• The Autism Wars
    • ABA as a scientific discipline v. ABA as a field
  • Evidence Basis
    • Research Designs Used in ABA
    • Where Most of the Evidence Basis for Autism Treatment Comes from
  • Applied Behavior Analysis
  • Concluding Comments
    • Who’s Right?
    • A Medical Analogy to Autism Therapy

The Autism Wars

Currently the field of autism and behavior intervention is embroiled in philosophical, political and business wars regarding intervention for autism.  The field of ABA is strongly tied to the legacy of B.F. Skinner and his theories, and the most common and widely known autism intervention methodologies strongly bear his footprint.  Most autism treatment methods currently in use were developed by Behavioral Science Psychologists.  There is now an official board that certifies and qualifies Behavior Analysts called the Behavior Analyst Certification Board (BACB), and several professional associations for Board Certified Behavior Analysts (BCBA).  The ‘Behaviorist‘ discipline currently dominates the field of autism intervention.

While I believe it is absolutely necessary to certify the qualifications of those practicing ABA in order to protect the public from quackery and unqualified or unethical practitioners, I think it is highly unfortunate that many in the field have become politically partisan and politically active, almost to the point of religious devotion.  This, along with some more down-to-earth business interests is what I believe is behind initiatives in many states to exclude payment for services for what they consider “alternatives to ABA” such as Floortime and RDI, as well as to lump such valid therapies with quackery (there’s definitely quackery out there).

Infighting over exclusive rights to autism treatment can deprive parents of choices of intervention and inhibit innovation.  Further, the efforts are based on what I believe are mistaken notions of Skinner’s philosophy of the science of human behavior “Behaviorism,” and what the founders of Applied Behavior Analysis (Baer, Wolf, and Risley 1968) really intended. For the sake of parents I would like to see the wars end so they can have choices. For the sake of practitioners, I would like to see a rapprochement in the field that I think is entirely possible.

We at SCS believe that there is no single intervention that works with every child and that no intervention can stand alone for any one child. A clinician that uses one treatment to solve all problems is bound to overextend the method.

We hold that ABA can function as a Master Theory of Behavior, along with other Master Theories such as Systems Theory.  Behaviorism is not an exclusive theory.  Until recently, it’s linear, algorithmic way of analysis bogged down heavily when it came to analyzing long chains of reciprocal and constantly changing behaviors such as conversations and spontaneous social interaction.  Current advances look to Systems Theory and other models of multivariate influence (e.g., Contextualism; Ethology and Evolutionary Psychology, Interpersonal Neurobiology to name a few) for analysis, but no one has published a means of considering these variables in a functional analysis of behavior – yet.

Master Theories of Behavior are strong enough to cohere with many other scientific disciplines related to behavior, from Anthropology to Zoology.   No field of psychology, behavior or development in humans and animals has ever repealed Skinner’s laws of operant conditioning.  Those fields of study can only add more sophistication, elaboration and nuance to them and should be welcomed with open arms and minds.

Evidence Basis

To determine whether any technique is “evidence-based,” one must conduct an Applied Behavioral Analysis of the technique and the results (below).  In other words, the technique or the body of techniques involved in a method must be valid (applied), reliable (behavioral) and analytic (high quality).  I explain what I mean by that below.  What I also try to emphasize in the rest of this article is that there is a difference between ABA as a process, and ABA as a field and a body of practitioners that call themselves Behaviorists.

More than anything, ABA is a process.  It emanated from the desire to make psychology a true science.  It utilizes the same practices of experimentation and empirical evaluation as the so called “hard sciences” such as biology or physics.  Like the rest of science, it is a constant struggle for validation.  Unlike faith, it relies on challenges to claims practitioners or advocates of one theory or method or another make in order to strengthen its validity and reliability.

Currently, there is widespread belief that only the techniques and methods developed by behaviorists meet the standards of ABA.  There are several methods out there, namely DIR/Floortime, RDI, the Miller Method, SCERTS that meet the standards of ABA outlined by the founders of the field and ABA’s current dimensions.

At SCS, we are not conducting research or experiments.  We are however, intelligent consumers of research literature and are very well aware of the research foundations of the methods we use.  In the sections that follow, we will explain the types of research there are available in order to comment on where the evidence basis comes from for the therapies mentioned here.

An aspect of experimental design (below) that we always use is that called “Baseline Logic.”  That is, we always have to start with an educated guess or a prediction about how a technique will work when applied.  We never really know because we’re dealing with human beings – very complex phenomena indeed.  We then have to try it out and verify the results.  Finally, we have to see if it works consistently, and if so or if not, what variables account for the consistency or lack thereof.

Randomized Controlled Trials, Experimental Control and Evidence Basis

Since the founding of the field of Behavioral Psychology as far back as the beginning of the 20th Century, there have been several laboratory and applied techniques developed to determine the evidence basis of behavior change techniques and packages usually referred to as “methods.”

