Science and Semantics: the Case of Vagueness

Joan Weiner

Abstract

Hilary Putnam, as most philosophers know these days, cannot tell an elm from a beech. Nonetheless Putnam, along with those of us who share his ignorance, continues to make confident assertions about elms and beeches. For there are people who can distinguish elms from beeches. When we say that elms are good shade trees, we mean to be talking about the trees that these people call elms. Language is a social phenomenon and there is a division of linguistic labor. We do not take polls to determine the meaning of such terms as 'elm', 'tiger' and 'lemon'. We defer to experts—people who are carrying out relevant research in the special sciences. And one reason we defer to experts is that we think that the meaning of such terms is dependent not just on what people think but, in part, on nonpsychological nonlinguistic facts about, for example, biology.

Although Putnam's views on this issue have been immensely influential, these views have been widely ignored in the philosophical literature on vagueness. No doubt it is because it has seemed obvious to so many writers that science has no purchase on the linguistic phenomena of import to the philosophical problems of vagueness. The views that notoriously lead to paradox—that a single grain of sand cannot make the difference between a heap and a non-heap, that a single hair cannot make the difference between a bald and non-bald person, etc.—are simply widely shared intuitions. And, while these intuitions can be (and often are) challenged, the challenges are typically based, not on empirical evidence but, rather, on fuller discussions of the intuitions involved. When empirical evidence is mentioned, it is typically evidence about competent speakers with no particular specialized knowledge. This may not seem surprising. After all, how could we find empirical evidence that a particular object is or is not a heap, or that the movement of a single grain of sand can make (or destroy) a heap? There is no science of heaps; no experts to whom deference is owed.

But the vagueness of natural language pervades scientific research as well. Indeed, some predicates that figure in the literature on vagueness, for instance ‘bald’ and ‘thin’, are used in scientific research. Nonetheless since, for many vague predicates, empirical research tells us nothing, one might suppose that empirical research will have nothing to tell us about the correctness of a general account of the semantics of vague predicates—that the consequences of an account of the semantics of scientific and non-scientific vague predicates will be neutral towards that research. But, as I shall argue, this is a mistake. In this paper, I focus on two sorts of strategies for developing semantic proposals: the supervaluationist strategy and degree theory strategies. The supervaluationist strategy, as well as some varieties of degree theory strategies, yield accounts of semantics that are not consistent with accepted scientific methodologies. And while there is a variety of degree theory that fits with our scientific methodology, it has little to tell us about vagueness.

The supervaluationist directs our attention to precisifications—ways of sharpening the bounds of a predicate. Given a particular precisification, each person is either obese or not obese. On the supervaluationist view a sentence containing a vague predicate is true just in case it is true given any admissible precisification. In drawing our attention to precisification, the supervaluationist highlights an important feature of our understanding of the meaning of (at least some) vague terms. Precisification is an essential part of our attempts to investigate such things as obesity.

Suppose we want to determine whether obesity increases risk of heart disease. The most efficient strategy is to begin with a case-control study. The first step is to identify a number of individuals who suffer from heart disease (cases) and a number of individuals who do not (controls). The next step is to determine the proportion of each group that is obese. The results will come from comparing the proportions in the two groups. In order to find these proportions, each person in the study must be classified as obese or not. The researcher cannot, as we can in our everyday use of 'obesity', afford to view some individuals as unclassifiable—if so, she would be unable to determine proportions. There is, however, always the possibility that there will be subjects in her study who are not determinately classifiable given the everyday understanding of obesity. What should she do about this?

She cannot simply hope for the best and decide that, should she be unlucky and find such individuals among the subjects in her study, she will make a decision when the time comes. This strategy may bias the study. Instead she will decide, antecedent to beginning the study, on a sharp distinction between the obese and non-obese. Moreover, such precisification is required even if, as sometimes happens in such research, she decides to exclude borderline cases. For the exclusion of borderline cases requires an exhaustive tripartite classification of people as obese, borderline, or non-obese.

On the supervaluationist proposal, to say that obesity increases risk of heart disease is to say that, given any admissible precisification of obesity, the group of obese individuals is at increased risk of heart disease. The epidemiologist’s introduction of precisifications—and her use of a variety of distinct precisifications—fits in important ways with the supervaluationist view. But there are also important ways in which the epidemiologist’s methodology is, on the supervaluationist account, simply wrong. Admissible precisifications need not (and generally do not) speak with one voice. Suppose that, for some precisifications that seem from our pre-investigation perspective to capture our understanding of ‘obesity’, the epidemiologist has good evidence that obesity does increase risk of heart disease. And suppose that, for other admissible precisifications, she has good evidence that obesity does not increase risk of heart disease. Can the epidemiologist conclude that obesity increases risk of heart disease? The supervaluationist answer is that she cannot. But this answer is at odds with the practices of epidemiology. We can see by considering how actual epidemiological studies work that, provided the evidence is of the right sort, the epidemiologist can (and should) conclude that obesity increases risk of heart disease.

‘Obesity’ may not mark out a natural kind, but it is a subject of scientific research. Obesity, we assume, is some weight related characteristic associated with increased morbidity and mortality. Part of the epidemiologist’s task is to say what obesity is. If there is a particular point at which increased weight has adverse effects on health, this is a good place to draw the line between the obese and non-obese. The result may not exactly fit the usage of the untutored competent speaker. But untutored competent speakers once classified whales as fish. Insofar as we think that epidemiological research has taught us something about the effects of obesity, the supervaluationist proposal does not fit our views about the meaning of ‘obesity’. Nor can this result be remedied either by regarding the epidemiologist as giving us approximate truth or by regarding supervaluationism as giving us an idealized semantics.

Another strategy for dealing with the semantics of vague predicates is to appeal to some sort of degree theory. This strategy takes a number of forms. On some, sentences in which vague predicates appear are associated with degrees of truth. Vague predicates can be associated, not with a determinate set, but with a fuzzy set—a set for which membership comes in degrees. On others, vague predicates are associated with both a relation between objects and degrees and a standard or norm. Some degree theory approaches conflict with our views about good scientific practice. Others fit these practices but conflict with the views about vagueness that motivate this literature. Due to time restrictions, I will limit my discussion to the former.