DIFFERENTIATION LEARNING AS A FUNCTION · 428 E. JAKUBOWSKA AND K. ZIELIRSKI the pair of...

ACTA NEUROBIOL. EXP. 1976, 36: 427-446

DIFFERENTIATION LEARNING AS A FUNCTION OF STIMULUS INTENSITY

AND PREVIOUS EXPERIENCE WITH THE CS+

Ewa JAKUBOWSKA and Kazimierz ZIELIRSKI

Department of Neurophysiology, ~ e n c k i Institute of Experimental Biology Warsaw, Poland

Abstract. Discrimination between 50- and 70-dB white noise was investigated in rats under three training procedures employing a conditioned emotional response (CER) technique. When a CER was initially established to the CS+ and then the CS- was introduced, clear generalization of the CER from CS-b to CS-- was observed at the beginning of differentiation learning. When both CSi were simultaneously introduced from the beginning of training, subjects acquired the CER to both intensities. Only then, did differential response to intensities occur as an effect of CER extinction to the CS-. Habituation of the white ncise intensity used during differentiation learning as the CS+ exerted only a small effect on the course of training. Under all training procedures differentiation of the 50-dB CS+ and 70-dB CS- was more difficult than between' the 70-dB CS+ and 50-dB CS-. This finding was related to both slower. acquisition of the CER when the less intense CS was paired with shock and to a marked decrease of suppressive properties of the weak CS+ together with CER extinction to the more intense CS-. Even by the end of differentiation learning intensity relations between CSi during criterion sessions exerted a strong effect on the magnitudes of suppression ratios.

INTRODUCTION

Within stimulus intensity dynamism theory Hull (5) postulated that, independent of the discriminability of the stimuli, the difficulty of differentiation learning is dependent on the intensity relationship between

428 E. JAKUBOWSKA AND K. ZIELIRSKI

the pair of conditioned stimuli 1. Using Hovland's (4) data, Hull predicted greater generalization of response strength from the weak CSf to the more intense CS- and less generalization from the intense CST to the weaker CS-, resulting in more difficult differentiation learning in the former situation. At that time the only experimental data supporting Eiull's prediction were obtained by Antoinetti (2) on brigthness differentiation. In a number of subsequent studies both on brightness (18, 19) and on auditory intensity differentiation (3, 12, 15, 17, 22) it has been shown that g-no go differentiation training is easier, and the final difference in the probability of conditioned responses on positive and on negative trials greater, when the more intense stimulus from the pair is used as the CSf. Similar differences in differentiation learning, de- pending on the choice of CS' were also observed, when discriminative stimuli differed in (i) the proportions of noise to silence within a fixed repeating cycle ( l l ) , (ii) the stimulus quality within the same modality (click-tone differentiation; 7, 8, 25), or (iii) modality itself (light-noise differentiation; 16). Differentiation learning has been clearly facilitated when a stimulus that has been confirmed by a between-% comparison as more effective for acquisition of the strong conditioned response (CR) is used for CS+, and less effective stimulus as CS- (7, 8, 16, 22, 25). However, it should be noted that in a few experiments the "relative intensity" of the discriminative stimuli exerted no effect on the course of differentiation learning (13, 14, 21).

Stimulus intensity (or quality) effects on differentiation le,arning were investigated using two training procedures. Most often the CR has been initially conditioned to the CSf and only 'then were negative trials introduced and differentiation training initiated. In other experiments both the CS+ and CS- were presented from the very beginning, and differentiation training was not preceded by CR pretraining. On the basis of some theoretical considerations it may be expected that the amount of generalization of the CR from the CS+ to the CS- will be greater when the former procedure is used. According to Konorski (6, p. 346- 347), such generalization emerges from the cruder elements of CS+ take part in the original conditioning, since precise examination of the stimulus to-be-conditioned is not necessary during a single stimulus train-

' In the go-no go differentiation task two conditioned stimuli (CSi) are presented in random succession or in predetermined order on separate trials. When classically conditioned reflexes are studied, one CS is always followed by the reinforcing unconditioned stimulus (US) and the second CS is never paired with the US. The conditioned stimulus is called positive (CSf) if it is associated with the presentation of the US, independent of whether the "reinforcing" agent is attractive or aversive. The other conditioned stimulus is termed negative (CS-).

DIFFERENTLATION LEARNLNG AND STIMULUS INTENSITY -1 ?!I

Tng procedure. Accordingly, there is a high probability that the set of stimulus elements, used as a cue in original conditioning, would also be attributed to the CS- introduced later, resulting in a strong generalization. Only then, as a result of the total perceptual reorganization of both stimuli is generalization of the CR to the CS- gradually dimi- nished. In the course of differentiation learning, the attention of the subject is shifted from crude to specific elements of the stimulus. On the contrary, Konorski (6, p. 347-348) expected very little generalization in the case when both discriminative stimuli are simultaneously introduced into training. He assumed that with such a procedure, sets of elements common to both stimuli are of no use for training since they form neural connections with both US and no-US centers. Consequently, the ,animal should attend only to the specific aspects of each stimulus from the beginning of training.