The “Gold Standard” is the Randomized Controlled Trial or RCT.  This is primarily a laboratory technique, which is one reason why so few of the ‘methods’ out there have RCT studies completed and replicated.  It is very often used in the pharmaceutical industry to determine the effectiveness of new drugs.  Experimenters of an intervention or technique divide a larger group of potential “subjects” into a “Study Group” and a “Control Group.”  The larger or whole group must be matched in ways that allow “apples-to-apples” comparisons.  The similarities of this group have to be believably relevant in ways that depend on what is being tested.  For instance, a drug being tested for the treatment of diabetes might restrict the whole group to individuals of a certain age, health status, diagnosis, and whether or not they are on other medications.  The validity of the results of the experiment will depend upon the homogeneity of this group.

The Study Group receives the intervention and is then Compared to a Control Group.  Subjects are assigned randomly to one or the other group, to prevent “cherry-picking” those most likely to benefit from being assigned to the Study Group. In the studies of behavior change methods and in particular, the larger treatment packages or “methods,” the Control Group is usually homogeneous by comparison, but they cannot be completely “controlled” in terms of other variables.  Most of these studies involve children.  They go to different schools, have different parents, etc., but they may be evenly matched by diagnosis, IQ, health conditions, etc.  There are several ways to do this: recruit a larger group of individuals and screen them to make up a larger homogeneous group, or use children who happen to be on the waiting list for a clinic.  Either way, the “apples-to-apples” homogeneity of the group affects the believability of the results.

There are numerous problems and complications here.  First, there is the finding of a truly homogeneous larger group.  No two human beings have exactly the same learning characteristics, temperaments, histories, etc., despite similarities in diagnosis, IQ and other supposedly common traits.

Cost is a major factor, because in order to do the screening, the individuals have to be tested and carefully matched.  Sometimes, in order to save costs, individuals may receive their testing and diagnosis from outside clinicians, and that can seriously affect the validity of the study.

Then there is the issue of ethically depriving children of a beneficial treatment.  This is tricky and not that serious of a concern.  Rarely do children get nothing.  Even if they are not in school – they still have parents and environments in which they can learn.  Those that are in school are getting that, and the major landmark studies have compared their Study Groups to children that are in school.

Finally – and this is huge – usually the Control Group is not receiving “Intensive” “High Fidelity” treatment.  Intensive treatment involves 1:1 teaching with a highly trained and consistent technician, and a lot of control over the environment.  This is vastly different than a typical special education classroom or home environment.  So is there really that much homogeneity in the larger group?

Random Assignment

In other sciences where such laboratory testing is more feasible (it is always expensive), the group is divided in two and numbers are picked at random to assign subjects to one group or another. So for instance, when testing a drug, one group will get the experimental drug, and other would get a placebo.

In behavioral studies, where behavior change techniques and methods are studied, the subjects of the experimental treatment are chosen by lottery or some other method for randomization.  The other half become the control group.


Blinding has to do with keeping out sources of bias or subconscious effects on the outcomes.  In psychological or behavioral studies, it’s difficult to blind the technicians because they have to be trained to administer the experimental method.  The parents and the children know whether they are involved in the method or not as well, so there is no “placebo.”  What can be done is to withhold the evidence of the study from the researchers until the the study is over.  Even that doesn’t completely achieve the ultimate in objectivity, as those behind the experiment usually want it to work.  They are going to try harder than the clinicians or Teachers or Parents of the Control Group, because they aren’t trying to prove anything and outside Teachers and Clinicians may not even know that they have Control Group members in their midst.

The vast majority of the evidence basis comes from smaller studies in which component techniques (component elements of packaged treatment methods) have been studied thousands of times.

Single and Multiple Subject Designs

These studies test a technique on a single person or a group as individuals.  When several individuals are tested, the results are reported as individual findings, or as the number of individuals that responded in one way or another.  The majority of research in ABA is on single subjects or very small groups of individuals.

Reversal and Alternating Treatment Designs

Reversal designs involve systematic administration and removal of the technique in order to prove the relationship between the technique under study and the changes in behavior observed.  The initial phase [A] is called the “baseline” or “baseline measurement phase, and that is to establish the rate or characteristics of the behavior before treatment.  The next is the treatment or “intervention phase” [B] in which the procedure is tested and measurements are taken, and finally, treatment is withdrawn [A] to see if behavior returns to it’s original state or baseline [A]-[B]-[A].  The simple [A]-[B]-[A] isn’t really strong enough, so most studies employ another phase of withdrawal of treatment just to make sure [A]-[B]-[A]-[B] or they go on alternating treatment with withdrawal of treatment.  Most studies employ some variations on the treatment each time it is re-administered.  The repeated “reversals” between treatment and withdrawal of treatment is the reason the design is called a “Reversal Design,” a pretty effective means of showing that the intervention is indeed the reason the behavior changed.  An applied clinician would look at reducing dependence on any treatment that is not naturally occurring in the environment, as the “treatment” might represent some level of artificial support.  We hope that skills learned will generalize and the effects remain even after specialized treatment conditions or supports are withdrawn.

