Previous studies have shown that by 11 but not by 10 months infants recognize words that have become familiar from everyday life independently of the experimental setting. This study explored the ability of 10-, 11-, and 12- month-old infants to recognize familiar words in sentential context, without experimental training. The headturn preference procedure was used to contrast passages containing words likely to be familiar to the infants with passages containing words unlikely to have been previously heard. Two stimulus words were inserted near the beginning and end of each of a set of simple sentence frames. The ability to recognize the familiar words within sentences emerged only at 12 months of age. The contrast between segmentation abilities as they emerge as a result of everyday exposure to language, as assessed here, and those abilities as measured in studies in which words are experimentally trained is discussed in terms of memory-based mechanisms.