Some experimental data cited by Konorski supported these pre- dictions, but direct comparisons of the course of differentiation learning in the two training procedures have not been reported. The major aim of the present study was to compare the course of differentiation learning and the amount of generalization between CS+ and CS- in the two training procedures - with or without conditioning of the CR to the CS+ before differentiation learning. To control for the possible effects of familiarity with the to-be-CS+, an additional training procedure was introduced in which the stimulus value was repeatedly presented but without US pairing before differentiation learning started.

MATERIAL AND METHODS

The Ss were 38 experimentally naive, male hooded rats, approxima- tely 3 mo. old at the beginning of the experiment. The apparatus consisted of four modified Skinner boxes with an electrifiable grid floor, a single bar on one of the walls, and a food-tray under it. The additional equipment providing for automatic programming and recording of the experiment was in an adjoining room. The CSi were white noise of 50- or 70-dB intensity and 3 min duration, delivered to the experimental box from a Grasson-Stadler Noise Genenator (Model 901A) via a loud- speaker placed below the floor of each box. The US was scrambled electric shock of 1 mA intensity and 0.5 s duration provided by a shock generator.

Prior to any training the rats were reduced to 75010 of ad lib. body weight and maintained a t that weight on a 24 h feeding rhythm during the entire experiment. Daily portions of food were given immediately after each experimental sesion which lasted 2 h and took place each day

-130 E. JAKUBOWSKA AND K. ZIELINSKI

at the same time. The same preliminary training for all rats was used. The first day consisted of initial presentiation of 40 "free" 45 mg food pellets on 1 min variable interval (VI) schedule ("magazine training"), followed immehately by a period with a continuous reinforcement schedule until 120 food pellets were delivered in a single session. Then followed five daily 2 h sessions of bar-pressing under a 2.5 rnin VI food- reinforcement schedule, resulting in acquisition of stable on-going bar- pressing behavior for food. The rats were then divided into six groups, which differed in the treatment procedures prior to differentiation training and in stimuli used as CS ' and CS- during differentiation training (Table I).

Pre-discriminative arrangements

1 Stimuli used during 1 Number of presentations of a given No differentiation stimulus during pre-differentiation training

On the Pretest day (P-day), which was the day following prelirni- nary training, those CSi that were used in subsequent stage of the experiment were presented during 3 min periods four times each without US application. The food-reinforcement schedule and occurrence of the CSi were programmed independently throughout the whole experiment. Gro- ups la and l b were submited to the differentiation training immediately after P-day. ''Habit~~ation pretraining" given to Groups 2a and 2b consisted af five additional pretest sessions in which only the white noise intensity that served further. as CS+ was presented. In Groups 3a and 3b CER pretraining began the day after P-day and continued for 5 days. During each daily session the 70 dB white noise in Group 3a and 50 dB in Group 3b was presented four times respectively during 3 min periods, the last 0.5 s of which coincided with 1 mA shock.

Differentiation training was the same in all groups a and in all groups b. For groups a termination of the 70 dB white noise CS coincided with shock application and the 50 dB white noise stimulus was not accompanied by the nociceptive stimulus, whereas for groups b these

DIFFERENTIATION LEARNING AND STIMULUS INTENSITY 431

relations were reversed. Each CS was presented four times during the 2 h session with variable intertrial intervals. Stimuli onsets were a t 18.5, 29.0, 43.0, 53.5, 61.0, 74.5, 92.0 and 106.0 min from the beginning of the session. Positive and negative stimuli were presented according to one of the following six stimuli arrangements:

During the first six consecutive differentiation sessions stimuli arrangements were applied in the same order from 1 to 6 for all rats and, if necessary, the stimuli arrangements were repeated.

The magnitude of the CER was measured by computing the "supression ratio" described by Annau and Kamin (1). The ratio is B/(A + B), where B represents number of bar-presses emitted during the 3-min action of CS and A number of responses during the 3-min immediately before the CS-onset. The criterion used as an index of the differentiation was three consecutive days without overlapping in suppression ratios between positive and negative trials for each session. In such a case, if suppression ratios for four positive and four negative trials are considered independent "samples", the Mann-Whitney U statistic (20) has a value of 0 and the probability of occurrence of such behavior in a single session by chance is as small as P < 0.014 (one-tailed test). Differen- tiation training lasted untii a given rat reached criterion, but not longer than 35 sessions. Additionally, daily suppresion ratios were computed for each rat by summating responses emitted during the appropriate intervals each of four positive and four negative trials.

