A clearer understanding of the strengths and weaknesses of measures of legislative preferences is essential for resolving substantive disputes about the composition of standing committees in legislatures. In criticizing interest group ratings as measures of legislative preferences, two recent works make important contributions in this regard (Hall and Grofman 1990, Snyder 1992a). However, some key methodlogical issues remain unclear or unresolved. This paper first formalizes and inspects Hall and Grofman’s claim that interest group ratings bias tests of the preference-outlier hypothesis to no-difference results when noncommittee members treat committee proposals with deference. It then examines Snyder’s model of artificial extremism which leads to a similar conjecture about no-difference results. Given a formal notion of deference and empirically plausible conditions, Hal and Grofman’s claim is shown not to hold. While Synder’s claim (as Snyder also notes) is shown not to hold in general, artifical extremism does tend to produce the conjectured bias on average. The bias is small, however, and the change that it leads to faulty inferences is also small.