Quine Skinner: Behavioural Laws and Reductionism

“Yet, Skinner and Quine do not have only different aims. If one examines Quine’s views about causal explanation in psychology, their behaviouristic theories turn out to be in fact incompatible…Even if the physiological variables between stimulus and response were to be completely specified, Skinner maintains, the laws are to be found on a behavioural level; physiologists and neuroscientists can at best fill the temporal and spatial gap between a stimulus and a response. Quine, on the other hand, defends the opposite view. He believes that behaviour ultimately requires a physiological (or better, a neurological) explanation instead.” (Verhaegh ‘The Behaviourisms of Skinner and Quine’ pp.36-38)


In his ‘The Behaviourisms of Skinner and Quine’ Verhaegh argued that Skinner and Quine held diametrically opposed views on the relation of behaviour to neuroscience. On Verhaegh’s picture; Quine believed that a true explanation is at the neuroscientific level, while the behavioural explanation is just a shallow stop gap, whereas Skinner believed that there are behavioural laws independent of what we discover in neuroscience. There is a lot to recommend Verhaegh’s interpretation of the data. Skinner did sometimes argue that neuroscientific explanations can only serve to plug up some gaps in behavioural knowledge, but that the functional laws were the most important thing:

“The physiologist of the future will tell us all that can be known about what is happening inside the behaving organism. His account will be an important advance over a behavioural analysis, because the latter is necessarily “historical”-that is to say, it is confined to functional relations showing temporal gaps. Something is done today which affects the behaviour of an organism tomorrow. No matter how clearly that fact can be established, a step is missing, and we must wait for the physiologist to supply it. He will be able to show how an organism is changed when exposed to contingencies of reinforcement and why the changed organism then behaves in a different way, possibly at a much later date. What he discovers cannot invalidate the laws of a science of behaviour, but it will make the picture of human action more nearly complete.” (‘About Behaviourism’ p. 237)

The above quote from Skinner’s 1974 ‘About Behaviourism’ is an interesting perspective on Skinner’s take on the relation between neuroscience and behavioural science. Skinner is arguing future neuroscientists will make important advances over behavioural science. This indicates that for Skinner; behavioural science isn’t entirely autonomous, and that behaviourists can learn something from neuroscientific studies. Skinner is arguing that behavioural science, like the science of natural selection is necessarily historical. If you want to establish a behavioural law you will need to do experiments that are historical in nature. These experiments will typically involve studying the three term contingency (antecedent, behaviour, consequence), to pick out a behavioural law. But with a sufficiently advanced neuroscience we may be able to discover the chemical laws that underlie the causal regularities discovered by the behavioural scientist. These discoveries in neuroscience won’t refute the discovered behavioural regularities but they will be an advance on our overall picture of the behaviour of organisms.

However it is difficult to see how Skinner’s above approach is incompatible with Quine’s approach. Consider the following statement of Quine’s (which Verhaegh quotes):

“An explanation, not the deepest one, but of a shallower kind, is possible at the purest behavioural level. One can hope to find, and I think one does find, behavioural regularities.” (Quine 2008 pp. 69-81)

On the face of it Quine and Skinner seem to be singing from the same hymn sheet; we can discover behavioural laws; but ultimately we should be able to discover more fundamental neuroscientific laws.

The obvious rejoinder to this is that while the above quote may indicate that Quine and Skinner were in agreement on this topic, a closer look at Quine indicates that he held views which are much stronger than the above quote indicates, in numerous different places he argued that behaviour is not the explanation, but something that must be explained by more fundamental sources e.g. physiology (Quine 1998 p. 94).

However, even the above claim by Quine finds resonance in the writings of Skinner:

Eventually, we may assume, the facts and principles of psychology will be reducible not only to physiology but through biochemistry to physics and subatomic physics.” (Skinner: Cumulative Record p. 302)

It should be noted that Skinner wasn’t always consistent in his views on this topic. As we saw above Skinner sometimes argued that behavioural laws are independent of neural discoveries (though they may be enriched by them). But above he is arguing that behavioural laws can ultimately be reduced to neuroscientific laws. The same inconsistency seems to dog Quine’s explanations of behavioural regularities. In some places he is arguing that behavioural regularities exist, but in other places he seems to think that such regularities are unimportant other than as pointers as to what is going on in the brain. There is obviously no contradiction in believing that ‘regularities occur’ and also believing that ‘such regularities are unimportant’. But there is a tension in the two beliefs.

