- Email: [email protected]
Stimulus specificity in avoidance learning
LEARNING
AND
MOTIVATION
16, 83-94
(1985)
Stimulus Specificity in Avoidance Learning GEORGE A. CICALA AND JULIAN L. AZORLOSA University
of Delaware
The present experiments address the issue of stimulus specificity in fear and avoidance learning. First, it was established that light stimuli are effective warning signals (WS) in shuttle avoidance. Then light stimuli were shown to produce conditioned suppression in the conditioned emotional response situation comparable to that produced by noise conditioned stimuli. Finally, the effectiveness of noise onset and noise offset as feedback signals was tested. This was assessed under conditions of immediate and delayed termination of a light WS. Delayed termination of a light WS interfered with avoidance learning and the introduction of noise offset as a feedback signal enhanced it. The only demonstration of stimulus specificity was the failure of noise onset to function as a feedback signal. o 1985 Academic
Press. Inc.
In a convincing series of experiments, Jacobs and LoLordo (1977, 1980) seriously questioned the efficacy of light stimuli to elicit conditioned fear and thereby control learned avoidance responses. The present series of experiments attempted to explore their contention by systematically testing light stimuli in aversive learning situations. Jacobs and LoLordo (1977, 1980) graphically demonstrated that light stimuli which have been paired with shock do not produce the enhanced wheel turn or running wheel Sidman avoidance responding so often found with a noise warning signal (WS). They interpreted this to mean that light stimuli are not important in the defensive learning system of the rat, and conclude that lights do not arouse a conditioned aversive state when previously paired with shock. For these reasons, they inferred that lights are ineffective warning signals in avoidance learning situations for rats. This analysis is important because it imposes yet another constraint on the learning process (see Seligman, 1970). Bolles (1970) has argued cogently for response specificity in avoidance learning situations. What Requests for reprints should be sent to Dr. George A. Cicala, Department of Psychology, 220 Wolf Hall, University of Delaware, Newark, DE 19716. 83 0023-%90/U
$3.00
Copyright 8 1985 by Academic Press. Inc. All rights of reproduction in any form reserved.
84
CICALA
AND
AZORLOSA
Jacobs and LoLordo are suggesting is conditioned stimulus specificity as well. Certain issues must be faced before it is concluded that stimulus specificity is the rule in avoidance learning. First, as Jacobs and LoLordo (1980) note, Welker and Wheatley (1977) demonstrated light onset to be an effective fear elicitor as indicated by the conditioned emotional response (CER) procedure. Rescorla (1969) demonstrated conditioned suppression of a bar-press response to a light stimulus while using retardation and summation procedures to study conditioned inhibition, and Rizley and Rescorla (1972) showed conditioned suppression to lights using a higher order conditioning paradigm. However, even if light stimuli were proven to be ineffective conditioned fear elicitors it is not altogether obvious that conditioned fear is necessary for avoidance learning (see Bolles, 1970; Seligman & Johnston, 1973). Finally, while there have been few experiments which have used a light stimulus as a warning signal in avoidance learning, Frontali and Bignami (1973) found light to be an effective WS in avoidance learning and Myers (1964) found lights to be effective in bar-press avoidance training. While it may be true that the effectiveness of a Pavlovian conditioned stimulus (CS) to modulate Sidman avoidance behavior has implications for the acquisition of avoidance responses (Rescorla & Soloman, 1967), it would appear that the more obvious approach would be the straightforward assessment of the effectiveness of different WSs in a standard avoidance learning situation. This is what the present experiment attempted to do. Additionally, two levels of shock were tested to reduce the possibility that the effects obtained would be specific to the shock level selected. EXPERIMENT
1’
Method Subjects. Twenty-four male Wistar rats between 95 and 110 days old and weighing about 350 g were used in this experiment. Apparatus. The experiment used a shuttle box with aluminum walls, 40 cm long, 17.5 cm wide, and 17.5 cm high, and a Plexiglas lid. The grid floor consisted of 0.23-cm stainless-steel rods on 1.25cm centers. Each half of the grid floor was mounted on a separate axle and connected to a mercury switch. These switches enabled the experimenter to monitor shuttle responding. Two 28-V light bulbs and a 20-cm speaker were located 27.5 cm above the box. The shuttle apparatus was enclosed in a sound-attenuating, dark box. The WS was either white noise or light. Readings from a sound meter indicated an average ambient sound level of 42 dB which increased to I This experiment was originally reported in the Bulletin 1984, 22,(l), 70-72.
of the Psychonomic
Sociery.
STIMULUS
SPECIFICITY
85
70 dB upon presentation of the noise WS. The light WS was provided by the two 28-V light bulbs and a 28-V power supply. The US was a IOO- or 150-V ac scrambled shock from a 150-a fixed-impedance source (Campbell & Masterson, 1%9). An Apple II computer was programmed to control presentation of the WS, deliver the US when appropriate, and monitor all shuttle box crossings. A different stimulus control, data recording program was used for each experimental condition. Permanent copies of these were stored on a floppy disk and accessed via an Apple II disk drive. An Epson printer was interfaced with the computer to produce a permanent record of all shuttle data. Procedure. Four equal groups of rats were given shuttle avoidance training. Each group received either light or noise as a WS and IOO- or 150-V shock as a US, yielding a 2 x 2 factorial design. Except for presentation of the WS, the subjects were in quiet darkness. The first of 100 training trials began 1 min after the subject was placed in the shuttle box. At the beginning of each trial the WS was presented. A failure to shuttle within 10 s resulted in shock onset, after which a shuttle response produced the cotermination of the WS and shock. A shuttle response during the 10-s WS terminated that stimulus, defined an avoidance response, and prevented shock onset. The intertrial interval (ITI) was 30 s. Results
and Discussion
Avoidance responding as a function of 10 trial blocks is shown in Fig. 1, which suggests that all groups acquired the response equally well reaching asymptote within 50 trials. A 2 x 2 x 10 ANOVA with trial blocks as a repeated measure showed no significant main effect of either WS or shock level, while the main effect for trial blocks was highly significant [F(9, 180) = 13.75, p < .OOl]. There were no significant interactions between these variables. It thus appears that the light stimulus used was an effective WS for rats in a standard shuttle avoidance situation. Indeed, the light was as effective as a more commonly employed noise ws. A more careful analysis of Fig 1 suggests one difference between noise and light trained groups which appeared on the first block of trials. To evaluate this difference, a 2 x 2 ANOVA for the first trial block was performed post hoc. This revealed a significant main effect for WS [F( 1, 56) = 11.428, p < .Ol]. Inspection of Fig. 1 suggests an interaction between WS and shock during the first trial block, but this was not significant. An examination of individual subject scores during the first trial block suggested that these differences were due to the spontaneous shuttling during the WS interval by noise-trained subjects. For these subjects, 23 avoidance responses (AR) occurred during the first five trials
86
CICALA
FIG.
