The purpose of this article is to outline the types and frequency of explanatory talk that occur in naturalistic conversations of low-income families of preschoolers. Thirty-one families participated in the study, tape-recording family mealtimes when their children were 3, 4, and 5 years old. A total of 75 transcripts were collected and analyzed for the presence of nine categories of explanatory talk, including intentional, causal, evidential, definitional/descriptive, procedural, and consequential. Explanatory talk consisted of conversation concerning some connection between objects, events, concepts, and/or conclusions that one speaker is pointing out to another. The most frequent type of explanations fell into intentional categories, which accounted for more than half of all segments of explanatory talk.It is a well-documented finding that children from low-income families do not achieve as well in school as their middle-class peers (Coleman, Campbell, Hobson, McPartland, Mood, Weinfeld, & York, 1966; National Assessment of Educational Progress [NAEP], 1981[NAEP], , 1985. Both NAEP reports indicated that middle-class children read and write better than children from low-income families. The Coleman report provided evidence that the gap between the reading abilities of children of different socioeconomic status groups expands as the children advance through the grade levels. The blame for this disparity in school achievement has often been leveled at parents, particularly mothers, because of their style of interaction with their children. A major assumption behind this view is that middleclass ways of talking with children support literacy development, while working-class ways inhibit it.Bernstein (1962,1972) specifically targeted social class differences in the style of interaction that occurs between parents and children at home. According to Bernstein, language constrains what and how a child learns, forming a basis for future learning. Studying families in Britain, Bernstein posited that children from working-class families were only exposed to restricted codes, styles of talk that are specific to the current physical context. These codes are limited, stereotyped, condensed, inexact, and nonspecific. Sentences are short and syntactically simple. On the other hand, mid-