An expanded G₄C₂ repeat in C9orf72 represents the most common known genetic cause of amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD). However, the lower limit for pathological expansions is unknown (the suggested cutoff is 30 repeats). It has been proposed that the expansion might have occurred only once in human history and subsequently spread throughout the population. However, our present findings support a hypothesis of multiple origins for the expansion. We report a British-Canadian family in whom a ∼70-repeat allele from the father (unaffected by ALS or FTLD at age 89 years) expanded during parent-offspring transmission and started the first generation affected by ALS (four children carry an ∼1,750-repeat allele). Epigenetic and RNA-expression analyses further discriminated the offspring’s large expansions (which were methylated and associated with reduced C9orf72 expression) from the ∼70-repeat allele (which was unmethylated and associated with upregulation of C9orf72). Moreover, RNA foci were only detected in fibroblasts from offspring with large expansions, but not in the father, who has the ∼70-repeat allele. All family members with expansions were found to have an ancient known risk haplotype, although it was inherited on a unique 5-Mb genetic backbone. We conclude that small expansions (e.g., 70 repeats) might be considered “pre-mutations” to reflect their propensity to expand in the next generation. Follow-up studies might help explain the high frequency of ALS- or FTLD-affected individuals with an expansion but without a familial history (e.g., 21% among Finnish ALS subjects).