I propose an overview of work I have been engaged in on this topic over the last several years -- an update of the talk I gave in Manheim six years ago. My collaborators are Mick Couper and Roger Tourangeau for some of the studies and Michael Schober for others.

I will discuss how some features -- like progress indicators -- seem not to help while others -- like providing running tallies when a series of answers must sum to a total like 24 hours or 100% -- clearly do help. I will talk about recent work in which by programming web questionnaires to intervene when respondents exhibit certain signs of trouble -- taking too long to answer or taking too little time to answer -- the qestionnaire can provide help or otherwise motivate respondents to improve their perormance. Finally I will talk about how the interactivity of web questionnaires can blur the distinction between interviewer- and self-administration. To illustrate this I will talk about our studies on virtual intervieweers or animated interviewing agents that speak questions and vary in their visual realism. We are currently investigating the effect of race and gender of virtual interviewers on answers and asking what happens if respondents can choose a virtual interviewer with the characteristics they want.