A variation on this has to when an experimenter wants to compare more than one treatment on a single behavior in a single subject.  In this case, you have two treatments [B] and [C], so the design would look like this: [A]-[B]-[C]-[B]-[C].  The experiment starts with the initial baseline measurements; tries the first technique, then; tries the other technique, and then; alternates the two techniques to see the differences in the effects.  There is a real problem of “pollution” here, or what researchers call “confounds,” where no one is sure which of the variables being tested accounts for which part of the behavior change.  Needless to say, [B] and [C] would have to be distinctly different from each other to show any clear functional relationship between the intervention and the observed changes in the behavior.

Multiple Baseline and Changing Designs

These forms are actually the most common forms of research design found in ABA.  In the Multiple Baseline Design, the researcher wants to know how a technique will work with more than one behavior, or with different persons, or in different settings.  Major aspects of a therapy such as the practice of following the child’s lead for instance, can be tested in multiple ways.  Similarly, the way that Guided Participation teaching works in different persons and in different settings would be beneficial to know.

The probable reason for the popularity of this research design is the very same precaution I mentioned above in regards to reversal designs; that is, we hope that the technique results in beneficial skills that generalize.  We don’t want reversals.  We want more enduring changes.

The Changing Criterion design is good for testing techniques for advancing larger domains of behavior and development.

Where Most of the Evidence Basis Claims come from…

Ideally, larger therapy packages should undergo some form of Randomized Control testing.  But given the cost and difficulties involved, the RCT standard can delay useful therapies and inhibit innovation.  As a result, most packages accomplish something short of this – either by testing a small population with patients in their own clinic or more extensively – claiming evidence basis by virtue of the fact that the package represents an aggregation of the extremely well tested and replicated techniques listed below.

Therapy packages are usually aggregation of the following principles and techniques that have been tested over and over using the above research designs:

Component Techniques used in Aggregated ABA Intervention “Packages”

For definitions, please refer to the Glossary of ABA Terms

  • Techniques of …
    • Modeling
    • Shaping and Fading
    • Behavioral Chaining
    • Prompting Techniques (numerous and varied types have been tested)
    • Stimulus enhancement or stimulus embedding (cuing techniques, e.g., exaggeration or amplification of stimuli; visual schedules, checklists and the like)
    • Response Interruption, Response Delay, Sabotage and other “engineered opportunities (called “productive uncertainty” techniques in Guided Participation teaching or “Stopping the Action” in RDI

Importantly, these techniques have been applied to teach an extremely wide range of behaviors, including those that make up the fuzzy, hard to teach and measure response classes such as intersubjectivity, attunement, empathy, dynamic and abstract thinking, etc. so valued in relationship-based interventions.

We maintain that after studying Floortime and RDI thoroughly, and to a lesser extent the Miller Method, that these methods represent different aggregations of time tested techniques.  The main difference is in the jargon used and the heuristics (ways of explaining and understanding) involved.  Further, when we remove jargon and speak plainly to parents, that the differences appear more like similarities. 

Finally, it is important to distinguish between being ope-minded and considering interventions under the rubric of ABA, and the famous admonitions against “eclectic” interventions so roundly criticized by the field of ABA.    I believe this criticism started with Lovaas’s 1987 study in which DTT was compared to the so-called eclectic interventions controls received in their non-intensive special education classes.  Special education classrooms are not well known for being systematic.  The following passage describes one of the “defining characteristics” of ABA, the characteristic of being conceptually systematic

“Although Baer et al. did not state it explicitly, a defining characteristic of applied behavior analysis concerns the types of interventions used in an effort to improve behavior.  Although there an infinite number of tactics and specific procedures that can be used to alter behavior, almost all are derivatives and/or combinations of a relatively few basic principles of behavior.  Thus Baer et al.  recommend that research reports of applied behavior analysis be conceptually systematic, describing the procedures for behavior in terms of the relevant principle(s) from which they were derived.”

They go on to maintain that “…conceptual systems are needed if a technology is to become an integrated discipline instead of a collection of tricks. (Cooper, Heron, Heward; 2007)

Our extensive experience and review of RDI and DIR/Floortime reveals to us that they not only can be explained in terms of Skinner’s principles, but also in terms of their very internally consistent conceptual systems.  RDI is grounded in Lev Vygotsky’s theories that have been well studied for 70 or more years, and DIR in the principles of child development.  Both utilize systems theory and attachment theory, also with extremely well supported bodies of evidence in science. 

