In her biography of the mathematician John Nash, Sylvia Nasar (1998) described the treatments that Nash underwent for schizophrenia including insulin coma therapy (ICT), in which insulin was delivered intravenously to lower his blood sugar levels and produce a coma. The use of insulin was not predicated on a well-developed theory of the neuropathological basis of schizophrenia. Rather, its application derived from the observations of a psychiatrist named Max Sakel, who, for reasons that are unclear, was using insulin to treat withdrawal in individuals addicted to morphine. Sakel noted that if those individuals experienced coma or convulsions, they seemed to do better afterwards (Healy, 2002).
Sakel extended the treatment to schizophrenia. During ICT, the fasting patient with schizophrenia was given a dose of insulin that induced a hypoglycemic coma. After about 30 minutes, the medical staff would attempt to bring the patient out of the coma, using an intravenous injection of dextrose. The major risk of ICT was that some patients did not come out of the coma. This "irreversible coma" was the major cause of death associated with the treatment.
ICT was used for years after Sakel's initial publication in 1935, with virtually no challenge, although other treatments were being used, such as electroconvulsive shock therapy. Then, in 1953, the psychiatrist Harold Bourne (1953) published a paper in The Lancet arguing that "there was no sound basis for the general o pinion . . . that insulin coma therapy counteracts the schizophrenic process in some specific manner" (Bourne, 1953, p. 964). He noted that there were many other variables-including the greatly increased medical and nursing care given to ICT patients-that might produce the positive treatment effects in ICT. The psychiatric community reacted negatively to Bourne's paper, and it was only after the drug chlorpromazine (Thorazine) came along that ICT was abandoned.
The Moral of the Cautionary Tale
Why is the story of ICT relevant for speech-language pathology? It is a compelling illustration of the potentially harmful (even fatal) consequences of initiating treatment that lacks a sound theoretical basis. ICT appeared to work, but no one knew why and the effects did not last long, so it had to be repeated on a daily basis for weeks. Bourne made another interesting observation about ICT. Initially, there was a great deal of drama associated with ICT for the staff administering the therapy, because it was risky. However, as the routine began to supplant the drama, and the staff became less enthusiastic about ICT, the schizophrenia remission rate declined considerably. This suggests a placebo effect. That is, because the therapists believed ICT would be effective, the patient also believed it. As this belief changed, the treatment became less effective.
Bourne provided a convincing argument that these other factors, not the insulin, produced the positive treatment outcome. There was no sound theoretical basis for the use of ICT, and patients died as a result of a treatment that was later shown to be ineffective for treating schizophrenia. However, it is not enough to base a treatment on a theory, because not all theories are equal. The theory that serves as the basis for developing a treatment must be a "good" theory, as I will illustrate in the following comparison of two theories of aphasia.
It is generally agreed that good theories must meet certain criteria (Fawcett, 1992; Lum, 2002). The first of the three most important criteria is that the terms used in the theory must be clearly defined and used consistently. A term like "selective attention" lacks precision and consistency. Does "selection" mean not taking in unnecessary information at all? Or does it mean taking in all information but allowing only necessary information into conscious awareness? Conversely, a term such as "speech articulation" is much more precise and consistently used.
The second criterion is that the theory must be supported by our observations. In 1943, based on limited observation, Leo Kanner theorized that autism was caused by remote, highly-educated parents who did not provide their children with enough love and affection. Kanner did not observe other socioeconomic groups to determine if they also had children with autism. His theory was thus not based on thorough observation. As a result, parents suffered greatly, convinced by professionals that they caused the autism. Conversely, a good theory is that language is mainly localized to the left hemisphere in most people. This theory is supported by many years of observation of the consequences of brain damage and, more recently, by observations from functional neuroimaging.
The third criterion of a good theory is that it can be falsified or disproved. No theory can be proved, but a good theory allows for the design of experiments that could potentially disprove it. This is important because when experiments-and the resulting data-are limited only to those that provide support for a theory, there is no opportunity for revision, improvement, or strengthening of the theory.