AND AZORLOSA
0
Ught
100”
n
Llttht
150”
()
Nolcc
100”
0
Nolcc
MO,,
1. Mean percentage avoidance as a function of the trial blocks.
prior to the first avoidance failure and the occurrence of the first shock. This represents a very high rate of spontaneous avoidance, since only 60 ARs were possible. Only 4 of the 12 noise-trained subjects failed to make an AR on the first trial. These data are in sharp contrast to those obtained in the light-trained groups. For these subjects, only 3 ARs occurred prior to the first avoidance failure and 9 of the 12 rats were shocked on the first trial. Thus it appears that differences in avoidance during the first trial block were due to either the greater tendency of noise-trained animals to shuttle during the WS or to a heightened tendency of light-trained animals to freeze during the WS or both. This result may suggest important differential motorogenic effects of the stimuli employed. Noise stimuli may naturally elicit more fear than lights, and this fear may produce escape attempts prior to the first shock presentation. Despite the clear suggestion that these behavioral differences to light and noise stimuli are nonassociative, the heightened operant level of responding in noise-trained subjects might have led to superior learning. However, it did not. The fact that light and noise stimuli are equally effective warning signals in avoidance learning would not be predicted by Jacobs and LoLordo (1980). Their finding that lights do not become conditioned fear stimuli when measured on a Sidman baseline appears irrelevant to their effectiveness as warning signals in shuttle avoidance.
STIMULUS
SPECIFICITY
87
Indeed, their finding that light CSs failed to produce suppression in the Sidman situation is especially curious when it is noted that Welker and Wheatley (1977), which is cited by Jacobs and LoLordo (1980), demonstrated the suppression of hunger motivated bar pressing by a light CS as did Rescorla (1969, Experiment II) and Rizley and Rescorla (1972). This effect has been demonstrated in numerous studies in the blocking literature (e.g., Mackintosh, Bygrave, & Picton, 1977). These studies did not include controls that might enable one to conclude that it was the light shock contingency which was responsible for the suppression. This shortcoming is understandable since the suppression of bar pressing by light stimuli was incidental to the major purposes of those experiments. Rescorla (1969) was interested in conditioned inhibition using retardation and summation tests. Welker and Wheatley (1977) were interested in light increases and decreases as conditioned fear elicitors. Rizley and Rescorla (1972) were studying higher order conditioning. That light stimuli can become conditioned fear elicitors is of considerable theoretical importance. If it is true that light CSs are not effective elicitors of conditioned fear but are effective WSs in avoidance training, this would suggest the appropriateness of minimizing or eliminating conditioned fear as a motivator in avoidance learning (see Bolles, 1970; Seligman & Johnston, 1973). Experiment 2 was designed to compare the effectiveness of light and noise WSs as conditioned fear stimuli using the CER procedure. EXPERIMENT
2
Me thud Subjects. Eighteen male Wistar rats served as subjects. They were approximately 100 days old and weighed between 300 and 350 g. During the experiment the rats were maintained on 24 water deprivation, receiving water only during experimental sessions. Apparatus. The experiment was conducted in a chamber (31 cm long, 15.5 cm high, and 15.5 cm wide) having two aluminum walls with the other two walls and top constructed of Plexiglas. The grid floor consisted of 0.23-cm stainless-steel rods. At one end of the chamber was a circular hole (diameter-l cm) located 5.5 cm above the floor. This provided access to an electrically insulated drinking spout which was connected through a drinkometer circuit to the grid floor. Drinking closed the circuit. This was detected and recorded by the Apple II computer. The CS was either the onset of white noise or the room light. The noise was delivered through a 17-cm speaker located 48 cm above the floor. Noise onset increased the ambient sound level from 48 to 81 dB (scale B). The room light consisted of two 28-V light bulbs located 27.5 cm above the box. The US was a 100-V scrambled shock delivered by a fixed impedance source (Campbell & Masterson 1969). The entire apparatus was enclosed in a sound-attenuating box.
88
CICALA
AND
AZORLOSA
Procedure. The subjects were divided into three equal groups. All groups received one session per day for 3 consecutive days. On the first day the rats were placed in the chamber and allowed to drink for 10 min. On Days 2 and 3, the rats received five training trials followed immediately by four test trials. A training trial consisted of a 10-s presentation of the CS, during the last 2 s of which the US was presented. The CS and US coterminated and were followed by a 95-s ITI. A test trial consisted of a 10-s baseline interval, a 10-s CS interval, and a 10-s feedback interval. No stimuli other than the CSs were presented. An IT1 of 75 s followed each trial. Thus the test and training trials were of equal length. One group received noise as a CS while another received light. To demonstrate that the light-shock contingency was critical in producing suppression, a third group was run as a control. For this group, training consisted of a random presentation of the light CS and the shock US. This was achieved by randomly generating numbers for use on a VI 105 schedule. That is, during training the CS and US were not deliberately paired or unpaired and were each presented five times on an average of once every 105 s. Thus this group received the same number of CS presentations and the same number of shocks during the same time period as the experimental groups. As it turned out, the CS and US never overlapped. Testing of this group was identical to the light CS group. Results and Discussion The data from this experiment are presented in Fig. 2. They state unambiguously that noise and light stimuli impressively and equally suppress licking behavior when these stimuli have previously been paired with shock. A 3 x 2 x 4 ANOVA was performed on these data with days and trials treated as repeated measures. This analysis revealed highly significant effects of CS (noise CS vs light CS vs light unpaired) [F(2, 15) = 24.57, p < .OOl]. A post hoc Sheffe’ comparison (a = .05) showed no significant difference between light CS and noise CS groups and a significant difference between both these groups and the light (unpaired) group. This further supports the conclusion that both the experimental groups were equally suppressed while the control was not suppressed. While it is possible, and under certain circumstances desirable, to run additional groups such as an unpaired noise control and a truly random control using both CSs, the present data leave little doubt suppression produced by the light CS was caused by the prior CS-US pairings. Since suppression under these circumstances has typically been interpreted as due to the mediating conditioned emotional response, fear, this is our preferred interpretation and the most likely interpretation of the suppression demonstrated by Rescorla (1969), Rizley and Rescorla (1972), and others.
