Stimulus Generalization along the Light Intensity Dimension as a Function of Discrimination Training

Transcription

1 Western Michigan University ScholarWorks at WMU Master's Theses Graduate College Stimulus Generalization along the Light Intensity Dimension as a Function of Discrimination Training Marilla D. Svinicki Western Michigan University Follow this and additional works at: Part of the Psychology Commons Recommended Citation Svinicki, Marilla D., "Stimulus Generalization along the Light Intensity Dimension as a Function of Discrimination Training" (1968). Master's Theses This Masters Thesis-Open Access is brought to you for free and open access by the Graduate College at ScholarWorks at WMU. It has been accepted for inclusion in Master's Theses by an authorized administrator of ScholarWorks at WMU. For more information, please contact maira.bundza@wmich.edu.

2 STIMULUS GENERALIZATION ALONG THE LIGHT INTENSITY DIMENSION AS A FUNCTION OF DISCRIMINATION TRAINING by Mari I I a D:.' Svi n i cki A Thesis Submitted to the Faculty of the School of Graduate Studies in p a rtia l fu lfillm e n t of the Degree of Master of Arts Western Michigan U niversity Kalamazoo, Michigan July 1968

3 ACKNOWLEDGEMENTS The encouragement and assistance of Dr. Richard W. M alott and Marilyn K. Malott in the planning and completion of th is study is g ra te fu lly acknowledged. The in te lle c tu a l stim ulation which resulted from our interactions is immeasurable. Their e ffo rts in the preparation of the fin a l manuscript have been invaluable. I wish also to thank Dr. John Michael and Dr. Roger U lrich fo r th e ir c r it ic a l comments and encouragements in the completion of the paper. F in a lly, I want to thank Penny Z lutnick fo r her e ffo rts in the production of the fin a l typed copy. Mari I la D. Svin icki

4 M A STER S THESIS M SVINICKI, M a rilla Doreen STIMULUS G ENERALIZATIO N ALONG THE LIG H T IN T E N S ITY DIMENSION AS A FUNCTION OF DISCRIM INATIO N TRAINING. Western Michigan U niversity, M.A., 1968 Psychology, experim ental U n iv e rsity M icrofilm s, In c., A n n A rb or, M ichigan

5 TABLE OF CONTENTS PAGE ACKNOWLEDGEMENTS... INDEX OF FIGURES... INDEX OF T A B L E S... i i iv v INTRODUCTI O N... 1 METHOD... 5 Subjects... 5 A p p a ra tu s... 5 Procedure... 6 RESULTS... 9 DISCUSSION CONCLUS I O N REFERENCES...24

6 INDEX OF FIGURES Fi qure 1 Changes in response rates as a function of treatments. This fig u re shows daily response rates (responses per minute) fo r each phase of the study. Circles represent responses made in the presence of the positive stimulus while tria n g le s represent responses made in the presence of the negative stimulus. Dashed lines represent mean response rates fo r each period and fo r S+ and S-. 2 Discrim ination ra tio changes as a function of sessions and treatments. 3 A comparison of generalization gradients obtained following (1) single stimulus tra in in g (solid lin e ), (2) discrim ination 1 (tria n g le s with dotted lin e ), and (3) d iscrim ination 2 (squares with dashed line). The in te n s itie s are: -6 = 0.00 fc; -5 = 0.03 fc (S-- in te n s ity ); -4 = 0.06 fc ; -3 = 0.11 fc (S~ 2 in te n s ity ); -2 = 0.23 fc; fc; 0 = 0.88 fc (St in te n s ity ); t l = 1.75 fc; +2 = 3.5 fc; and t3 = 7.0 fc. f.ag.e i v

7 INDEX OF TABLES Tab I e 1 Maximum responses fo r each generalization gradient. Each cell contains the to ta l number of responses made by a subject at the maximum point fo r a s p e c ific g eneralization gradient. 2 Areas and medians fo r each generalization gradient. Values represent to ta l percent under each generalization gradient and the position of the median of to ta l responses fo r each gradient. Page 17 19

