Skip to content

Some observations on the reinforcement of verbal operants.

Greenspoon, in 1950, described an experiment in which the rate of saying ‘plural nouns’ was increased by a procedure that included the use of a verbal reinforcing stimulus, ‘mmm-hmmm;’ he found that the Ss were unable to report either the reinforcing stimuli or the change in their behavior. That is, Ss were ‘unaware’ of the experimental manipulation of their behavior. This study reports two brief experiments concerned with similar verbal behavior and similar reinforcing stimuli: we were trying to find out more about some of the conditions under which the effect described by Greenspoon could be obtained, and to clarify the relationships between changes in the Ss’ verbal behavior and the S’s’ ability to report either the reinforcing stimuli or the change in behavior. In addition, we wanted to experiment with verbal behavior using other than saying-plural-nouns as a response.



Procedure. The Es, 15 in number (4 women and 11 men), were students in psychology. The Ss, 17 in number (5 women and 12 men), were fellow students. Experimental sessions were, with two unimportant exceptions, conducted in S’s living quarters with only E and S present.

The instructions given S were: “This is a study of the vocabulary of college students. Say words. Do not repeat. Do not count. Do not say sentences.” Half the Es used as a reinforcing stimulus a casual ‘mmm-hmmm’ or ‘good;’ the other half wrote down ‘significant words’ on the data sheet. (The Es were instructed to make an obvious writing movement.) The responses reinforced were ‘plural nouns’ and ‘adverbs.’ S kept at the task until he had said 800 words in all; no reinforcement was given for the first 100 words; one response (either ‘plural nouns’ or ‘adverbs’) was reinforced during the next 300 words; no reinforcement was given during the next 100 words; the alternate response (either ‘adverbs or ‘plural nouns’, respectively) was reinforced during the last 300 words. The number of plural nouns, adverbs, and other words was recorded minute by minute by ticking off on data sheets. S was asked at the end of the experimental session to tell all he could about the experiment.

Results. The total rate of saying words did not differ significantly under the various conditions of reinforcement and non-reinforcement (ranked analysis of variance, p.>0.50). The median rate of saying words was 21 per min., with a range across Ss of 8 to 36.

The rate of saying ‘plural nouns’ increased under the conditions of reinforcement in 13 of 14 Ss. (Two Es, who started reinforcing ‘adverbs,’ terminated the experiment when their Ss failed to give any. Thus, they did not undertake to reinforce ‘plural nouns.’) The probability that such a result could be obtained in the absence of a systematic effect of the experimental variable is 0.002 (Wilcoxon’s matched pairs test, using signed ranks). The median increase in rate was 95%, ranging from 6% to 2800%. Quite evident, both kinds of reinforcement, saying ‘mmm-hmmm’ and ‘good,’ and writing the word down, were effective. In the first case, there was an increase in rate in 7 of 7 Ss; in the second case, in 6 of 7 Ss. When reinforcement was withheld for ‘plural nouns’ following the period during which they were reinforce, the number decreased in 8 of 11 Ss (the experimental design did not furnish us with this information in all cases); this result could be obtained by chance in the absence of an effect of the experimental variable only once in 10 times. It is probable that the extinction period of 100 words was to short to indicated the full effect of withholding reinforcement.

The rate of saying ‘adverbs’ increased under conditions of reinforcement in 6 of 7 Ss who said any adverbs at all; this is statistically significant at p <0.025 (signed ranks test). Nine of the full 16 Ss had an initial rate of zero for saying ‘adverbs’ hence reinforcement was impossible, and hence this portion of the experiment could not be carried out; the two Es mentioned before stopped at this point and did not carry out subsequent parts of the experiment. The median amount of increase in the rate of saying ‘adverbs’ under conditions of reinforcement (for those Ss with an intial rate greater than zero) was 500%, with a range from 35% to 1800%. All three Ss from whom reinforcement for saying ‘adverbs’ was withdrawn showed a decrease in the rate of saying ‘adverbs’ during this extinction period.

When the Ss were questioned at the end of the experimental session, 12 of 14 mentioned the reinforcing stimuli; 6 of 7 reported the fact in one way or another, that E had said ‘mmm-hmmm’ and ‘good’ during parts of the experiment (p<0.13, sign test), and 6 of 7 reported that they noticed that E wrote down certain words during the experiment. These Ss would say something like “I noticed that for a while you liked nouns, and then you didn’t care what I said,” or “Sometimes I’d hit the ‘jackpot,’ and you wrote it down.” One, who showed positive conditioning, thought that it was ‘bad’ for a word to be written down. Six Ss reported correctly the response-class that had been reinforced; two Ss reported a subclass or a class related to the reinforced class, and four Ss did not report in any way the reinforced class when questioned. One anthropology student, for example, gave the names of plural nouns, reported that “nouns were what E liked.”