RESULTS

Training prior to differentiation learning

The first presentation of the white noise on P-day interfered with bar-press responding, although neither 70- nor 50-dB stimuli produced statistically significant reduction of on-going behavior, when compared with the 3-min period preceeding the first pretest trial (Wilcoxon mat- ched-pairs, two-tailed test, Groups la , 2a and 3a similarly combined as Groups lb, 2b, and 3b). In the next pretest presentations of the 70- or 50-dB white noise intensities, the tendency to increase bar-pressing rate during stimulus ,action was observed (see Table 11), although this accele-

432 E. JAKUBOWSKA AND K. ZIELINSKI

ration of responding during the future CS action reached statistical significance only on the 4th pretest trial with the 50-dB stimulus (P < < 0.05, Wilcoxon test, two-tailed).

In Groups 2a and 2b, that underwent habituation pretraining, mean and median daily suppression ratios in all pretraining sessions slightly exceeded the value of 0.50 (see Fig. 1). Comparison of the numbers of bar-presses emitted before and during the action of the 50- or 70-dB white noise showed that this increase in response rate was significant, F (1, 11) = 6.40, P < 0.05. All other sources of variation and their inter- actions did not reach an acceptable level of significance (mixed design, type VI, Lindquist, 10).

T A B L E 11

Median differences in bar-pressing rate between the 3-min period before and 3-min period during the stimulus action on the Pretest Day (A-B index). Groups 2a and 3a were

combined similarly as Groups 2b and 3b

As seen from the Fig. 1, CER acquisition was more rapid with the more intense CS. Comparisons of the rates of responding before and during each CS presentation by the Wilcoxon test showed that the initially significant suppression of on-going bar-press behavior during the CS, occurred in both groups on the 7th trial (2nd day of CER pretraining). However, when daily suppression ratios were compared, by the 3rd day of CER acquisition there was significantly greater conditioned suppression observed in the group trained with the 70-dB CS than in the group trained with the 50-dB CS (P < 0.013, Mann-Whitney, two-tailed test). On the 4th and 5th CER days daily suppression ratios were nearly at the same level in both groups.

Between the 1st and the 3rd days of CER pretraining in both groups, there was a significant decrease of the base-line bar-pressing rate (Wil- coxon test). The bar-pressing rate in the intertrial intervals increased slightly during the two last days of CER acquisition, but these changes were not significant when compared with the base-line bar'-pressing rate observed on 3rd CER day (the same test; see Fig. 2).

White noise intensity

Consecutive presentations - _-

(dB) 1 1 1 2 1 3 1 4 . - . p p p - - -

I I 70 1 0 I -1.0 / -3.0 - 3 . 5 50 , 6.0 ' 2.5 ,

- -

6.5 - 6.5 -

DIFFERENTIATION LEARNING AND STIMULUS INTENSITY

7 0 dB and US

5 0 dB and US

0 7 0 d B no US

day Pre t ra in lng D i f f e r e n t i a t i o n t r a i n i n g

Fig. 1. Median daily supression rations during P-day, pre-differentiation training and differentiation learning in Groups la, l b (upper part), Groups 2a, 2b (middle part), and Groups 3a, 3b (at the bottom), until the day on which the first rat in

a given group reached the differentiation criterion.

E. JAKUBOWSKA AND K. ZIELINSKI

- Groups a

--- G r o u p s b

Fig. 2. Median number of bar-press responses per minute emitted in pre-CS periods during P-day, pre-differentiation training and differentiation learning in Groups la, l b (upper part), Groups 2a, 2b (middle part), and Groups 3a, 3b (at the bottom), until the day on which the first rat in a given group reached the

differentiation criterion.

DIFFERENTIATION LEARNING AND STIMULUS INTENSITY -1 3 5

Rate of differentiation learning and final level of responding

All rats except one subject each from Groups l b and 2b reached the criterion of differentiation learning within 35 days of training. An additional index of differentiation learning rate included the number of sessions of differentiation training until no overlap, in CER ratios between pcsitive and negative trials for a single session occurred. To enable between-groups comparisons of the number of session to the differentiation criterion, the maximum length of training, 35 sessions, was taken as a measure for those r,ats that did not stabilized their discriminative behavior for the three consecutive days. Group means of these two indices of differentiation learning rates are given in Table 111.