8 The phenomenon of a s h ift in the peak of a generalization gradient a fte r discrim ination tra in in g was o rig in a lly described by Hanson (1959). Working with the wavelength continuum, Hanson demonstrated th a t discrim ination tra in in g between two stim u li on the same continuum produced a change in the gradient as compared to the gradient th a t is produced by tra in in g on one value only. The sp e cific form of the change was a s h if t in the gradient away from the positive stimulus (S+, the stimulus associated with reinforcement) in a d irection opposite to the negative stimulus (S-, the stimulus associated with e x tin c tio n ). Hanson also found th a t the smaller the difference between the S+ and S-, the greater the s h if t in the peak. Bloomfield (1967) has observed the same phenomenon when tra in in g and testing along the line angle dimension. Malott and Malott (1967) obtained a peak s h if t while investigating the tone in te n s ity continuum. In an attempt to account fo r th is phenomenon, Hanson alluded to the explanation of a s im ila r phenomenon, tra n sp o sitio n, supplied by Spence (1937). Transposition referred to the behaviors displayed by animals trained to discrim inate between two stim uli varying along a dimension such as in te n s ity or size. For example, when trained to choose the shorter of two line lengths and tested with the o rig in al short line and one even shorter, subjects showed a preference fo r the shortest lin e. Spence attempted to explain these transposition data in terms of the in h ib ito ry gradient formed around the S- resu lting from the e xtin c tion of responding in the-presence of th a t stimulus. This gradient 1

9 2 in conjunction with the positive gradient around the S+ might then produce a s h if t in the peak or area of the composite gradient as compared to each individual gradient. This summation of gradients theory, however, was not complete enough of its e lf to account for a ll of the data which Hanson presented. Hanson found that responding during generalization tests a fte r discrim ination tra in in g was as high as responding during tests p rio r to any discrim ination tra in in g. Hanson pointed out that summation of the theoretical gradients of Spence would produce a decrement in tota l number of responses. Later in v e s tigators such as Terrace (1966) and Reynolds (1961) showed th a t the phenomenon of behavioral contrast, an increase in response rate during S+ with a corresponding decrease in the rate during S-, could supplement the notion of summation of gradients to produce the type of gradients described by Hanson. One purpose of the present study was to investigate the effects of discrim ination tra in in g along the lig h t in te n sity dimension, a continuum not previously studied with regard to these phenomena. E a rlie r studies in the area of the generalization of lig h t in te n sity have investigated the theory of stimulus in tensity dynamism as postulated by Hull (1949). Stimulus in te n sity dynamism proposes th a t, other things being equal, a subject w ill respond at a higher rate to s tim u li of greater in te n sity. According to Mednick and Freedman (1963), th is theory would p redict gradients extending from stim uli less intense than the tra in in g value to the tra in in g value to be slowly increasing over very low in te n s itie s with a rapidly accelerating rate ju s t p rio r to and including the tra in in g value. For the

10 range extending from the tra in in g value to stim uli of greater in ten sif the theory would p redict rapid acceleration of rates u n til an asymptote was reached. The most common form fo r studies of stimulus in te n sity dynamism to take was the conditioning (e ith e r operant or respondent) of a response in the presence of one stimulus value (e ith e r high or low in te n sity) on the end of a range of te s t values, and the te s tin g fo r production of th a t response by other values. The general re su lts fo r studies using a d ire c t lig h t source support H u ll's theories (Brown, Fink and Patton, 1953; F rick, 1948). More recent investigators, however, have presented other arguments to explain the phenomenon a ttrib u te d to stimulus in ten sity dynamism. Perkins (1953) and Logan (1954) have suggested that the phenomenon a ttrib u te d to stimulus in te n s ity dynamism is in re a lity a function of an in h ib ito ry response b u ilt up in the absence of the tra in in g stimulus and is s im ila r to the in h ib ito ry gradient developed during d iscrim ination tra in in g as hypothesized by Spence (1937). This absence of the tra in in g stimulus can be interpreted as lying on the in te n s ity continuum and being a stimulus of zero in te n s ity. Hence, the in h ib ito ry generalization gradient developed around the no-stimulus value (zero in te n sity) interacts with the p o sitive generalization gradient developed around the tra in in g value to produce a steepening of the lower in te n s ity side of the gradient and a re la tiv e fla tte n in g of the higher in te n s ity side. M alott and Malott (1967) have presented data supporting the position th a t stimulus inten sity dynamism can be a ttrib u te d to the e ffe cts of the Hanson peak s h ift discussed e a rlie r. Using the tone

