Some 4,500 verbal exchanges among residents, and between staff and residents were recorded during structured observations in eight British purpose-built Homes for old people. These interactions were initially categorised as 'supportive/acceptive' or as 'other'. Subsequently, the 'other' data from four of the Homes was further classified as 'instrumental' and 'unclassifiable'. The participants in the exchanges were noted (resident or staff) as was the mental status assigned to residents ('confused' or 'rational'). The data suggest that interactions among residents are more likely to be supportive/acceptive than are those between staff and residents; that, similarly, interactions among rational residents are more likely to be supportive than those between rational and confused residents, or than those among confused residents. Interactions between rational and confused residents are more likely to be supportive than those among confused patients. Further, the data indicate that interactions between staff and rational residents, and between staff and confused residents do not differ in quality; predominantly, interactions between staff and residents are instrumental. These interpretations of the data are discussed in relation to the care of residents in institutional settings.