Group comparison--at best offers only preliminary findings, can hide key causational factors
Is there a single etiological factor to autism?
not only poses one of the biggest puzzles in neuroscience and
cognition, but also one of its most difficult research problems.
It must involve the brain. But the brains of those with autism are remarkably normal. Subtle differences have been claimed. But nothing specific. People can have remarkably unusual brains (as above) without autism..
claims of difference are made a close look at the data shows that most
individuals with that trait in fact are not autistic. If you get a
group of 30 people with autism, and a group of 30 without
and find everyone in the autism group has factor X while only 3 in the
normal group do, it might be tempting to think factor X is tied to
autism. Because autism is so rare (about one to two cases per 1,000
people for autism), you would need to collect together a large group if
it was to contain individuals with autism-- 15 to 30,000. Suppose that
factor X occurs in one in every 30 nonautistic person (one nonautistic
person in the group of 30 has it), then such a group would contain 500
to 1,000 individuals with factor X. By simple arithmetic,
one person in 15 or 30 with factor X, also has autism. This is the
problem of specificity.
The real population
problem of specificity can be seen in the two dot collections. Circles
represent individuals with autism, squares those without autism. An
empty circle or square represents an individual with factor X. In the
group comparison, all the individuals with autism have factor X, and
this is only found in one individual (colored red to highlight them).
But most individuals with factor do not have autism as can be seen in
the second dot collection. Very few individuals have autism, but
because most people are not autistic, the few that do have factor X,
far out number those with autism. This illustrates that that in terms
of absolute numbers, a large percentage of a small number can be
smaller than a small percentage of a large one.
does not mean that most autism research--which uses a comparison group
design--lacks usefulness. Where so little is known, any clues are
important. But such research is only that--provisional clues as
to where more definite findings might be discovered.
Doubts have even been expressed as to whether a causal factor might exist, or that autism might instead be a composite condition of unrelated cognitive defects (Gillberg and Coleman 1992, p. 283; Waterhouse, Fein et al. 1996; Ozonoff 1997, p.875; Happé, Ronald et al. 2006; Happe and Ronald 2008). This could relate to the absence of suitable candidates. There is notably little interest in subdividing or dropping the category of autism, though this logically should follow the absence of a single etiological factor. There are several reasons in particular to remain open to single etiological approaches.
First, the evidence for multiplicity is not overwhelming. The work, for example, of Happé that argues against the existence of a single deficit shared by the central three symptoms of autism is not based on evidence upon those with autism but the inheritance of less extreme deficits in the general population that are inherited independently of each other (Happé, Ronald et al. 2006; Happe and Ronald 2008). There are methodological problems with this claim. One that it does concern the actually symptomology of those with autism: it may be that the inheritance of related but less extreme parallels involves separate processes to those involved in autism. A deficit, for instance, in extracting and consolidating higher-order information might affect diverse faculties without more limited impairments being so general.
Second, while a theory of developmental delay or retardation may have a single etiological factor, any theory of atypical development must be more complex and at least binary: both a theory of a core deficit factors and a theory of capacity factors that shape the cognitive adaptation of remaining intact processes. Any individual will differ in regard to the degree of their core deficit, and separate to this, their capacity to adapt using their remaining intact processes. Two individuals even with similar autism might therefore differ. For instance, it might be that one individual has a relatively minor core deficit but a severe incapacity to adapt intact processes to mitigate that impairment, while the other might have a severe core deficit but strong mitigating and compensating capacities. The possibility for heterogeneity even from a single core etiological factor, as a result, is large particularly when the diversity of inherited, environmental and therapeutic factors is considered that might amplify or reduce its cognitive effect.
Third, no reason exists to assumed the present understanding of normal cognition is sufficiently complete in regard to the full range of the information processes that underlie and enable everyday human cognition and its normal development. All major areas of scientific investigation even after they have developed to an apparently mature stage of full professional and academic development continue to discover major and unexpected additions to their topic phenomena (for example, the Kuiper Belt for astronomy in 1992; gases for chemistry in the late eighteenth century, and the noble elements in 1894; plate tectonics for geology in the 1960s; and in biology, viruses in 1898, archaea in 1977, and extremophiles in the 1980s). It is thus quite plausible that the failure to identify a core factor behind the atypicalness of cognition in autism is due to an incompleteness in the inventory of the core information processing operations carried out by the brain by present cognitive science.
