When shares are traded infrequently, beta estimates are often severely biased. This paper reviews the problems introduced by infrequent trading, and presents a method for measuring beta when share price data suffer from this problem. The method is used with monthly returns for a one-in-three random sample of all U.K. Stock Exchange shares from 1955 to 1974. Most of the bias in conventional beta estimates is eliminated when the proposed estimators are used in their place.