Bertrand Russell played a formative role in the development of B.F. Skinner as an intellectual. Skinner noted in his autobiography ‘The Shaping of a Behaviourist’, that it was Russell writings on behaviourism that led to Skinner becoming a behaviourist. Skinner though, despite his admiration for Russell, disagreed with him on a variety of different topics. A key area of disagreement between Skinner and Russell was on the use of the notion of ‘meaning’. In his ‘An Inquiry into Meaning and Truth’ Russell introduced meaning as an explanatory explanation in verbal behaviour:
“Just as jumping is one class of bodily movements, and walking another, so the uttered word ‘dog’ is a third class of bodily movements…Words, spoken, heard, or written, differ from other classes of bodily movements, noises, or shapes, by having meaning”
(An Inquiry into Meaning and Truth pp. 22-23)
Skinner argued that Russell’s appeal to meanings in his explanation was superfluous:
“Our subject matter is verbal behaviour, and we must accept this in the crude form in which it is observed. In studying speech, we have to account for a series of complex muscular activities which produce noises. In studying writing or gesturing, we deal with other sorts of muscular responses. It has long been recognized that this is the stuff of which languages are made, but the acknowledgement has usually been qualified in such a way as to destroy the main point…Bertrand Russell asserts that “just as jumping is one class of movement…so the word ‘dog’ is [another] class,” but he adds that words differ from other classes of bodily movements because they have “meaning.” Here something has been added to an objective description.” (‘Verbal Behaviour” p. 13)
Skinner’s point is wildly at odds with our everyday phenomenology. It seems to be just obvious that our words have meaning. To say that a word like ‘dog’ means something; doesn’t seem to be adding anything to the objective description. If a person says ‘dog’ and means by that what everyone else would describe as a cat we would say that the person did not understand the meaning of the term ‘dog’. So Skinner’s views are very counterintuitive from a commonsense perspective, and Russell’s views are much more congenial from the intuitive perspective. Obviously though scientific debates are not decided by commonsense, if they were we could rule out discoveries in relativity theory and quantum mechanics by fiat.
Russell on Meaning:
Russell’s views on the nature of meaning and propositions changed throughout his career. I will not here track Russell’s various changes of minds on the topic over the course of his entire intellectual development. Rather, I will be concerned with Russell’s take on meaning in his IMT, and how this relates to Skinner’s position in his ‘Verbal Behaviour’.
Russell’s theory is highly dependent on his postulation of the existence of propositions. For Russell a proposition is something that can be said in any language. He gives the example of the sentence ‘Socrates is Mortal’ and ‘Socrate est mortel’; these are sentences spoken is different languages which express the same proposition. Russell defines a proposition as “all the sentences which have the same meaning as some given sentence” (IMT p. 10). Given that his definition of proposition relied on undefined terms such as ‘meaning’ and ‘sentence’, Russell tried to offer definitions of them. He defined sentences as either a single word or a combination of words put together by syntactic rules.
Russell explicated ‘meaning’ in terms of natural language. He broke languages down into languages on different logical hierarchies; his reasoning for doing this was to avoid the semantic paradoxes. The most basic form of a language is what Russell calls the object language. Russell describes the object language as follows:
“We can now partially define the primary language or object-language as a language consisting wholly of ‘object-words’, where ‘object-words’ are defined, logically, as words having meaning in isolation, and psychologically, as words which have been learnt without its being necessary to have previously learnt any other words” (IMT p. 62)
The object language excludes the concepts of ‘true’ and ‘false’ as these terms when applied to the nth language belong to the nth+1 language (ibid p. 60). The object language also excludes the logical connectives ‘or’ ‘and’ ‘negation’ etc, as these words have no meanings in isolation and must be applied to words of our object language which are created prior to the secondary language.
Russell’s picture of the child learning his first words in the object language involves the child learning to associate certain sounds with objects in the environment. And understanding certain sounds in the environment used in the absence of those objects, which typically result in the absent object appearing. The child then learns to mouth the sounds in the presence of the object, and then to mouth the sound in the absence of the object, in order to get the object brought to him (IMT p. 63). Russell’s description of the child mouthing his first words is very similar to Skinner’s explications interms of Tacts and Mands .