Applied Behavior Analysis

Applied Behavioral Analysis gives us some basic ways of understanding learning: the relationship between the learner and the role the environment plays in learning, as well as ways to measure both behaviors and the effectiveness of techniques.  Yes, ABA is the only way if you look at it for what it really is and can do, how inclusive and flexible it can be, and, if you separate it out from the political, partisan and business aspirations of many of it’s practitioners. What ABA really is is a way to judge the results of theories of behavior, interventions for behavior, and research and claims about causes of behavior and behavior change techniques.  The field of ABA was officially started in 1968 with a paper entitled “Some Current Dimensions of Applied Behavior Analysis.”  If you read the paper, it is clear that as the principles learned in behavioral science laboratories were starting to be used in applied practice with humans in real-life human contexts, and the authors felt the need for some of the rules and standards of the field had to be redefined.  By this time, Skinner had already published his seminal papers and books defining the “experimental analysis of behavior,” a set of principles learned from thousands of experiments with animals.  He elaborated on his lifetime of research in his book about the philosophy regarding the scientific study of behavior, “About Behaviorism” (Skinner, 1974; 1976).  On page 1, he wonders aloud about the “science of human behavior”…

… Is such a science is really possible?  Can it account for every aspect of human behavior?  What methods can it use?  Are its laws as valid as those of physics and biology?  Will it lead to a technology, and if so, what role will it play in human affairs?”

At this time, how can we answer Skinner’s questions?  Here are my answers…

Is a science of human behavior really possible?

That answer is of course.  By 1974 the science was well underway, following the principles he learned from laboratory research.  It was coming out of the laboratories and into many aspects of human society.  Very serious applications of the science were already undergoing experimentation with individuals with disabilities, and the first systematic, packaged treatment for autism, Discrete Trial Therapy (“DTT;” Lovaas, et al.) emerged around that time.  Since then, DTT is one of the very few autism methodologies to undergo “gold standard” Randomized Controlled Trials (RCTs) to prove both it’s effectiveness (Lovaas et al. 1987) and replication (McEachin, Smith, and Lovaas, 1993).  Many other replication studies have been conducted since.

While many, including many in the field of behavior and autism research question DTT’s methods and effectiveness, it was an astounding advance: Lovaas proved that children with autism and intellectual disabilities could learn.

 At this time, behavioral science grounded in the principles and theories of B.F. Skinner has been applied in almost every aspect of human life, and many of his principles of behavior, especially those involving contingency, reinforcement and punishment (below), are widely accepted across many diverse fields of human and animal psychology.  I use it to validate any method and it still is very instrumental in my understanding of how behavior works, although in my later career, I got very heavily into attachment, child development, information processing, neuropsychology and systems theory.  Those theories also go a long way towards explaining how behavior works and are very compatible with [most aspects of] Skinnerian behavioral science.  Keep in mind that the science has evolved quite a bit since 1974.

Can it account for every aspect of human behavior? 

I would answer with a very qualified, “Yes.”  The qualification has to do with the consideration of “private events” as Skinner called them (i.e. feelings, emotions, moods, perceptions, cognitions, thoughts, beliefs, etc.).  Skinner clearly maintained that private events influence behavior.  He just was very wary of using private events as explanations for behavior.  He called the hypothetical explanations of behavior that could not be observed publicly, “mentalistic fictions” (a very harsh and unfortunately dismissive term that is currently overused).  We now have fMRI machines, QEEGs, reams of data on mental processing and neurocognitive science, as well as years of empirical research in a broad and inclusive field that is now called, “Developmental Psychopathology” (the study of typical development and the mechanisms that cause deviation from it) that includes many empirical studies of thinking and emotion – all of which add up to a “consilient” (referring to the linking together of principles from different disciplines in order to form a comprehensive theory; see, E.O. Wilson, “Consilience: the Unity of Knowledge 1998) mountain – no, a world of data, no longer so easily dismissed as sources of mentalistic fiction.  I argue strenuously here and elsewhere throughout this site that it is a huge mistake for the field to ignore the evidence from other fields in Functional Analysis of Behavior (see the Criticism of Functional Analysis by John Stewart, Ph.D.).   A few leading behavioral researchers, notably Steven Hayes and his disciples (BCBAs! many of them) that are looking with very open minds at contributions from the above-mentioned fields and talking about apost-Skinnerianworld.

What methods can it use?

This is a very broad question, but it can be narrowed down to methods that are observable, measurable and replicable, or in other words, adhering to the typical empirical methods and standards of other fields of scientific inquiry.

Given the current state of the Autism Wars, otherwise sane, smart and extremely well educated, scientifically trained professionals are engaged in a fool’s errand: trying to prove that their philosophical or disciplinary orientation or their method is the best way to intervene for autism, or worse – the only way.  That is not science – that is religion.  First of all, Skinner didn’t know much about autism.  If he did, you can’t find it in his writings.  No one really knew much about it until after he died anyway – if you believe we know much about it now.  Secondly, this thinking is one of many forms of cognitive bias applicable here.  Specifically, it takes a lot of time and devotion to learn a particular discipline or an autism intervention method.  No one in their right mind would pursue a field or treatment that they thought was wrong – not for the money we make.

Are its laws as valid as those of physics and biology? 

