A Touch of Alyricism

Dedicated to the equally fascinating topics of autistic advocacy and the 'sisterly sophistries' of radical gender feminism. Other topics may occasionally crop up. Contactable at alyric@gmail.com


Polemicist since Grade 8

Tuesday, June 13, 2006

Teetering on the Tabula Rasa (in draft)

In 2004, Ipsos Reid Canada ran a poll on behalf of Families for Early Autism Treatment, British Columbia (FEAT BC). [1] The poll asked ordinary Canadians two questions:

1. “Whether or not they believe that children who suffer from autism should be covered by their provincial healthcare program to receive an Intensive Behavioural Intervention (IBI but more commonly known as Applied Behavioural Analysis (ABA), which has been scientifically proven to effectively treat autism.”

2. After being read a statement including the results of the Auton case and this statement: “It is estimated that IBI, a scientifically proven effective treatment for children, which costs a family approximately $60,000 a year, would save taxpayers 1-2 Million dollars over the lifetime of an autistic child”, Canadians were asked their opinion on: “even though not obliged to cover the cost of IBI, the provincial government should do so anyway”.

Twice in this poll ABA is given the ‘scientifically proven’ seal of approval. The question is, does ABA deserve this frequently cited accolade? Should Science even appear in the same sentence to lend an air of verisimilitude to the proceedings? Certainly, behaviorist literature is sprinkled with ‘science’ or ‘scientific in every second paragraph, somewhat reminiscent of the significance given to the presence of ‘democratic’ in a country’s title.

The belief of some of the greatest minds of the 20th Century is that Radical Behaviourism (hereafter behaviourism) [2], the wellspring of ABA, is a suspect enterprise and not close kin to Science as most folk think of it. Experimental does not equate automatically to ‘scientific’. The belief of a number of others working in the field of autism intervention strategies is that it is also not proven to be especially effective. Behaviourism likes to style itself as ‘scientific’ for much the same reason that anyone claims to be scientific: people will regard your work as the result of objective methods based on sound theoretical principles and will assume that work to have the same reliability and replicability that is usual in scientific endeavours.

What’s true in these theories is not new and what is new in these theories is not true. - Hans Eysenck [3].

Eysenck levelled this charge at Freudian psychoanalysis. The charge could have been levelled with equal accuracy at behaviourism. There is no doubt that behaviour is influenced by the environment, but problems arise with the inference that this influence can be codified into a system, a science even, and those problems multiply exponentially when such a system is applied to complex human behaviour. Behaviourism owes its existence to BF Skinner who developed its principles (I use the term very lightly here) and its working vocabulary. Skinner’s work with rats and pigeons formed the basis for the stimulus – response – reinforcer unit of basic behavioural analysis. It was axiomatic for this generation of behaviorists that the basic principles of animal experimentation could be applied across species and to any variety of human complex behaviour. Skinner recognised that if these principles could be applied to complex human behaviour, they would have to explain that most human of all complex activities: language. His classic text, Verbal Behaviour gave a blueprint for this application. Here, Skinner noted that:

the basic processes and relations which give verbal behavior its special characteristics are now fairly well understood ... the results have been surprisingly free of species restrictions. Recent work has shown that the methods can be extended to human behavior without serious modification - Skinner 1957 as quoted by Chomsky

This bold claim did not go unchallenged. Noam Chomsky on behalf of linguists everywhere and other grateful humans delivered a scathing critique not only of the book but also on the methods of behaviourism in general. He pointed out in great detail the lack of precision in Skinner’s basic definitions of stimulus, response and reinforcer. While conceding that the laboratory definitions may have had some specificity when restricted to animal experimentation in Skinner boxes (the epitome of a controlled environment), Chomsky made it clear that in the world of complex behaviour such as language, it is far from easy to match up the response to the stimulus in a way that could meaningfully be interpreted as prediction or control, regardless of Skinner’s definition of verbal behaviour, which is quite explicit. The verbal operant as defined by Skinner is a class of responses of identifiable form functionally related to one or more controlling variables. The basic problem, is identifying what controls what. For example, the ‘stimulus’ chair might elicit the response, ‘chair’ or ‘red’ or ‘square’ or ‘my mother-in-law’ because she owns one just like it. The ‘lawful’ relationship only becomes apparent after the event or as behaviorists like to say, the controlling stimulus is inferred post hoc. By way of example, ‘red’ as a response indicates that it is under the control of ‘redness’. This has to be a unique variety of empiricism bearing no resemblance to its scientific counterpart.