When Russell discussed object words he said that the child learns the meaning of an object word by hearing it said frequently in the presence of the object. Russell claimed that this process can be accounted for as a form of association. The child learns to associate the sound with the object and the object with the sound. This is similar to Skinner’s discussion of Tacts, except Skinner doesn’t rely on a process of association; rather Skinner works with a system of differential reinforcement which is used to increase the probability that the child will say the word in the presence of the object.
Russell’s account of meaning is that the meaning of the object word is the object it is associated with. Eventually the child will be able to understand the object that the word means even in the absence of the object. The sense that Russell uses meaning in this account isn’t on the face of it particularly esoteric. The meaning of the word ‘Mama’ is the object typically associated with the sound.
On this Russelian picture we don’t just learn the meaning of proper nouns by direct association; we also learn the meaning of general nouns, of verbs, and of prepositions, as a result of direct conditioning to objects in the world. (‘An Inquiry into Meaning and Truth’ p. 69). Russell though doesn’t offer any help as to how people abstract out verbs, prepositions and nouns through direct association with objects in the world. Russell just asserts that these different functional features of language are learned through direct conditioning and distinguishes them from elements of the meta-language (truth and false predicates, logical particles), that are not part of the object language.
Skinner criticisms of Russell on Meaning:
As we saw above Russell’s theory of meaning for the object language is relatively straight forward and minimalist; Russell doesn’t postulate explanatory fictions willy-nilly. Yet despite this fact Skinner is critical of Russell for unwarrantedly going beyond the empirical facts when describing linguistic behaviour.
Skinner argued that Russell’s account of meaning relied on the notion of a proposition and that Russell hadn’t properly defined what a proposition is. As we saw above Russell described a proposition as something that could be said in any language. To this Skinner replied that this supposed explanation doesn’t tell us what a proposition is; or what it is made of (‘Verbal Behaviour’ p. 8). This criticism of Skinner’s is open to the objection that when a scientist proposes a theoretical entity to explain otherwise inexplicable behaviour, the onus isn’t on the scientist at the outset to settle the ontological status of the theoretical construct. Biology would be in a terrible state today if when the gene was first postulated it was rejected out of hand because we didn’t have a precise characterisation of its physical structure. In the case of the gene; we had a theoretical entity to explain otherwise inexplicable facts of heredity, and we later developed an accurate model which filled about the actual details about the physical structure of the gene. By parity of reasoning if Skinner believed that postulating genes as an explanatory explanation prior to characterising its precise physical nature was ok, then there is little reason to hold proponents of propositions to a higher standard than we held proponents of genes.
Skinner though could cast a serious doubt on the above analogy by arguing that propositions, unlike genes, do not explain otherwise inexplicable facts? One area where propositions were used was as an explanatory tool to explain different languages of the world being able to say the same thing. Skinner though rejected the claim that propositions were a necessary tool to explain the various different languages of the world:
“The audience variable is important in interpreting the traditional notion of “proposition.” If we define a proposition as “something which may be said in any language,” then instead of trying to identify the “something,” we may ask why there are different languages. The answer is that different contingencies of reinforcement involving single state of affairs are maintained by different verbal communities. A proposition is not “free to be expressed in any one of many forms,” for the form is to be determined by other variables, among them the audience. If there were only one standard and consistent verbal community, a proposition could be, though perhaps not happily, identified with “the response which expresses it.” When there are many different communities and as many different audiences, the “something” common to all of the resulting alternative “expressions” cannot be identified with a verbal form. The only common factor is among the controlling variables…there is no true synonymy in the sense of choice of different forms. When all the features of the thing described have been taken into account and when the audience has been specified, the form of response is determined.” (‘Verbal Behaviour’ pp. 174-175)
Skinner’s argument above is that the proposition isn’t a necessary theoretical postulate. We can explain the different languages in the world not by appeal to them all sharing the same internal propositional structure. Rather we can explain the different languages of the world in terms of different types of audience control.