STIMULUS
89
SPECIFICITY
; ;
.l-
2 3
: ; .” m
.5 -
Day
2
.4 -
s
I
I 1
2
3
, 4
lrlrls
FIG. 2. Mean suppression ratios as a function of trials. N, Noise CS: L, Light CS: L(U), Light unpaired. EXPERIMENT
3
Experiments 1 and 2, taken together, demonstrate that light stimuli are effective warning signals in avoidance learning and that they can become fear elicitors as measured by response suppression in the CER situation. While these data argue against the conclusions of Jacobs and LoLordo (1980) they do not suggest the role fear might play in avoidance learning. For example, when a light WS occurs in avoidance learning it may, as the present data imply, elicit fear which is available to motivate the response. The termination of light following the occurrence of the avoidance response might reduce fear, thereby reinforcing that response. This is, of course, classical two-factor theory. Alternatively, while WS onset may elicit fear, fear may be irrelevant to avoidance learning, the WS serving as an informational stimulus signaling the appropriateness of the avoidance response, and warning signal termination (WST) might serve a safety signal function as Bolles (1970) suggests. In any event, the present data would argue for the functional equivalence of light and noise stimuli in fear and avoidance learning situations, and would not support the WS specificity assumption. The discovery that lights are functionally equivalent to noises in the
90
CICALA
AND
AZORLOSA
avoidance situation provides an exciting opportunity to study some related theoretical problems in avoidance learning. The problems relate to the effectiveness of a feedback signal (FB) when WST is delayed. Bower, Starr, and Lazarovitz (1965), Bolles and Grossen (1969), and Owen, Cicala, and Herdegen (1978) have all shown, with noise WSs, that when response contingent WST is delayed, learning is retarded. Further, this retardation is eliminated by the introduction of light offset as a FB signal. These investigators suggested that light served as an effective reinforcer because it was a conditioned fear inhibitor. Owen et al. (1978) showed that the introduction of an F3 signal caused the immediate release from conditioned suppression of a drinking response, a finding which supports this interpretation. Since Experiment 1 demonstrated that light can serve as an effective WS, it is possible to determine whether noise will serve as an effective FB signal and therefore presumably function as a conditioned fear inhibitor. Jacobs and LoLordo (1980) have addressed this issue also in the Sidman avoidance situation. Specifically, they showed that noise offset but not noise onset was a good FB signal as measured by the attenuation of the Sidman avoidance response. From this demonstration they concluded that “Since an increased intensity of auditory stimulation that had repeatedly followed shock and preceded a shock-free period does not become a conditioned inhibitor of fear, that stimulus will not serve to reinforce an avoidance response which produces it” (p. 452). The third experiment was designed to test this generalization even though, as Owen et al. (1978) have indicated, fear inhibition may not play a role in avoidance situations which do not provide exteroceptive FB stimuli. An additional purpose of Experiment 3 was to determine whether delaying the termination of a light WS would have the same detrimental effect on avoidance learning as has been demonstrated with a noise WS (Kamin, 1956). If this effect is obtained, it would support the functional equivalence of lights and noises in avoidance situations, in that it would show response contingent termination of a light WS to be an effective reinforcer of the avoidance response. Method Subjects. Sixty male Wistar rats were used. They were approximately 90 days old at the beginning of this experiment. Apparatus. The shuttle box described in Experiment 1 was used. The light WS was identical to that of Experiment 1. A white noise stimulus, used in the feedback conditions, increased the sound level from 42 to 81 dB (scale B). A 100-V scrambled shock, described above, served as the US. Procedure. The paradigm employed was similar to that of Experiment 1. Subjects were randomly assigned to one of six groups. All groups
STIMULUS
SPECIFICITY
91
received a light WS. Failure to make a shuttle response in 10 s resulted in the presentation of the US, shock. In three groups, an escape or avoidance response resulted in the termination of the WS (Immediate WST). For the remaining three groups, the WS remained on for an additional 10 s after a shuttle response was made (Delay WST). Subjects also received one of three feedback conditions. Two groups (Noise Onset) received a response contingent FB signal. Immediately after an escape or avoidance response, the 81-dB white noise was presented for IO s. Two other groups (Noise Offset) received continuous white noise (81 dB) during training which was terminated for the 10-s period following an escape or avoidance response. The final two groups (No FB) received no feedback. This yields a 2 x 3 design with two WS conditions (Immediate vs Delay) and three feedback conditions (Noise Onset vs Noise Offset vs No FB). As in Experiment 1, the IT1 was 30 s. The first 10 s of this constituted the feedback interval and the light WS remained on during this period in the Delay WST condition. All groups received 100 trials of avoidance training. Results
and Discussion
The avoidance scores obtained throughout acquisition were subjected to a 2 x 3 x 10 ANOVA with the 10 trial blocks treated as repeated measures. In support of the results of Experiment 1, the main effect for trial blocks was found to be highly significant jF(9, 486) = 41.22, p < .OOl], substantiating the effectiveness of light onset as a WS. Additionally, the main effect for Immediate vs Delay WST was highly significant [F( 1, 54) = 9.93, p < .Ol]. The upper half of Fig. 3 presents the main effects of light termination conditions across feedback conditions. Collapsing across feedback conditions is appropriate since none of the interactions even approached significance. Figure 3 clearly suggests that the immediate termination of a light WS upon the occurrence of a shuttle response produced markedly better performance than when the light remained on for 10 s following that response. This effect using a noise WS was originally demonstrated by Kamin (1956) and subsequently by many others (e.g., Cicala & Owen, 1976). To demonstrate that the termination of a light WS is an important source of reinforcement in avoidance may further support the thesis that lights and noise are functionally equivalent warning signals. Unfortunately, these results do not help us to understand why immediate WST facilitates avoidance learning when either light or noise WSs are used. It may be that WST provides a supporting stimulus as suggested by Jacobs and LoLordo (1980). That is, the termination of a noise WS results in quiet and the termination of a light results in darkness. This is consistent with the consummatory stimulus analysis of avoidance learning proposed by Masterson and Crawford (1982). Alternatively, light ter-
92
CICALA
AND AZORLOSA
BlOCkC of 10 Trialc 3. Mean percentage avoidance as a function of trial blocks. Upper panel: Delay WST vs Immediate WST. Lower panel: Noise Onset FB signal vs Noise Offset FB signal vs No Feedback. FIG.
