~OURNAL OF VERBAL LEARNmG AND VERBAL BEHAVIOR ~, 4 8 4 - 4 8 7
(1966)
The Effects of Extinction on Two Concurrent Verbal Responses I CARL DANSON2 AND REED LAWSON
The Ohio State University, Columbus, Ohio After reinforcing two verbal responses concurrently in a free responding situation, 40 college Ss were informed of one response-reinforcement contingency, but not the other. The responses were then extinguished in two successive phases: (1) the informed response was extinguished while the uninformed was still reinforced, and (2) the uninformed response was extinguished as the informed response continued extinction. The major finding was the interaction between the two responses which occurred during the extinction periods. The decreasing frequency of one response resulting from extinction was accompanied by a rapid and marked increase in the other response. The present interaction appears to be similar to the phenomenon of behavioral contrast. T h e a t t e m p t s of psychologists such as G r e e n s p o o n (1955) a n d s u b s e q u e n t workers to e m p h a s i z e the free o p e r a n t characteristics of v e r b a l b e h a v i o r a d d e d increased v i t a l i t y to an issue t h a t has n e v e r been d o r m a n t : the role of " a w a r e n e s s " in h u m a n learning (e.g., E r i k s e n , 1962). One difficulty w i t h formulating a clear answer is the difficulty of assessing S's awareness. A m o n g the n u m e r o u s a p p r o a c h e s to this p r o b l e m is t h a t suggested b y W e i n s t e i n and L a w s o n ( 1 9 6 3 ) . I n this technique, Ss are specifically i n f o r m e d during the e x p e r i m e n t of the correct r e s p o n s e - r e i n f o r c e m e n t contingency. T h e i m m e d i a t e effect of this proced u r e is h i g h l y predictable. W h a t is interesting are s u b s e q u e n t effects w h e n a d d i t i o n a l procedures are introduced. T h e p r e s e n t p a p e r extends the direct-inform a t i o n research in two ways. First, the question is asked: I f S is being c o n d i t i o n e d conc u r r e n t l y to two sets of contingencies and is i n f o r m e d a b o u t o n l y one of t h e m , w h a t h a p 1 The main results of this paper were presented at the 1964 MPA Convention and were based upon a thesis submitted in partial fulfillment of the M.A. degree at The Ohio State University by the first author. Now at California State College, Long Beach. 484
pens to the o t h e r ? Second, and p e r h a p s m o r e i m p o r t a n t , w h a t h a p p e n s to the two response classes w h e n the i n f o r m e d class is ( w i t h o u t w a r n i n g ) s u b j e c t e d to extinction? METHOD
Subjects and Experimenter. The Ss were 40 volunteer undergraduates, 19 male and 21 female, who were enrolled in different sections of an introductory psychology course at The Ohio State University. The E was a male graduate student in experimental psychology. Responses and Rein]orcements. The correct response classes were plural nouns and words ending with the letters ING. The verbal stimulus "good" or the sound of a buzzer constituted the reinforcements. Recording and Experimental Setting. Both classes of correct responses were counted separately as well as all irrelevant resPonses. The E transcribed the number of responses in these three categories every minute on a prepared data sheet. All apparatus was kept out of sight of the Ss. The S and the E faced each other across a table. Procedure. The Ss were given the instructions typical to this kind of task--"say words that come into your mind," but no counting, connected discourse, or proper nouns. The Ss then responded for 6 rain while E obtained a baseline for the two response classes. After a brief pause an acquisition period began. The Ss were then reinforced following the emission of a correct response for a 25-rain period. During this period of acquisition, one of the t w o
EXTINCTION OF FREE VERBAL OPERANTS