11 in te n s ity continuum and giving tra in in g on a single in te n sity value, they found f l a t generalization gradients when te stin g over a range of in te n s itie s. However, when discrim ination tra in in g was given between tone on (60 decibels) and tone o ff (0 decibe ls), they found th a t the subjects produced what in i t i a l l y appeared to be an example of stimulus in te n s ity dynamism. Testing over a wider range on the continuum, however, revealed the presence of decreasing rates of response at very high in te n s itie s. They interpreted these data as showing the e xistence of a generalization gradient with the peak shifte d away from the zero in te n s ity. From these data one could p re dict th a t, had the studies mentioned e a r lie r in support of H u ll's theory involved testin g over a wider range of stimulus values, they would have revealed a s h ift in the peak of the generalization gradient. P ierrel and Sherman (1960) have provided data which show that rats given d iscrim ination tra in in g between a low in te n s ity tone as the S+ and a high in te n sity tone as the S- w ill produce a generalization gradient with a peak shifted in the d ire ctio n of dimmer in te n s itie s. This study lends strong support to the position th a t stimulus intens ity dynamism could be re a lly a peak s h if t phenomenon. A second purpose of the present study was to te s t with lig h t inte n s ity the theory that the e ffe cts a ttrib u te d to stimulus in te n sity dynamism could be in re a lity a s h if t in the generalization gradient as a function of tra in in g.

12 5 METHOD Subjects The subjects were three experimentally naive White King barren hen pigeons, housed in individual cages. They were fed at the end of each experimental session or once daily when sessions were not conducted so th a t at the s ta rt of each experimental session, they would be at 70$ of th e ir free-feeding weight. Purina Pigeon Grains were used to maintain weight and served as the re in fo rce r. G rit and water were continuously available in the home cage. Apparatus A Lehigh Valley Electronics pigeon te s t chamber #1519c was used with the houselight o ff and the window covered. Only one response key was operative; i t consisted of a transparent p la s tic paddle behind a hole H in. in diameter. A d iffu sin g Plexiglas plate, located d ire c tly behind the key, was tra nsii Iuminated from the rear by a 120 v o lt, 6 watt G.E. lamp. The lamp was mounted 4 in. from the rear of the key in a fla t-b la c k rectangular enclosure which directed the lig h t through a Kodak Wratten f i l t e r, series V, on to the Plexiglas plate. The in te n s ity of th is d ire c t lig h t source was measured by placing a Gossen Luna-pro exposure meter against the fro n t of the key so tha t the lig h t coming from the lig h t source f ille d the meter's aperture. The in te n s ity of the food magazine lig h t was dimmed by placing a re s is to r in series with the lig h t. The in te n s ity of the lig h t coming from the magazine was 1.2 footcandles (fc ), when measured with the exposure meter held d ire c tly in fro n t of the opening to the magazine.

