We studied 50 unrelated pedigrees with a family history of retinoblastoma (Rb) (165 carriers of a RB1 mutation) to delineate the spectrum of RB1 germline mutations in familial Rb and to identify genotype-phenotype correlations as well as putative modifiers. Patients were followed at Institut Curie and they were examined by an ophthalmologist, a pediatrician, and a geneticist. All cases of familial Rb were determined via genetic counseling. Clinical features included disease status, laterality, age at diagnosis, mutation type, follow-up, and disease-eye ratio (DER). To eliminate mosaic cases, first-generation carriers displaying low-penetrance (LP) Rb were excluded from the analysis. Complete penetrance was the rule for nonsense and frameshift mutations (25 families) and high penetrance was observed for large rearrangements (eight families). Promoter (two families) and missense (two families) mutations displayed heterogeneous phenotypes and LP. Variable penetrance was observed for splice abnormalities (13 families) and was explained by in/out of frame mutations or respect of functional domains. Surprisingly, two families with the LP g.45867G>T/IVS6+1G>T mutation presented data that conflicted with the data reported in previous publications, as unaffected carriers had paternally inherited mutant alleles. Moreover, RNA analyses suggested that the lack of penetrance in unaffected carriers could be explained by an increase in expression levels of the wild-type allele. This observation prompted us to define a new class "3" of LP alleles. We believe this is the first large-scale study of familial Rb with a high level of homogeneity in the clinical and genetic analysis of patients and their relatives, thereby allowing for reliable intrafamilial genotype-phenotype correlations. Our analysis suggests in some cases the influence of modifier factors probably involved in mRNA level regulation and/or pRB pathway regulation.