Willingness to support public programs for risk management often depends on individual subjective risk perceptions in the face of uncertain science. As part of a larger study concerning climate change, we explore individual updated subjective risks as a function of individual priors, the nature of external information, and individual attributes. We examine several rival hypotheses about how subjective risks change in the face of new information (Bayesian updating, alarmist learning, and ambiguity aversion). The source and nature of external information, as well as its collective ambiguity, can have varying effects across the population, in terms of both expectations and uncertainty. In this paper, we use a particular example of a public policy problem--climate change risks. We focus on estimating the weights that individuals place on the opinions of government scientists and environmental groups as they update their subjective beliefs about the nature of this problem. We test four rival hypotheses about how people update their beliefs in the face of new and sometimes conflicting information. These hypotheses include (1) 2 strict Bayesian updating, (2) alarmist learning, (3) ambiguity aversion, and (4) cross-sectional homogeneity/heterogeneity in the updating process.For the climate change example, this research can be viewed as a precursor to understanding how willingness to pay to prevent climate change will evolve over time. 4 How do people combine alternative information about future climate prospects, in terms of both expectation and uncertainty, based on evidence or conclusions that they pick up from different sources? Individuals will have different views about the credibility of these different sources, and sources deemed more-credible can be expected to have a greater influence.In this paper, we focus on estimating the weights that individuals place on the opinions of government scientists and environmental groups in forming their opinions about future climate conditions in the absence of mitigation. The task is thus narrowly delineated. 5 The quasi-experimental exercise we examine here is a stylized version of what goes on in the real world, but it is sufficient to explore richer models than have previously been examined for the process whereby individuals update their priors in response to external information. Our key results are demonstrated for a convenience sample, so further research is clearly warranted.Nevertheless, the present data produce some very clear findings concerning our four main hypotheses: people do not use strict Bayesian updating, they are not, on average, alarmist learners, but there is considerable ambiguity aversion and systematic heterogeneity in the manner in which different people respond to multiple alternative sources of climate information.Section 1 provides a very simple model in the context of the climate change example and elaborates upon the four hypotheses. Section 2 describes the relevant portions of a survey instrument that was administered to a convenienc...