mination may reinforce the avoidance response by reducing fear. Despite the fact that Owen et al. (1978) show that fear dissipation following the termination of a noise WS is a relatively slow process and is therefore an unlikely reinforcer of the avoidance response, it is an empirical question whether light stimuli function in the same way. That a light WS elicits conditioned fear was demonstrated in Experiment 2. What must be determined in future experiments is whether termination of a light WS produces a rapid reduction of fear which might reasonably reinforce the avoidance response. The bottom half of Fig. 3 shows avoidance learning as a function of feedback conditions following the response collapsed across the two conditions of WST. Inspection of this figure suggests that the termination of noise effectively facilitated avoidance responding while the onset of noise produced responding indistinguishable from the No Feedback condition. This is suggested by a significant main effect for Feedback condition [F(2, 54) = 4.84, p < .Ol]. Post hoc Sheffe’ comparisons confirmed that the no feedback and noise onset groups were not significantly different
STIMULUS
SPECIFICITY
93
from one another and that both these groups responded at a significantly lower rate than did the noise offset group (p < .Ol). This effect is in total agreement with that obtained by Jacobs and LoLordo (1980) who demonstrated that noise offset, when it followed shock termination and predicted a period free from shock, significantly decreased Sidman avoidance responding. Under the same conditions, noise onset did not decrease avoiding. Jacobs and LoLordo interpret this to mean that noise offset and not noise onset can function as a conditioned fear inhibitor. From this they predict that noise offset would be an effective fear feedback signal in avoidance learning and noise onset would not. This is what the present experiment demonstrated. Since this pattern of results was predicted from the Jacobs and LoLordo (1980) demonstration, it is tempting to interpret it, as they do, to the failure of noise onset to acquire fear inhibitory properties. This argument suggests that noise offset reinforces the avoidance response by inhibiting fear, a property which cannot be acquired by noise onset. However, it may be that noise onset can become a good fear inhibitor but the parameters of the present avoidance learning situation are not such as to provide an opportunity for such learning. For example, it may be that the onset of noise stimuli elicits more innate fear than the onset of light, and a greater number of inhibition training trials than is provided in shuttle avoidance would be necessary to demonstrate this potential. Indeed, D’Amato, Fazzaro, and Etkin (1968) demonstrated, using barpress avoidance, that response contingent noise onset reinforced the avoidance response in the absence of immediate WST. Bar-press avoidance, being vastly more difficult than shuttle avoidance and taking many more trials to learn, may have provided a sufficient number of noise-no shock pairings to have acquired fear inhibitory properties. The results of these experiments do not in any way suggest that stimulus specificity deserves a place in models of avoidance learning. Indeed, what they suggest is that lights and noises are equally effective WSs in shuttle avoidance training (Experiments 1 and 3), that they are equally effective fear elicitors as measured by conditioned suppression (Experiment 2), and that their termination following the shuttle response is critical to the learning of it as an avoidance response. That noise onset may not be an effective conditioned fear inhibitor may have implications for the issue of stimulus specificity in classical inhibitory conditioning, but such conditioning may not play an important role in avoidance learning which does not provide specific feedback stimuli. REFERENCES Belles, R. C. (1970). Species-specific defense reactions and avoidance learning. Psychological Review,
77, 32-48.
Boiles, R. C.. & Grossen, N. (1969). Effects of an informational stimulus on the acquisition
94
CICALA
AND AZORLOSA
of avoidance in rats. Journal of Comparative and Physiological Psychology, 68, 9099. Bower, G., Starr, R., & Lazarovitz, L. (1965). Amount of response produced change in the CS and avoidance learning. Journal of Comparative and Physiological Psychology, 59, 13-17. Campbell, B. A., & Masterson, F. A. (1969). Psychophysics of punishment. In B. A. Campbell & R. M. Church (Eds.), Punishment and aversive behavior. (pp. 3-42). New York: Appleton-Century-Crofts. Cicala, G. A., & Owen, J. W. (1976). Warning signal termination and a feedback signal may not serve the same function. Learning and Motivation, 7, 356-367. D’Amato, M. R., Fazzaro, J., & Etkin, M. (1968). Anticipatory responding and avoidance discrimination as factors in avoidance conditioning. Journal of Experimental Psychology, 77, 41-47. Frontali, M., & Bignami, G. (1973). Go-no go avoidance discrimination in rats with simple “go” and compound “no go” signals: Stimulus modality and stimulus intensity. Animal Learning and Behavior, 1, 21-24. Jacobs, W. J., & LoLordo, V. M. (1977). The sensory basis of avoidance responding in the rat. Learning and Motivation, 8, 448-466. Jacobs, W. J., 8~ LoLordo, V. M. (1980). Constraints on Pavlovian aversive conditioning: Implications for avoidance learning in the rat. Learning and Motivation, 11, 427-455. Kamin, L. J. (1956). The effects of termination of the CS and avoidance of the US on avoidance learning. Journal of Comparative and Physiological Psychology, 49, 420424. Mackintosh, N. J., Bygrave, D. J., 62 Picton, B. M. B. (1977). Locus of the effect of a surprising reinforcer on the attenuation of blocking. Quarterly Journal of Experimental Psychology, 29, 327-336. Masterson, F. A., & Crawford, M. (1982). The defense motivation system: A theory of avoidance behavior. The Behavioral and Brain Sciences, 5, 661-696. Myers, A. K. (1964). Discriminated operant avoidance learning in Wistar and G-4 rats as a function of type of warning stimulus. Journal of Comparative and Physiological Psychology, 58, 453-455. Owen, J. W., Cicala, G. A., & Herdegen, R. T. (1978) Fear inhibition and species-specific defense reactions may contribute independently to avoidance learning. Learning and Motivation, 9, 297-313. Rescorla, R. A. (1969). Conditioned inhibition of fear resulting from negative CS-US contingencies. Journal of Comparative and Physiological Psychology, 67, 504-509. Rescorla, R. A., & Soloman, R. L. (1967). Two-process learning theory: Relationships between Pavlovian conditioning and instrumental learning. Psychological Review, 74, 151-181. Rizley, R. C., & Rescorla, R. A. (1972). Associations in second-order conditioning and sensory preconditioning. Journal of Comparative and Physiological Psychology, 81, l-11. Seligman, M. E. P. (1970). On the generality of the laws of learning. Psychological Review, 77, 406-418. Seligman, M. E. P., & Johnston, J. (1973). A cognitive theory of avoidance learning. In F. J. McGuigan & D. B. Lumsden (Eds.), Contemporary approaches to conditioning and learning. New York: Wiley. Welker, R. L., & Wheatley, K. L. (1977). Differential acquisition of conditioned suppression in rats with increased and decreased luminance levels as CS + S. Learning and Motivation, 8, 247-262. Received June 11, 1984 Revised August 30, 1984
AND
MOTIVATION
16, 83-94
(1985)
Stimulus Specificity in Avoidance Learning GEORGE A. CICALA AND JULIAN L. AZORLOSA University
of Delaware
The present experiments address the issue of stimulus specificity in fear and avoidance learning. First, it was established that light stimuli are effective warning signals (WS) in shuttle avoidance. Then light stimuli were shown to produce conditioned suppression in the conditioned emotional response situation comparable to that produced by noise conditioned stimuli. Finally, the effectiveness of noise onset and noise offset as feedback signals was tested. This was assessed under conditions of immediate and delayed termination of a light WS. Delayed termination of a light WS interfered with avoidance learning and the introduction of noise offset as a feedback signal enhanced it. The only demonstration of stimulus specificity was the failure of noise onset to function as a feedback signal. o 1985 Academic
Press. Inc.
