INTRODUCTION
We should not feel triumphant about the advances in cognitive neuroscience, genetic s, or brain imaging or our general understanding of the etiological aspects of non-syndromic autism made over the past 30 years. Hence, we are pessimistic about the prospect of big future breakthroughs in our mechanistic understanding of autism. The reason for such pessimism is that recent research practices and methodological norms have overwhelmingly favored the production of type 2-like errors, thus not detecting mechanisms that account for the nature and existence of autism. Recent meta-analytical studies indicate that case-control effect sizes have decreased by up to 80% for neurocognitive constructs (emotional recognition, planning, capacity of cognitive perspective-taking, brain size, and EEG characteristics) that distinguish autistic from nonautistic people (Rodgaard et al., 2019).The gradual 30-fold increase in the prevalence of people diagnosed as autistics over the last 50 years coincides with the inclusion of individuals who are increasingly distant from the initial description (Fombonne, 2018; Hollin, 2017), resulting in increasing heterogeneity. The number of signs required to provide autism diagnosis decreased by a factor of two between 2004–2005 and2014 for children diagnosed at school age in Sweden(Arvidsson et al., 2018). The evolution in the demarcation of autism and the detection of the difference between autistic and non-autistic individuals has been accompanied by minimal replicability of structural and functional results in brain imaging. In genetics, the most important results are those that have ruled out an important causal role of entire classes of genetic abnormalities (such as deletions: Douard et al. (2021). Concerning interventions, the major findings have been the negative results that show the minimal or dubious effectiveness of intervention techniques (Brignell et al., 2018; Sandbank et al., 2020). Some researchers have suggested breaking down the autism spectrum into subgroups to treat this ailment. However, meta-analyses of studies attempting to create subgroups for the current autism spectrum report that the number of possible clusters may be impractically large, with most of the doubtful clinical value (Wolfers et al., 2019). Concluding that this demonstrates the validity of the spectrum category (Fombonne, 2020) may miss the point. The current dilemma may instead be explained by the current definition of autism spectrum not allowing the detection of subgroups because it gathers unrelated and dissimilar sets of individuals. Proponents of a dimensional position see such drift as progress. The sharing of diagnostic signs with other multiple psychiatric and neurodevelopmental conditions and the existence of common predisposing factors between autism and these same conditions could suggest that such a categorical distinction has become obsolete(Constantino & Charman, 2016). However, although categories are plagued by the problem of boundary, dimensions suffer from a problem of choice. The use of dimensional measures to treat the reification of diseases substitutes the grouping of individuals into a category deemed to be arbitrarily circumscribed, with the classification of individuals according to the measure of a dimension, of which the choice is even more arbitrary.Another intrinsic limit of the dimensional approach is the uncontrollable increase in the number of dimensions when the complexity of objects increases, or the “curse of dimensionality” (Feczko et al., 2019). Such assimilation confuses the possibility of measuring the same variable, such as reciprocal socialization (Constantino et al., 2003)or empathizing/systemizing pairs (Baron-Cohen, 2009) in all individuals of a group, and its explanatory value in a mechanistic model. There may be several causes at the center of the current situation: the standardization of inclusion strategies for individuals identified as autistic in research, the blind application of methodological rules, such as requiring a large sample size and the search for representativeness, a misuse of the pleiotropism analogy to autism without identified variants, premature and over-extended use of“autistic traits” as autistic, and a misconceived sign/specifier distinction. They will be discussed in this order.