There many behavioural laws that have been experimentally and observationally studied over the last few decades. An extinction burst is a clear behavioural regularity. Applied Behavioural Analysis is the most effective scientific treatment that currently exists for managing challenging behaviour. In a hospital setting, where some patients with severe learning difficulties exhibit dangerous challenging behaviour, such as, a child punching themselves repeatedly in the head; analysts must try to discover what reinforcements are maintaining such behaviours. To do this Skinner’s three term contingency is typically applied. The analyst will carefully record the instants before the behaviour occurred, the behaviour itself, and the consequences which immediately follow the behaviour. Through this process he can discover which procedures are reinforcing the behaviour.  By removing these reinforcers the analyst can extinguish the behaviour.

The process of functional extinction has been verified in many studies and across many species (‘Applied Behaviour Analysis’ p. 473). By removing the reinforcers controlling the behaviour, the analyst can make the behaviour extinct. However, prior to extinction there is an increase in the said behaviour occurring, and this is called an extinction burst (Lerman, Iwata and Wallace (1999), Goh and Iwata (1994). The occurrence of extinction bursts are well established in basic behavioural research.

When Quine says that there are behavioural regularities but that the fundamental regularities occur at the physiological level it is hard to parse what he means. In the case of extinction bursts we have clear regularities; understanding the physiology better would add to our knowledge of what is going on. But it is hard to see how the underlying physiology is any more real than the behavioural regularity which has been discovered, and which can be predicted and controlled using behavioural science. When we discover behavioural laws, as Quine admits that we do, then these laws are real patterns that have been discovered, we can learn more about the underlying causal sequences that make these patterns occur, but such real patterns are more than just pointers towards the underlying physiology they are law like facts in their own right.

Thus far we have seen that Quine and Skinner are both a bit inconsistent in their views on the relation of relation of behaviour to physiology. There is a side of Skinner, and of Quine, which comes close to endorsing a kind of crude reductionism; where the ultimate explanation is at the physiological level; with the eventual aim being to give our explanations in terms of basic physics. However, this preference for the underlying physiology as the real explanation is much more prominent in Quine’s philosophy than in Skinner’s. The general thrust of Skinner’s philosophy is that there are real behavioural laws and while neuroscientific data enrich our behavioural laws; they cannot supplant them.

Quine seems to acknowledge that we have behavioural laws but argues that these laws are just pointers we can use to get at the real data; the neuroscientific data. Quine’s position on this subject isn’t entirely inconsistent with Skinners. Both admit that behavioural laws exist, and both admit that the underlying physiology can enrich our behavioural laws. To the extent that they disagree it is on the status of the behavioural laws; Skinner takes the importance of these behavioural laws seriously, while Quine argues they are mere pointers to the real data; the underlying physiological facts.

Where Quine and Skinner’s views diverge it is pretty obvious that Skinner’s views on the nature of laws of behaviour are more accurate than Quine’s are. The laws of behaviour that are studied by behavioural scientists do much more than merely point towards underlying physiological states they are tools that are useful in the prediction and control of the behaviour of both human and non-human animals.  The success of disciplines such as Applied Behavioural Analysis are clear evidence that Quine’s dismissal of behavioural laws as mere pointers towards underlying physiology is very wrong headed.

Aside from Behavioural Analysts using and discovering behavioural laws; behavioural laws have proven to be useful tools for neuroscientists to use. In the years since Quine and Skinner were writing, conditioning has become a vital tool which neuroscientists use to understand the circuitry of the brain. Classical conditioning has proven more useful than operant conditioning in these experiments:

And work by my laboratory and the laboratories of colleagues using Pavlovian fear conditioning was very successful in achieving, in a few short years, what instrumental avoidance conditioning had failed to do- identification of the brain areas and connections between them that constituted what came to be known as the brain’s fear system.” (Le Doux: Anxiety p. 31)

While classical conditioning has proven a useful tool for neuroscientists to use when trying to understand how the brain works, these studies have also revealed some useful information about the neuroscientific basis of types of classical conditioning. In his lab, neuroscientist Joseph LeDoux has done some ground breaking work on the neuroscience of fear and has used classical conditioning as a tool. His research not only helps us understand fear but also helps us understand the circuitry of fear conditioning:

