I agree with several of the points made by Regan and Colyvan (this issue), but they have made important errors in their reading of the proposal by me and my coauthor Burgman (Todd & Burgman 1998). Regan and Colyvan raise two issues. The first is that we obtain fuzzy membership functions from probabilities and that this is not correct ("demonstrably fallacious," in their language). The second is that this excursion is not necessary. I deal with the issues raised in order. Regan and Colyvan correctly highlight some of the differences between probabilities and fuzzy memberships. In fact, we did the same thing in our discussion. But there is nothing new in this criticism of probabilities being used for fuzzy membership functions (for a survey see Hisdal [1988]). Kandel and Byatt (1978) point out that there "is a view of probability that frankly admits a subjective component ... there is an element of human judgement even in the seemingly most objective procedures for determining quantitative probabilities." Many if not most data sets will have elements of subjectivity; this is particularly so in many areas of data collecting for conservation biology. Turksen (1988) states that a fuzzy measure interpreted as a subjective measure expresses the fuzzy membership function, that probabilities can equal fuzzy membership functions as stated in our proposal. Regan and Colyvan provide a contradictory view. They should have directed their critique at the method provided by Turksen (1988) and how that may or may not apply in Todd and Burgman (1998) and explored the consequences of their view for the whole field of study supported by Turksen's (1988) work. As a consequence of some admittedly poor wording by Todd and Burgman (especially regarding the reference to Bosserman and Ragade [1982]), their criticism misses the point. We assume that subjective probabilities will be used to describe fuzzy membership, in the sense that the probability distributions will be used as a model for constructing fuzzy membership functions. Our proposed approach is not necessarily wedded to the equality alluded to by Regan and Colyvan. I agree with Regan and Colyvan that the categories of Millsap et al. (1990) are (arbitrarily) sharp. The case would have been better made if we had omitted the words "believed to have a. ..," as in the original manuscript. Regan and Colyvan appear to have misread our proposal with regard to the point they make that the machinery of fuzzy set theory is unnecessary. The proposal accounts for information provided as expert opinion rather than as measured variation. It suggests a means by which information of this kind might be combined with information given as probability distributions, together with point data. Regan and Colyvan do not acknowledge this. The whole point of the proposal was to present a coherent method by which to deal with different types of information without changing the original criteria so that biologists and policymakers could use the method without having to come to terms with a new set of criteria. Regan and Colyvan state that a number of fuzzy settheoretic operators are available and that they need to be selected carefully. Their criticism of our proposal seems unjustified because we (1998) made the same point (pp. 971-972). We provided "[t]wo examples that have some intuitive appeal for risk assessment" (emphasis added). We did not claim that the algebraic product and sum have more appeal than the standard min and max operators. Rather, they were mentioned simply as examples worthy of investigation. Moreover, Regan and Colyvan are dismissive of the standard approach (the min and max operators) and concentrate on the algebraic product and sum (an example of other types of operators) as though this were the only one we proposed. The algebraic product was used simply to show that, depending on the operator used, different membership values may be achieved. In their criticism of the use of the algebraic product, Regan and Colyvan failed to acknowledge that it does not matter which operator is used; the rank order of fuzzy sets (largest degree of membership to smallest) remains the same. Paper submitted February 11, 2000; revised manuscript accepted March 15, 2000.