The laws are – the methods derived from them – not always.  For instance, Skinners theories of language development have come under serious attack.  He did not have the same background in scientific experimentation for his theories on language and cognitive development that he did on Operant Behavior.  Recent operant methods have looked beyond his theories, and I happen to think that his focus on the speaker ignores foundations of joint attention and has led to the widespread mistake of teaching speech without real (equal or more) emphasis on joint attention.  That is just one example of methods and traditions being different from the laws of the science.  Sometimes the only difference is in the jargon.

Will it lead to a technology, and if so, what role will it play in human affairs?

It’s already led to many technologies.  Again, most branches of psychology accept as fact Skinner’s laws of contingency, reinforcement and punishment.

Defining ABA

I would like to offer a heuristic for understanding A-B-A.  I propose viewing ABA in terms of validity, reliability and quality control.  Validity questions have to do with whether or not the technique does what is says that it does.  They also have to do with whether the technique brings real benefit or not.  Reliability questions have to do with the consistency of the results, and Quality Control questions provide a check on the first two. Here is an example of how that works.  Let’s look at a car for instance.  You can use a car as a storage bin or a planter or a show piece, but most of us want it to be a machine that provides transportation.  If it cannot provide transportation, it is not valid.  A lawnmower is supposed to be able to cut the grass, and an autism therapy should improve a person’s quality of life and increase important and needed skills. Back to the car.  You can have a very reliable car that isn’t valid.  I had one of those.  I could rely on it to breakdown every day.  Because my car wasn’t valid, it had to be changed.  I’d replace parts here and there as a means of quality control.  My quality control efforts were meant to make the car more valid and reliable.  My quality control consisted of not only changing how the car works by replacing parts, but I also could see whether the “fixes” were working by road testing the car.  Sometimes, the car would work in the driveway but not on the road, so road-testing became an important aspect of quality control.

What ABA is then is not a method per se – but a system for insuring that treatment for behavior is valid, reliable and high quality.

“Applied” – The Validity Axis of ABA

Baer, Wolf and Risley looked first at the mechanisms developed from “analysis of individual behavior” in the laboratories and “…the possibility of their application to  problem behavior.”  They maintained that “…A society willing to consider a technology of its own behavior apparently is likely to support that application when it deals with socially important behaviors, such as retardation, crime, mental illness, or education.”  So the social validity of ABA  – the real benefits to the individual and society was a prime consideration. The problem at hand was how laboratory research, that had different constraints than applications in the real world on human beings could be applied and gain “social approval.”

The authors point out that in the laboratory, it may be important to find out what can improve a behavior as well as what can make it worse and what happens if treatment is withheld.  Using the example of disruptive behavior in the classroom, they make the point that experiments to see what can make the behavior worse might be questionable in an applied setting.  For instance, applying repeated electric shocks to dogs in the laboratory helped discover the behavioral principles of “learned helplessness” – a very useful concept that has helped us understand why people remain in situations that are clearly not good for them (such as domestic violence).  But such an experiment – using continuous electric shocks on people, would be unethical in a laboratory let alone in an applied setting.

Importantly, the “Applied” aspect of ABA meant that it was more important to benefit the individual than to choose behaviors to study simply because they are easy to measure or because the authors of a paper seek to prove a technique can work.  I often joke about applications of ABA that can teach a chicken to play the piano or a rat to pedal a unicycle across a tightrope.  Sure, it works, but how does that benefit a human?

The authors elaborate,

“…examining behaviors which are socially important, rather than convenient for study.  It also implies, very frequently, the study of those behaviors in their usual social settings, rather than in a “laboratory” setting. But a laboratory is simply a place so designed that experimental control of relevant variables is as easy as possible. Unfortunately, the usual social setting for important behaviors is rarely such a place. Consequently, the analysis of socially important behaviors becomes experimental only with difficulty.  …Thus, analytic behavioral applications by definition achieve experimental control of the processes they contain, but since they strive for this control against formidable difficulties, they achieve it less often per study than would a laboratory-based attempt.  Consequently, the rate of displaying experimental control required of behavioral applications has become correspondingly less than the standards typical of laboratory research. This is not because the applier is an easy-going, liberal, or generous fellow, but because society rarely will allow its important behaviors, in their correspondingly important settings, to be manipulated repeatedly for the merely logical comfort of a scientifically sceptical audience.”

The above statement is extremely important in light of autism treatment methods emerging and currently under study.  Along with advances in the understanding of what autism is and what mechanisms cause the deviations from typical development characteristic of autism spectrum and other disorders or relating and communicating, we are indeed looking at more difficult aspects of behavior to measure.

