How do the millions of hours spent in each election on door-to-door canvassing affect voters' behavior? Data from the SRC election surveys of 1952, 1956, 1960, and 1964 are analyzed by maximum-likelihood methods to estimate the effects of such canvassing. Personal contact is found to be effective in increasing turnout, but not effective in influencing voter preferences for presidential, congressional, or local-office candidates. Repeated contacts are also found to be relatively ineffective. Gerald Kramer is Associate Professor of Political Science, and a staff member of the Cowles Foundation for Research in Economics, at Yale.