None of the Ss made any statement to the effect that this was a conditioning experiment. Of those who exhibited skepticism of it as a vocabulary test, some made such statements as, “You were trying to find out how I thought.”



Procedure. In the second experiment, the Es were the same students. The Ss were again, with one exception, fellow students; (13 men and 2 women); the one exception was an older physician. The experiments were, with two exception, again conducted in S’s living quarters, with E and S alone.

S’s instructions were the same as before. ‘Mmm-hmmm’ and ‘good,’ both spoken casually, were the reinforcing stimuli. Half the Es reinforced words that had anything to do with ‘travel;’ the other half reinforced words that had anything to do with ‘living-things.’ The Ss said 600 words. No reinforcement was given for the first 100 words, reinforcement was given for the selected response-category each time it occurred during the next 300 words and no reinforcement was given during the last 200 words. The numbers of words dealing with ‘travel,’ ‘living-things,’ and all other words were recorded for S. At the end of the experimental session, he was asked to “tell what you can about the experiment.”

Results. The total rate of saying words did not differ significantly under conditions of reinforcement and non-reinforcement (ranked analysis of variance, p> 0.75). The median rate of saying words was 19.25 words per min., with a range from 7.6 to 49.

The rate of saying ‘travel-words’ and ‘living-things-words’ increased under the appropriate condition of reinforcement in 14 of 16 Ss. The probability that this result could be obtained in the absence of an effect of the experimental variable is less than 0.005 (signed ranks test). The median increase in saying these words was 175%, with a range from 13% to 7600%. Under the control conditions, i.e. saying ‘travel-words’ when ‘living-things-words’ was being reinforced, and vice versa, the rates did not change significantly. When saying ‘travel-words’ and ‘living-things-words’ are considered separately, the results are 7 of 9 Ss for ‘travel-words’ gave a median increase of 92%, and 7 of 7 Ss for ‘living-things-words’ gave a median increase of 600%. When reinforcement was withheld (i.e. and extinction condition), the rate of saying these words decreased in 13 of 16 Ss, with 2 Ss continuing at the same rate. This is statistically significant with p<0.005 (signed ranks). The median amount of decrease was 45%, with the range of decrease from 9% to 80%.

At the end of the experimental sessions, 13 of 15 Ss reported in one way or another the reinforcing stimulus. Ten of the 13 Ss who reported the reinforcing stimulus could not identify correctly the response class reinforced, although 5 of these 10 Ss reported a subclass of or a class partially overlapping the one the E was reinforcing. Examples of this are “geographic locations,” or “words which have to do with transportation” when ‘travel words’ were being reinforced, or “biological terms” when ‘living-things-words’ were reinforced.



Despite the absence of instruction, the verbal behavior of most Ss is under the experimental control of an E who systematically reinforces specific classes of verbal response; this occurs whether or not S ‘notices’ the reinforcing stimuli, although most do so. S will work for an ‘mmm-hmmm’ or ‘good,’ or a response by E to S’s behavior, such as singling out a word for writing-down. S seem to be acting under self-imposed instructions. Thus, S might say “I knew you liked cities, states, and countries, and so I said them,” or, “I knew you seemed to want me to say words that had to do with the names of places.”

Both the class of responses chosen by Greenspoon and the classes chosen in these experiments prove to behave as responses. Under conditions of reinforcement, their rate of occurrence increases (provided, of course, that they have an initial rate of occurrence grater than zero). S may, however, often respond with a subclass of, or a response class partially overlapping with, the class arbitrarily singled out for reinforcement by E. For such Ss, the response is saying a more narrowly specified class of words.

Some Ss who were conditioned remained unable to report either the occurrence of reinforcement, or the response-class that they had been giving at an enhanced rate. They had not noticed either the environmental event on which their behavior was conditional, nor that may one response had occurred more frequently than others. Although showing a change in rate, i.e. being conditioned, seems a necessary condition for Ss to report either the reinforcing stimulus or the response-class, the converse is not true. Most interesting, there is no striking difference between the behavior of Ss who become aware of these variables and the behavior of those who do not.

In summary, (1) the Greenspoon effect is easily reproduced, (2) under our conditions, most Ss, who show verbal conditioning, are able to report the occurrence of reinforcing stimuli, and of these, most are able to report the response conditioned, and (3) saying ‘adverbs’ is an example of verbal behavior that has a very low operant level.


Accepted for publication June 28, 1955. These experiments were performed at Harvard University.

Joel Greenspoon, The effect of a verbal stimulus as a reinforcement, Proc. Ind. Acad. Sci. 59, 1950, 287.

Statistical procedures utilized in this study may be found in Fredrick Mosteller and R.R. Bush, Selected quantitative techniques, in G. Lindzey (ed.), Handbook of Social Psychology, 1954.

Where N is sufficiently large, and rankings are possible, Wilcoxon’s signed ranks test is used; in all other cases the sign test is used.

Creative Commons License is licensed by Admin under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 United States License.