We report findings from a corpus-based investigation of three young children growing up in German-English bilingual environments (M = 3;0, Range = 2;3–3;11). Based on 2,146,179 single words and two-word combinations in naturalistic child speech (CS) and child-directed speech (CDS), we assessed the degree to which the frequency distribution of CDS predicted CS usage over time, and systematically identified CS that was over- or underrepresented in the corpus with respect to matched CDS baselines. Results showed that CDS explained 61% of the variance in CS single-word use and 19.3% of the variance in two-word combinations. Furthermore, the bilingual nature of the over or -underrepresented CS was partially attributable to factors beyond the corpus statistics, namely individual differences between children in their bilingual learning environment. In two out of the three children, overrepresented two-word combinations contained higher levels of syntactic slot redundancy than underrepresented CS. These results are discussed with respect to the role that redundancy plays in producing semiformulaic slot-and-frame patterns in CS.