Mean (and range) numbers of sessions of differentiation to the first day in which no overlap in CER ratios between positive and negative trials occurred (index A) and until the criterion of three consecutive days without overlap

was reached (index B)

Groups 1 Index A 1 Index B

1 a 7.3 (4-1 3) 12.3 (6-26) I b 14.2 (6-26) 22.7 (1 5-35) 2a 7 2 (5-1 1) 10.5 (7-1 5) 2b 12.6 (5 20) 20.3 (1 1-35) 3 a I 3.7(1-8) 9.1 (4-1 8) 3b

-- --

7.6 (3-15) 15.3 (I 1-21)

Source of I df I Values of F statlstlcs varlatlon I I - - - --

Procedures / 2, 32 4 26' 2 13 S t ~ r n u l ~ 1, 32 1 11.92** 18 02*** Lnteract~on 2, 32 1 0.53

-- 0 68

--

The analysis of variance indicates that the rate of differentiation learning was strongly affected by intensity relations between the CSi. In all training procedures, the differentiation of the 70-dB CS+ and 50-dB CS- was more rapidly acquired than the differentiation of the 50-dB CS+ and 70 dB CS-. Only a small part of the variability was controlled by the training procedures. EIowever, the length of differentiation learning until the first day with no overlap in CER ratios on positive and negative trials was significantly affected by the training procedures, F (2, 32) = 4.26, P <. 0.025. This finding was du.e to the rapid learning of differentiation after CER pretraining.

436 E. JAKUBOWSKA AND K. ZIELII(ISKI

For the 36 rats that attained the required criterion of differentiation, the magnitudes of suppression ratios during the three criterion sessions were compared independently for positive and negative trials. As seen from Table IV, suppression of bar-press responding on positive

Mean daily suppression ratios during three consecutive days without overlap on positive and on negative trials

Groups 1 Poslt~ve t r~als Negative tr~als

.- -- ; D a y 1 -- D a y 2 D a y 3 , D a y l / D a y 2 D a y 3 - - -- -

I

l a 0 . 1 3 9 0.140 0.123 0.388 0 .4470 .466 Ib 0.277 0.232 0.245 0.509 0.510 0.519 2a 0.029 0.010 0.020 0.439 0.429 0.444 2b 0.125 0.108 0.11 3 0.497 0.526 0.558 3a 0.055 ' 0.068 I 0.091 0.415 0 435 1 0.479 3b 1 .

-- - - C.201 0.203 0.209 0.564 0.539 0.542

~ ~p

Source of variation df / Values of F statistics ~ ~ . - ~ - - - -- - -

Procedures B 2, 30 ' 5.58* * 0.94 Stimuli C 1 1, 3 0 , 15.03** * 26.91 * * * BC 2, 30 0.51 1

I 0.10

Days A 2, 60 0.24 i

4.58* AB 4, 60 0.47 0.55 AC 2, 60 0.18 1.34 ABC 1 4, 60 0.22

- - - - - - - - - . . 1 1.79

trials was more pronounced in rats trained with the 70-dB CS+ than the 50-dB CS+ for all training procedures. The CS+ evoked the least suppression in groups which learned differentiation immediately after P-day, which resulted in a significant effect from the procedure variable on the magnitude of suppression ratio for positive trials, F (2, 30) = 5.58, P < 0.01. It is interesting to note that in spite of similarities between Groups la and 2a as well as between Groups l b and 2b during the course (discussed below) and for the length (Table 111) of differentiation learning, the CER response on positive trials was stronger by the end of learning in those rats which had prolonged perceptual experience with the future CS+. Daily suppression ratios to the 70-dB CS+ were significantly less in Group 2a than in Group la, and also less to the 50-dB CS+ in Group 2b than in Group l b on each criterion sessions (P < 0.05, Mann-Whitney test).

Group means cf suppression ratios to the 50-dB CS- were lower than 0.5 in all cases, which shows that bar-pressing was slightly sup-

DIFFERENTIATION LEARNING AND STIMULUS INTENSITY 43 '7

pressed on negative trials in Groups la, 2a, and 3a, whereas suppression ratios to the 70 dB CS- were greater than 0.5 in 8 cases out of 9 indi- cating some enhancement of responding on negative trials in Groups lb, 2b, and 3b. A trend of decreasing suppression and increasing of enhancement of bar-pressing during the CS- presentations was observed over consecutive days, resulting in a significant effect of days, P (2, 60) = 4.58, P < 0.025.

The course of differentiation learning

The differences between groups observed during criterion sessions were more pronounced a t the beginning of differentiation learning than later. In Fig. 1 the median daily suppression ratios and ia Fig. 2 the medan numbers of bar-presses during the 3-min periods before CS action (baseline level of responding) are presented until the day on which the first rat in a given group reached the criterion of 3 days with no overlap on positive and on negative trials. Suppression ratios for the first 4 days of differentiation training were subjected to analyses of variance (mixed design type 111, 10) and the results are given in Table V. Additionally the numbers of bar-presses before and during each presentation of the CS+ or CS- were compared, and the significance of the Wilcoxon's T

Mean daily suppression ratios on positive and on negative trials during the first 4 days of differentiation learning