"It is on the whole an inappropriate methodology in developing improved cognitive images of complex, unstable systems with changing parameters and cumulative structures, where rare events are significant. Humans are a supreme example of systems of this kind" K. E. Boulding (1984)

Likewise, reinforcement has a reasonably precise definition in the extreme minimalist environment of the Skinner box. Stimuli, which are reinforcing, increase the frequency of a response. However, in usage with ‘verbal behaviour’ or other complex human behaviours, the term loses every bit of specificity, becoming a series of synonyms for like, need, want and so on; and this is how it is used in behaviourist experimentation, which is puzzling considering that the reinforcer is pivotal to behavioural analysis. Nevertheless, in one experiment, the reinforcers are “empirically identified food reinforcers (candy for Jimmy and chips for Rob)”. In another “Edible items that had been previously established as reinforcers were presented to the child after approximately every third correct mand”. There is no data provided on the capacity of these reinforcers to do their job, that is, to increase the frequency of a desired response. Howard et al seemingly forgot that behaviourism is supposed to be a scientific enterprise and elected for the politically correct route. In their experiment, “Children were taught to select their own reinforcers, record their own performances and sequence their learning activities as appropriate for each child”.
The behaviour analyst has a major problem with the inner man and the notion that innate factors, which are unobservable have an impact on behaviour. They claim the feat of allowing private behaviour but not what folk psychology, that is, most of us, see as the mind. Most ignore it altogether. As Skinner stated:

"The practice of looking inside the organism for an explanation of behavior has tended to obscure the variables which are immediately available for a scientific analysis."

With that in mind, examine the schematic representations of functional units of analysis, in Murray Sidman's paper on equivalence relations.
Note that half of the behaviour going on is of unknown origin. It simply is. The assumptions of behaviour analysis are twofold: innate factors are of less significance than environmental factors and that behaviour that can be seen but is unexamined is of no significance either in itself or its effect on the target behaviour. Neither is supported by data of any kind and neither enhance the credibility of behaviourist claims to be able to either scientifically predict or control behaviour. Both encourage the experimenter to fit the data to predetermined outcomes and ignore any that are discordant. This is not how the scientific method works, but appears to be very common in behaviourist experimentation.

"Men dare not avow, even to their own hearts, the doubts which they entertain on such subjects. They make a merit of implicit faith; and disguise to themselves their real infidelity, by the strongest asseverations and the most positive bigotry." - David Hume

In Clark and Green 2004, there are only two experimental subjects. The method informs us that the subjects are profoundly retarded, yet one of them proceeds to perform a match to sample of two sets of stimuli without a mistake. This tour de force requires no explanation since none is given, though an independent observer might question the entire foundation of the experiment, which concerned researching effective ABA teaching methods, based on that datum. Murphy et al have a complicated experimental set-up designed to showcase the possibilities of combining Skinnerian applied behavioural analysis with Relational Frame Theory (RFT). The three experiments are designed as methodological replications with minor variations. Notably, two of three participants have a lot of difficulty performing one section, reverse derived manding. The first study subject managed the task without problem, but not the second or the third, though some explanation was attempted to explain the third’s performance in terms of comorbid ADHD. No mention was made of the second subject’s difficulties with this task or more importantly, why this section was dropped in its entirety from Experiments 2 and 3. This is not quibbling. Reliable demonstration of reverse derived functions supposedly supports the concept of mutual entailment, a pillar of RFT. However, these peccadillos fade into insignificance compared to Sallows and Graupner’s efforts. Their experiment was designed as a replication of the classic 1987 Lovaas ABA, which demonstrated a 48% recovery rate from autism (based on survival in mainstream classes and IQ in the normal range). Essentially, the experiment didn’t work as intended. The control group outperformed the experimental group 2:1 in terms of achieving ‘recovery’ on Lovaas criteria. So, the authors, with no explanation, combined the results of both groups and the data was interpreted de novo as the differences between ‘rapid’ and ‘moderate’ learners and what factors might be the correlates for the two. While this interpretation of the research data is both extremely interesting and of considerable value, there is no explanation given for why this experiment failed in its primary aim and so spectacularly. The data suggest that behaviourism and its techniques have little to do with ‘recovery’. Indeed the hypothesised correlates of certain characteristics with rapid or moderate learning rates may suggest even more strongly that these characteristics are the major determinants of progress independent of any intervention. There is an inherent deficit in ethics here with ABA programs, that bears further examination and this will be addressed later. The other non-scientific feature of Sallows and Graupner’s work is the disappearance without trace, or explanation, of the second control group. It appears in earlier reports of the work but not in the final report.

