The legitimization argument
It will help to begin by laying out the legitimization argument more explicitly. Some new terminology will smooth the way: the as-if practice for Xs is the practice of talking as if there are Xs, and as if they have certain sorts of features, or stand in certain sorts of relations (to one another or to other things). I will have in mind throughout cases in which the Xs are models, mathematical structures, fictional objects, or some combination of these (though I suspect my arguments would generalize to other kinds of putative entity). For example, Alma, a philosopher offering an account of scientific representation, might say,
Physicists often use mathematical structures to represent the systems they are investigating via relations of isomorphism; for example, they will use a function from an interval on the real line to vectors in an infinite-dimensional Hilbert space to represent the evolution of the state of an electron over time.
In uttering this sentence, Alma is talking as if there are mathematical structures, and as if they can stand in relations of isomorphism to physical systems; she is thus engaging in the as-ifpractice for mathematical structures. Similarly, Lester, a philosopher offering an account of scientific modeling might say,
One way scientists model real pendula involves exploring the ways in which they are similar to and differ from such fictional objects as the simple pendulum.
Lester is talking as if there are fictional objects, and as if they can stand in relations of similarity to physical systems; he is thus engaging in the as-if practice for fictional objects. One important note: the label “as-if practice” should not be taken to imply that the ways of talking to which it applies are mistaken or misleading; it is meant to leave that question open.
We can now set out the legitimization argument. Consider Iris, a philosopher of science who is offering an account of some aspect of scientific practice—scientific representation, modeling, explanation, confirmation... . Suppose that in offering her account, Iris is engaging in the as-if practice forX’s (models, mathematical structures, or fictional objects). Iris insists, however, that she can bracket questions about the existence and nature of X’s for the purposes of her work as a philosopher of science. To support this insistence, she offers the legitimization argument:
- (1) Any adequate account of ordinary and scientific discourse will legitimize our ordinary and scientific ways of talking about X’s.
- (2) When I engage in the as-if practice for X’s in offering my account of this aspect of scientific practice, I am engaging only in ordinary and scientific ways of talking about X’s. So, from (1) and (2):
- (3) Any adequate account of ordinary and scientific discourse will legitimize my use of the as-if practice for X’s in offering my core account. So, from (3):
- (4) Autonomy thesisfor X’s: I don’t need to answer questions about the existence and nature of X’s in order to provide a satisfying account of the aspect of scientific practice with which I’m concerned. So, from (4):
- (5) For the purposes of developing a satisfying account of this aspect of scientific practice, I can adopt the bracketing strategy for X’s: I can put aside questions about the existence and nature of X’s as questions I don’t need to address (even if, in developing my account, I engage in the as-if practice for X’s).
This seems to be the sort of justification many philosophers of science have in mind when they claim that they can set aside questions about the existence and nature of models, mathematical structures, or fictional objects even when considering core accounts in which talk of such putative entities plays a central role.
Before presenting my objections to this argument, I should address two questions about claims (1) and (3): First, what does it mean to call something an “adequate account of ordinary and scientific discourse” in this context? And second, what is it for an account of ordinary and scientific discourse to “legitimize” our ordinary and scientific ways of talking about X’s ? It is worth noting that any difficulties that might be involved in providing full answers to these perfectly good questions are, after a certain point, grist for my mill—they would simply provide additional grounds for doubt about the argument, over and above the grounds I will identify in the next section. Still, my own arguments in what follows will assume that the relevant notions are to be taken in a certain sort of way, at least, and so it would be a good idea to be explicit about what I am assuming along those lines before moving on.
First, then, I am assuming that an account of our ordinary and scientific discourse will answer various semantic questions about the utterances involved in that discourse. On the assumption that we engage in the as-if practice for Xs as part of our ordinary and scientific discourse, this will mean answering some central semantic questions about the utterances we produce when engaging in the as-if practice for Xs—our as-if utterances about Xs, as we might call them. And that will involve answering some questions about the existence and nature of Xs.
An ideal account of our ordinary and scientific discourse, I take it, will then have all the following features: it will make sense of the ways we ascribe truth values to the utterances making up the discourse (when we do), and of patterns in any other, less direct ways we have of treating such utterances as true, or false, or lacking a truth value; it will make sense of our epistemic attitudes to the utterances making up the discourse (making sense, for example, of the fact that we regard some utterances as knowable and others as unknowable, some as known and others as unknown); it will be internally consistent; it will employ an acceptable ontology, or at least not employ an unacceptable one; and it will both display internal coherence, and cohere with our best accounts of other kinds of discourse, at least where it seems it should.
Adequacy, of course, is a lower bar. Perhaps some of the features just listed are features that even a merely adequate account must have (internal consistency, for example), and perhaps adequacy is then a matter of having a sufficiently large proportion of the other features, or having enough of them to a sufficiently great degree (making sense of most of our ascriptions of truth values to utterances, for example). But saying this much is, I think, enough for the purposes of the arguments that follow.
What is it for an account of our ordinary and scientific discourse to legitimize our ordinary and scientific ways of talking about Xs (step ) or, more specifically, our engagement in the as-if practice for Xs (step )? I will focus on just the latter half of this question, as the as-if practice is the part of our ordinary and scientific way of talking about Xs that matters here. Now the most obvious way of at least making a start on legitimizing our engagement in the as-if practice would be to show that our as-if utterances are true (or at least those we seem to produce assertorically). But the proponent of the legitimization argument surely does not want to begin by assuming that any adequate account of our ordinary and scientific discourse will do that; surely she means to leave open the possibility that our as-if utterances about models, mathematical structures, or fictional objects are useful fictions, say. Otherwise, premise (1) of the argument is a highly controversial claim (and, one might add, a claim it might be hard to support without abandoning the bracketing strategy, which the argument is intended to justify). So I will take it that the proponent of the legitimization argument means to allow that legitimization could take any one of a number of forms. More specifically, I will take it that an account of our ordinary and scientific discourse could legitimize our ordinary and scientific ways of talking aboutXs in part by showing that our as-if utterances about Xs, or a sufficiently large proportion of them
- • are true, or
- • are useful fictions, in the sense of being useful falsehoods, or
- • though unknowable, are useful, or
- • though strictly meaningless, are apt, in the sense that they obey the rules of the (useful) game we are playing when we produce them, or
- • fall into one of the preceding categories.
