Samuel Zack for their help i n carrying out the experiments. The authors thank Kathleen Anger and John Mueller for helpful suggestions on early versions of this manuscript. 'Requests for reprints should be sent to Douglas Anger, Psychology Department, University of Missouri, Columbia, Missouri 65201.
896
J. L. FLANAGAN & D. ANGER
The high responding during conditioning and the low responding by the end of the first extinction do not necessarily indicate stimulus control or a discrimination, since they may just be due to the presence of reinforcements in one case and their absence in the other. Consequently, other evidence of a discrimination, such as peak shift or negative stimulus control, are needed to demonstrate a discrimination in the repeated extinction situation. Honig, Thomas, and Guttman (1959) did not obtain peak shift or contrast after massed conditioning and extinction, and Weisman and Palmer (1969) failed to find an incremental gradient around a stimulus paired with massed extinction. Admittedly the results are meager and negative, but at present they indicate that one alternation of massed conditioning ( S + ) and extinction (S-) is insufficient to produce a discrimination between exteroceptive stimuli. The objectives of the present study were to determine how the decrease between the first two extinctions was influenced by several variables, chiefly the number of conditioning and extinction trials, since those influences may provide additional evidence concerning the source of the decline. Rats were used in a discrete-trial procedure that permitted only one bar press during each S+ (stimulus during which responses were reinforced) by terminating S+ after the first press. The bar was withdrawn between trials. This procedure was used because it produces a high degree of stimulus control of the response, and seems to minimize influences from the preceding responding (differential reinforcement of different IRTs) over which the experimenter has limited control (Anger, 1956; Ferster & Skinner, 1957; Morse, 1966). Those conditions along with the continuous reinforcement (crf) employed are similar to the conditions of Perkins and Cacioppo (1950) and Jenkins (1961) who reported large decreases between their first two extinctions. After finding a large decrease with 30 crf conditioning trials and 350 extinction trials, changes were made from that base condition in: ( 1 ) number of reinforced trials during conditioning, ( 2 ) number of extinction trials, ( 3 ) feeding during extinction. Concerning the feeding variable, Rescorla and Skucy ( 1969) have shown that low feeding is confounded with low reinforcement in most extinctions. Consequently, in one group of the present study the low-feeding of extinction was reduced by administering food pellets between trials to determine whether the response decline between extinctions persisted when feeding continued during extinction. Those pellets will be called gifts, since that brief term implies little contingency, though not a complete absence of any contingency ( A negative contingency is present, because the gifts never follow closer than 30 sec. after bar-pressing).
METHOD Subjects The subjects were 70 male Sprague-Dawley retired-breeder rats from Car-
EXTINCTION PERSISTING THROUGH RECONDITIONING
897
worth Farms, about 15 mo. old at the experiment start. They were housed in individual cages and their feeding adjusted to maintain a body weight of 75% of their free-feeding weight, with only slight occasional feeding adjustments during the experiment. The animals were fed 3 hr. after the end of the experimental session. Water was available at all times except during the experimental session. Four Skinner boxes of the same design were used. The experimental chamber was 31.8 cm long, 20.3 cm wide and 21.6 cm high. The ceiling, the two long side walls, and the floor were 0.24-cm stainless-steel rods centered 1.3 cm apart; the front and back walls were sheet stainless steel. Eight sections of the rods that formed the floor (two rows of four sections) were supported separately, and were so hinged and counterweighted that when more than 43 gm. of the rat's weight was on a section, it moved down about 0.16 cm and operated a microswitch. These switches recorded the activity and location of the rat. The bar was made from thin sheet stainless-steel, and was 1.11 cm thick, and 5.24 cm wide, but tapered to 3.97 cm wide at the rounded front edge. The front edge could be pressed down 0.64 cm, and a microswitch operated near the middle of this excursion. A magnet and the weight of the bar were so adjusted that 15 gm. weight on the front edge moved the bar but after 0.16-cm movement, then 8 gm. weight completed the depression. The top of the bar was 10.2 cm above the floor and its center was 4.4 cm from rhe left wall. A motor moved the bar either out of reach of the rat or to a position where the bar projected 1.3 cm into the cage. A 5.1-un square hole to the right of the bar was covered by a very light stainless-steel door. The door was hinged at its upper edge, and rats easily pushed it open to reach a small food chamber into which all pellets (P. J. Noyes 0.045 gm.) were dropped. Door openings were recorded by a switch. The S+ was a 4000-8000-Hz band of white noise at 73 db above reference level produced by a speaker on the right side of the cage. Each apparatus was enclosed in a sound attenuating box, and two such units were enclosed in a sound attenuating booth. A blower introduced air into the boxes and provided a masking noise. Cages were illuminated through a diffuser in the box roof by a 60-wart bulb in series with 60 ohms. The temperature was 75' F in the colony room and booths. Electromechanical control circuits were in a separate room and were inaudible in the booth room. O n the first day of magazine training the food chamber contained two Noyes pellets when the subject was placed in the experimental cage. Afcer the rat had eaten those pellets, 40 pellets were delivered about 1 min. apart, and