In a convincing series of experiments, Jacobs and LoLordo (1977, 1980) seriously questioned the efficacy of light stimuli to elicit conditioned fear and thereby control learned avoidance responses. The present series of experiments attempted to explore their contention by systematically testing light stimuli in aversive learning situations. Jacobs and LoLordo (1977, 1980) graphically demonstrated that light stimuli which have been paired with shock do not produce the enhanced wheel turn or running wheel Sidman avoidance responding so often found with a noise warning signal (WS). They interpreted this to mean that light stimuli are not important in the defensive learning system of the rat, and conclude that lights do not arouse a conditioned aversive state when previously paired with shock. For these reasons, they inferred that lights are ineffective warning signals in avoidance learning situations for rats. This analysis is important because it imposes yet another constraint on the learning process (see Seligman, 1970). Bolles (1970) has argued cogently for response specificity in avoidance learning situations. What Requests for reprints should be sent to Dr. George A. Cicala, Department of Psychology, 220 Wolf Hall, University of Delaware, Newark, DE 19716. 83 0023-%90/U
$3.00
Copyright 8 1985 by Academic Press. Inc. All rights of reproduction in any form reserved.
84
CICALA
AND
AZORLOSA
Jacobs and LoLordo are suggesting is conditioned stimulus specificity as well. Certain issues must be faced before it is concluded that stimulus specificity is the rule in avoidance learning. First, as Jacobs and LoLordo (1980) note, Welker and Wheatley (1977) demonstrated light onset to be an effective fear elicitor as indicated by the conditioned emotional response (CER) procedure. Rescorla (1969) demonstrated conditioned suppression of a bar-press response to a light stimulus while using retardation and summation procedures to study conditioned inhibition, and Rizley and Rescorla (1972) showed conditioned suppression to lights using a higher order conditioning paradigm. However, even if light stimuli were proven to be ineffective conditioned fear elicitors it is not altogether obvious that conditioned fear is necessary for avoidance learning (see Bolles, 1970; Seligman & Johnston, 1973). Finally, while there have been few experiments which have used a light stimulus as a warning signal in avoidance learning, Frontali and Bignami (1973) found light to be an effective WS in avoidance learning and Myers (1964) found lights to be effective in bar-press avoidance training. While it may be true that the effectiveness of a Pavlovian conditioned stimulus (CS) to modulate Sidman avoidance behavior has implications for the acquisition of avoidance responses (Rescorla & Soloman, 1967), it would appear that the more obvious approach would be the straightforward assessment of the effectiveness of different WSs in a standard avoidance learning situation. This is what the present experiment attempted to do. Additionally, two levels of shock were tested to reduce the possibility that the effects obtained would be specific to the shock level selected. EXPERIMENT
1’
Method Subjects. Twenty-four male Wistar rats between 95 and 110 days old and weighing about 350 g were used in this experiment. Apparatus. The experiment used a shuttle box with aluminum walls, 40 cm long, 17.5 cm wide, and 17.5 cm high, and a Plexiglas lid. The grid floor consisted of 0.23-cm stainless-steel rods on 1.25cm centers. Each half of the grid floor was mounted on a separate axle and connected to a mercury switch. These switches enabled the experimenter to monitor shuttle responding. Two 28-V light bulbs and a 20-cm speaker were located 27.5 cm above the box. The shuttle apparatus was enclosed in a sound-attenuating, dark box. The WS was either white noise or light. Readings from a sound meter indicated an average ambient sound level of 42 dB which increased to I This experiment was originally reported in the Bulletin 1984, 22,(l), 70-72.
of the Psychonomic
Sociery.
STIMULUS
SPECIFICITY
85
70 dB upon presentation of the noise WS. The light WS was provided by the two 28-V light bulbs and a 28-V power supply. The US was a IOO- or 150-V ac scrambled shock from a 150-a fixed-impedance source (Campbell & Masterson, 1%9). An Apple II computer was programmed to control presentation of the WS, deliver the US when appropriate, and monitor all shuttle box crossings. A different stimulus control, data recording program was used for each experimental condition. Permanent copies of these were stored on a floppy disk and accessed via an Apple II disk drive. An Epson printer was interfaced with the computer to produce a permanent record of all shuttle data. Procedure. Four equal groups of rats were given shuttle avoidance training. Each group received either light or noise as a WS and IOO- or 150-V shock as a US, yielding a 2 x 2 factorial design. Except for presentation of the WS, the subjects were in quiet darkness. The first of 100 training trials began 1 min after the subject was placed in the shuttle box. At the beginning of each trial the WS was presented. A failure to shuttle within 10 s resulted in shock onset, after which a shuttle response produced the cotermination of the WS and shock. A shuttle response during the 10-s WS terminated that stimulus, defined an avoidance response, and prevented shock onset. The intertrial interval (ITI) was 30 s. Results
and Discussion
Avoidance responding as a function of 10 trial blocks is shown in Fig. 1, which suggests that all groups acquired the response equally well reaching asymptote within 50 trials. A 2 x 2 x 10 ANOVA with trial blocks as a repeated measure showed no significant main effect of either WS or shock level, while the main effect for trial blocks was highly significant [F(9, 180) = 13.75, p < .OOl]. There were no significant interactions between these variables. It thus appears that the light stimulus used was an effective WS for rats in a standard shuttle avoidance situation. Indeed, the light was as effective as a more commonly employed noise ws. A more careful analysis of Fig 1 suggests one difference between noise and light trained groups which appeared on the first block of trials. To evaluate this difference, a 2 x 2 ANOVA for the first trial block was performed post hoc. This revealed a significant main effect for WS [F( 1, 56) = 11.428, p < .Ol]. Inspection of Fig. 1 suggests an interaction between WS and shock during the first trial block, but this was not significant. An examination of individual subject scores during the first trial block suggested that these differences were due to the spontaneous shuttling during the WS interval by noise-trained subjects. For these subjects, 23 avoidance responses (AR) occurred during the first five trials
86
CICALA
FIG.