1 Positive trials 1 Negative trials Groups 1st , 2nd 3rd I 4th 1st 2nd 3rd 4th

day day day I day day / day / day I day

I a 1 0.550 I b 1 0.528 2a 0.529 2b I 0.509 3a 0.059 3b 0.138

Source of variatron df I -- Procedures B Stimuli C BC Days .A AB AC A BC

0.294 I 0.027 0.057 0.508 1 0.319 1 0.078 0 . 6 0.350 0.338 0.526 1 0.482 , 0.410 0.378 1 0.038 1 0.074 0.379 ' 0.332 0.085 0.444 1 0.311 : 0.315 0.487 I 0.468 0.332 0.058 0.060 I 0.078 0.138 0.254 0.301 0.212 0.219 0.334 ' 0.193 0.304 1 0.425

- - -- - - - - - Values of F statistics

438 E. JAKUBOWSKA AND K . ZIELIRSKI

statistics estimated separately for each CS presentation and group is shown in Fig. 3. These data provide information on the suppressive properties of the CSi a t the beginning of differentiation learning.

Let us analyse the course of differentiation learning for each training procedure.

C 9 Group la .

CS-

Group 1 b =.+ n CS-

--

Group Pa CS+ I Cs- ,nm, I I I I I I I I I I ~ V

Group 2b : ~ . + - . " ; L ~ ~ T ~ # ~ I ~ T ~ I I

Group 3a C s+ CS-

CS+ Group 3b

CS -

Fig. 3. Significance of conditioned suppression (tested by Wilcoxon one-tailed test) on positive (upper rows) and negative trials (lower rows) during the last pretraining day and in the course of differentiation learning until the day on which the first ra t in a given group reached differentiation criterion. Rectangle denotes a significant change in bar-pressing rate during a given CS- ,;r CS-

presentation a t P < 0.031. When differentiation learning began immediately ~ f t e r P-day (Groups

la and lb), the rats did not discriminate the conditioned stimuli at the beginning of learning. The intensity of the stimulus paired with shock was responsible for rapid (Group la) or slow (Group lb) acquisition of suppression both on positive and negative trials (Fig. 1, Table V). As seen from Fig. 3 the 70-dB CSr evoked significant suppression of the on-going bar-pressing behavior on the 6th, 8th, 9th, 11th and on all next CER trials, whereas the 50-dB CS' resulted in somewhat consistent CER responding only by the 8th day of differentiation learnin%. The first significant differences between Groups la and l b in the amount -

of conditioned suppression on positive as well as negative trials occurred as early as the 2nd day of differentiation training (h'iann-Whitney test). These group differences were largest on the 3rd day of differentiation learning and further decreased, but were observed until the end of training. There was a clear tendency for more rapid differentiation of the 70-dB CS t and 50-dB CS- than the 50-dB CS+ and 70-dB CST. The

DIFFERENTIATION LEARNING AND STIMULUS INTENSITY .430

first significant difference between daily suppression ratios on the positive and negative trials (group data) occurred on the 5th day of differeri- tiation training in Group la, whereas in Group lb it was found only on the 12th differentiation session. Before differential behavior developed, alternating periods of increasing and decreasing conditioned suppression elicited by both CSi were noted for both groups. These fluctuations were clearly visible in the learning curves of individual rats and were more pronounced in Group lb.

At the beginning of differenti,ation learning a marked decrease of the on-going bar-press response rate was observed, and Groups la and l b did not differ in this respect in spite of clear differences in the amount of suppression evoked by the sporadic stimuli (Fig. 2). Partial restoration of the base-line response rate in Group l b was observed concurently with acquisition of the differentiation between CS+ and CS-.

The course of differentiation in rats given "habituation pretraining" (Groups 2a and 2b) was very similar to that observed in Groups la and lb. At the beginning of the differentiation learning no discrimination of the CS+ and the CS- was observed and the amount of conditioned suppression depended on the strength of the CSf. In Group 2a con&- tioned stimuli started to elicit the suppression response more consisten- tly by the 3rd differentiation session. However in Group 2b systematic suppressive responding began only with the 6th day of differentiation, which was slightly earlier than in Group l b (Fig. 3). Before differential behavior developed, paralleled fluctuations in the amount of suppression to CS+ and CS- were observed in both groups. The first significant difference between daily suppression ratios on positive and negative trials occurred on the 5th day of differentiation training in Group 2a, and on the 7th session in Group 2b (Mann-Whitney test). Similarly in Groups 2a and 2b as in Groups l a and l b a marked decrease of the base-line response rate at the beginning of differentiation learning was observed.