898
J. L. FLANAGAN
&
D. ANGER
40 more pellets were given on the following day. After the first few pellets, they were not delivered while the rat's head was in the food chamber. If, by the end of the second day, the rat did not immediately go to the food chamber when the magazine operated, then a third day of magazine training was given. Only three subjects needed a third day. Conditioning occurred on the next day. Three additional magazine trials trial the white noise were given first, then 50 sec. after the third magazine stimulus (S+) came on, and the bar came into the cage. The bar stayed in and the S+ stayed on until a bar press occurred. When a press occurred, the S+ terminated immediately, and the bar was withdrawn 6 sec. later. The next S+ began 60 sec. after the last Sf onset when the response latency to the last Sf was less than 10 sec., but when the response latency to the last S+ exceeded 10 sec., then the next S+ began 50 sec. after the last response. These conditions lasted for 30 trials (30 bar introductions). Beginning on the 31st trial, if a response did not occur within 10 sec. of the introduction of the Sf and bar, then the S+ was terminated and the bar was withdrawn. If a response did occur within 10 sec., then the S+ went off immediately, and the bar retracted 2 sec. later. These conditions were in effect for the remainder of the first conditioning and during all extinction trials. The two conditionings were identical; the whole sequence just described was repeated on the second conditioning. On a reinforced trial one pellet was delivered after the first response of that trial. Only the first response during a trial was ever reinforced, and since the S+ terminated at the time of the first response, only one response during the Sf was possible. Since the bar stayed in for 2 sec. following the response, it was possible for the rat to make more than one response per trial. However, extra responses were never reinforced, and their frequency was quite low in nearly all rats. In the following results and discussion, all references to responses or responses per trial refer only to the first responses. If the response latency of a rat exceeded 700 sec. on any of the first five trials of the first conditioning then the rat was shaped to bar-press. Shaping was only done on the first five trials and only on trials where the latency exceeded 700 sec. About a third of the rats had shaping, usually only for one trial. Four rats that exceeded a 700-sec. latency between Trials 5 and 30 were dropped from the experiment. Two rats were discarded because they failed to respond on over five trials after the Sf was limited to a 10-sec, duration during the reinforcement phase on the first day of conditioning. The 70 rats were divided into five groups of 12 and two groups of 5 . The different groups will be referred to by a pair of numbers, the first designating the number of conditioning trials (crf), and the second designating the number of extinction trials. Thus 30-350 refers to the group with 30 conditioning trials on both conditionings and 350 extinction trials on the first extinction. The three groups with different conditioning trials were 30-350 (base group),
EXTINCTION PERSISTING THROUGH RECONDITIONING
899
90-350, and 270-350. The three groups with different extinction trials were 30-50, 30-170, and 30-350 (same base group). The two groups with five rats were the group with gifts (30-35OG) and a control group, 30-50R, that received the same conditionings and extinctions as 30-50 but were not tested for five days ("rest") in order to provide the same time between the two conditioning periods as occurred in the 30-350 group. Group 30-350G received the same conditioning and extinction treatment as Group 30-350. However, during extinction (but not conditioning) gifts (single pellets) were delivered while the bar was retracted. Delivery of these gifts was independent of the occurrence of bar presses on the preceding trial. Gifts were only delivered at one particular time in the cycle, the gift-time, which was 40 sec. after the last bar introduction (therefore all gifts occurred at least 30 sec. after the bar had been withdrawn). Twenty-four gifts were given per hour (per 60 trials) at variable intervals. The delivery of gifts was independent of all responses of the rat, except that delivery of a scheduled gift was delayed until the next gift-time if the food-chamber door was opened during the 15 sec. preceding the gift-time. Such a delay prevented the door-open time from increasing to high levels. Such increases have sometimes been observed without such a delay. For each group the last 30 trials of conditioning and the first 50 trials of extinction were given on the same day. On this day the experimental session was 80 trials long. On all other conditioning and extinction days the experimental session was 60 trials ( 1 hr.) long. A nonresponse during extinction was recorded as a 10-sec. latency for that trial. All transitions from the end of the first extinction to the beginning of the second conditioning occurred between days. After the second conditioning only 50 extinction trials were given.
