1. Introduction
People successfully use language to communicate. This works even when they have never previously spoken to each other. People even use and understand sentences never before uttered. We probably don’t know you and you probably don’t know us, but we suspect you’ve had no problem following this text so far. All these facts deserve an explanation, and traditional views of language have one to give: we share a common language. The language contains strings we call sentences. These strings are made up of phrases that are made up of morphemes. And all these parts of our language have the meanings they do because of the meanings of their parts and the order they are placed in. You understand the foregoing sentences because you have mastered this language and, so, know its rules of composition (albeit tacitly) and the meanings of a great many of its words. The explanatory success of this story about language suggests a picture according to which meanings are stable and determinate.
On the other hand, people who have mastered a language often end up in arguments with others who have mastered that language. Many of these arguments are about matters that seem “merely” semantical. Moreover, it is beyond dispute that the meanings of terms change over time. And, it seems such changes sometimes happen over the course of a single conversation. These facts demand explanation just as the previous ones do, but these are not so nicely explained by a theory that tells us meanings are stable and determinate.
One trick, then, is to pull off the construction of a view that addresses these apparently competing considerations.
In Living Words (hereafter LW), Peter Ludlow presents a theory of word meanings as extremely dynamic and underdetermined (cf. Ludlow Reference Ludlow2014, 2–3). To give a feel for his notion of underdetermination, Ludlow has us consider a dispute over whether the racehorse Secretariat is an athlete. This dispute, we are told, hinged upon whether horses can be satisfiers of ‘is an athlete’ (Ludlow Reference Ludlow2014, 78). Let us grant to Ludlow for the moment that this is a disagreement about the semantics of ‘athlete’. Then here are a few ways one might characterize the dispute:
(1) Everyone is right, but the participants on either side mean something different by the word ‘athlete’ – they are speaking overlapping but different languages.
(2) Everyone is speaking the same language, and one side in the dispute is right while the other is wrong. There is a fact as to whether Secretariat satisfies ‘is an athlete’. It’s just that some semantic facts are highly non-obvious, even to those who are masters of the language being used.
(3) Everyone is speaking the same language, but the meaning of the term ‘athlete’ is underdetermined. There are no facts about the language which yield an answer as to whether Secretariat satisfies ‘is an athlete’. The dispute is directed at changing the semantic value of ‘athlete’ so as to rule Secretariat in or out.
In LW, Ludlow takes a view like 3 as his starting point and positions it against views like 1 and 2. On Ludlow’s view, with affinity to that of Davidson (Reference Davidson1986) in “A Nice Derangement of Epitaphs”, there are no (natural) languages in the traditional sense. But there are microlanguages – codes constructed on the fly to serve the purposes of language users (Ludlow Reference Ludlow2014, 3). The terms of these microlanguages have meanings that are both underdetermined and highly prone to change. Thus, the term ‘athlete’ is not a term of English at all – since nothing answers to the term ‘English’. The string, ‘athlete’, appears in many microlanguages, and this string has many different meanings, all of them underdetermined and subject to change.
LW’s contribution to the literature consists partly in a defence of this seemingly radical view regarding underdetermination and dynamicity. But it also presents a research project designed to explain apparent linguistic stability by importing insights from truth conditional semantics, syntax and the study of thematic structure. In this manner, LW is somewhat conservative in rejecting an ‘anything goes’ theory of language. LW contains a good deal of interesting detail and argumentation, as well as some case studies that are interesting independently of their role in Ludlow’s development of his view. (For example, the history of the term ‘hacktivist’ is nicely informed by Ludlow’s own involvement with the hacktivist community.) LW introduces a number of useful distinctions and draws on a great deal of academic and non-academic work ranging from case studies in Conversational Analysis to a TV-broadcasted conversation between Donald and Ivana Trump.
In this review, we lay out some of LW’s theses and detail some points of contention with them. Section 2 is devoted to the former task, and Section 3 is concerned with the latter. (Of necessity, we have been somewhat selective in our choosing.) Section 4 concludes with some broad reflections about linguistic stability and flexibility, and about the contributions and challenges to the philosophy of language that LW offers.
