Experiments in the measurement of the size of social networks

Abstract This paper reports the results of a series of experiments conducted to measure the effects of alternative procedures for determining the size of social networks. The three methods tested are recall, recognition, and numerical estimation. Respondents were selected from work, church, and social groups. The major findings were that recall estimates are substantially below those obtained by recognition methods and become worse for larger groups. On average, numerical estimates are reasonable approximations of network size estimates based on recognition, but there is substantial variability in the accuracy of individual reports. Surprisingly, there is no evidence that recall is better for reports on number of close friends than of friends or acquaintances. There are order effects on numerical estimates of network size and better estimates are obtained by first asking about close friends. Finally, there is some evidence that respondent characteristics such as age affect recall of network size.