“One of the targets suggested by the tracing studies was the amygdala. When we lesioned this area, or disconnected it from the auditory system, the fear conditioned responses were eliminated. Within the amygdala, we also found an area that receives auditory CS input (the lateral amygdala, LA) and connects with an area (the central amygdala, CeA) that sends outputs to downstream targets that separately control freezing and blood pressure conditioned responses. Further, we were able to locate cells in the LA input region that received both the auditory CS and the shock US. This was an especially important discovery because the integration of the CS and the US at the cellular level was thought to be required for fear conditioning to occur. After the circuit and cellular changes involved in the process was identified, we turned to the molecular mechanisms in the LA that underlie the learning and expression of conditioned fear, many of which were the same as those discovered by Kandel and others invertebrates.” ( Le Doux: ‘Anxious’ p. 30)

Research like this is important because it provided experimental evidence of the underlying circuitry involved in fear conditioning. This is only a small piece of the puzzle; classical conditioning is a much more general process than the conditioning that occurs in fear conditioning. There is more research needed into how general the neural processes are which underlie classical conditioning in general. But research is proceeding at break neck speed and we can only hope that these general problems will eventually be solved:

“Numerous studies by my laboratory and others have confirmed that when the CS is paired with an aversive US, LA neurons do respond more strongly to the CS. Further, we and others have identified many molecules that contribute to the induction of these changes during learning and the stabilization of these changes in the storage of memory. Once the associative memory has been formed, the CS can, on its own, strongly activate LA neurons.” (ibid p. 95)

Why Classical Conditioning is more useful than Operant Conditioning is not entirely clear. In general classical conditioning is a type of learning that is useful in helping an animal passively learn from environmental experiences, and operant conditioning is more suitable for an animal to learn as it actively moves about its environment. The different functions of classical conditioning and operant conditioning may explain their relative uses for neuroscientists. A passive form of learning would obviously be more useful to in studies involving neuroscientific instruments.

                                     The Evolution of Conditioning

When discussing the work of the Le Doux lab he made a distinction between Classical and Operant conditioning in terms of their utility for neuroscientific research. This distinction is well established in the literature; since about Skinner’s time. But there is some evidence that while Operant and Classical Conditioning are not identical they may both rely on the same underlying neural architecture. In their recent paper; ‘Classical and Operant Conditioning: Evolutionary Distinct Strategies’ Bronfmann et all argue that classical conditioning and operant conditioning are different facets of the same underlying associative learning system (Bronfmann et al. p. 34). They suggest three criteria to use to help discover whether operant and classical conditioning are separate capacities or if they rely on the same underlying architecture; (1) Functional Distinctiveness, which can be inferred by double dissociations, (2) Taxonomic distinctiveness: members of one animal taxa will have one system (CC), while members of another animal taxa will have another system (OC), (3) Adaptive evolutionary distinctiveness:  distinct forms of learning should have distinct evolutionary rationales (ibid p.35)

In answer to their first question they note that there has been some experimental research indicating dissociation where through brain damage a creature can learn through operant conditioning and not classical conditioning (Brembs et al 2008, Lorenzetti 2006, Ostland, et al 2007). However they note that there are only a few experiments indicating this dissociation is possible and that these studies haven’t been replicated. So before drawing any large scale conclusions more research is needed. On question two they claim that there no evidence of any animal who possesses one type of conditioning but not the other. Again research is in its infancy and more research is recommended. On the third question they argue that given that OC and CC are paradigm domain general learning processes it is unlikely that theorists will be able to construct a plausible evolutionary rationale of them being selected for in different way.

On the whole then a theorist who wanted to argue for two distinctive processes underlying OC and CC could appeal to the few experiments indicating that dissociation of the  OT and CC is possible. But overall there would be very little evidence to support their views on the topic. So Bronfmann et al argue that despite the consensus in behavioural science there is little evidence to suggest that we should adopt an absolute distinction between classical and operant conditioning.

With new evidence chipping away at the neuroscientific nature of conditioning, with cross comparative and experimental data being use to discover if classical conditioning and operant conditioning use similar underlying neural circuitry, and even some data on the evolution of conditioning we are learning much more about conditioning than either Quine or Skinner knew. And so far everything we have learned seems to support the less reductive position than the one Quine proposed. We are learning more and more about conditioning and its neural basis; but this hasn’t come close to reducing the behavioural regularities to mere pointers to underlying states. Rather despite what we have learned behavioural analysts are today are still using behavioural laws (some of which were discovered by Skinner), to shape the behaviour of human and non human animals. There is no reason at present to follow Quine in treating behavioural laws as some kind of shallow explanation.