2. Some of LW’s theses
A basic premise of the book, and a starting point in Ludlow’s development of his view, is that words (construed as terminal lexical nodes) have meanings that are underdetermined and dynamic.Footnote 1 The underdetermination of word meanings, according to LW, brings with it the consequence that the ranges of words are underdetermined. ‘Range’ is a technical term employed in LW. We can think of ranges as playing a role in the theory much like that typically played by extensions. The important difference between a range and an extension is the underdeterminacy of the former and the determinacy of the latter. On Ludlow’s view, we can correctly say, as with an extension, that all the satisfiers of a predicate are in its range.
That a range R is underdetermined has the consequence that, for some x, it is not settled whether x is in R or outside of R.Footnote 2 Take, for example, a microlanguage 
containing the predicate ‘is an athlete’. As a function of the meaning of ‘athlete’ plus the relevant (non-syntactic, non-semantic, and non-pragmatic) facts, some things will be in the range of ‘athlete’. Then, for something, call it ‘Zappa’, the meaning of ‘athlete’ simply does not settle whether Zappa is in the range of ‘athlete’. And, so, neither an utterance of ‘Zappa is an athlete’ nor an utterance of ‘Zappa is not an athlete’ will be true in .Footnote 3
Note that, we have been avoiding use of the term ‘indeterminacy’ in favour of ‘underdetermination’. This is to respect the distinction drawn in LW between Quinean indeterminacy and meaning underdetermination. Ludlow points out that if a term is indeterminate in the Quinean sense, there may well be perfect agreement among users of a language with respect to the correct usage of a given -term. For example, all the members of a linguistic community might agree about the correct conditions of usage and assent surrounding sentences containing ‘gavagai’. But even perfect agreement of this sort is consistent with indeterminacy. Ludlow’s notion of semantic underdetermination is different. If the meaning, and hence the range, of ‘gavagai’ is underdetermined in Ludlow’s sense, then, there will be possible circumstances under which relevantly informed speakers of the language will disagree or simply express bewilderment as to whether something is a gavagai (cf. Ludlow Reference Ludlow2014, 85). Nonetheless, like Quinean indeterminacy, underdetermination is not properly characterized epistemically. When it is underdetermined whether a is an F, there is no fact that settles whether a is an F, other than a decision by users to change the meaning of F. (Hence what Ludlow means by ‘underdetermination’ is very different from what is typically meant in discussions of underdetermination of theory by data, for example.)
All this allows LW to explain why there are disagreements like the one over Secretariat’s status as an athlete: facts about the meanings of a microlanguage’s parts are largely controlled by the users of that microlanguage (Ludlow Reference Ludlow2014, 83). But, in the case of the Secretariat dispute, the control exerted over the meaning of ‘athlete’ does not suffice to resolve whether or not Secretariat is an athlete.
Of course, that a group of conversants can’t determine whether or not something is an F doesn’t mean that ‘that [demonstrating the object] is an F’ is semantically underdetermined, nor that the range of F is underdetermined. They may simply agree that they have too little non-semantic information to tell either way. Or, they may allow that the meaning of F in their microlanguage is partly determined by experts who are not currently partaking in their microlanguage. Putnam’s famous example involving ‘elm’ and ‘beech’ serves our purposes nicely: the fact that we can’t tell elms and beeches apart does not lead us to think that there is no fact as to whether the tree we are looking at is in the range of ‘is an Elm’ or ‘is a Beech’ (or neither). Semantic deference to experts seems to explain this case (Putnam Reference Putnam and Gunderson1975).
Notice that semantic deference cannot resolve all such disputes. One of the case studies Ludlow takes up is that of the recent controversy among astrophysicists over whether Pluto is a planet (Ludlow Reference Ludlow2014, 41). Many of the participants characterized the dispute as being about the definition of ‘planet’. Let us grant that that’s right for the sake of discussion. So in this case we have a metalinguistic dispute among the experts. And, we take it, we should not say that some of the astrophysicists knew the meaning of ‘planet’ and some did not. So there seem to be cases in which disagreement over whether a is an F is coherent, appears to be semantical, and appeal to the semantic knowledge of experts doesn’t settle the issue.