A clear example can be made between a technique that allows for the easy recording of data versus a technique that by virtue of what it values and the manner in which it is supposed to be applied, makes precise data recording a challenge.  For instance, the way trials work in a technique such as Discrete Trial Training make it very easy to record data.  That is not why the method works that way, it just is that way.  Most ‘trials’ consist of a Therapist or Teacher prompt, a response from the individual, feedback or a ‘consequence’ from the Therapist/Teacher, the Therapist/Teacher records the result of the trial and begins another trial.  This is an enormous advantage for reliability and quality control – the other two axes, but it doesn’t necessarily insure validity.  Validity has to come from the usefulness of the skills taught.

At SCS, validity has to meet an important social criterion – it has to lead to a better quality of life for the individual and for the family.

One set of skills we seek to improve in autism is the individual’s ability to engage in long chains of reciprocal, emotionally present and engaged social interaction – something referred to as “reciprocity.”  The skill sets involved in reciprocity may seem fuzzy, but that are definitely noticeable when present or absent such as intersubjectivity, “experience-sharing,” warmth, empathy and the like are part of this skill set and obviously noticeable as demonstrations of behavior to even the casual observer – are much more challenging for data recording than a method that uses “trials” that focus on a single contingency at a time. I note here that terms such as “intersubjectivity” and “reciprocity” are not measurable without breaking them down into their sub-component skills.  But the relationship-based techniques, the techniques that value these fuzzy skill sets have done the work of breaking these large behavioral classes into constituent parts (see Systematic, below).  The point is, a therapy can be more “applied” because it focuses on the right skills and uses effective techniques, but it may not be superior to a therapy that is perhaps less valid, but more “convenient…”

The non-applied researcher also may study bar-pressing because it is a convenient response for study; easy for the subject, and simple to record and integrate with theoretically significant environmental events. By contrast, the applied researcher is likely to study eating because there are children who eat too little and adults who eat too much, and he will study eating in exactly those individuals rather than in more convenient ones (Baer, Wolf, Risley, 1968).

“Behavioral” – The Reliability Axis of ABA

The ability to demonstrate actual behavior change is the emphasis of this axis.  ABA strongly values direct observation of overt, measurable behaviors as the only reliable way of judging progress.  Behavioral results can only be “chalked up” when behavior change can be clearly and reliably observed.

Thus it usually studies what subjects can be brought to do rather than what they can be brought to say; unless, of course, a verbal response is the behavior of interest. Accordingly a subject’s verbal description of his own non-verbal behavior usually would not be accepted as a measure of his actual behavior unless it were independently substantiated. …Application has not been achieved until this question has been answered satisfactorily.

Therefore, a person’s testimony or verbal report of progress is not a valid measure of progress, nor is the comment of an observer sufficient data to analyze progress.  More reliable measures are needed. 

Under the “Behavioral” heading of this seminal article, the authors revisit the “applied” or validity axis again, because by nature these axes are overlapping and inseparable…

Since the behavior of an individual is composed of physical events, its scientific study requires their precise measurement. As a result, the problem of reliable quantification arises immediately. The problem is the same for applied research as it is for non-applied research. However, non-applied research typically will choose a response easily quantified in a reliable manner, whereas applied research rarely will have that option. As a result, the applied researcher must try harder, rather than ignore this criterion of all trustworthy research. Current applied research often shows that thoroughly reliable quantification of behavior can be achieved, even in thoroughly difficult settings (Baer, Wolf, Risley, 1968)..

Due to the fact that newer therapies address the more important yet more difficult to teach and record deficits of autism and other disorders of relating and communicating (e.g., intersubjectivity, reciprocity, warmth, empathy, etc.), they are at a disadvantage in the rapid calculation of data.  Videotape has become almost indispensable, but I can tell you first hand, it takes generally three hours to quantify the kind of behaviors we observe in an ongoing reciprocal and highly dynamic social interaction for every one hour of tape!

The question then becomes, how long do the people who need the most valid treatments have to wait before the standards of “evidence-basis” are met, and what criteria should that be?

(Baer, Wolf, Risley, 1968) …analytic behavioral applications by definition achieve experimental control of the processes they contain, but since they strive for this control against formidable difficulties, they achieve it less often per study than would a laboratory-based attempt. Consequently, the rate of displaying experimental control required of behavioral applications has become correspondingly less than the standards typical of laboratory research.

Analytic behavioral application is the process of applying sometimes tentative principles of behavior to the improvement of specific behaviors, and simultaneously  evaluating whether or not any changes noted are indeed attributable to the process of application-and if so, to what parts of that process (Baer, Wolf, Risley, 1968).

I note strongly here that the packaged treatments claiming ‘evidence basis’ most vociferously have not demonstrated “apples to apples” comparison.  They measure intensive autism treatment methods against control groups that do not receive intensive intervention, such as children in regular education or non-intensive special education classrooms.  It is fair to ask whether any treatment wouldn’t be superior if it included multiple hours of 1:1 instruction when compared to controls that do not receive 1:1 instruction?  Further, the active ingredients of the method (versus the inactive or neutral ingredients) have yet to be identified, so again, is it just the intensity?  All that we really know at this time is that intensive is better than non-intensive, but we do not know whether one method is better than another.  A systematic comparison between methods of both ABA traditional, “post-Skinnerian, and Relationship-Based methods has yet to be done.