This list of varieties of legitimization may well not be exhaustive; in what follows, I will assume only that everything on it is indeed a variety of (partial) legitimization.
Why, even on this more inclusive notion of legitimization, should we believe premise (1)? The best way of making this premise seem plausible, I suspect, is as follows: “It is clear that engaging in our ordinary and scientific ways of talking about Xs is (at least) a useful practice, and one with some reasonably well-defined internal rules. Given that this is clear, any account of our ordinary and scientific discourse, in order to be deemed adequate, will have to capture that fact—namely, the fact that talking in our ordinary and scientific ways about Xs is a well-defined and useful practice—and so will have to legitimize those ways of talking in at least that sense" This line of support might strike us as more persuasive for some values ofX than others: I take it that it is clearer in the case of mathematical structures that our ordinary and scientific talk about Xs is a well-defined and useful practice than it is in the case of models or fictional objects. In all three cases, however, this way of supporting premise (1) seems promising enough that, at least for the purposes of this discussion, I will put aside any further doubts of this sort.
We might also ask about the support for premise (2). What is required, note, is that in any particular case in which (i) a core account is offered which engages in the as-if practice for Xs, and (ii) the legitimization argument is invoked to justify the bracketing of questions about the existence and nature of Xs nonetheless, the core account in question must engage in the as-if practice for Xs only in ways which are part of our ordinary and scientific ways of talking about Xs. Whether this requirement is satisfied in this or that particular case will be a moot issue, however, if my main line of criticism succeeds.
-  See n. 4 for some philosophers who write in these sorts of ways.
-  (i) Perhaps there are other features an ideal account of our ordinary and scientific discourse should have, too;we need not settle that here. (ii) For the sake of brevity, I will write as though utterances themselves can havetruth values, and can properly be treated as knowable or unknowable, known or unknown, and so on; whatfollows could easily be rephrased to accommodate the idea that it is rather the propositions expressed by ourutterances which have such features.
-  This is not to say that an ideal account will deliver the judgment that our ascriptions of truth-values are correct,or that we are right to treat utterances as true, false, or lacking a truth value in the particular ways that we do;it might entail that we are mistaken in some or all cases, and perhaps go on to explain why we do not realizeour mistakes.
-  Which, similarly, is not to say that an ideal account will deliver the judgment that our epistemic attitudes arecorrect.
-  After all, what we find acceptable ontologically in the end might in part be a matter of seeing what the ontological commitments of an otherwise-adequate account of ordinary and scientific discourse are.
-  I want to allow this might only amount to making a start on the task of legitimizing our engagement in theas-if practice for two reasons. First, to talk about showing that our as-if utterances are true is implicitly tolimit attention to declarative as-if utterances. This is the right place to put the emphasis in the present context,because declarative utterances take center stage when we construct our core accounts; but legitimizing theas-if practice for Xs overall might involve dealing with other kinds of as-if utterance—questions, commands,and so on. Second, full legitimization of the as-if practice might involve treating things other than individualutterances: e.g., it might involve showing that the inferences we perform when engaging in our ordinary andscientific ways of talking about Xs (or a sufficiently large proportion of them, anyway) are good, in some senseof “good.”
-  Cf., for example, the view that (virtually) all our ethical utterances are false, but valuable nonetheless; or, rathercloser to home, the parallel view about our mathematical utterances.
-  I have in mind a view on which our as-if utterances are taken to be truth-valued, but on which we cannot knowtheir truth values. It might nonetheless be thought useful in one way or another to engage in the practice ofproducing utterances: in the service of producing an empirically adequate and explanatorily powerful accountof some aspect of scientific practice, say (perhaps given an appropriately extended notion of empirical adequacy, and noting that it will be controversial whether explanatory power is compatible with unknowability).Here van Fraassens constructive empiricism—a view about scientific theorizing about the world—is serving asa model of a view one might take about certain sorts of philosophical theorizing about science.
-  Cf. old-time semantic instrumentalism about theoretical utterances in the sciences.
-  That is, the legitimizing account could show that all our as-if utterances fall into the same one of the precedingcategories, but without establishing which; or it could merely establish the weaker claim that any given as-ifutterance falls into one of the preceding categories.
-  Perhaps this amounts to legitimization of the fifth sort just listed.
-  Except this one: Faced with this way of supporting premise (1), one might wonder why we should be confidentthat there will be an adequate account of our ordinary and scientific discourse which captures the (putative)fact that talking in our ordinary and scientific ways about Xs is a well-defined and useful practice. After all,in searching for an adequate account we are at least hoping to find an account which employs an acceptableontology, and which coheres with our accounts of other kinds of discourse. This makes adequacy sensitive toour ontological beliefs, and our beliefs about other kinds of discourse; and some of those beliefs may be wrong,even assuming that they we have not adopted them without justification. The response to this worry, I take it, is to say that if there is no account of our ordinary and scientific discourse which both captures the (putative) fact that our ordinary and scientific ways of talking about Xs is awell-defined and useful practice and coheres with our ontological beliefs, or our beliefs about other kindsof discourse, then given the (putative) self-evidence of the (putative) fact, we have good reason to reject the