AND AZORLOSA
0
Ught
100”
n
Llttht
150”
()
Nolcc
100”
0
Nolcc
MO,,
1. Mean percentage avoidance as a function of the trial blocks.
prior to the first avoidance failure and the occurrence of the first shock. This represents a very high rate of spontaneous avoidance, since only 60 ARs were possible. Only 4 of the 12 noise-trained subjects failed to make an AR on the first trial. These data are in sharp contrast to those obtained in the light-trained groups. For these subjects, only 3 ARs occurred prior to the first avoidance failure and 9 of the 12 rats were shocked on the first trial. Thus it appears that differences in avoidance during the first trial block were due to either the greater tendency of noise-trained animals to shuttle during the WS or to a heightened tendency of light-trained animals to freeze during the WS or both. This result may suggest important differential motorogenic effects of the stimuli employed. Noise stimuli may naturally elicit more fear than lights, and this fear may produce escape attempts prior to the first shock presentation. Despite the clear suggestion that these behavioral differences to light and noise stimuli are nonassociative, the heightened operant level of responding in noise-trained subjects might have led to superior learning. However, it did not. The fact that light and noise stimuli are equally effective warning signals in avoidance learning would not be predicted by Jacobs and LoLordo (1980). Their finding that lights do not become conditioned fear stimuli when measured on a Sidman baseline appears irrelevant to their effectiveness as warning signals in shuttle avoidance.
STIMULUS
SPECIFICITY
87
Indeed, their finding that light CSs failed to produce suppression in the Sidman situation is especially curious when it is noted that Welker and Wheatley (1977), which is cited by Jacobs and LoLordo (1980), demonstrated the suppression of hunger motivated bar pressing by a light CS as did Rescorla (1969, Experiment II) and Rizley and Rescorla (1972). This effect has been demonstrated in numerous studies in the blocking literature (e.g., Mackintosh, Bygrave, & Picton, 1977). These studies did not include controls that might enable one to conclude that it was the light shock contingency which was responsible for the suppression. This shortcoming is understandable since the suppression of bar pressing by light stimuli was incidental to the major purposes of those experiments. Rescorla (1969) was interested in conditioned inhibition using retardation and summation tests. Welker and Wheatley (1977) were interested in light increases and decreases as conditioned fear elicitors. Rizley and Rescorla (1972) were studying higher order conditioning. That light stimuli can become conditioned fear elicitors is of considerable theoretical importance. If it is true that light CSs are not effective elicitors of conditioned fear but are effective WSs in avoidance training, this would suggest the appropriateness of minimizing or eliminating conditioned fear as a motivator in avoidance learning (see Bolles, 1970; Seligman & Johnston, 1973). Experiment 2 was designed to compare the effectiveness of light and noise WSs as conditioned fear stimuli using the CER procedure. EXPERIMENT
2
Me thud Subjects. Eighteen male Wistar rats served as subjects. They were approximately 100 days old and weighed between 300 and 350 g. During the experiment the rats were maintained on 24 water deprivation, receiving water only during experimental sessions. Apparatus. The experiment was conducted in a chamber (31 cm long, 15.5 cm high, and 15.5 cm wide) having two aluminum walls with the other two walls and top constructed of Plexiglas. The grid floor consisted of 0.23-cm stainless-steel rods. At one end of the chamber was a circular hole (diameter-l cm) located 5.5 cm above the floor. This provided access to an electrically insulated drinking spout which was connected through a drinkometer circuit to the grid floor. Drinking closed the circuit. This was detected and recorded by the Apple II computer. The CS was either the onset of white noise or the room light. The noise was delivered through a 17-cm speaker located 48 cm above the floor. Noise onset increased the ambient sound level from 48 to 81 dB (scale B). The room light consisted of two 28-V light bulbs located 27.5 cm above the box. The US was a 100-V scrambled shock delivered by a fixed impedance source (Campbell & Masterson 1969). The entire apparatus was enclosed in a sound-attenuating box.
88
CICALA
AND
AZORLOSA
Procedure. The subjects were divided into three equal groups. All groups received one session per day for 3 consecutive days. On the first day the rats were placed in the chamber and allowed to drink for 10 min. On Days 2 and 3, the rats received five training trials followed immediately by four test trials. A training trial consisted of a 10-s presentation of the CS, during the last 2 s of which the US was presented. The CS and US coterminated and were followed by a 95-s ITI. A test trial consisted of a 10-s baseline interval, a 10-s CS interval, and a 10-s feedback interval. No stimuli other than the CSs were presented. An IT1 of 75 s followed each trial. Thus the test and training trials were of equal length. One group received noise as a CS while another received light. To demonstrate that the light-shock contingency was critical in producing suppression, a third group was run as a control. For this group, training consisted of a random presentation of the light CS and the shock US. This was achieved by randomly generating numbers for use on a VI 105 schedule. That is, during training the CS and US were not deliberately paired or unpaired and were each presented five times on an average of once every 105 s. Thus this group received the same number of CS presentations and the same number of shocks during the same time period as the experimental groups. As it turned out, the CS and US never overlapped. Testing of this group was identical to the light CS group. Results and Discussion The data from this experiment are presented in Fig. 2. They state unambiguously that noise and light stimuli impressively and equally suppress licking behavior when these stimuli have previously been paired with shock. A 3 x 2 x 4 ANOVA was performed on these data with days and trials treated as repeated measures. This analysis revealed highly significant effects of CS (noise CS vs light CS vs light unpaired) [F(2, 15) = 24.57, p < .OOl]. A post hoc Sheffe’ comparison (a = .05) showed no significant difference between light CS and noise CS groups and a significant difference between both these groups and the light (unpaired) group. This further supports the conclusion that both the experimental groups were equally suppressed while the control was not suppressed. While it is possible, and under certain circumstances desirable, to run additional groups such as an unpaired noise control and a truly random control using both CSs, the present data leave little doubt suppression produced by the light CS was caused by the prior CS-US pairings. Since suppression under these circumstances has typically been interpreted as due to the mediating conditioned emotional response, fear, this is our preferred interpretation and the most likely interpretation of the suppression demonstrated by Rescorla (1969), Rizley and Rescorla (1972), and others.