Chomsky’s other argument against behaviourist claims to adequately explain and describe language was the obvious disparity between the time required for a history of reinforcement to develop a repertoire of language and what happens in real time, a phenomenon known as ‘poverty of stimulus’. If a repertoire in one’s language depends on the history of reinforcement in that language, then most people might be able to manage a declarative sentence in that language in their lifetime, Shakespeare would be out of the question. This notion is particularly questionable in small children whose rapid rate of language acquisition in a remarkably short time tends to rule out behaviourist explanations.

Behaviourism has of course moved on since Skinner and has attempted to answer Chomsky’s poverty of stimulus critique by postulating shortcuts to learning through theories of equivalence relations and later through RFT. However it doesn’t mean that the approach has become any more scientific.

All three of the main theories are adaptable to any outcome, thus making any empirical evidence to the contrary unlikely: - Clayton & Haye (1999) as quoted by Sidman.

This criticism was levelled at equivalence relations. The concept arose when in
Murray Sidman’s 1971 study, a developmentally disabled subject who had learned to match spoken words to pictures and spoken words to printed words, spontaneously matched printed words to pictures in the absence of training to do so. Sidman saw a way to explain both rapid language acquisition and the ‘emergent’ or novel nature of most of it. From this ah ha! erlebnis sprang a new wing of radical behaviorist experiments of language acquisition. In all the enthusiasm, no one seemed inclined to acknowledge that this phenomenon is hardly new, it goes back to Harlow’s 1949 learning set theory, and is not particularly behaviorist, even though it has been commandeered as such. Tacking on the behaviourist accoutrements of stimulus, response and reinforcer doesn’t seem to have developed the concept very far. Sidman, 2001 acknowledges that thirty years after the initial study, it is still not known, in behaviorist terms, where equivalence relations are formed, though he is optimistic about it. But then, the behaviorist’s capacity to satisfactorily explain ‘verbal behaviour’ is still in the foetal stages and after 50 years gestation, may yet be stillborn. According to Leigland:

Perhaps a direct analysis [of verbal behavior] is beyond our reach, in part because it may be beyond the ethical bounds of human experimentation, in part a casualty of myriad embedded interactions, non-linear dynamics, and massive complexity.

This opinion was bolstered by the participants at a behaviorist symposium in 2005. Sautter et al said: Skinner’s framework has not yet fully impacted the experimental literature, followed by Sundberg: There is much more research that needs to occur before it could be said that the necessary and sufficient empirical supports exist. Both are, however, optimistic.

Gina Green, a major advocate for ABA as the ‘scientifically proven’ treatment regime for autistic children is rather fond of pointing to the 50 years of research supporting behaviorism. There is certainly 50 years worth but the jury is still out on exactly what it supports.

The latest development in behaviourist lore, RFT, attempts to broaden the scope of equivalence relations to create multiple relational frames, of which equivalence is only one example. Others are described by a specific feature such as co-ordination, opposition, comparison and so on. According to the authors, language is merely the art of framing things relationally, an interesting concept.

RFT is in the main, a controversial topic in behaviourist circles, mostly for its willingness to discard large amounts of Skinnerian received wisdom, such as Skinner’s definition of verbal behaviour and not least for the messianic fervour of its architects. Jose Burgos related: I found the concepts, logic, and justification of RFT to be unintelligible and when not, incoherent, misguided, trivial and sterile. And as is now a commonplace with behaviourist propositions, RFT is not yet empirically justified, but the authors of the concept are optimistic.