13 6 White masking noise was provided by a Grason-StadIer noise generator and was presented through a speaker in the chamber. A fan provided ve n tila tio n and additional masking noise. Programming during tra in in g was done autom atically with solid state d ig ita l switching c irc u itry. Intensity changes during tra in in g were co ntro lle d by a re sisto r series and a relay system which a lte r nated between two re s is to r values to produce inten sity changes. During generalization tests, the key lig h t in tensity was manually contro lle d with an A d ju st-a -vo lt variable transformer. Time in each stim ulus, responses and reinforcements were recorded with e le c tro mechanical counters. A styrofoam p icn ic freezer, painted fla t-b la c k inside, was used as a dark-adaptation chamber. V en tila tio n was provided by a baffled fan system. Subjects were transferred from the dark-adaptation chamber to the experimental chamber in a container having a fla t-b la c k in te rio r. The room lights were o ff. Precautions were taken to avoid exposure to any lig h t during this time. Procedure P rio r to each experimental session, the subjects were placed in the dark-adaptation chamber fo r one hour. They were then transferred to the experimental chamber. Through the use of standard operant conditioning procedures, they were trained to eat from the food magazine and to peck the response key tra n s ii Iuminated with a lig h t of 0.88 fc. When the key peck response was established, the subjects received 200 reinforcements (50 each day fo r 4 sessions) on a continuous reinforcement (CRF) schedule. Reinforcement consisted of a

14 3 sec presentation of the food magazine accompanied by the magazine lig h t. The average interval between the successive periods of the a v a ila b ility of reinforcement was then increased each day according to the c r it e r ia listed below. If the subject received: 50 reinforcements in 10 min on CRF 50 reinforcements in 20 min on random interval (Rl) 8 sec 50 reinforcements in 30 min on RI 16 sec 50 reinforcements in 40 min on Rl 32 sec the schedule was increased to the next value the follow ing day u n til an Rl 64 sec schedule was reached. With the Rl 64 sec schedule, the p ro b a b ility of reinforcement fo r the f i r s t response in each successive 4 sec period was 1/16 (Farmer, 1963). Sessions were conducted seven days per week and lasted u n til 50 min had elapsed or 50 reinforcements had occurred, whichever came f i r s t. Following 15 days of tra in in g with the Rl 64 sec schedule (a period which produced stable response rates in a ll subjects as determined by visual inspection of a graph of response rates as a function of sessions), the subjects were tested in e xtinctio n fo r generalization along the lig h t in te n sity dimension. preceded by 10 min of warm-up exactly like tra in in g. The te st was The stim uli used fo r the generalization te s t were selected so tha t each stimulus above the lowest one was approximately twice as intense as the one preceding it. The in te n sitie s used were: 0.00 fc, 0.03 fc, 0.06 fc, 0.11 fc, 0.23 fc, 0.45 fc, 0.88 fc (S+ in te n s ity ), 1.75 fc, 3.5 fc and 7.0 fc. During the te s t, these 10 stim uli

15 8 were presented in 10 blocks of 10 stim uli each, each value appearing once in each block. The stimulus appeared on the key fo r 20 sec followed by a tim e-out (period of complete darkness) fo r 10 sec. The in te n s ity was changed during tim e-out. The number of responses occur- ri ng in each stimulus presentation was recorded. Training with the single stimulus on an Rl 64 sec schedule continued fo r seven days follow ing the generalization te s t. Discrimination tra in in g was then begun. The positive stimulus (S+), the stimulus associated with reinforcement on an Rl 64 sec schedule, was the same in te n sity (0.88 fc) as had been present during in it ia l tra in in g. The negative stimulus (S -), the stimulus associated with e x tin c tio n, was set at 0.03 fc. This value was the second lowest te s t value, the lowest being 0.00 fc. The S+ remained on the key u n til the end of the f i r s t reinforcem ent. The S- was then presented and remained on u n til 30 sec of no responding occurred. The S+ then reappeared and the stim u li were alternated in th is manner throughout the session. Each day a discrim ination ra tio consisting of the rate in S+ divided by the rate in S+ pi us the rate in S- was computed. When th is value reached 0.90 or greater, a second generalization te s t was conducted. This te s t was identical to the f i r s t generalization te s t. Following the second generalization te s t, a new S- was s u b s t i tuted in the d iscrim ination procedure described above. The S- was changed from 0.03 fc to 0.11 fc. The S+ remained at its value of 0.88 fc. When the discrim ination ra tio reached 0.90 or b e tte r, a th ird generalization te s t, identical to the f i r s t two, was conducted.

