A correspondent asked, “From a behavioral perspective, what is a pronoun?” This was my response (lightly edited):
Several possible responses came to mind. A) It’s not a behavioral concept, so I have nothing to say “from a behavioral perspective.” B) The question should be rephrased as, “What are the characteristic controlling variables for the verbal response “pronoun”? C) The question should be rephrased as “What contingencies affect a group of speakers so that verbal forms emerge that “hang together” in such a way that a special term (“pronoun”) proved useful? Only the third seems in the spirit of the question. A definitive answer is probably out of reach, but characteristic contingencies are suggestive.
First, I think it will help to clarify the distinction between nature and our models of nature. For example, for some purposes it is helpful to say that the average height of human males is normally distributed with a mean of 70 inches and a standard deviation of 3 inches. From this statement, others can be deduced, such as that 84% of males are shorter than 73 inches; only 2.5% of males are taller than 76 inches, and so on. These statements are truths about a statistical model, not about nature itself. For many practical purposes the model is a close enough approximation, but we should never forget that the normal distribution and its parameters are just models. Even the notion of height as a single, measurable, rigid quantity is a model. We use the reading on a scale as though it represented a fixed thing, but our height varies, within a small range, according to lots of variables.
Sentence, clause, phrase, and the various parts of speech are terms that apply to our models of verbal behavior; they are not themselves behavioral units. So concepts like subject, predicate, preposition, and pronoun that apply to such models will not find precise translations into behavioral terms. To put it another way, terms in models are essentialistic and have no perfect counterpart in selectionist domains. Nevertheless, there is a phenomenon to be explained. At a minimum, we can specify the variables that control the tact pronoun, just as we might for ether, or ghost, but to do so would be to dodge the question, for it was presumably inspired, not by the self-evident practices of the verbal community, but by the prevalence of certain forms of responses in verbal behavior. We can’t just turn away from the challenge by insisting that pronouns are not behavioral terms. But we need not worry if our analysis is incommensurate with those of the linguist.
So what might be a behavioral translation of pronoun? Here is one suggestion:
We call something a verb because of the autoclitic frames in which it plays a part. In Skinner’s lexicon, an autoclitic frame is a set of fixed terms intermingled with other terms that vary from one circumstance to another. (Prepositional phrases are simple examples: on the X; next to the X; under X.) Unlike many other words, verbs tend to require or imply frames that specify the roles played by variable terms. For example, a word like agree implies the existence of one who agrees and a position or person that is agreed with. Sarah agrees with Tom would be one example of the frame X agrees with Y. If we hear lots of examples of this frame, with different variables for X and Y, the frame will start to hang together as a unit glued together by the intraverbal relationship among the elements of the frame. The phonetic properties of the terms Sarah and Tom would not be part of the unit, because they occur together too infrequently. However, the prosodic stress with which Sarah and Tom are uttered is invariant from one X and Y pair to another. So perhaps the frame is better described as X (stressed) agrees with Y (stressed).
If Sarah were the only person in our world who ever agreed with anything, then Sarah agrees with Y would presumably emerge as an autoclitic frame, and if Tom were the universal object of agreement, then X agrees with Tom would emerge as a unit. As it is, X and Y are variable terms, and we must look to the context to complete our autoclitic frame each time it is uttered. Thus, proper nouns seldom become part of autoclitic frames because the role they play in such frames is commonly played by other terms as well. The same argument can be made for generic nouns as well. The pizza delivery guy agrees with Y would occur too rarely to become a verbal unit of analysis.
Perhaps in a small verbal community, such as a family, proper nouns do sometimes become part of frames. Sarah’s husband may have several nearly identical autoclitic frames in his repertoire, under different sources of control: X agrees with Y, and Sarah agrees with Y. Verbal units will often differ from one speaker to another, and that is particularly true of intraverbal sequences.
However, the words we call pronouns vary much less frequently than proper nouns from circumstance to circumstance. He agrees with me might emerge as an intraverbal chain (not a frame) because it can remain constant over a wide variety of examples. He agrees with Y might emerge as a relatively simple autoclitic frame. Autoclitic frames are much more complicated than intraverbal chains because the frame and its variables are under separate sources of stimulus control. Nevertheless, they have to be emitted smoothly, in sequence, as the speaker speaks. In fact, it isn’t at all clear how stimulus control shifts as we speak. But pronouns can become part of larger verbal operants and will tend to do so to the extent that they are repeatedly used in certain constructions. Even when they are not part of larger units, pronouns are economical. The proper noun Sarah will tend to be weaker than the pronoun she, because in any given circumstance, the pronoun is more likely to have been appropriate.
So I am suggesting that pronouns emerged out of a tendency for behavior to migrate toward economy. Pronouns are placeholders in autoclitic frames freeing the speaker from the task of filling in variable terms and are appropriate to a very wide variety of social contexts.
Skinner himself alluded to speakers’ tendency toward economy, and his remarks are relevant to the present context:
[A speaker’s] repertoire of speech-sounds may become simplified … Above the level of the speech-sound we observe simplifications of larger operants in the dropping of unaccented syllables, especially at the beginnings or ends of forms, in the resort to nicknames and other expressions “for short,” in the dropping of one of two or more identical syllables (haplology), and so on. If the response photo is as effective as photograph, the shorter form is likely to be stronger. Slight changes in these directions are accumulated historically in well-known examples of linguistic change. … Verbal behavior under pressure of time is likely to show telescoping, omissions, a reduction in the range of pitch variation, and so on. Memorized speech which deteriorates when the contingencies are relaxed is exemplified by standard ritualistic verbal behavior—for example, the mumbling of the truth, the whole truth, and nothing but the truth. Ritualistic prayers are subject to this deterioration. In the Middle Ages, religious people commonly “gabbled” through prayers and other services to get them over with quickly. “They left out the syllables at the beginning of words, they omitted the dipsalma or pause between verses…they skipped sentences, they mumbled and slurred….” (Verbal Behavior, pp. 210-211)