The theory, popular some years ago, that language disorders in children were caused by "minimal brain damage" (MBD), is an example of a theory that does not permit the design of experiments that might disprove it. MBD was considered to be sufficient to cause language problems, but so minimal that it could not be observed with any available technology. Since there was no way to observe "minimal" damage in living children, there was no way to design a study that could show there might be some children with language problems who did not have MBD. If an investigator could design such a study, the theory could be revised to account for the finding that some children with language disorders had MBD and others did not. If the investigator found that none of the children with language disorders had MBD, the theory could be abandoned. Finally, if the investigator found that the theory held up, despite a good attempt to disprove it, the theory would be strengthened.
The theory that the outer hair cells of the cochlea are the source of otoacoustic emissions (OAEs) is one that could be disproved. One could find cases where loss of the outer hair cells does not affect OAEs. If the theory is disproved, then we can look for another source of OAEs. If we cannot disprove it, despite legitimate attempts to do so, this does not prove the theory. This is because there is still the possibility, with the design of new experiments, that we might disprove it. Every failed attempt to disprove a theory provides further support for the theory, and this is the paradox of a good theory. A good theory is not one that has been shown, over and over again, to be true, but one that has been shown, over and over again, not to be false.
It was difficult to find a theory of aphasia that met all these criteria, but one that generally has the features of a good theory is the localizationist theory of aphasia. It has its roots in 19th century aphasiology and has been further elaborated by others (e.g., Goodglass and Kaplan, 1983). It holds that the speech and language difficulties that an individual with aphasia demonstrates are dependent on the site of lesion in the brain. Damage to Broca's area, for example, will produce the pattern of deficits described as Broca's aphasia and damage to Wernicke's area will produce a characteristic pattern of deficits described as Wernicke's aphasia.
The first criterion of a good theory is that the terminology is clear and consistent. Goodglass and Kaplan provided a description of the speech and language characteristics associated with each type of aphasia. For example, in Wernicke's aphasia, there is an impairment of auditory and reading comprehension with fluent speech that contains paraphasic errors. In actual practice, because some terms are a bit subjective, there is some variability in classifying aphasia by these features. For example, behaviors like speech fluency occur along a continuum, so it is difficult to determine precisely where fluent speech ends and nonfluent speech begins. Despite these problems, localizationist theory meets the criterion of using consistent terminology, as the different aphasia syndromes are described consistently.
The second feature of a good theory is that it is supported by thorough observation. Although not every individual with aphasia can be classified into one of the localizationist types of aphasia, observations made over the last 100 years have shown that many of them can. Thus, this theory meets the criterion of being supported by our observations of the primary deficits in aphasia.
The third feature of a good theory is that it is falsifiable. Localizationist theory also meets this criterion. In order to disprove the theory, one can search for individuals who have lesions to Broca's area who do not have Broca's aphasia, individuals who have lesions to Wernicke's aphasia who do not have Wernicke's aphasia, and so forth. If a number of such individuals are identified, then one can decide to revise or reject the theory.
On the other hand, a theory of aphasia that does not meet the criteria of a good theory is the resource theory of aphasia, which was originally conceived to explain why human cognitive abilities are limited. For example, it can be difficult or even impossible for humans to perform two or more tasks simultaneously. According to the original version of resource theory, humans have limited attention, restricting our ability to process information (Broadbent, 1958).
Although details of the theory have changed over the years, the basic idea has remained the same, that is, we have limited attention and we can process only a portion of the information that is available to us. It has been suggested that persons with aphasia have even more limited attentional resources and/or that they have difficulty in efficiently assigning their attentional resources to processing different aspects of language (e.g., Caplan & Waters, 1994; McNeil & Kimelman, 1986; McNeil, Odell, & Tseng, 1991; Murray, Holland, & Beeson, 1997a, b; Tseng, McNeil, & Milenkovic, 1993).