The case of ‘planet’ is also useful for introducing some of Ludlow’s ideas about meaning dynamicity. There was not always a controversy among the experts over whether Pluto is a planet – all agreed that it is. Now, however, nearly all agree that it is not. This suggests that the meaning of ‘planet’ has changed. On Ludlow’s view, meanings are dynamic because they are changed by discourse participants. In LW, the process of meaning change is called ‘modulation’ (‘modulation’ is also used to describe the outcome of such modulation processes) (Ludlow Reference Ludlow2014, 7). The activity that discourse participants engage in to settle on modulations is called ‘meaning litigation’. In addition to the case of ‘planet’, Ludlow gives a number of examples of purported meaning litigation and attendant modulation, with some being clearer cases of these processes than others. Some of the cases – those concerning the meanings of ‘planet’, ‘rape’ and ‘person’ – play out over a period of many years (Ludlow Reference Ludlow2014, Chap. 2). Other cases discussed by Ludlow play out over the course of a single conversation – cases such as a conversation involving Donald and Ivana Trump in which, according to Ludlow, the meanings of the terms ‘flash’, and ‘selling a candy bar’ are modulated (Ludlow Reference Ludlow2014, 31).Footnote 4
One interesting feature of LW is that it delves into some of the finer points regarding the nature of modulation. Ludlow notes, for example, that one may modulate a meaning by making it more explicit (‘explicifying’) without actually changing the range of things to which it applies. For example, we can stipulate that as a matter of the meaning of ‘moustache’, cats worn on faces don’t count as moustaches without ruling out any borderline or otherwise controversial cases of being moustachioed.
The most called upon distinction, however, is between narrowing (broadening) and sharpening (unsharpening). Narrowing/broadening is a change to a word’s meaning such that the word’s range comes to exclude/include certain things as satisfiers of the word. Sharpening is a matter of making borderline cases determinate by mitigating vagueness (Ludlow Reference Ludlow2014, 84). Interestingly, a word’s meaning can be narrowed without being sharpened, since there may be no borderline cases affected by the modulation. To take an example of Ludlow’s, we can narrow the range of ‘athlete’ by asserting that horses cannot be athletes. But this stipulation does not sharpen the meaning or the range. While it may have been underdetermined whether horses can be athletes, there is no relevant scalar dimension along which horses constituted a borderline case (Cf. Ludlow 2014, 87).
LW also delves into describing the techniques we use to modulate word meanings. Some of these techniques involve actions that don’t appear to have metalinguistic import. An example used by Ludlow involves an assertion of ‘Ludlow is bald’ in a microlanguage in which it is underdetermined whether Ludlow is bald.Footnote 5 Ludlow claims that assertions like these are metalinguistic proposals to sharpen (or, in some cases, broaden) a predicate by including the previously borderline case in its range.
This raises an interesting issue about the transparency of word modulation. Surely speakers sometimes intend to use a word without changing its meaning and simply assert that a contentious case falls within the word’s range without altering the range itself. Ludlow admits this possibility but within the confines of the view presented in LW, it is likely that many speakers will be frequently in error about the sort of act that they are undertaking. To take an extreme example of an argument that is clearly not about modulation, two people who disagree about whether or not Barack Obama was born in the United States aren’t suggesting ways to modulate the predicate ‘born in’. While Ludlow can allow this, it needn’t be obvious that people arguing over whether or not Secretariat is an athlete will take themselves to be engaged in a fundamentally different type of disagreement. However, besides speaker agreement, it’s not clear what determines the nature of the inquiry as metalinguistic or ‘empirical fact’ (Ludlow Reference Ludlow2014, 63). And there is no reason to think in the cases at hand that the speakers will agree the issue is purely empirical.Footnote 6 While we won’t pursue this line of thought here, it’s worth noticing that this result makes the view in LW much more radical than it may have seemed at first.