The course of differentiation learning for groups in which CER to the CS+ was first established and then negative trials were introduced, differed in many respects from those observed under the other two training procedures. As seen from Fig. 3, the first day of differentiation learning showed significant suppression to the CS- on each trial both in Groups 3a and 3b. Rapid extinction of conditioned suppression to the negative CSi then occurred (Fig. I), however, in Group 3b it was accompanied by marked decrease of suppressive properties of the 50-dB CS+ (Table V). The first significant difference between daily suppression ratios on positive and negative trials occurred during the 3rd day of differentiation in both Groups 3a and 3b. The rapid extinction of con-

3 - Acta Neurobiologiae Experimental is

ditioned suppression to the CS- in Group 3b, together with a concurrent decrease of suppression to CS;, resulted in significantly different suppression ratios between groups for positive trials beginning at the 3rd differentiation session. No such differences were observed on the previous days. These group differences in the amount of suppression, both cn positive and negative trials, remained significant even during criterion sessions (Table IV). For neither group did the introduction of CS- produce changes in base-line response rate (Fig. 2).

Data presented in Table V indicate that the significant effect of the training procedures on suppression ratios during positive trials was due to low values in Groups 3a and 3b, which acquired the CER before the differentiation learning. In contrast to all others, these groups showed weakening of the suppression over the first 4 days of differentiation learning resulting in a significant days vs. procedures interaction both on positive and negative trials. From the very beginning of learning intensity relations between CS'I- and CS- exerted a very strong effect on the behavior of rats both on positive and negative trials.

DISCUSSION

In all theories of differentiation learning the discriminability of the conditioned stimuli is considered the main factor responsible for the difficulty of training. However, it may be expected that some other experimental variables influence the acquisition of differential responding to the two external stimuli, one of which is paired with the unconditioned stimulus while the other is not. According to Konorski '(G), introduction of both CS+ and CS- from the very beginning of training prevents generalization of the acquired response and results in more successful differentiation training than in a procedure in which the response is first conditioned to the CS+ and the CS-- is introduced only afterwards. It is important to add that these differences are expected to occur at the beginning of differentiation learning (6, p. 346-348).

Data presented in this paper for all training conditions indicated that generalization of the suppressive responding occurred from the CS+ to the CS- and also to the intertrial interval stimulus situation (ITS). Let us consider the course of differentiation of the 50-dB CS+ and the 70-dB CS- when both of these stimuli were introduced at the same time (Group lb). At the beginning of training a decrease of the on-going bar- pressing rate was observed. Then significant response suppression during the white noise presentations occurred, which was accompanied by a gradual restoration of bar-pressing for food during intertrial intervals. Only afterwards was differential responding to the CS+ and the CS-

DIFFERENTIATION LEARNING AND STIMULUS INTENSITY 411

observed. A similar sequence of changes in the rats' behavior was also evident in other groups in which the CER was not established before introduction of negative trials. The decrease in base-line bar-pressing earlier than the significant suppression to the sporadic stimuli gives support to some extent for Lashley's hypothesis (9) that conditioning consists in the successive establishment of new associations with stimuli (or aspects of stimuh) not noticed on earlier stages of training. Further, restoration of the base-line responding, the strengthening of conditioned suppression during CS+ and weakening of the CER during CS- are in agreement with Pavlovian theory of differentiation. This assumes that a wide range of associations established at the beginning of training as a result of stimulus generalization is narrowing during the course of differentiation learning. Theoretical considerations developed on the basis of experimental data obtained by the traditional procedure employing conditioning of the response to the CS+ before the introduction of the CS-, are applicable to the procedure in which positive and negative trials were presented from the very beginning of differentiation learning.

The lack of differences in responding to the 50-dB and 70-dB noise intensities observed a t early stages of differentiation learning in Group la and also in Group l b seems to indicate that in this training procedure rats initially learned to react to the common element of both stimuli. This common element was the onset of the white noise independent of its intensity. However, the absolute intensity of the CSi exerted an effect on the course of le,arning from the very beginning of training. Despite nearly the same magnitudes of CER rates on CS+ and CS- presentations observed in Groups la and lb, by the 2nd day of differentiation learning there were marked between-group differences observed in the amount of suppression evoked by sporadic stimuli (P < 0.01, Mann-Whitney test). Also in the Groups 2a and 2b for which preliminary shaping of the CER to the CS; was not used a t the beginning of differentiation learning a lack of differential responding to the 50- and 70-dB white noise intensities was observed together with marked between-groups differences in daily CER ratios. These were statistically significant for both CSi by the beginning of the 3rd day of learning (Mann-Whitney test). In both training procedures the strength of the CER was greater when the more intense stimulus (i.e., 70-dB) was paired with shock. Collectively, these findings seem to indicate that in rats given no pre-differentiation CER training it was not the intensity of white noise but the increase in noise level at the onset of the sporadic stimuli that possed signalling properties at the beginning of differentiation training. Further, the absolute intensity of the CSi affected the

rapidity of learning and the strength of the CER evoked by the onset of both intensities of the white noise. It should be recalled that a similar distinction between signalling and non-specific arousal properties of CSi was proposed in earlier studies of the discrimination between two intensities of white noise, when one CS was of higher and the other CS of lower intensity than the white noise presented during intertrial intervals (23, 24). In these experiments employing a CER technique it was observed that an increase of the white noise intensity evoked one mode of behavior and a decrease of white noise intensity resulted in another mode of behavior, that w.as independent of the absolute intensity of noise before and during the CS action. However, the elicited response was performed with greater magnitude when the absolute intensity of the CS was greater.