Fourth, few of the “standard” findings in autism research so far identify what is specific to the atypicalism of autism rather than what is predominant but etiologically nonspecific. Research is nearly entirely based upon group comparisons in which overlap occurs between those with autism and those without (Mervis and Robinson 1999, p. 123). This leaves the actual nature of the atypically of autism unclear. As Volkmar and colleagues in a review noted, “Typical research designs involve a simple comparison between a group with autism and a control on the measure of interest. Significant results in such comparisons have generated sweeping theories of autism in the past, regardless of effect sizes obtained or the extent to which the variable of interest could be directly related to social adaptive functioning”, and “small effect sizes are unlikely to account for the very large gaps in social abilities and disabilities separating individuals with autism from their peers” (2004, p. 143). In the standard autistic-group matched control-group type experiment, usually a few individuals in the control nonautistic group exist that perform at the same level as the autistic group. However, the existence of these nonautistic “outliners” is statistically ignored when the means of the two groups are compared. But the existence of these nonautistic outliners can be etiologically important. Estimates of the prevalence of autism range from 34 to 60 per 10,000. Taking the higher estimate, a sample population will need to be the size of 5,000 to contain 30 individuals with autism. Thus, a research group of 30 individuals with autism will match in terms of sampled prevalence with one of nearly 5000 nonautistic individuals. If one individual (and usually there are more) exists in a typical control group of 30 implicitly sampled from this 5000 population, then by arithmetic the vast majority individuals that perform like those in the autistic group will not in fact be labeled as autistic. Indeed that for every individual that has a autistic-like performance that in the control group, there will be for those 30 individuals with autism, 200 individuals in the prevalence population from which the control is implicitly sampled.
Statistical studies into the replicability of research in biomedical research where investigations problems are less daunting than those faced with autism also suggests a need for caution. The conclusion of one study of clinical trials, genome and epidemiological studies is that “Most research findings are false for most research designs and for most fields” (Ioannidis 2005). The situation of present autism research therefore is not yet one where sufficient information exists to rule in or out the existence of a single or multiple etiology.
Fifth, etiological theories of biomedical conditions develop in two directions: bottom up and top down. They can be proposed in regard to the mechanisms of defective processes— empirically (bottom-up), or from an understanding of normal processes and the likely results that arise after they are impaired—theoretically (top-down). Top-down approaches may turn out to be particularly critical to furthering the understanding of the atypicality of autism due to the limited etiological quality, noted above, of much empirical research. This suggests a continuing need to identify the existence of potential sources of atypical cognition in a top-down manner from developments in the cognitive science of normal cognition.
Gillberg, G. and M. Coleman (1992). The biology of the autistic syndrome. New York, Cambridge University Press.
Happe, F. and A. Ronald (2008). "The 'fractionable autism triad': a review of evidence from behavioural, genetic, cognitive and neural research." Neuropsychol Rev 18(4): 287-304.
Happé, F. G., A. Ronald, et al. (2006). "Time to give up on a single explanation for autism." Nature Neuroscience 9: 1218-1220.
Ioannidis, J. P. (2005). "Why most published research findings are false." PLoS Med 2(8): e124.
Mervis, C. B. and B. F. Robinson (1999). "Commentary: Methodological issues in cross-syndrome comparisons: matching procedures, sensitivity (Se), and specify (Sp)." Monographs of the Society for Research in Child Development 61: 115-130.
Ozonoff, S. (1997). Casual mechanisms of autism, A unified perspective from an information-processing framework. Handbook of autism and pervasive developmental disorders. D. J. Cohen and F. R. Volkmar. New York, John Wiley: 868-879.
Volkmar, F. R., C. Lord, et al. (2004). "Autism and pervasive developmental disorders." J Child Psychol Psychiatry 45(1): 135-170.
Waterhouse, L. W., D. Fein, et al. (1996). "Neurofunctional mechanisms in autism." Psychological Review 103: 457-489.