The first criterion of a good theory is clear, consistent terminology. "Resources" is a vague term. As a result, different investigators have used resources to describe various cognitive abilities, including attention, working memory, and a "central executive" that controls all cognitive processes. Since the terms, particularly the most important term in the theory, "resources," are not always defined the same way, resource theory does not meet the criterion of using consistent terminology.
The second criterion of a good theory is that it is supported by our observations. One of the main approaches used to test resource theory has been to ask individuals with aphasia to perform two tasks simultaneously. The reasoning is that if individuals with aphasia perform more poorly than people without aphasia, this shows that individuals with aphasia have more limited attentional resources and/or that they have more difficulty dividing their resources between two tasks.
In fact, studies have shown that people with aphasia have more problems performing two tasks than those without the disorder, thus suggesting that they may have problems with attention. However, individuals with Alzheimer's disease, right hemisphere damage, and schizophrenia also have problems performing two tasks simultaneously, and resource theory also has been used to explain these findings. Thus, it would appear that many people with brain damage have limited cognitive resources or difficulty in allocating their resources to different tasks. But the critical question, posed by Navon (Shuster, 2004), is whether aphasia is primarily a disorder of attention as resource theory would suggest. I would argue that it is not. Aphasic problems do not disappear or even improve substantially in a quiet room with minimal attentional demands. Thus, resource theory does not meet the second criterion of being supported by our observations of the primary deficits of aphasia.
Resource theory also does not meet the third criterion of a good theory in that it is not falsifiable. An experiment cannot be designed to try to disprove it. Any language deficit that is observed in a study of individuals with aphasia can be attributed to a resource problem. If a person cannot retrieve words, we can conclude that this is due to limited resources or problems in allocating resources to the process of word retrieval. Similarly, if a person cannot produce grammatically complete sentences, understand discourse, read single words, or write the letters of the alphabet, we can conclude that this is due to limited resources or problems in allocating resources to these processes. Since we cannot design an experiment to disprove it, we cannot revise, improve, or strengthen resource theory.
Although localizationist theory has the features of a good theory and resource theory does not, localizationist theory may not necessarily be the best theory of aphasia, or be problem free. For example, it must account for the fact that all individuals cannot be classified into one of the aphasia types, and, if it cannot, it may need to be rejected altogether. However, even though it needs to be revised, localizationist theory is a better theory than resource theory, precisely because it permits revision (or rejection).
The "goodness" of a theory has important clinical implications. The better theory, localizationist theory, provides some guidance regarding targets for treatment. For example, localizationist theory would suggest that individuals with Broca's aphasia should receive treatment for agrammatism. In fact, agrammatism has received much attention in the treatment literature, and the target population for this type of treatment is overwhelmingly individuals with Broca's aphasia.
Conversely, resource theory does not provide a clear idea regarding what to treat and what to expect as a result of treatment. Does the clinician treat a "resources deficit" by targeting attention, working memory, or the central executive? What aspects of aphasia should the clinician expect to improve as a result of working on attention-word retrieval? agrammatism? understanding written words? all areas of deficit? A good theory can guide us in developing treatments that are more likely to produce real functional gains.
It is critically important that we base our treatment on good theory. As ICT illustrated, a therapy that is not based on good theory may appear to be effective, when it is some other factor that is producing the positive change. Good theories must meet certain criteria. The first criterion of a good theory (clear and consistent terminology) relates to the issue of reliability. This criterion ensures that the theory is applied and studied in the same manner by all investigators and clinicians. The second criterion of a good theory (that it is supported by our observations) relates to the issue of accuracy. Thorough observation helps to ensure that the theory accurately describes all the phenomena of interest. The third criterion of a good theory (falsifiability) relates to the issues of revision and improvement. A theory that can be falsified can be revised, improved, and strengthened by our failed attempts to disprove it. For our clients' sake, we must base our treatments on the best available theories.