The bottom line of the “behavioral axis” is that no matter which method one applies, the individual and the environment must show some demonstrable change.  It cannot be left to opinion, beliefs, or testimonials.

“Analytical” – the Quality Control Axis of ABA

In their paper, Baer, Wolf and Risley mention “believability” as a means of judging intervention qualitatively.  In their words, an intervention is “analytical,” or at least ongoing analysis of progress must show…

…a believable demonstration of the events that can be responsible for the occurrence or non-occurrence of that behavior.

The “events”  in which they are talking about refer to the actions of the interventionist and the particular changes he or she makes in order to influence behavior or development.

An experimenter has achieved an analysis of a behavior when he can exercise control over it.  By common laboratory standards, that has meant an ability of the experimenter to turn the behavior on and off, or up and down, at will.  …Applied research, as noted before, cannot often approach this arrogantly frequent clarity of being in control of important behaviors. Consequently, application, to be analytic, demonstrates control when it can, and thereby presents its audience with a problem of judgment.  The problem, of course, is whether the experimenter has shown enough control, and often enough, for believability (Baer, Wolf, Risley, 1968).

Here again, there is overlap with the notion of value to the individual and society they proposed as a measure of how “applied” an intervention might be…

…Application typically means producing valuable behavior; valuable behavior usually meets extra-experimental reinforcement in a social setting; thus, valuable behavior, once set up, may no longer be dependent upon the experimental technique which created it.

If statistical analysis is applied, the audience must then judge the suitability of the inferential statistic chosen and the propriety of these data for that test. Alternatively, the audience may inspect the data directly and relate them to past experience with similar data and similar procedures. In either case, the judgments required are highly qualitative, and rules cannot always be stated profitably.  However, either of the foregoing designs gathers data in ways that exemplify the concept of replication, and replication is the essence of believability.   At the least, it would seem that an approach to replication is better than no approach at all.

Here, our bottom line is that any intervention can be judged on the directness of its role in behavior change and how stable and useful that change will be.  An intervention cannot be analytical if it isn’t valid (teaches skills that are not the most useful) and reliable (the behavior changes are durable, useful in the environments in which the individual must function and can continue to be elaborated upon).

Concluding Comments

Who’s Right?

For many years, intervention is autism was (still is) predominated by Discrete Trial Training (DTT), a method devised by Lovaas et al. that features a very strong clinician lead.  But in the late 1980s and 1990s you saw the emergence of natural environment teaching, the “Natural Language Paradigm” (Koegel, Koegel, & Surrat (1992) and it’s outgrowth, Pivotal Response Therapy (Koegel, O’Dell, & Koegel, 1987; Laski, Charlop, & Schreibman, 1988) – that maintained that the most important thing to do was to “follow the child’s lead.”  Whereas in DTT, the clinician chose all of the learning materials, maintained specific control over the environment, and used heavy doses of positive reinforcement (below).  Pivotal Response and many of the newer ABA therapies now look to harnessing the child’s natural interests and motivations and teaching within naturally occurring routines.

Method bodies developed from other theories of development and behavior also disagree on this.  DIR/Floortime is a method that starts with a very assiduous follow of the child’s lead, whereas RDI on the other hand emphasizes that Guides (usually parents and other Teachers) take a strong hand in deciding the roles and role responsibilities of the Student or Learner.

Whether the child, parent or Clinician is in the lead is just one aspect of intervention.  There are many variables and great diversity among behaviorist traditions and methodologies developed from behaviorism, and those that come from outside the field of ABA (“outside the field” referring to practitioners that do not call themselves Behaviorists and that do not use behavioral terminology and jargon, and that come from other fields of psychology or development).

The point is that there can not only be disagreement within scientifically valid approaches, we expect this disagreement because no method as yet is self-contained and can work for every person.  The process of ABA on the other hand, is valid for everyone in that the process merely spells out means of scientific evaluation that can and should be applied to all methods.

We prefer to look at the strengths, weakness and indications and contraindications (below) of the available technologies in an individualized way, rather than to think in terms of who’s right and who is not.  ABA allows us to do that with the therapies we use.  Lovaas published his landmark study in 1987, and we can say that the field of autism intervention started officially at that time.  What this means to me is that we are still in the embryonic stages of the field.  It is way too soon to make such determinations, and judging by the newer developments in the field, it seems as if there will eventually be a great deal more rapproachment between the now diverse points of view as the evidence continues to come in.

Way too often, upon getting a diagnosis, parents are often asked to choose sides.  Unfortunately, they have to deal with a fractured professional community that purports scientific non-bias, but is in fact becoming more like religion, politics and business.  This is why we choose to take the road less traveled – to study them all as extensively as we can and to remain “orthodox agnostics.”  We cannot know what is best for your child unless we look at the individual, the family, the resources and the contexts in which we are supposed to help.