When sets of functional analyses exist within a behavioral domain, broader generalizations can be made to explain complex forms of behavior. These generalizations are theories, but they are analytic abstractive, not hypothetico-deductive: “Analytic abstractive theories are simply organized sets of behavioral principles that are used to help predict and influence behaviors in a given response domain.”Hayes & Berens 2004.

Theory in behaviourism is fundamentally different to the concept that the scientific community recognises. Apparently, Skinner was quite happy that this new ‘science’ would progress more rapidly unhampered by the manufacture of coherent hypothetico-deductive theoretical concepts. Behaviourism could thrive by adhering to inductive strategies tied to the purely functional description of what is occurring - a monument to Pragmatism. Science operates by other means, for other purposes and under other philosophical constraints Science seeks to explain phenomena, to answer why to the best of our knowledge this or that is so. Behaviourism, as Zuriff clarifies does something different.

SIB [Skinner Inspired Behaviour] explains events by showing that they instantiate behavioural principles. These principles, in turn, are explained as derived from more fundamental principles. Ultimately, explanation reaches its bedrock in the form of primitive principles such as the law of operant conditioning with no further explanations forthcoming.

Only a behaviourist or a naïve inductivist could call this Science. Mathematically, the probability of a generalisation from a finite number of observations over time being ‘true’ is zero (the inductivist fallacy)[4]. Intuitively, if one looks at the generalisations from elastic data generated from imprecise definitions in the behaviourist’s arsenal, the probability of any of it being ‘true’ is also zero. Science proceeds from explanatory theories, which predict testable hypotheses automatically as a function of the logical coherence of the theory. Behaviourism proceeds from a purely Pragmatic position - it’s true if it works. Burgos says –

Under Jamesian Pragmatism, anything goes, even nonsense, as long as it is useful to someone.

This philosophical difference more than any other factor underlines the non-scientific nature of behaviourism. Pragmatism says that inconsistencies in experimental data can be safely ignored if they add nothing useful to the experimental outcome. Science says that inconsistencies in experimental data point to incoherence in the explanatory theory and as a matter of logic, may never be ignored.

"Behaviorism is indeed a kind of flat-earth view of the mind." Arthur Koestler

Behaviourism has a unique view of the person. Behaviourists claim to acknowledge private, covert behaviour with the caveat that it is controlled by the current environment or a previous one all the way back to Cromagnon man. Man is a mere marionette on the strings of his reinforcement history. As Skinner said: To man qua man, we readily say good riddance. The things to be got rid of include creativity, self determination, personal responsibility, free will, volition, judgement, anything in fact that characterises man as an agent of his own destiny. Here’s the Skinnerian view of some of this list from the architect himself:

On self: - The self is a repertoire of behaviour appropriate to a given set of contingencies.
- Self knowledge is of social origin. It is only when a person‘s private world becomes important to others that it is made important to him.

On feeling: - I miss you could almost be thought of as a metaphor based on target practice, equivalent to “My behaviour with respect to you as a person cannot reach its mark”, or “ I look for you and fail to find you”.

On will Like “idea”, “will” is used almost interchangeably with behaviour.

On the mind: - Mental life and the world in which it is lived are inventions. They have been invented on the analogy of external behaviour occurring under external contingencies.

On ethics: Ethos and mores refer to the customary practices of a group.

Behaviourists have objected to a very common perception that they view the human as the blank slate, the veritable tabula rasa, infinitely malleable, just waiting for the history of reinforcement to etch a path. In the light of the above, I don’t see how they can avoid the charge. A number of behaviorists have intimated that it is perfectly possible to use behaviorist principles of analysis and leave the philosophical view outlined above strictly alone. This stratagem is not realistic. Your 3 or 4 term contingency analytic unit has enough trouble with post hoc attribution of stimulus control not to mention observable behaviour, which remains unexamined, compounded with a 'technical' vocabulary of less precision than standard English. Admitting that the subject of the analysis could have sufficient autonomy to behave on a whim would render any claims to behavioural prediction and control simply deus ex machina.

This philosophy has specific consequences for autistic children. If ethics are merely a social product, it follows that it is perfectly acceptable to change the individual to accommodate the society. Further, this primacy of the social imperative has I would think, several negatively reinforcing consequences for autistic individuals arguably the archetype of autonomy. The definitive review of ethics or lack thereof in behaviourism can be found here.