RESULTS Since one group had only 50 extinction trials, to enable comparison of all groups, the measures employed throughout will be the number of responses and average latency during the first 50 extinction trials. Fig. 1 shows the average number of extinction responses for each group during the first 50 extinction trials of both extinctions. The solid line and circles labeled E l show the first extinction results; while the dashed line and crosses labeled E2 show the second extinction results. The left panel of this figure shows the effect of changing extinction trials while holding conditioning trials constant. The middle panel shows the effect of changing conditioning trials with extinction trials constant. The right panel shows the effect of adding gifts with conditioning and extinction trials constant. The base group (30-350, large circles and crosses) is repeated in each panel for comparison. Separate analyses of variance using both extinction responses and mean latencies as the dependent measures were carried out for each of these t h e e comparisons.
J. L. FLANAGAN & D. ANGER
I C30 E 50
G -
30 170
-
C - CONDITIONING
30 350
-
TRllLS
I
. 30 350
-
90 350
-
270 350
. 30
30
350 350 +
-
0. G I F T S
E~EXTINCTION TRIALS
FIG. 1. Number of trials with a response during the first 50 trials of the first extinction ( E l ) and second extinction (E2). Groups differing in number of extinction trials are shown in the left panel; groups differing in conditioning trials are in the middle panel; groups differing in gifts are in the right panel. The base group of 30 conditioning trials and 350 extinction trials is repeated in each panel for comparison, and is distinguished by a larger circle and cross.
Change in Number of Extinction Trials The effect of number of extinction trials was not significant ( F = 1.47, > .2). However, the difference between successive extinctions was significant ( F = 26.43, df = 1/33, j~ < .0001), as was the interaction of number of extinction trials with repetition of extinction (F = 5.04, d f = 2/33, p < .01). Between-group comparisons were made for the three groups at each level of extinction (horizontal comparisons in left panel of Fig. 1 ) . There were no significant differences between any of the groups in che number
df = 2/33, p
-
-
E- E X T I N C T I O N
IRIaLS
0 . GIFTS
F I G . 2. Mean latency/response in sec. during the first 50 trials of the first extinction ( E l ) and second extinction ( E 2 ) . Groups differing in number of extinction trials are shown in the left panel; groups differing in conditioning trials are in the middle panel; and groups differing in gifts are in the right panel. The base group of 30 conditioning trials and 350 extinction trials is repeated in each panel for comparison, and is distinguished by larger symbols. The ordinate scale is inverted to simplify a comparison with Fig. 1.
EXTINCTION PERSISTING THROUGH RECONDITIONING
90 1
of extinction responses in the first extinction. In the second extinction, Group 30-350 had fewer second extinction responses than Group 30-50 ( q = 4.54, d f = 66, p .01; Tukey's ratio, Kirk, 1968, p. 268). However, Group 30-170 did not differ significantly from either Group 30-50 or Group 30-350. Withingroup comparisons between the first and second extinctions (vertical comparisons in Fig. 1 ) showed that both Group 30-170 and Group 30-350 ( q = 5.02, df = 11, p < .01; q = 6.86, df = 11, p < .01 respectively) had fewer responses on the second extinction than on the first extinction but there was no difference between extinctions for Group 30-50 (q < 1.00). The average latency is shown in Fig. 2 where latency was used as the dependent measure. The ordinate scale is inverted to simplify comparison with Fig. 1. Not only are the two figures quite similar but the significant differences were essentially the same. The mean latency/response over five trial blocks during the initial and repeated conditioning-extinction series is shown in Fig. 3 for the three groups that differed in the number of extinction trials. Fig. 3A shows the group receiving 50 extinction trials (Group 30-50), Fig. 3B shows Group 30-170, and Fig. 3C shows Group 30-350. In Groups 30-170 and 30-350 shown in Figs. 3B and 3C it can be seen that the first and second extinction curves separated after the first 10 extinction trials.