In the present study the absolute intensity of the CS paired with shock influenced not only the performance of the CER response, but the process of acquisition of the CER response. This statement is document- ed by the observation of clear between-groups differences in CER ratios within the same training procedure much earlier than the emer- gence of differential responding to the two intensities of white noise in each group. At this stage of learning rats mastered only the differentiation between the background stimuli situation and the sporadic stimuli, so that onset of the white noise was a signal for freezing behavior and offset of the noise was a signal for bar-pressing responses. Only on the next stage of differentiation learning did the rats acquire differential responses to the CS+ and the CS-. Signalling properties of the onset per se had to be extinquished and either the particular amount of change in noise level caused by the CS+ onset or the definite absolute intensity of the CS+ became a cue for the fear response.

Extinction of the CER to the 50-dB CS- produced little or no changes in the amount of suppression on positive trials, whereas extinction of the CER to the 70-dB CS- was related to a weakening of suppressive properties of the 50-dB CS+. This was clearly observed in Group 2b and very pronounced in Group 3b. Changes in CER ratios on positive trials were related to the acquisition of inhibitory properties by the 70-dB CS- and may be considered as generalization of the more intense CS- to the less intense CS+.

In Group l b the influence of the 70-dB CS- on the suppressive properties of the 50-dB CS+ was manifested in the pronounced retar- dation of CER acquisition. Comparison of Group l b and Group 3b indicated that acquisition of conditioned suppression to the 50-dB noise- plus-shock and 70-dB noise without shock was much slower and the final level of the CER much weaker than the conditioned suppression

DIFFERENTIATION LEARNLNG AND STIMULUS INTENSITY 443

,acquired during the CER pretraining stage to the 50-dB CS+. On the other hand, the conditioning of suppression to the 70-dB noise-with-shock and the 50-dB noise without shock (Group la) was as rapid as the conditioning to the 70-dB CSI. during the CER pretraining stage in Group 3a, and the asymptotic levels of suppression were similar in both groups.

Thus, there are two inferences that may account for more difficult differentiation of the weaker CSI- and stronger CS- rather than vice versa. The first, which was operational in the stage of differentiation between sporadic stimuli and the ITS, is characterized by slow conditioning of the CER to white noise presentations when the weaker CS is paired with shock. The second, clearly observed a t the stage of differentiation of the tw'o whlte noise intensities, involves a strong generalization of the inhibitory properties of the more intense CS- to the less intence CSt, resulting in a marked decrease of the suppression ratios on positive trials. The first inference was especially evident when CS+ and CS- were introduced at the same time (Group lb), and the second was prevalent when differentiation learning was preceded by conditioning of the CER to the CS" (Group 3b).

Habituation of the to-be-CS+ before differentiation learning exter- ted no effect on the course of learning when the CS' was more intense than the CS- (cf. Groups 2a and la) and slightly shortened acquisition of the CER response when intensity relations were reserved (cf. Groups 2b nad lb). I t may be hypothesized that during such pretraining extinction of the orienting response to the onset of while noise occurred and then, during differentiation learning, animals did not attend to this aspect of the CSi, which caused less generalization of the acquired responses. Even during criterion sessions, suppression ratios on positive trials were less in groups trained under this procedure than in others, whch resulted in the greatest difference in suppression r'atios on positive and negative trials in Groups 2a and 2b.

All of the above considerations indicate that differential responding to two discriminable stimuli is a complex task which is mastered in several stages. Intensity relations between stimuli exerted a strong effect on each stage of differentiation learning. When the more intense CS- and less intense CSf were used, not only the generalization of CER from CS- to CS-, but also the generalization of the extinction from CS- to CS+ are observed. Under different training procedures one or another process is more pronounced, however in e'ach case, generalization of responses conditioned to different stimuli is evident.

REFERENCES

1. ANNAU, 2. and KAMIN, L. J. 1961. The conditioned emotional response as a function of intensity of the US. J. Comp. Physiol. Psychol. 54: 428-232.

2. ANTOINETTI, J. A. 1950. The effect of discrimination training upon generalization (unpublished), Quoted, Hull (5).

3. BLUE, S. 1967. Stimulus intensity dynamism during two-valued auditory discrimination acquisition: a systematic replication. Psychon. Sci. 8: 29-30.