So much for a quick look at Ludlow’s views about underdetermination and dynamicity. But how, according to LW, are we able to communicate with others without first engaging in processes of modulation? To answer, Ludlow gestures at the notion of lexical entrainment that comes to us from the literature on conversational analysis (see e.g. Clark Reference Clark1992). He further posits that these processes of entrainment allow us to come to new conversations with ‘residual values from previous entrainments [and] … the expectation that we and our discourse partners are already entrained with each other’ (Ludlow Reference Ludlow2014, 40).
We wish to note a difficulty with this story. Ludlow often emphasizes the ease with which we change the meanings of the words we use in the course of our conversations. But if these changes come so easily, shouldn’t we expect that the sets of word/meaning pairs that distinct entrained groups converge upon will be quite different from each other? That is, why don’t the word/meaning pairs that small groups converge upon rapidly ‘drift away’ from those of other groups? Ludlow’s theory of modulation and residual semantic values at more localized levels appears to be in tension with the idea of convergence at more global levels. More details about the interaction of entrainment, modulation and convergence seem called for.
Putting aside worries about entrainment, a natural question to ask is whether modulation is norm-governed or an anything-goes free-for-all. Are there better and worse modulations? These questions are especially prescient given Ludlow’s endorsement of meaning egalitarianism – the doctrine that there are no privileged word meanings.Footnote 7 Ludlow offers several norms of meaning litigation (which we discuss further in Section 3.3). These principles are thought by Ludlow to be both descriptive and normative. This is to say that they are principles to which we often conform and also serve as evaluative criteria for the quality of modulations.
In this section, we have given a thumbnail sketch of Ludlow’s view, according to which all, or almost all, word meanings are dynamic and underdetermined. Underdetermination guarantees that for some things there is no semantic fact that settles whether those things are satisfiers of a given predicate. Dynamicity guarantees that meanings can and do undergo frequent modulation, even within the space of a single conversation. These modulations are evaluable relative to principles of meaning litigation. We turn now to more critical remarks.
3. Critical discussion
3.1. Ranges
Recall that Ludlow has introduced ranges and characterized them only negatively as like extensions but as ‘open and underdetermined’ (Ludlow Reference Ludlow2014, 3). This is somewhat disappointing. It would be nice to have a positive and more informative characterization of ranges. How could such a characterization go?
One (non-novel) idea is to use a few resources from set theory (or some other mathematically precise theory) to model ranges. On Ludlow’s view, for a microlanguage 
and predicate P of , if an utterance of ‘Pa’ is true, then a is in the range of P, and if an utterance of ‘Pa’ is false, then a is outside the range of P. Given this idea, a natural way to characterize the underdetermination of ranges is as follows:
For any P, there is an x such that x is not in the range of P and x is not outside the range of P.
Now consider the set of things such that they are in the range of P, call it ‘IN’; the set of things that are outside the range of P, call it ‘OUT’; and the set of things that are not in In and not in Out, call it ‘INnOUT’. This is all fine and coherent, we think, and allows us to assign a perfectly determinate set of sets (or, alternatively, a partial function) to each predicate P, which can serve as a model of its range. It turns out, however, that Ludlow is resistant to the use of what he calls the move to ‘precise mathematical machinery’ in semantic theories:
If you believe in meaning underdetermination then you can and should be cautious about the move to precise mathematical machinery in the semantics because, if the artifacts (like precise extensions or functions from possible worlds to truth values) of these models are taken too seriously, the theory will misfire precisely where it counts. (Ludlow Reference Ludlow2014, 128)
So modelling ranges as above isn’t consonant with Ludlow’s view.
What alternatives are there? It seems that in order to characterize underdetermination we must go higher order. That is, we must hold that for a predicate P with an underdetermined range there are objects x for which the right thing to say (in a metametalanguage now) is:
“Px’ is underdetermined’ is underdetermined.
The move is forced upon us at this stage, we think, because otherwise there is a definable set of things x for which Px is underdetermined. (It’s just those things for which “Px’ is underdetermined’ is true.) But since Ludlow’s view is that using mathematical objects to model ranges is too misleading, we must avoid a theory of ranges that delivers such a set.