16 RESULTS The changes in response rates fo r both S+ and S- during in it ia l conditioning and discrim inations 1and 2 are shown in Fig. 1. (Dashed lines show mean response rates fo r S+ and S- during each experimental phase.) A ll subjects show stable response rates, both p rio r to and immediately follow ing the f i r s t generalization te s t (represented by the f i r s t v e rtic a l divid in g lin e ). Following the introduction of discrim ination 1 (second ve rtic a l lin e ), the response rates in the presence of the S+ increased above th e ir previous mean rates while the rates in S- ra p idly decreased. The introduction of discrim ination 2 (th ird v e rtic a l line) produced a s lig h t overall rise in the S+ response rate fo r subject B-5-7 while the S- rate was i n i t i a l l y high and declined rapidly. For both subjects B-5-8 and B-5-9, th is discrim ination d ra s tic a lly lowered the S+ rate and raised the S- rate. These rates then changed so th a t both subjects reached S+ response rates higher than any previously shown and S- rates near zero. Because of the i n i t i a l l y low rates produced by B-5-9 in discrim ination 2, the mean S+ rate fo r tha t phase is low. The actual response rate at the end of tra in in g, however, is higher than any seen previously. The re g u la rity of changes in the S+ and S- rates is fu rth e r illu s tra te d in Fig. 2. This figure shows the changes in the d is crim ination ra tio as a function of sessions. For a ll birds the ra tio is a monotonicai Iy increasing function fo r each discrim ination. The fig ure shows th a t the discrim ination learning was an o rd e rly, regular process.

17 Fig. 1. Changes in response rates as a function of treatments. This figure shows d a ily response rates (responses per minute) for each phase of the study. C ircles represent responses made in the presence of the p ositive stimulus while tria n g le s represent responses made in the presence of the negative stimulus. Dashed lines represent mean response rates fo r each period and fo r S+ and S-.

18 11 CO I CO _o CXI. 0 TRAINING SESSIONS - if) O C D ^ O c 3 in N lh 3 3 d S3SNOcJS3d

19 12 Fig. 2. D iscrim ination ra tio changes as a function of sessions and treatments.

20 13 1.On B-5-7 i f B-5-8 c c U ^ 1.0 -r Q B-5-9 SESSIONS

21 14 Figure 3 is a comparison of the three generalization gradients produced by each subject. Median gradients are also included to summarize the resu lts. The to ta l number of responses occurring at the peak of each gradient is given in Table 1. All of the gradients are based on a large number of responses and are therefore re lia b le. For B-5-7 and B-5-8, the gradients produced a fte r tra in in g on one stimulus value are re la tiv e ly f l a t on the le ft or dimmer side of the S+. If a re lia b le peak fo r these gradients e x is ts, i t is located between 0.00 fc and the tra in in g stimulus. Since the response rates are so sim ila r, the s lig h t amount of v a r ia b ility prevents determining the exact location of the peak. Subject B-5-9 shows some generalization decrement in both directions with the peak at the tra in in g value. All subjects responded near zero at the 0.00 fc in te n sity. The median curve fo r the f i r s t generalization gradient follows the form of the f i r s t two subjects, being f l a t over in te n s itie s below the S+ and dropping in rates fo r in te n s itie s above the St. The second generalization te s t (follow ing discrim ination 1) shows a sharpening of the slopes on both sides of the St value. Only one subject (B-5-8) shows what may be a s h if t in the peak a fte r tra in in g on the f i r s t discrim ination problem. For a ll subjects there is a more c le a rly defined peak. The median gradient shows a sharpening of the gradient over the dimmer in te n s itie s, a somewhat f la t area around the St value and a sharp increase in the slope of the gradient over the highest values re la tiv e to the f i r s t gradient. For the th ird gradient, B-5-7 and B-5-8 show a s h ift of the peak to the more intense values, which had previously been rather low.