Behaviorism is not Science, has never been Science and does not look as if it will make the grade at some point in the foreseeable future. It remains a proposition short of a paradigm.

If, however, I treat every training goal as a "conditioning process," then I am no longer open to other possibilities, as to either what is going on or what needs to be done. Hild - on dog training.

The second half of Mr Wright’s claim embedded in the Ipsos Reid Poll was that ABA was ‘proven’ to be effective. Behaviorists will claim that indeed it is and point the reader firmly in the direction of the 1987 landmark study conducted by Ivar Lovaas, which found that after a minimum two years intensive ABA, 9/19 children were ‘best outcome’ in that their IQ was normal and they were surviving in mainstream classes. This study has never been replicated. Some studies come close to Lovaas’ findings but most do not. All of them, including the Lovaas study, have major shortcomings and these have been exhaustively documented elsewhere. The best critique of the Lovaas study was conducted by the Centre for Health Services and Policy Research of the British Columbia Office of Health Technology Assessment (no small irony there). Herbert et al have compiled a comprehensive review of all interventions for autism. I plan to make only a few points here.

It is incredibly difficult to test for effectiveness of an autism intervention and no surprise that only one, Smith, Groen & Wynn, 2000 [5], qualifies as scientific, in that this one had truly random assignment to experimental and control groups. The results were not nearly as impressive as Lovaas. Two of fifteen met the Lovaas criteria for ‘best outcome’ (13%). There were no statistically differences between the experimental and control groups on language, socioemotional functioning or adaptive functioning.

There are factors in studies of this type, which compromise the claims of any intervention, notably the use of IQ as an outcome measure. As noted here:

The lack of validity of using IQ scores with children with autism has been noted in several research studies (Lincoln, Allen, Kilman, 1995; Lord & Schopler, 1989; Lord, Schopler, & Revecki, 1982). These studies indicated that, while IQ scores for a speaking autistic child were likely to remain in the same range, the converse for a nonverbal child was not true. In fact, in one study, 50% of the nonverbal autistic 3-year-olds showed increases of more than 30 points when reassessed 5 to 8 years later.

Most of these effectiveness studies are conducted starting with a non-verbal child and ending with a child in the process of developing language. IQ gains could be pure serendipity. The problem appears to be in the realm of accurately assessing intelligence in non-verbal autistic children. The charge was levelled at Lovaas by Schopler that he specially selected his experimental group on the basis that the older group were echolaic and therefore showing good signs of a capacity to develop language and likely to have good outcomes overall.

Universally, there is a problem with treatment integrity. No one thinks it important to document accurately exactly what people are doing. Sallows and Graupner apparently used an array of techniques, some not usually associated with ABA, though they claim to be assessing the effectiveness of ABA. Their study, as has been mentioned earlier, is the best evidence to date that it really may not matter what the intervention is, since ‘less’ and ‘less professional supervision’ seem to have vastly better outcomes.

A significant confound of these studies is that small variations in initial factors can have large effects over time. This is a particular problem with the Howard et al study. There are statistically significant differences between the experimental and control groups in some factors, notably non-verbal intelligence between the experimental group and the AP control group, which the researchers reduced to non-significance by combining means for both control groups, an interesting statistical manoevre. However, for this factor and many others, small initial differences can lead to very large differences over time and ‘control’ becomes meaningless because it assumes that the magnitude of the difference does not change over time.

The major obstacle in these studies seems to be in matching control groups to experimental groups in terms of an equal starting position. On occasion, the obvious differences appear to be too much in favour of the experimental group. The classic example is the gender ratio of the Lovaas study. It is well known, and was then, that autism in females seems to be a more severe syndrome on average with less favourable outcomes. Lovaas had a ratio of boys to girls of 5.3:1 in the experimental group and 1.4:1 in control group 1. Other factors can also be very different between the control and experimental groups. Howard et al went to considerable trouble to document the educational attainment levels of the parents in both groups. They neglected the same criteria in their therapists. The experimental group had undergraduate and some post-graduate psychology students as therapists. The control group therapists were described as classroom aides, whatever that means.