Let us take now an instance of “Px’ is underdetermined’ that is underdetermined. For illustration’s sake, let’s suppose “Secretariat is an athlete’ is underdetermined’ is such a sentence. Ludlow’s theory explains why some sentences like this are underdetermined. The theory tells us it is because there are not enough metalanguage semantic facts to make the metalanguistic sentence “Secreteriat is an athlete’ is underdetermined’ true or false. But that also means that in the example case Ludlow’s theory entails: “Secretariat is an athlete’ is underdetermined’ is underdetermined. Hence, Ludlow’s theory has parts that are neither true nor false; so Ludlow’s theory is not true.
This line of thought is similar to Williamson (Reference Williamson2002) arguments concerning vagueness. If one likes the thought that some people are indeterminate with respect to ‘bald’ and to ‘not-bald’, one either has to acknowledge a sharp line between the bald and the indeterminately bald or needs to account for the non-sharpness of the line between the two categories by positing higher-order vagueness. But if we take a borderline case of a borderline case of someone bald, the sentence ‘He is a borderline case of baldness’ won’t come out true, pace the theory. The same seems to hold, mutatis mutandis, for underdetermination and Ludlow’s theory. Either ranges are perfectly determinate and can be adequately modelled by sets and functions, or Ludlow’s theory is not true. Perhaps there is a way to avoid the dilemma we’ve tried to lay out. But it is hard to see how that would go.
Of course, Ludlow might deny that his claims about ranges and underdetermination call for further characterization at all. It is interesting to note that ranges never actually show up in the explicit semantics he gives for microlanguages (cf. Ludlow Reference Ludlow2014, 101–114). But this makes us wonder whether one should posit ranges at all.
Interestingly, Ludlow considers an objection similar to the first horn of the dilemma, though it detours through modality:
I’ve said that word meanings are underdetermined, but suppose someone really pressed on this and said that, no, actually even the dynamic theory is forced into accepting a definitive extension for each use of the term ‘heap’. Why? Because even though no one in fact reasons analogically from canonical cases of something in the range of a predicate, for example ‘planet’ or ‘heap’, to everything that would end up in the range, we can ask what would be in the range if they had done this … We take the answers from all the closest worlds where I make the determination, and what we get is a fixed extension. (Ludlow Reference Ludlow2014, 127)
Ludlow’s answer is to claim that ‘would have’ is underdetermined and so, presumably, the truth of some of the counterfactuals is underdetermined. But it’s not obvious how to apply this sort of answer to our dilemma.
One final point about ranges. Ludlow claims that all word meanings and ranges are underdetermined (Ludlow Reference Ludlow2014, 5). This thesis applies not just to words like ‘athlete’ but also to the terms of mathematics and logic, terms like ‘straight line’, ‘three’ and ‘entailment’ (Ludlow Reference Ludlow2014, 6).Footnote 8 However, notice that many predicates of many microlanguages have empty ranges – e.g. ‘wizard’.Footnote 9 But a predicate with an empty range surely has a determinate range. Further, we can find predicates with ranges that are both non-empty and determinate. Think, for example, of the complex predicate ‘fisherman in Alaska with argyle socks on’. There may be only one person who satisfies this predicate, and no other people even come close. Worse still, this means that for any number n we can cook up cases such that in context some predicate has a determinate extension with exactly cardinalty n. (Suppose two Alaskan fishermen have argyle socks on. Suppose three do …) LW does not discuss these kinds of cases, but they look like counterexamples.
3.2. The T-schema and bivalence
Ludlow’s story about underdetermination seems to lead to T-schema violations. Ludlow calls our attention to an argument, attributed to David Braun, to this effect (Ludlow Reference Ludlow2014, 111). Consider a microlanguage in which ‘Secretariat is an athlete’ is underdetermined.Footnote 10 Relative to this microlanguage we can argue as follows:
P1. (An utterance of) ‘Secretariat is an athlete’ is true if and only if Secretariat is an athlete.
P2. The right-hand side of P1 is underdetermined.
C1. The left-hand side of P1 is false. (By P1 and P2)
C2. Contradiction against P1.