Obviously in the sixty years since Skinner wrote ‘Verbal Behaviour’ there has been a lot of experimental data gathered on the degree to which the audience controls the form of language (e.g. Brown and Hanlon 1970, Choinard and Clark 2002, Ochs and Schieffelin 1986, etc). Furthermore there have been models developed that accept that the linguistic environment plays a big role in the structure of language, but also argues that this role is constrained by genetic instructions on the form language must take (Chomsky 1981). Models such as Chomsky 1981, don’t postulate the existence of propositions in the sense that Russell uses the term , but nor do they place almost total control of the structure of language to the linguistic audience a la Skinner. It is not the purpose of the present piece to evaluate how Skinner and Russell’s views stand up vis-a-vis contemporary linguistic science. My primary point is simply that Skinner offered an alternative conception of language that suggested that propositions weren’t a necessary postulate in explaining our linguistic behaviour. To this degree, Skinner conception showed that Russell’s adoption of propositions as explanatory tools, wasn’t dictated by necessity, but instead only seemed necessary because the alternative of different audiences controlling different linguistic shapes wasn’t even considered.
Russell further argued that a proposition was all the sentences that have the same meaning as some given sentence. Skinner correctly noted that this explanation was vacuous because we still entirely lacked an explication of what the supposed meaning of said “given sentence” was (‘Verbal Behaviour’ p. 8). Skinner noted that instead of relying on the undefined and poorly understood notion of ‘meaning’ to explain the proposition we could instead drop the unnecessary posit of a proposition and focus on the external contingencies which shaped linguistic behaviour. It was in this sense that Skinner believed that Russell was going beyond the facts when he tried relied on notions such as ‘meaning’.
A clear objection to the above argument of Skinner’s is that he hasn’t shown that the notion of ‘meaning’ isn’t necessary to explain linguistic behaviour, rather he has just shown that ‘meaning’ isn’t sufficient to justify the postulation of ‘propositions’. When discussing Russell’s conception of the object language above we noted that his explication of meaning was fairly prosaic and one that most people uncontaminated by philosophy would accept as obviously correct.
For Russell the meaning of a word in the object language was simply derived from the pairing of the sound with a corresponding object or event in the mind independent world. The child would eventually learn to remember that the sound typically signified the said object by remembering which object is typically associated with the sound.
Skinner though had serious difficulties with Russell’s crude referential explication of meaning. Skinner notes that when we move from simple cases like simple proper nouns, referential semantics begins to seem much less plausible. Skinner cites a series of words such as ‘atom’, ‘gene’, ‘minus one’, ‘the spirit of the times’, ‘nevertheless’, ‘although’ and ‘ouch’ which resist any simple attempt of definition interms referential semantics (‘Verbal Behaviour’ p. 8). Russell though wouldn’t be overly worried by Skinner’s examples. In his IMT Russell wasn’t arguing that ALL our words were derived from association with mind independent objects. Rather, Russell was arguing that some of our words were learned by association with objects in the world, and the rest of our concepts were built up in terms of various different ways of combining these sensory words into more complex meanings. Russell considered his IMT a way of using the insights of philosophers like Hume and Berkeley’s insights with into perception, with the logical analysis preferred by the logical positivists. So given Russell’s aim in IMT it is no surprise that his account of meanings bears a strong similarity to Hume’s account of simple and complex ideas.
Russell’s account of how we combine our basic object words to produce more complex ones is extremely vague. He speaks of us leaning new words based on dictionary definitions; but notes that we understand these definitions interms of words we have already learned through association with objects in the world. This is a battle that is still raging today; even contemporary psychologists such as Susan Carey who argues that we learn our new concepts by combining our primitive concepts through analogy hasn’t come even close to filling in the details. So Russell can be forgiven for being a bit vague on the details when he wrote IMT eighty years ago.
