Relationships between the measures of intellectual and physical status in the fragile X syndrome and the size of amplification of the fragile X-specific fragment, equivalent to the number of CCG repeats within the FMR1 locus, were studied by a maximum-likelihood scoring technique for analysis of pedigree data. This allows for estimation of random effects (genetic and environmental variance) concurrently with other (fixed) effects in a quantitative trait. FMR1 expression is usually shut down in males penetrant for the fragile X syndrome who have hypermethylated CCG amplifications of > or = 0.6 kb. The assumption of the step versus curvilinear function representing this relationship was tested by the likelihood-ratio criterion. The maximum-likelihood parameters were based on the most appropriate model for each measure. The results were indicative of the presence of a curvilinear relationship between the amplification size and the two intellectual scores, the Peabody Picture Vocabulary Test and Block Design Test, measuring verbal and spatial abilities, respectively. Reasons for the unexpected curvilinear regression between the amplification size and intellectual scores were explained further by methylation analysis of fragile X males with amplifications of 0.6 < delta < or = 1.2 kb who appeared to be responsible for the curvilinearity of the relationship. Four of these showed unmethylated status of the amplified bands in lymphocytes, which were presumably transcriptionally active. Removal of the aberrant individuals led to the anticipated step function between amplification and intellectual scores. For the combined anthropometric score, as well as for several single physical measures, the step function was the most appropriate model regardless of the inclusion or omission of the aberrant individuals in the pedigree sample.