possible combinations of responses and reinforcements, i.e., plural nouns--"good" and ING--buzzer, or plural nouns--buzzer and ING-"good," was randomly assigned to the Ss, with the restriction that an equal number of Ss experienced each combination. Reinforcement was administered on a continuous reinforcement schedule for each correct response of either class. At the end of the acquisition period, Ss were stopped and informed about one of the contingencies in effect (e.g., "Whenever you say a plural noun, I say, 'Good'"). Nothing was said about the other relationship. Then they were asked to resume the task. Both contingencies continued to be reinforced as before. The selection of the informed contingency by E was based upon the relative frequency of emission during acquisition. The S was either informed of his more frequently or less frequently emitted response, Hi and Lo groups, respectively, which was randomly determined with the restriction that an equal number of Ss were assigned to each group. To insure against a possible bias during acquisition, S's group assignment was not known by E until S was to be informed. The informed condition (a 10rain period) was continuous with the remaining sections of the experiment. The two ensuing 15-min periods of extinction consisted of (a) the extinction of the informed contingency but not of the uninformed and (b) the continued extinction of the informed contingency and the extinction of the uninformed relationship. Thus, each S served 1 hr 11 min in the experiment, excluding the time for instructions, during which two verbal response classes were concurrently reinforced and then differentially identified and extinguished. RESULTS The different response categories which were recorded each minute were combined into 5-rain blocks. A median transformation of the 25-min acquisition period wa~ performed to obtain homogeneity of variance. Each S was assigned a score from 0 to 5 according to the number of times his 5-min block scores were above the median. A 2 X 2 X 2 analysis of variance was computed in order to assess the effects of the responses and reinforcements which were used in the experiment. A blocks factor, a replication, was used in the design as a statistical control and consisted of the first 20 Ss in block 1 and the second 20 Ss in block 2. Since it was
485
not significant, the conclusion that the Ss in the experiment were affected uniformly before the administration of treatments following acquisition is justified. The only significant effect was the words factor (plural nouns ~ I N G words, p * ( . 0 1 ) , although other effects m a y have been obscured b y the median transformation. A t-test of difference scores of response frequencies between the operant level and tl~e final 5-rain block of acquisition was computed. The calculation assessed the effects of reinforcement on the more frequently emitted response for the 40 Ss in the experiment and was significant, t ( 3 9 ) - - 2 . 7 1 , p ~ .01. A test of significance was not performed on the less frequently emitted response since no difference between the base rate and final acquisition period was observed. I n other words, Ss tended to learn plural nouns better than I N G words regardless of the reinforcer. F u r thermore, Ss tended to learn on|y one contingency at the a p p a r e n t expense of the other. Figure 1 shows the successive changes of t h e mean ranks of responding during the 5min blocks of the experiment excluding the early acquisition data. The first two points of the graph display the operant level and final performance before one of the two contingencies was identified. The high- and lowfrequency responses, regardless of the subsequent informed or uninformed c]assification, are shown in order to exhibit the difference between the two responses which existed at the termination of acquisition as well as to provide a basis for observing the effects of the subsequent treatments. The lines representing the concurrent changes in the identified and unidentified responses are not continuous with the earlier section of the graph because each is composed of both the highand low-frequency responses. The general effect of informing Ss, represented b y the break in the graph, was a pronounced increase in the emission of the identified response. W h e n the identified response was extin-
486
DANSON AND LAWSON IO
High Frequency ..... ~
Low Frequency
. Informed
.
. . . . . . . Uninformed
z
/v~l.
,:
t~ t~
.1"
z O Q.
