This study was a part of a U.S.-Japan joint project, titled "A cross-cultural study of the influence of socializing agents upon cognitive functioning, communication styles, and educability in children," co-directed by R. D. Hess of Stanford University and H. Azuma of the University of Tokyo. This paper surveyed the stylistic measures of maternal communication with children and their relation with measures of children's cognitive development one and two years later. Sixty-four U.S. mothers and fifty-five Japanese mothers were observed on a referential communication game. The mother's task was to explain one target picture among four similar ones so that her four-year-old child could choose it from the same set of pictures. The protocols of their interaction were analyzed in four general categories: "preparation" or how the mothers start their explanation, "description" or how they explain each target picture, "feedback" or what kind of feedback they give to the child's errors, and "re-explanation" or how they re-explain the target after those errors. 1) Internal correlatians. In the category of "feedback," there seemed to be three specific styles in the U.S.; to negate child's errors explicitly, to explain how errors were different from the target, or to give no feedback. It was speculated that the U.S. mothers used explanation to avoid negative expressions. In Japan, only two styles were identified; to give feedback (either negation or explanation), or not. Giving diverse explanation was related to repetitive re-explanations in the U.S., while it was related to modified re-explanations in the U.S., while it was related to modified re-explanations in Japan. This suggested that there could be two culturally different ways to pursue the task: U.S. mothers tended to explain the target fully before they had their children choose it and Japanese mothers tended to modify their explanations according to children's errors. 2) Correlations with child's communication-game performance. Structured preparation showed good effects only in Japan. In both countries, modifying original explanations after errors was related to better performance. Totally renewed explanation was correlated with poorer performance in the U.S., while repetitive re-explanation showed the same negative effect in Japan. 3) Correlations with child's cognitive development measures. In Japan, giving organized pre-viewing remarks seemed to have a possibility to help children develop cognitive skills. In the U.S., the descriptive explanation style seemed effective. This variable of descriptiveness showed the clearest cultural differences between the two countries. Giving explicit negative feedback to errors was negatively correlated with cogntive measures in both countries. As for re-explanation styles, Japanese data showed a complex pattern; the modifying type was correlated with higher spatial and numerical abilities whereas the renewing type was with higher verbal ability. In the U.S., the renewing type showed general negative pattern. Verbal ability seemed to be rather promoted by the modifying type.