Observational studies have shown that caregiver sensitivity predicts child language skills. These studies, however, have entirely relied on between-family designs (single parent-child dyad per family), which cannot rule out the contribution of shared family confounds (e.g., genetics, books in home). The current study investigates whether observed caregiver sensitivity predicts changes in child receptive language using a sibling comparison design. Participants were 890 Canadian children (51.7% male; 52.4% White) nested within 447 families from diverse socioeconomic backgrounds with children between the ages of 2 and 3.5 years (Wave 1) and 3.5 and 5 years (Wave 2). Independent observers provided ratings of maternal sensitivity with each sibling using several coding protocols (i.e., Coding of Attachment-Related Parenting and the Parent-Child Interaction System). Child receptive language was assessed using the Peabody Picture Vocabulary Test. Maternal sensitivity predicted within-person change in receptive language. That is, the sibling that receives comparatively more sensitivity from the caregiver showed more development in language over time when compared to their sibling. The obverse association, child language to later maternal sensitivity, was not observed, pointing to a unidirectional association of maternal sensitivity on child receptive language. Our sibling comparison design rules out the role of shared family confounders, which provides a strong test of causal processes within an observational design.