A Medical Analogy to Autism Therapy

Even though Behavioral therapies (they’re all interventions for behavior and development) are considered “non-medical” or educational approaches, I have learned that is proper to see interventions as similar to “medicines.”  Like medicines, all of the treatments have “indications” (what the intervention is supposed to work for and is supposed to be able to do, and more importantly what it should not be used for), “benefits” (the judgment of whether the intervention actually works in the individual), and “side-effects.”

Yes, side effects.  In medicine, the worsening of a condition or the creation of another problem is called an “iatrogenic” effect. All of the methods have potential for iatrogenic effects, so one of the best checks is to not to try to use a method that is the best match for the characteristics of the Learner and his or her family, and, when there isn’t a match, try something else.

Back in the mid 1980s, I was one of the army of graduate students that participated in accumulating data for Discrete Trial Therapy, before the landmark paper was published (Lovaas et al., 1987).  While I was not part of the official clinical group involved in that paper, I was one of many Education Majors that participated in the “Mark Twain Project” through the Los Angeles County Office of Education (LACOE).  At the time, LACOE was in consultation and collaboration with the researchers at UCLA, where the bulk of the research was done.  I remember one 12 year old girl with autism in particular that was my “subject” (as we referred to the children back then).  There was no relationship-basis to that version of the model, so my role was rotated among 4 or 5 interventionists.

When I seated her at the table with me, she began giggling.  In retrospect (I still think about her often – and you’ll see why), I think she was giggling nervously.  There was no joy in this laughter.  The supervisors dismissed it as “self-stimulatory” behavior, which was considered “off-task” behavior and therefore subject for punishment.  I was instructed to put her in a dark closet, wait for the laughter to stop, and then seat her back at the table.  When she got a right answer, I gave her a raisin and said “Good job!” I gave her corrective feedback (“Try again” or a prompt or the correct “response”) depending on what kind of response she gave.  I was teaching her vocabulary words from flashcards.

A few years later, I became a Special Education Teacher for LACOE, and she was in my classroom.  Somewhere in the meantime, she had developed a behavior of biting her knuckles.  I could see layers of scar tissue on her knuckles -she obviously bit herself very hard.  She almost never spoke in school – despite the extensive work on her speech and her so-called “mastery” of the DTT “language” programs  – all extensively quantified in her 5-inch thick notebook.  She only requested food – that was the only time she ever spoke voluntarily.  She had been taught to answer questions, so if you wanted to know what she knew, you would have to ask her.

She bit her knuckles whenever we presented her with the endless worksheets that represented the extension of the same manner of teaching. She was doing “multiple trials” on her worksheets.  She would complete 10 or so math problems, or write her name 10 times in a row.  She would rush through the work and then ask for some food.  She was getting quite overweight.

The knuckle biting, and probably the giggling and maybe even the obesity were iatrogenic effects.

I visited her home and saw how she did there.  She didn’t even recognize or acknowledge me, but she spoke regularly to her family at home.  She also did several other very intelligent things that frankly astounded me – but nothing that had anything to do with what we were working on in school.  She was much more self-sufficient and cooperative.  She never had DTT at home.  Her father told me that she started biting her hands during the DTT therapy and for a brief time they stopped treatment.

That was a pivotal moment in my own development.  I began to wonder about the emotional lives of these kids.  I picked up a copy of “First Feelings” by Stanley Greenspan (Greenspan and Greenspan 1985) because in my education as a Special Education Teacher, my training was almost entirely behavioral, and we spent no time whatsoever studying the emotional development of children.  This was around 1987, because I remember that the Lovaas paper was published that same year.

DTT has changed.  There are no more dark closets.  Therapists are playful.  They try to avoid food rewards. You get away from the table a lot sooner, but there is still the belief among many practitioners that being ready to sit at a table for instruction is the same thing as being ready to learn.

I still use DTT!  It is very good for teaching procedural skills.  It features task analysis – the identification of component steps (sometimes micro-steps) of a task, and examines the order of the steps.  It provides for teaching the last steps first or the first steps last – whatever works (see Behavioral Chaining procedures).  I use it especially for children that have difficulties with the visuospatial maps of their bodies and therefore have trouble imitating other people.  Procedures require repeated practice and “overlearning” to the point where they become automatic.  I tend to do the teaching using a Guided Participation teaching approach – showing, helping, and gradually backing off, and when you see me or one of our Therapists do it – it’s hard to tell whether it is Guided Participation teaching or just a teaching trial.  This would only be true if the lesson involved a singular procedure, such as an imitation of a movement, or tying one’s shoes or pulling up one’s pants.  Guided Participation is different in that teaching emphasizes the social and emotionally attuned role of the Teacher and the Learner.

On the other hand, if a method chooses to use other terms to describe events, or a system of teaching that is different and that did not evolve from traditional behavioral therapies, it can be just as much or more “applied,” “behavioral,” and analytic.”