22 Fig. 3. A comparison of generalization gradients obtained follow ing (1) single stimulus tra in in g (so lid lin e ), (2) d is crim ination 1 (tria n g le s with dotted lin e ), and (3) discrim ination 2 (squares with dashed lin e ). The in te n s itie s are: -6 = 0.00 fc; -5 = 0.03 fc (S i in te n s ity ); -4 = 0.06 fc; -3 = 0.11 fc (S 2 intens it y ) ; -2 = 0.23 fc; -1 = 0.45 fc; 0 = 0.88 fc (St in te n s ity ); t = 1.75 fc ; t2 = 3.5 fc ; and t3 = 7.0 fc.

23 16 A < * -o c\j OJ 00 ini CQ 00 + CM + INTENSITY + o ( CM i 00 I I L f) i CD i o o o o o o o CD 0M O CD c m (la lflh IX V la j JO %) 3 -LVd 3 S N O ds 3 d

24 17 Table 1 Maximum responses fo r each generalization gradient. Each ce ll contains the to ta l number of responses made by a subject at the maximum point fo r a s p e c ific generalization gradient. Subject 1 Generalization gradient 2 3 B B B

25 The th ird subject shows a fla tte n in g of the gradient around the S+ and a bimodal peak position which centers around the S+. The form of the gradient for subject B-5-7 is again much sharper than the gradients from the f i r s t two tests. For the other two subjects, there is a rise in the re la tiv e rates at higher in te n s itie s and an increase in the slope at the lower in te n s itie s. The le ft hand column under each phase in Table 2 shows the progressive change in the area under the generalization gradient as a function of successive discrim inations. (These values were obtained by summing the percentages fo r each stimulus value on a gradient.) There is a s t a t is t ic a lly s ig n ific a n t decrease in the area under the curve as a function of the discrim ination tra in in g (F = 7.98, df = 2/4, p < ). This indicates a d e fin ite sharpening of the gradient as a function of discrim ination tra in in g. The second column in Table 2 shows the placement of the median stimulus value (the point at which 50$ of the gradient f a lls above and 50$ f a lls below). For B-5-7 and B-5-8, the median point s h ifts to the rig h t toward the higher in te n s itie s with each discrim in atio n. The median point s h ifts fo r B-5-9 fo r the f i r s t discrim ination but not fo r the second discrim ination. Overall there is a s t a t is t ic a lly s ig n ific a n t s h ift of the median value to the higher in te n s itie s (F = 10, df = 2/4, p< 0.05).

26 19 Table 2 Areas and medians fo r each generalization gradient. Values represent to ta l percent under each generalization gradient and the position of the median of to ta l responses fo r each gradient. Subject Genera Iiz a tio n gradient Area Med i an Area Med i an Area Med i an B B B Med i an

27 20 DISCUSSION The f i r s t generalization gradient obtained a fte r tra in in g on only one stimulus value is p a rtic u la rly f l a t over the in te n s itie s dimmer than the S+ while re la tiv e ly sharp over higher in te n s itie s. Blough (1959) obtained broad yet peaked generalization gradients over the lig h t in te n sity dimension when he trained subjects on one stimulus value. His stim uli covered a much wider range than the present study (0.14 mi I I a Iamberts to 820 mi I I a Iamberts) thus making comparison of the two studies d i f f i c u l t. He did fin d, however, that pigeons showed a preference fo r the lower-middle values of his testin g range. It is possible th a t the values covered by the present study lie w ithin his range and the gradients of the present study represent parts of his gradients. The size of stimulus could have produced the difference in gradient shapes between the present study and Blough*s. Blough used a pinpoint of lig h t on the key while the present study had the e n tire key illum inated. Heinemann and Rudolph (1963) have shown th a t larger stimulus areas will produce f la t t e r lig h t in te n sity gradients than smaller areas. Their smallest stimulus (which produced the sharpest gradient) was somewhat larger than the one in the present study. There are, however, several other d i f ferences, such as th e ir use of reflected lig h t and the stimulus not being on the key but around i t, which could explain th e ir gradients as compared to those of the present study. The re la tiv e ly low response rates over the higher in te n s itie s, as compared to low in te n s itie s and the fa ilu re to produce peaks at

