Opinion pollsters, political scientists, and democratic theorists have long been concerned with the normative and methodological implications of nonattitudes (Converse 1964). Of the proposed remedies to the weak and labile attitudinal responses proffered by an uninformed and disinterested public, perhaps the most ambitious to date has been Fishkin’s concept of the deliberative poll (Fishkin 1991, 1995, 1997). Combining probability sampling with information intervention and increased deliberation affords a unique insight into what might be considered the true "voice of the people." Yet, while deliberative polling draws heavily on the general notion of political sophistication (Luskin 1987), empirical analyses have tended to focus almost entirely on how the process of deliberation impacts on marginal totals of attitude items at both the individual and aggregate level (Fishkin 1997; Luskin, Fishkin, and Jowell 2002; Sturgis 2003). Little attention, in contrast, has been paid to outcomes that relate to other dimensions of opinion quality, such as attitude constraint. Constraint refers to the level of consistency between attitudes within an individual belief system that arises from a combination of logical, social, and psychological factors (Converse 1964). In this article we analyze data from five deliberative polls conducted in the United Kingdom in the 1990s in order to investigate the impact of political information and deliberation on attitude constraint. Across a broad range of issue areas we evaluate the extent to which the deliberative process impacts on statistical associations among attitude items between the first and subsequent waves of the polls. We conclude by discussing the implications of our results for the validity and reliability of survey measures of the attitude and the broader utility of the deliberative polling method as a tool of social scientific inquiry