i
\ "~
\
L.,
II
"%\
\ z
5 '\
z t~
/
\ 4
\'
ol
t t 1 2 1 2 3 1 2 3 Operant Final Informed Level
Acquisition
Extinction Extinction
5 MINUTE BLOCKS WITHIN EACH CONDITION
FIC. I. Successive changes of the mean ranks of the two concurrent verbal responses as a function of the experimental treatments (see text for explanation). guished, however, its decrease in frequency was accompanied by a rapid increase of the unidentified but still reinforced response. Another interaction of responses of a smaller magnitude occurred during the last phase of extinction. Accompanying the extinction of the unidentified response was an increase in frequency of emission of the identified but then extinguished response. Thus, a change in the schedule of reinforcement from continuous reinforcement to extinction of one response resulted in a concomitant increase in the other response. The interaction occurred both when the response which increased in frequency of emission remained under the same reinforcement conditions as had previously existed (extinction phase 1) and when it continued to be extinguished (extinction phase 2). Due to the marked heterogeneity of variance, a separate analysis of the identified and
unidentified response for both the Hi and Lo group was performed by means of a Friedman two-way analysis of variance. The observed value of 22 for each of the four tests was significant (each p ~ .01). Since each response in the two groups was significant and the response changes of both groups were essentially the same, the data were combined for Fig. 1. The combined data do, however, obscure one between-group difference. When extinction phase 1 was initiated, the unidentified response in the two groups existed at two different response strengths. For the Lo group, the unidentified response was the higher frequency response in acquisition, a response which Ss in general learned. For the Hi group the unidentified response was the lower-frequency response, a response of low strength which showed no previous change. Apparently, a low-frequency or "latent" response was affected in the same manner as the high-frequency response since both showed a similar interaction. DISCUSSION Although, to our knowledge, a similar interaction of verbal responses has not been reported in the literature, the decrease of one response and the concomitant increase of another has been found in studies dealing with operant discrimination in lower organisms (e.g., Herrick, Myers, and Korotkin, 1959). These investigators found that in the development of a discrimination there is an increase in the responding in the presence of the SD and a decrease in the presence of the Sa. Skinner (1938) has referred to such a phenomenon as "behavioral contrast." Contrast is a specific instance of an interaction between responses in which responding in the presence of one stimulus is in the opposite direction from responding in the presence of the other stimulus. Reynolds (1951) concluded that rate of response in the presence of a given stimulus is in part determined by the frequency of reinforcement in its presence rela-
EXTINCTION OF FREE VERBAL OPERANTS
tive to the frequency during the other stimuli which control the behavior. Thus, an increase in relative reinforcement results in an increase in responding to that stimulus. Although in the present experiment an external stimulus was not correlated with different responses, the resulting interaction would seem to be an instance of behavioral contrast. A problem concerning the present data was revealed by the marked heterogeneity of variance. Part of the variability appears to be attributable to the initial difference in operant level between the two responses. An overall statistical analysis was further hampered by the differential treatment effects, "i.e., the resulting interactions, and by the large differences both among Ss and in terms of momentary fluctuations of a S's behavior. Although statistical significance was attained., the initial establishment of a stable level of responding would be a prerequisite for a more accurate assessment of subsequent manipulations. Within the boundaries of this restriction, the obtained interaction has been explained without the necessity of resorting to "higherorder" constructs, e.g., awareness, and, at least at the extremes of continuous reinforce-
487
ment and extinction, it appears that the occurrence of a verbal response is partially determined by its frequency of reinforcement relative to the frequency of reinforcement of another response. One of the implications of this result is that a decrement in a "correct" response may be the consequence of a decreasing frequency of reinforcement of that response or an increasing frequency of reinforcement of another response. REFERENCES
ERIKSEN, C. W. (Ed). Behavior and awareness.
Durham: Duke University Press, 1962. GREENSPOON, J. The reinforcing effect of two spoken
sounds on the frequency of two responses. Amer. J. Psychol., 1955, 68, 409-416. HERRICK, R. M., MYERS, J. L., ANDKOROTKIN,A. L. Changes in SD and S~ rates during the development of an operant discrimination. J. comp. physiol. Psychol., 1959, 69., 359-364. REYNOLDS, G. S. Behavioral contrast. J. exp. anal. Behav., 1961, 4, 57-71. SKINNER, B. F. Behavior o] organisms. New York: Appleton-Century-Crofts, 1938. WE~NSTEIN, W. K., AND LAWSON, R. The effect of experimentally-induced "awareness" upon performance in free-operant verbal conditioning and on subsequent tests of "awareness." J. Psychol., 1963, 56, 203-211. (Received February 23, 1965)