The sense of fundamental Quine typically appeals to is one that relies on a strong sense of physicalism:

“Nothing happens in the world, not a flutter of an eylid, not the flicker of a thought, without some redistribution of micro-physical states. (Quine ‘Goodman’s Ways of World Making’ p. 98)

Quine’s above statement that all forms of behaviour depend on some kind of underlying microphysical process is relatively uncontroversial. It is hard to imagine a behavioural scientist who would object to the claim that any behavioural laws discovered will have an explanation in terms of underlying physical processes. Likewise, it is hard to imagine an evolutionary scientist who would deny that all examples of natural selection have underlying physical causes. But it obviously doesn’t follow that because a process is causally dependent on underlying physical states that the process is shallow piece of information that will ultimately be explained away.

There are real patterns that exist in the world that will be missed out on if we try to understand something at the wrong level of abstraction. If we stick to just understanding a portion of the world in terms of subatomic particles and forces acting on them then our explanation will be incomplete; such an explanation will be entirely blind to things like sexual selection. The fact that sticking entirely too fundamental physics will blind one real patterns at the evolutionary level obviously doesn’t mean that physics is irrelevant to evolutionary theory.

Physics can provide constraints to what type of creatures can be built by natural selection; see for example work on scaling laws and invariants in animal locomotion Bejan and Marden (2006), and Trevisian et al. (2006) on the physics of bird songs. Now suppose one adopts the Quinean approach to evolutionary explanations of the origins of life, and argues that explanations interms of natural selection are shallow, the real explanation is at the level of basic physics or chemistry. As we saw above if we adopted this approach in its reductionistic sense we would be ignoring real patterns in the world and explaining away patently real phenomena. A less radical approach would be to accept that physics can constrain, and inform explanations in evolutionary science but not supplant.

When it comes to behavioural science and biology things are similar. The degree to which Quine and Skinner disagree on the status behavioural laws and their relation to neuroscience; isn’t always clear. But it is clear that any radical reductionism that tries to reduce behavioural laws to mere pointers to underlying neural states is untenable.

4 thoughts on “Quine Skinner: Behavioural Laws and Reductionism

  1. brembs

    Great comparison of Quine*’s and Skinner’s positions. However, further down, you write:
    “However they note that there are only a few experiments indicating this dissociation is possible and that these studies haven’t been replicated. ”
    Since then, there have been several studies sowing the classical/operant distinction in several animals:
    Aplysia: https://www.ncbi.nlm.nih.gov/pubmed/18786364
    Drosophila: https://www.ncbi.nlm.nih.gov/pubmed/24964149, https://www.ncbi.nlm.nih.gov/pubmed/19576773, https://www.ncbi.nlm.nih.gov/pubmed/18674907
    Mice: https://www.ncbi.nlm.nih.gov/pubmed/22021859

    Importantly, all of these studies reveal the same critical dependence on key molecular genetic components such as PKC and FoxP. The latter, in particular is instructive here, with regard to Skinner’s “Verbal Behavior”. Chomsky famously eviscerated Skinner’s argument that language acquisition and operant learning were “mere homonyms, with at most a vague similarity of meaning”. After the discovery of the FOXP2 gene in humans being involved highly specifically in articulation and speech, testing manipulations of that gene first in birds and then in an operant paradigm in flies, showed that FoxP appears to be a gene whose likely ancestral function was in operant learning, i.e., supporting Skinner’s account in “Verbal Behavior”. This function appears to be distinct from other forms of learning, such as classical conditioning, because a whole host of FoxP manipulations have been tested and none show any classical conditioning defects in any animal.

    Hence, we now have a growing body of very consistent evidence that operant and classical conditioning double-dissociate on the genetic level and it appears that this dissociation was already present in the Urbilaterian, last common ancestor of vertebrate and invertebrate animals, more than 530 million years ago. I have reviewed Bronfman’s PhD thesis and talked to his co-authors several times of the last few years – I did not read or hear any objections to these conclusions.

    1. surtymind Post author

      Thanks very much. I wasn’t aware of this data. It is fascinating and seems very important. I will have a read of the links you shared and get back to you. Looking forward to getting stuck into it.

  2. Rodrigo

    Hello, on the relationship between Skinner´s radical behaviorism and neuroscience, I quite like these two papers. Do you know them?

    Zilio, D. (2016). On the autonomy of psychology from neuroscience: A case study of Skinner’s radical behaviorism and behavior analysis. Review of General Psychology, 20(2), 155-170.
    Zilio, D. (2016) . Who, What, and When: Skinner’s Critiques of Neuroscience and His Main Targets. The Behavior Analyst , v. 39, p. 1, https://www.researchgate.net/publication/293958867_Who_What_and_When_Skinner's_Critiques_of_Neuroscience_and_His_Main_Targets


Leave a Reply to Rodrigo Cancel reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s