Even if Skinner accepted Russell’s vague account of how we learn complex meanings, he still had difficulties with Russell’s account of how we learn object words. Skinner noted that when we used words like ‘Cat’ it was unclear whether the word was referring to one particular cat, or the set of all cats etc (Verbal Behaviour p. 8). This objection though would have raised little difficulties for Russell:
“Thus, by the usual pleasure pain mechanism which is employed in training performing animals, children learn, in time, to utter noises appropriate to objects that are sensibly present…a child learning the object-language applies Mill’s Cannons of Induction, and gradually corrects his mistakes. If he knows a dog called ‘Caesar’, he may think this word applies to all dogs. On the other hand, if he knows a dog who he calls ‘dog’, he may not apply this word to any other dog. Fortunately many occurrences fit into natural kinds; in the lives of most children, anything that looks like a cat is a cat, and anything that looks like one’s mother is one’s mother. But for this piece of luck learning to speak would be very difficult.” (‘An Inquiry into Meaning and Truth’ pp. 66-72)
The first part of Russell’s account of using nouns like ‘dog’, ‘cat’, ‘mama’, etc is somewhat congenial with Skinner’s account. When Russell talks about the pleasure pain mechanism that is used in training performing animals, he is talking about the method of classical conditioning used by Pavlov and Watson in their labs. Despite the difference between classical conditioning and Skinner’s operant conditioning , it is obvious that operant conditioning would have been useful tool that fit perfectly with Russell’s account of how we learn the object language. So to an extent Skinner and Russell would have agreed that reinforcement for mouthed verbal behaviour would have played a role in helping the speaking subject understand words in the object language. Operant conditioning would help explain whether words such as ‘Dog’ were meant to refer to a particular type of animal; as opposed to being a name of a particular dog, as a result of the contingencies of reinforcement the child encountered in her environment.
But Russell’s second move to argue that the child generalizes sounds like ‘Dog’ to the whole species because of a propensity of the world to fit into natural kinds wouldn’t have been as congenial to Skinner. In ‘Verbal Behaviour’ in his chapter on Tacts Skinner discussed our process called generic extension. A clear example of generic extension would be a person using the word ‘chair’ to denote a new entity. Whether this new use gets accepted by the reinforcing community will be decided based on practical considerations such as, can we do the same things with it we can typically do with a chair? Skinner notes that if we want to discuss the “essence” of the word chair our best practice would be to look at the contingencies of reinforcement within the community (‘Verbal Behaviour’ p. 91). So control of the extension of the term ‘chair’ will be a social community issue. People will still be reinforced for using the term ‘chair’ in response to certain kinds of objects but the reinforcement and hence survival of how the term is going to be used is ultimately controlled by the socio-linguistic community. Skinner’s pragmatic approach focuses on sociolinguistic reinforcement to explain that words, so called essence, are not necessarily the result of correspondence with set patterns in the environment.
An obvious criticism of my explication of Skinner’s take is that it involved a bit of shuffling of the deck. Russell was speaking about words that pick out natural kinds in the environment such as ‘Cat’, or ‘Dog’ whereas I am speaking about an artefact such a ‘Chair’. However, I don’t think that this distinction makes much of a difference in this case. The standard usage of ‘fish’ picks out many different things, depending on the usage of the particular linguistic community. To a lot of people a Whale is a fish, and people will be reinforced for speaking this way. However in the biological community speaking of a Whale as a fish may elicit either negative reinforcement or punishment. So even when it comes to words about living creatures our usage in general will be controlled primarily by socio-linguistic reinforcement and the shared world we live in.
I am not here arguing that there is no fact of the matter on whether a Whale is a Fish. When one is making truth claims about the world; certain categorisations would be disastrous and hence they are not used by scientists. Given the pragmatic success of science people typically accept the scientific definition of terms, though they may not pay much heed to such definitions in their daily life. However to discuss this topic in detail I will need to delve into Russell and Skinner’s respective views on the pragmatic theory of truth. That topic though will have to be postponed until after my next blog-post where I compare Skinner and Russell’s respective take on how children acquire logical behaviour and how the developing child applies this behaviour to the object language.