28 21 the tra in in g value could be due to the magazine lig h t in te n s ity. F irs t, because the magazine lig h t was brig hte r than the key, the subjects had an opportunity to compare stim uli and associate dimmer stim uli with the key. Second, during tra in ing th is more intense lig h t could have produced a greater illum ination of the chamber during reinforcement. This general increase in illu m in a tio n during reinforcement could have become a discrim inative stimulus fo r moving to the food magazine. This response is incompatible with key pecking and produces lower key peck rates. During te stin g then, b rig h te r stim uli produced levels of illum ina tio n comparable to those during reinforcement and hence would re su lt in lower response rates. Future studies will be needed to investigate the e ffe cts of th is stimulus on generalization gradients. All of these studies in the area of lig h t in te n s ity employed n o n -d iffe re n tia l tra in in g at one stimulus value and produced no e v i dence to support the theory of stimulus inten sity dynamism. Studies of tone in te n s ity have produced s im ila r results (M alott and M alott, 1967). It would appear from these studies th a t the e ffe c ts of stimulus in te n s ity dynamism are not inherent in the organism and might, th e re fo re, require tra in in g to produce them. The introduction of discrim ination tra in in g produced d ra stic changes in the behavior of the subjects during both tra in in g and te stin g. The f i r s t of these changes is a decrease in the rate of responding during the S- component with a corresponding ris e in the rate during the S+ component. These changes in response rates during S+ and S-, known as "behavioral contrast" (Reynolds, 1961), have been

29 22 observed by other investigators working in the area of d iscrim ination tra in in g (Hoy, 1954; Herrick, Meyers and Korotkin, 1959; Terrace, 1966). Reynolds has demonstrated that the increase in the St rate is a function of the decrease in reinforcement density in the presence of the S- and not necessari ly a function of the decreased S- rate alone. In the present study, i t was noted that each successive g en eralization gradient was sharper than the previous gradients. This is supported by the work of other investigators who have shown th a t any discrim ination tra in in g, whether interdim ensionai, intradim ensionai, or extradimensionai, will sharpen the generalization gradient over the te st continuum (Thomas, Mariner, and Sherry, 1968). Another interesting re s u lt is the changing d is trib u tio n of responding with each new S+ - S- d iscrim ination. As the data presented show, there is a progressive s h if t in the position of the gradients to the higher in te n s itie s. These results are in line with the e a rlie r studies of generalization gradients a fte r discrim ination tra in in g discussed previously (Hanson, 1959; Terrace, 1966). The gradients produced following the f i r s t discrim ination in which the S- was d ista n t from the S+ are not shifte d as fa r as the gradients produced a fte r the second discrim ination problem in which the S- was close to the S+. As stated above, Hanson (1959) has described a sim ila r phenomenon. It is possible that changes in the gradients are only a product of continued tra in in g and/or te s tin g rather than a re su lt of the d iffe re n t discrim ination problems (Mountjoy and M alott, 1968). It