4. HOVLAND, C. I. 1937. The generalization of conditioned responses: 11. The sensory generalization of conditioned response with varying intensities of tone. J. Genet. Psychol. 51: 279-291.

5. HULL, C.L. 1952. A behavior system. New Haven: Yale Univ. Press. 372 p. 6. KONORSKI, J. 1967. Integrative activity of the brain. The interdisciplinary

approach. Univ. Chicago Press, Chicago. 531 p. 7. KOWALSKA, D., DABROWSKA, J. and ZIELINSKI, K. 1975. Effects of par-

tial prefrontal lesions in dogs on go-no go avoidance reflex differentiation and reversal learning. Acta Neurobiol. Exp. 35: 549-580.

8. KOWALSKA, D., DABROWSKA, .I. and ZIELINSKI, K. 1975. Retention of symmetrical go-no go avoidance differentiation after prefrontal lesions in dogs. Bull. Acad. Pol. Sci. Ser. Biol. 23: 487-494.

9. LASHLEY, K.S. 1942. An examination of the "continuity theory" as applied to discriminative learning. J. Genet. Psychol. 26: 241-265.

10. LINDQUIST, E. F. 1953. Design and analysis of experiments in psychology and education. Riverside Press. Cambridge. 393 p.

11. MacMILLAN, A. S. C., GRAY, I. A. and ISON, I. R. 1973. An apparent new in- stance of stimulus intensity dynamism during discrimination of duration of repeating auditory stimuli. Quart. J. Exp. Psychol. 25: 62-70.

12. 1LIOORE, J. W. 1964. Differential eyelid conditioning as a function of the frequency and intensity of auditory CS. J. Exp. Psychol. 68: 250-259.

13. PIERREL, R. 1958. A generalization gradients for auditory intensity in the rat. J. Exp. Anal. Behav. 1: 303-313.

14. PIERREL, R. and SHERMAN, I. G. 1960. Generalization of auditory intensity following discrimination training. J. Exp. Anal. Behav. 3: 313-322.

15. PIERREL, R., SHERMAN, I. G., BLUE, S. and HEGGE, F. W. 1970. Auditory discrimination: a three variable analysis of intensity effects. J. Exp. Anal. Behav. 13: 17-35.

16. ROSIC, N. and BIGNAMI, G. 1970. Scopolamine effects on go-no go avoidance discriminations: Influence of stimulus factors and primacy of training. Psychopharmacologia 17: 203-213.

17. SADOWSKY, S. 1966. Discrimination learning as a function of stimulus loca- tion along an auditory continuum. J. Kxp. Anal. Behav. 9: 219-225.

18. SHEVELEV, I.A. 1961. Effects of physical strength and signalling values of the visual conditioned stimulus on parameters of motor response in man (in Russian). Tr. Inst. Vyssh. Nerv. Deyat. 6: 12-23.

19. SHEVELEV, I. A. 1963. Differentiation of visual stimuli according to brightness with opposite change of their intensity and signalling value (in Russian). In E.A. Asratian (ed.), Nervnye mekhanizmy uslovnoreflektornoP deyatel'- nosti. Izdat. AN SSSR, Moscow, p. 41-48.

20. SIEGEL, S. 1956. Nonparametric statistics for behavioral sciences. McGraw-Hill Book Co., New York. 312 p.

DIFFERENTIATION LEARNLUG AND STIMULUS INTENSITY 445

21. SMITH, M. q. Jr., HAY, W. J. 1954. Rate of response during operant discrimination. J. Exp. Psychol. 48: 259-264.

22. ZIELINSKI, K. 1965. The influence of stimulus intensity on the efficacy of reinforcement in differentiation training. Acta Biol. Exp. 25: 317-335.

23. ZIELINSI~I, K. 1565. The direction o i change versus the absolute level of noise intensity as a cue in the CER situation. Acta Biol. Exp. 25: 337-357.

24. ZIELINSKI, K. 1966. Direction of change of the background noise intensity as a cue for the reflex based on alimentary - defensive antagonism. Bio- logical and physiological problems of psychology. Proc. XVIII Int. Congr. Psychol. (Moscow) vol. 1. (Abstr.).

25. ZIELIRSKI, K., CZARKOWSKA, J. 1973. Go-no go avoidance reflex differentiation and its retention after prefrontal lesion in cats. Acta Neurobiol. Exp. 33: 467490.

Accepted 19 March 1976

Ewa JAKUBOWSKA and Kazimierz ZIELI~JSKI, Nencki Institute of Experimental Blology, Pasteura 3, 02-093 Warsaw, Poland.

DIFFERENTIATION LEARNING AS A FUNCTION · 428 E. JAKUBOWSKA AND K. ZIELIRSKI the pair of...

Documents

Transcript of DIFFERENTIATION LEARNING AS A FUNCTION · 428 E. JAKUBOWSKA AND K. ZIELIRSKI the pair of...