When behaviorists laud the effectiveness of their techniques, they seem touched with curiously selective vision – they only look at the ‘good’ data. No one ever points to the significant minority of children out there in ABA-land who are not only not succeeding but getting worse. Behaviourists may like to claim that these children would regress anyway, but that doesn’t really let them off the hook. The as yet untested hypothesis is that for some children ABA is harmful and on that basis alone Ipsos Reid had no business encouraging Canadians to think that Governments should sanction this particular intervention for all children. Table 2 in Sallows and Graupner’s study represents the progress over time with IQ and other factors. Taking a look at the worst achievers, is there something in the ABA handbook that says it is acceptable to continue the same strategy year after year on the same downward spiral? Birnbrauer and Leach’s small and problem-plagued study is frequently cited showing the IQ gains made by 4 of the participants. No one mentions that the other 5 in the experimental group actually lost IQ points. When things aren’t going well there seems to be no consideration that something other than ABA should be tried. Within ABA there is such an air that this is the one true way that research from another camp in psychology however trailblazing, can be safely ignored. This is ethics violation at its worst.

The research results for ABA’s effectiveness as a treatment for autistic children are all over the map, which shoots down the pretensions of the discipline to scientific rigour. It doesn’t work all that well and certainly not for all autistic children. It’s an ideology, just another fashion in intellect in the Humanities and as balkanised as any other faculty in that area. Given the view of the person that this intervention is based on, I feel that it should come with a warning label – Keep away from small children.[6].

End Notes and Book References

1. Unsurprisingly, 85% of the 1000 respondents agreed with the Poll statements. Ask any section of a Western democracy if the government should pay to cover the costs of treating those poor suffering children and the answer will be yes, particularly if the respondent has the haziest idea of autism and no clue what an ABA program looks like. People trust their government implicitly to save a buck wherever possible. Given their trusting nature one would have to question why the affirmative response rate didn’t approach 100%.
Not all Canadians warmly welcomed this poll, especially autistic Canadians such as Michelle Dawson. Ms Dawson had an email exchange with Mr David Wright, Vice President of Ipsos Reid Canada, and set up a Quick Topic forum to discuss the issue. Mr Wright didn’t understand what the fuss was about. After all, he was on the side of the angels and all those poor suffering autistic children. He began to gather some inkling apparently, one year later when a random search for Ipsos Reid on Google turned up the Quick Topic forum. Much of that discussion centred round the use of this ‘scientifically proven’ ABA treatment to prevent effeminate boys turning into homosexuals. Now while it is perfectly acceptable these days to have autistic children in ABA programs, it is absolutely not OK to attempt to change someone’s sexual orientation using the same methods. Suspecting, no doubt that he’d been slipped a Mickey Finn by the poor suffering parents of FEAT BC, Mr Wright elected to utter threats of dire legal action to the softer target, Ms Dawson, and the Quick Topic forum was subsequently closed down.
Closing down a fairly innocuous discussion, censorship in other words, to protect your company’s image is certainly not cricket, but the real problem is the dissemination of more misinformation to 1000 innocent Canadians. Incidentally, I like to work with the paper copy of documents so in total innocence, I have a copy in PDF of the original Quick Topic forum discussion. Send me an email if you would like a copy.

[2] There are a great many varieties of behaviourism. The one under examination in this essay is Radical Behaviourism because it underpins behavioural analysis and therefore underpins ABA interventions in autistic children. The 20th Century thinkers that come to mind include Ayn Rand, Noam Chomsky, Arthur Koestler and the philosopher A J Ayre.

[3] Eysenck, H, Decline and Fall of the Freudian Empire, Harmondsworth, Middlesex, England ; New York, N.Y., U.S.A : Viking, 1985 p 34.

[4] Chalmers, A.F, What is this thing called science?, St. Lucia, Qld. : University of Queensland, 1999. p
[5] I don’t have access to Smith, Groen & Wynn but I think I’m justified in using Dr Gernsbacher’s review. She is the President of the APA after all.

[6] I know of a number of reasonable parents who are really enthusiastic about this intervention, which says something I think about how little is really known about learning and how resilient most children are. It is also about where behaviourism is headed these days – into oblivion as ABA therapists use a hodgepodge of techniques and pay lip service to the tenets of the field. Now if we could persuade them to ditch the cognitive deprivation approach in favour of autistic learning styles as indicated by current research, all would be well. What they would lose on the merry go-round of pseudoscientific authority, they would win on the swings of ethics as well as efficacy.