30 would be desirable, the refore, to conduct an additional control experiment in which the order of the discrim ination problems was changed or tra in in g continued on one value with several tests conducted at varying in te rv a ls. CONCLUSION The present study does not support the notion of stimulus inten s ity dynamism. The proposal of Hilgard and Marquis (1961) that the e ffects of stimulus in te n s ity dynamism should in teract with the generalization decrement e ffe c ts to produce f la t t e r gradients on the higher in te n s itie s side of the S+ than over lower in te n s itie s finds no substantiation. Testing a fte r the in it ia l tra in in g showed no evi dence of the phenomenon. D iscrim ination tra in in g produced a sharpening of the in it ia l gradient on both sides of the tra in in g value instead of the increase on the higher side of the S+ value. The s h iftin g of the gradient to the higher in te n s itie s appears to follow the formulations and predictions of those th e o rists discussed e a rlie r who propose the peak s h if t phenomenon as a s u ffic ie n t explanation fo r e ffe c ts commonly a ttrib u te d to stimulus in ten sity dynami sm.

31 24 References Bloomfield, T. M. A peak s h if t on a l i n e - t i l t continuum. _J_. exp. Anal. Behav., 1967, 10, 4, Blough, D. S. Generalization and preference on a stimulus in te n s ity continuum. J_. exp. Ana I. Behav., 1959, 2, Brown, J. S. The generalization of approach responses as a function of stimulus in te n sity and strength of m otivation. J. comp. Psychol., 1948, 41, Farmer, J. Properties of behavior under random-intervai reinforcement schedules. J_. exp. Ana I. Behav., 1963, 6, Fink, J. B. and Patton, R. M. Decrement of a learned drinking response accompanying changes in several stimulus ch aracte ristic s. J_. comp. Phys io I. Psycho I., 1953, 46, F rick, F. C. Analysis of an operant d iscrim ination. J. Psychol., 1948, 26, Hanson, H. M. Effects of discrim ination tra in in g on stimulus generalization. J_. exp. Psychol., 1959, 58, 5, Heinemann, E. G. and Rudolph, R. L. The e ffe cts of d iscrim inative tra in in g on the gradient of stimulus genera Iiz a tio n. Amer. J. of Psychol , 76, 4, H errick, R. M., Meyers, J. L., and Korotkin, A. L. Changes in SD and SA rates during the development of an operant discrim ination. J_. comp. Phys io I. Psycho I., 1959, 52, Hull, C. L. Stimulus in te n s ity dynamism and stimulus generalization. Psychol. Rev., 1949, 56, Jenkins, H. M. and Harrison, R. H. E ffect of discrim ination tra in in g on auditory generalization. J_. exp. Psychol., 1960, 59, Kimble, G. A. (Ed.) Hilgard and Marqu?st Conditioning and Learning. New York: Appleton-Century-Crofts, Inc., Logan, F. A. A note on stimulus in ten sity dynamism (V). Psychol. Rev., 1954, 61,

32 25 M alott, R. W. and M alott, Marilyn K. A preliminary analysis of loudness in terms of stimulus generalization. Paper read at Midwestern Psychological Association convention, Chicago, Mednick, S. A. and Freedman, J. L. Stimulus generalization. In Contemporary Research in Learning, John R. Braun (Ed.). Princeton, N. J.: Van Nostrand Company, Inc., Mountjoy, P. T. and M alott, Marilyn K. Stimulus generalization in chickens reared in re stricte d portions of the spectrum. Psycho I. Rec., 1968 (in press). Perkins, C. C. The re la tio n between conditioned stimulus in te n sity and response contours in terms of stimulus generalization. J_. exp. Psycho I., 1953, 46, P ie rre l, R. and Sherman, J. G. Generalization of auditory in te n s ity following discrim ination tra in in g. J. exp. Anal. Behav., 1960, 3, Reynolds, G. S. Behavioral contrast. J. exp. Anal. Behav., 1961, 4, 1, Spence, K. W. The d iffe re n tia l response in animals to stimuli varying within a single dimension. Psycho I. Rev , 44, Terrace, H. S. Discrim ination learning and inhibition. Science, 1966, 154, Thomas, D. R., Mariner, R. W., and Sherry, G. The role of preexperimental experience in the development of stimulus control. J_. exp. Psychol., 1968 (in press).