Generic stereotypes

Generics are generalizations that are not explicitly quantified, like “Tigers are striped”. We use them to state a general fact about a kind, without explicitly specifying how many of its members we are talking about. Instead of saying that “Most tigers are striped”, for example, a generic generalization simply says that “Tigers are striped”. Although generics appear to be very simple generalizations, they have puzzled linguists and philosophers for decades. To understand why they are so puzzling, consider the following examples:

  1. Tigers are striped.
  2. Lions have manes.
  3. Turtles are long-lived.
  4. Humans are right-handed. (False)

Spelling out what these sentences say about the world (i.e. their truth conditions) has proven very difficult. This is partly due to the fact that different generics seem to mean different things. Clearly, none of them just express a universal generalization. Sentence (1) is true, for instance, yet there are some albino tigers without stripes. Perhaps then, (1) says that most tigers are striped. Yet this cannot be what all generics say, since (2) is also a true generic even though only adult male lions have manes. What’s even more striking is that (3) is a true generalization about turtles, even though 99% of turtles dies shortly after hatching. Still, it is a true statement that “Turtles are long-lived”. The puzzle is complete when one thinks about all the generic sentences that are false even though most members of the kind do have the generalized property, like (4). What do generic sentences say about the world such that “Turtles are long-lived” is true but “Humans are right-handed” is false?

According to some proposals, generics say that all normal members of the kind have the generalized property. Others believe that generics express something about the probability of a member of the kind having a property. The details of the most promising semantic theories are intricate and complicated, but none of them are without counter-examples. One goal of this project is to defend a new, epistemically motivated semantic account of generics.

Finding a correct semantic theory for generics is interesting in its own right, at least for philosophers of language and for linguists. But this debate also has wider implications. Some of the most objectionable stereotypes have a generic form. Consider statements like:

  1. Muslims are terrorists.
  2. Black people are lazy.

It is obvious that these statements are highly objectionable, perhaps even more so than explicitly quantified stereotypes. Why is that exactly? Perhaps one would want to answer that these generic stereotypes are false overgeneralizations. But given that generic sentences can be true even though only a minority instantiates the property (see 2 and 3 above), it is not clear that these sentences really overgeneralize. A second goal of this project is to spell out what exactly is said by uttering generic stereotypes like (5) and (6), and how one can respond to such statements. After all, if generics are not just claims about the majority of the kind, one will also not be able to falsify them simply by providing evidence of many counterexamples.

In my work on generic stereotypes, I argue that generic sentences have disjunctive truth conditions. This explains why generic stereotypes are so hard to falsify; they can be true based on several different conditions. Very roughly, (5) expresses the proposition that “most Muslims are terrorists” or ” at least some Muslims are terrorists by virtue of being Muslims”. However, I also argue that utterances of (5) by default implicate the conjunction of these two conditions. That is, saying that “Muslims are terrorists” by default implicates that “most Muslims are terrorists” and “at least some Muslims are terrorists by virtue of being Muslims”. ThisĀ  implicature explains why generic stereotypes strike us as particularly objectionable.

Due to their disjunctive truth conditions, generic stereotypes cannot be falsified by providing evidence that one of these conditions fails. Instead, I believe there are two strategies in responding to generic stereotypes. Either one must deny both the disjuncts, or one must argue that one of the presuppositions triggered by a generic stereotype is false.

In other work, I provide an account of the default implicature of generic stereotype utterances. Given my view that generics have disjunctive truth conditions, it is surprising that someone who says that “Muslims are terrorists” tends to be interpreted as saying that “most Muslims are terrorists” and “at least some Muslims are terrorists by virtue of being terrorists”. After all, given what is know about conversational implicatures, one would expect the expression of a disjunction to implicate that the speaker does not believe in the conjunction. I argue that generics are special, however, and that this implicature arises due to an assumption in the common ground that one can infer the ‘generic’ conjunction based on the ‘generic’ disjunction without any evidence to the contrary. That is, I hypothesize that people reason using an innate default inference rule such that when a property is a generic property of a kind in one way, it is a generic property in each way that is compatible with the nature of the kind and the property. Because this inference rule is present in the common ground, people who choose to utter a generic rather than an explicitly quantified generalization, are also interpreted as themselves implicating the conjunction. Hence why generic formulations of stereotypes strike us as particularly objectionable.