Lack of expressive and receptive language skills can have a negative effect on a developing child if not identified and remedied early in the child’s life. Current community and individual strategies to identify families with children who may need additional support are limited, and may not be sufficient to detect child language problems before they become entrenched. The present study explores the feasibility of using observed indices of parental language as a means of identifying families whose children are at risk of poor outcomes. Fifteen-minute speech samples taken from videotaped observations of 68 English speaking Welsh parent-toddler dyads interacting in the home during free-play were coded for 11 categories of parent language. Three complex measures were developed through factor analysis; parent prompts, encouraging and critical language. Two simple language indices (parent total words and total different words) were calculated for comparison. Two complex measures evidenced acceptable levels of inter-rater reliability, reasonable stability over time (p < 0.05) and some construct validity in terms of their association with socioeconomic disadvantage. ‘Parent prompts’ predicted toddler receptive and expressive language six months later (p < 0.05). In comparison the two simple measures were more reliable and stable over time and were just as strongly predictive of toddler language. The findings suggest that observed indices of parental language could prove useful in identifying high-risk families in need of specific support, such as parent training or other speech and language support, and the use of simple measures could be integrated into the assessment frameworks used by existing Early Years services. Further research is required to establish the feasibility of integrating such methods into current service delivery and to establish the overall cost for Early Years services of incorporating this measure.