STIMULUS
89
SPECIFICITY
; ;
.l-
2 3
: ; .” m
.5 -
Day
2
.4 -
s
I
I 1
2
3
, 4
lrlrls
FIG. 2. Mean suppression ratios as a function of trials. N, Noise CS: L, Light CS: L(U), Light unpaired. EXPERIMENT
3
Experiments 1 and 2, taken together, demonstrate that light stimuli are effective warning signals in avoidance learning and that they can become fear elicitors as measured by response suppression in the CER situation. While these data argue against the conclusions of Jacobs and LoLordo (1980) they do not suggest the role fear might play in avoidance learning. For example, when a light WS occurs in avoidance learning it may, as the present data imply, elicit fear which is available to motivate the response. The termination of light following the occurrence of the avoidance response might reduce fear, thereby reinforcing that response. This is, of course, classical two-factor theory. Alternatively, while WS onset may elicit fear, fear may be irrelevant to avoidance learning, the WS serving as an informational stimulus signaling the appropriateness of the avoidance response, and warning signal termination (WST) might serve a safety signal function as Bolles (1970) suggests. In any event, the present data would argue for the functional equivalence of light and noise stimuli in fear and avoidance learning situations, and would not support the WS specificity assumption. The discovery that lights are functionally equivalent to noises in the
90
CICALA
AND
AZORLOSA
avoidance situation provides an exciting opportunity to study some related theoretical problems in avoidance learning. The problems relate to the effectiveness of a feedback signal (FB) when WST is delayed. Bower, Starr, and Lazarovitz (1965), Bolles and Grossen (1969), and Owen, Cicala, and Herdegen (1978) have all shown, with noise WSs, that when response contingent WST is delayed, learning is retarded. Further, this retardation is eliminated by the introduction of light offset as a FB signal. These investigators suggested that light served as an effective reinforcer because it was a conditioned fear inhibitor. Owen et al. (1978) showed that the introduction of an F3 signal caused the immediate release from conditioned suppression of a drinking response, a finding which supports this interpretation. Since Experiment 1 demonstrated that light can serve as an effective WS, it is possible to determine whether noise will serve as an effective FB signal and therefore presumably function as a conditioned fear inhibitor. Jacobs and LoLordo (1980) have addressed this issue also in the Sidman avoidance situation. Specifically, they showed that noise offset but not noise onset was a good FB signal as measured by the attenuation of the Sidman avoidance response. From this demonstration they concluded that “Since an increased intensity of auditory stimulation that had repeatedly followed shock and preceded a shock-free period does not become a conditioned inhibitor of fear, that stimulus will not serve to reinforce an avoidance response which produces it” (p. 452). The third experiment was designed to test this generalization even though, as Owen et al. (1978) have indicated, fear inhibition may not play a role in avoidance situations which do not provide exteroceptive FB stimuli. An additional purpose of Experiment 3 was to determine whether delaying the termination of a light WS would have the same detrimental effect on avoidance learning as has been demonstrated with a noise WS (Kamin, 1956). If this effect is obtained, it would support the functional equivalence of lights and noises in avoidance situations, in that it would show response contingent termination of a light WS to be an effective reinforcer of the avoidance response. Method Subjects. Sixty male Wistar rats were used. They were approximately 90 days old at the beginning of this experiment. Apparatus. The shuttle box described in Experiment 1 was used. The light WS was identical to that of Experiment 1. A white noise stimulus, used in the feedback conditions, increased the sound level from 42 to 81 dB (scale B). A 100-V scrambled shock, described above, served as the US. Procedure. The paradigm employed was similar to that of Experiment 1. Subjects were randomly assigned to one of six groups. All groups
STIMULUS
SPECIFICITY
91
received a light WS. Failure to make a shuttle response in 10 s resulted in the presentation of the US, shock. In three groups, an escape or avoidance response resulted in the termination of the WS (Immediate WST). For the remaining three groups, the WS remained on for an additional 10 s after a shuttle response was made (Delay WST). Subjects also received one of three feedback conditions. Two groups (Noise Onset) received a response contingent FB signal. Immediately after an escape or avoidance response, the 81-dB white noise was presented for IO s. Two other groups (Noise Offset) received continuous white noise (81 dB) during training which was terminated for the 10-s period following an escape or avoidance response. The final two groups (No FB) received no feedback. This yields a 2 x 3 design with two WS conditions (Immediate vs Delay) and three feedback conditions (Noise Onset vs Noise Offset vs No FB). As in Experiment 1, the IT1 was 30 s. The first 10 s of this constituted the feedback interval and the light WS remained on during this period in the Delay WST condition. All groups received 100 trials of avoidance training. Results
and Discussion
The avoidance scores obtained throughout acquisition were subjected to a 2 x 3 x 10 ANOVA with the 10 trial blocks treated as repeated measures. In support of the results of Experiment 1, the main effect for trial blocks was found to be highly significant jF(9, 486) = 41.22, p < .OOl], substantiating the effectiveness of light onset as a WS. Additionally, the main effect for Immediate vs Delay WST was highly significant [F( 1, 54) = 9.93, p < .Ol]. The upper half of Fig. 3 presents the main effects of light termination conditions across feedback conditions. Collapsing across feedback conditions is appropriate since none of the interactions even approached significance. Figure 3 clearly suggests that the immediate termination of a light WS upon the occurrence of a shuttle response produced markedly better performance than when the light remained on for 10 s following that response. This effect using a noise WS was originally demonstrated by Kamin (1956) and subsequently by many others (e.g., Cicala & Owen, 1976). To demonstrate that the termination of a light WS is an important source of reinforcement in avoidance may further support the thesis that lights and noise are functionally equivalent warning signals. Unfortunately, these results do not help us to understand why immediate WST facilitates avoidance learning when either light or noise WSs are used. It may be that WST provides a supporting stimulus as suggested by Jacobs and LoLordo (1980). That is, the termination of a noise WS results in quiet and the termination of a light results in darkness. This is consistent with the consummatory stimulus analysis of avoidance learning proposed by Masterson and Crawford (1982). Alternatively, light ter-
92
CICALA
AND AZORLOSA
BlOCkC of 10 Trialc 3. Mean percentage avoidance as a function of trial blocks. Upper panel: Delay WST vs Immediate WST. Lower panel: Noise Onset FB signal vs Noise Offset FB signal vs No Feedback. FIG.
