This study analyzes how others engage rural and urban Mozambican infants during naturalistic observations, and how the proportion of time spent in different engagements relates to infants' language development over the second year of life. Using an extended version of Bakeman and Adamson's (1984) categorization of infant engagement, we investigated to what extent a detailed analysis of infant engagement can contribute to our understanding of vocabulary development in natural settings. In addition, we explored how the different infant engagements relate to vocabulary size, and how these differ between the two communities. Results show that rural infants spend significantly more time in forms of solitary engagement, whereas urban infants spend more time in forms of triadic joint engagement. In regard to correlations with reported productive vocabulary, we find that dyadic persons engagement (i.e. interactions not about concrete objects) has positive correlations with vocabulary measures in both rural and urban communities. In addition, we find that triadic "coordinated joint attention" has a positive relationship with vocabulary in the urban community, but a contrasting negative correlation with vocabulary in the rural community. These similarities and differences are explained, based upon the parenting beliefs and socialization practices of different prototypical learning environments. Overall, this study concludes that the extended categorization provides a valuable contribution to the analysis of infant engagement and their relation to language acquisition, especially for analyzing naturalistic observations as compared to semi-structured studies. Moreover, with respect to vocabulary development, Mozambican infants appear to benefit strongest from dyadic Persons engagement, while they do not necessarily benefit from joint attention, as tends to be the case for children from industrial, developed communities.