Surveys of egocentric networks are especially vulnerable to methods effects. This study combines a true experiment--random assignment of respondents to receive essentially identical questions from either an in-person interviewer or an online survey--with audio recordings of the in-person interviews. We asked over 850 respondents from a general population several different name-eliciting questions. Face-to-face interviews yielded more cooperation and higher-quality data but fewer names than did the web surveys. Exploring several explanations, we determine that interviewer differences account for the mode difference: Interviewers who consistently prompted respondents elicited as many alters as did the web survey and substantially more than did less active interviewers. Although both methods effects substantially influenced the volume of alters listed, they did not substantially modify associations of other variables with volume.