mination may reinforce the avoidance response by reducing fear. Despite the fact that Owen et al. (1978) show that fear dissipation following the termination of a noise WS is a relatively slow process and is therefore an unlikely reinforcer of the avoidance response, it is an empirical question whether light stimuli function in the same way. That a light WS elicits conditioned fear was demonstrated in Experiment 2. What must be determined in future experiments is whether termination of a light WS produces a rapid reduction of fear which might reasonably reinforce the avoidance response. The bottom half of Fig. 3 shows avoidance learning as a function of feedback conditions following the response collapsed across the two conditions of WST. Inspection of this figure suggests that the termination of noise effectively facilitated avoidance responding while the onset of noise produced responding indistinguishable from the No Feedback condition. This is suggested by a significant main effect for Feedback condition [F(2, 54) = 4.84, p < .Ol]. Post hoc Sheffe’ comparisons confirmed that the no feedback and noise onset groups were not significantly different
STIMULUS
SPECIFICITY
93
from one another and that both these groups responded at a significantly lower rate than did the noise offset group (p < .Ol). This effect is in total agreement with that obtained by Jacobs and LoLordo (1980) who demonstrated that noise offset, when it followed shock termination and predicted a period free from shock, significantly decreased Sidman avoidance responding. Under the same conditions, noise onset did not decrease avoiding. Jacobs and LoLordo interpret this to mean that noise offset and not noise onset can function as a conditioned fear inhibitor. From this they predict that noise offset would be an effective fear feedback signal in avoidance learning and noise onset would not. This is what the present experiment demonstrated. Since this pattern of results was predicted from the Jacobs and LoLordo (1980) demonstration, it is tempting to interpret it, as they do, to the failure of noise onset to acquire fear inhibitory properties. This argument suggests that noise offset reinforces the avoidance response by inhibiting fear, a property which cannot be acquired by noise onset. However, it may be that noise onset can become a good fear inhibitor but the parameters of the present avoidance learning situation are not such as to provide an opportunity for such learning. For example, it may be that the onset of noise stimuli elicits more innate fear than the onset of light, and a greater number of inhibition training trials than is provided in shuttle avoidance would be necessary to demonstrate this potential. Indeed, D’Amato, Fazzaro, and Etkin (1968) demonstrated, using barpress avoidance, that response contingent noise onset reinforced the avoidance response in the absence of immediate WST. Bar-press avoidance, being vastly more difficult than shuttle avoidance and taking many more trials to learn, may have provided a sufficient number of noise-no shock pairings to have acquired fear inhibitory properties. The results of these experiments do not in any way suggest that stimulus specificity deserves a place in models of avoidance learning. Indeed, what they suggest is that lights and noises are equally effective WSs in shuttle avoidance training (Experiments 1 and 3), that they are equally effective fear elicitors as measured by conditioned suppression (Experiment 2), and that their termination following the shuttle response is critical to the learning of it as an avoidance response. That noise onset may not be an effective conditioned fear inhibitor may have implications for the issue of stimulus specificity in classical inhibitory conditioning, but such conditioning may not play an important role in avoidance learning which does not provide specific feedback stimuli. REFERENCES Belles, R. C. (1970). Species-specific defense reactions and avoidance learning. Psychological Review,
77, 32-48.
Boiles, R. C.. & Grossen, N. (1969). Effects of an informational stimulus on the acquisition
94
CICALA
AND AZORLOSA
of avoidance in rats. Journal of Comparative and Physiological Psychology, 68, 9099. Bower, G., Starr, R., & Lazarovitz, L. (1965). Amount of response produced change in the CS and avoidance learning. Journal of Comparative and Physiological Psychology, 59, 13-17. Campbell, B. A., & Masterson, F. A. (1969). Psychophysics of punishment. In B. A. Campbell & R. M. Church (Eds.), Punishment and aversive behavior. (pp. 3-42). New York: Appleton-Century-Crofts. Cicala, G. A., & Owen, J. W. (1976). Warning signal termination and a feedback signal may not serve the same function. Learning and Motivation, 7, 356-367. D’Amato, M. R., Fazzaro, J., & Etkin, M. (1968). Anticipatory responding and avoidance discrimination as factors in avoidance conditioning. Journal of Experimental Psychology, 77, 41-47. Frontali, M., & Bignami, G. (1973). Go-no go avoidance discrimination in rats with simple “go” and compound “no go” signals: Stimulus modality and stimulus intensity. Animal Learning and Behavior, 1, 21-24. Jacobs, W. J., & LoLordo, V. M. (1977). The sensory basis of avoidance responding in the rat. Learning and Motivation, 8, 448-466. Jacobs, W. J., 8~ LoLordo, V. M. (1980). Constraints on Pavlovian aversive conditioning: Implications for avoidance learning in the rat. Learning and Motivation, 11, 427-455. Kamin, L. J. (1956). The effects of termination of the CS and avoidance of the US on avoidance learning. Journal of Comparative and Physiological Psychology, 49, 420424. Mackintosh, N. J., Bygrave, D. J., 62 Picton, B. M. B. (1977). Locus of the effect of a surprising reinforcer on the attenuation of blocking. Quarterly Journal of Experimental Psychology, 29, 327-336. Masterson, F. A., & Crawford, M. (1982). The defense motivation system: A theory of avoidance behavior. The Behavioral and Brain Sciences, 5, 661-696. Myers, A. K. (1964). Discriminated operant avoidance learning in Wistar and G-4 rats as a function of type of warning stimulus. Journal of Comparative and Physiological Psychology, 58, 453-455. Owen, J. W., Cicala, G. A., & Herdegen, R. T. (1978) Fear inhibition and species-specific defense reactions may contribute independently to avoidance learning. Learning and Motivation, 9, 297-313. Rescorla, R. A. (1969). Conditioned inhibition of fear resulting from negative CS-US contingencies. Journal of Comparative and Physiological Psychology, 67, 504-509. Rescorla, R. A., & Soloman, R. L. (1967). Two-process learning theory: Relationships between Pavlovian conditioning and instrumental learning. Psychological Review, 74, 151-181. Rizley, R. C., & Rescorla, R. A. (1972). Associations in second-order conditioning and sensory preconditioning. Journal of Comparative and Physiological Psychology, 81, l-11. Seligman, M. E. P. (1970). On the generality of the laws of learning. Psychological Review, 77, 406-418. Seligman, M. E. P., & Johnston, J. (1973). A cognitive theory of avoidance learning. In F. J. McGuigan & D. B. Lumsden (Eds.), Contemporary approaches to conditioning and learning. New York: Wiley. Welker, R. L., & Wheatley, K. L. (1977). Differential acquisition of conditioned suppression in rats with increased and decreased luminance levels as CS + S. Learning and Motivation, 8, 247-262. Received June 11, 1984 Revised August 30, 1984