For nearly a decade now, we have been conducting a new species of poll, a Deliberative Poll (Fishkin xxxx, xxxx; Luskin, Fishkin, and Jowell 1997; Fishkin and Luskin 1998). Reduced to essentials, the idea is to gauge the opinions of a random sample of citizens after they have had the opportunity of learning about, reflecting on, and discussing a set of policy issues. They are sent briefing materials, brought together in one place for a weekend, and then asked their opinions. There have now been nine regional Deliberative Polls, all in the US, and nine national ones, in the US, the UK, Australia, and Denmark. The first was in the UK, in April, 1994 (Luskin, Fishkin, and Jowell 1997), but easily the most prominent Stateside, and in ways the most ambitious, was the National Issues Convention of January, 1996, on which we focus here. 1

The project is partly and in origin an innovation in democratic practice, intended to address concerns in democratic theory. The value of democracy hinges jointly, we contend, on representativeness and deliberation. The democratic ideal is to aggregate the well-formed preferences of either the whole public or some much smaller set of people representing them by some form of majority rule, where by “well-formed” we mean something like “full-information.” We see deliberation as an indispensable means of informing preferences (Fishkin xxxx, xxxx). But how far and in what directions do well-formed preferences differ from those people actually hold? The rationale of Deliberative Polling presumes the existence of some sizable gap, but is there one, really? Perhaps the distribution of well-formed opinion can already be read from ordinary polls.