This present research analyses the linguistic environmental setting in homes of children under the age of two years from different social groups and looks at the extent to which the speech from other children contributes to shaping that environment. The corpus includes recordings of spontaneous speech from middle-class households in residential urban areas, from lower socioeconomic classes in marginalized urban areas and from impoverished semi-rural areas. Beta regressions were used to estimate whether place of residence could explain the proportion of words from other children that were directed, as well as those that were not directed, towards the child under study. The results showed that children from impoverished semi-rural areas and children from residential areas hear a higher proportion of non-directed words than children from marginalized urban areas. However, when child-directed speech is considered, the two lower socioeconomic groups, urban and semi-rural, hear a higher proportion of words than their residential peers. These results reveal heterogeneity within social groups. [Translation from Spanish by Jennifer Martin.]