Thus, any underdetermined sentence in a microlanguage will constitute a counterexample to the T-schema.
Given the T-schema’s importance to our understanding of truth, this result looks problemantic for Ludlow’s (truth conditional!) approach to semantics. His response is to deny that in the relevant contexts, sentences such as ‘Secretariat is an athlete’ are admissible in the microlanguage. Ludlow writes:
My view is that no sentence is admissible until it is sharp enough to assert a claim that is modulated enough for us to assert a claim that is clearly either true or false. Let’s state this as a principle.
Microlanguage S-admissibility: No utterance u of a sentence S is admissible in a microlanguage L, unless discourse participants (tacitly) agree that the terms of S are modulated so that an utterance of the sentence will be determinably either true or false. (Ludlow Reference Ludlow2014, 112)
The view is not equivalent to the principle. Notice that the discourse participants could agree that their terms are sufficiently modulated but be wrong. For example, suppose they believe that Secretariat is a human and thereby agree that Secretariat is an athlete. Nevertheless, in their microlanguage it may be underdetermined whether a non-human can satisfy ‘is an athlete’.
But even ignoring this complication, why does labelling some sentences ‘inadmissible’ cause any less disruption of the T-schema? Because admissibility governs which sentences are parts of the microlanguage. As Ludlow puts it:
Not only are word meanings dynamically narrowed or broadened in these cases, but there is also a question of which words (and hence sentences) make it into the microlanguage. (Ludlow Reference Ludlow2014, 112)
The idea seems to be that since the principle guarantees that underdetermined sentences aren’t admissible (and hence not in the microlanguage) no microlanguage will contain sentences that violate the T-schema.
The solution might seem ad hoc, but it has some independent motivation relative to the lines of thought in LW. Microlanguages are systems constructed for the purpose of communicating, and they can serve their purpose only when their users sufficiently harmonize the meanings of their words. It is natural, then, to think that failure of such harmony entails that the language users are not using a (single) microlanguage.
Unfortunately, given some natural assumptions, the principle won’t help avoid T-schema violations. Suppose we have a conversational background in which discourse participants do not agree that ‘Pat is a bachelor’ is determinably true or false. (We might suppose that Pat is not old enough to be legally married and that this is a source of the disagreement.) Thus, an utterance of ‘Pat is a bachelor’ is inadmissible, and, hence, not part of the microlanguage. Say, furthermore, that discourse participants agree, on purely semantic grounds, to the truth of (A)
(A) All bachelors are unmarried.
Then the discourse participants may well agree to the truth of (E):
(E) Pat is unmarried or Pat is not a bachelor.
In fact, it is quickly appreciable that (E) is an entailment of (A), so it seems our discourse participants must agree to the truth of (E) on pain of irrationality. But this agreement suffices to secure satisfaction of the S-admissibility principle. So the principle is powerless to rule (E) out of the microlanguage. But if (E) is part of the microlanguage, then ‘Pat is not a bachelor’ is part of the microlanguage, despite its failure to be determinably true or false.
Here’s the argument in full:
P1. ‘Pat is unmarried or it is not the case that Pat is a bachelor’ doesn’t run afoul of the S-admissibility principle.
P2. ‘Pat is unmarried or it is not the case that Pat is a bachelor’ contains the sentence ‘It is not the case that Pat is a bachelor’ as a part. (By Ludlow’s Syntax of L [Ludlow Reference Ludlow2014, 105])
P3. For S of the form ‘S1 or S2’, if S is a part of the microlanguage then S1 and S2 are part of the microlanguage. (By Ludlow’s Syntax of L [Ludlow Reference Ludlow2014, 105])
P4. ‘Pat is a bachelor’ is insufficiently modulated to receive a truth value in the microlanguage.
C. ‘Pat is not a bachelor’ is part of the micro language and has no truth value. (By P1–P4)
There are various ways to resist this argument, but all of them seem pretty unpalatable. First, one may deny P1 and claim that ‘Pat is unmarried or it is not the case that Pat is a bachelor’ is inadmissible. But, this goes against the spirit of Ludlow’s characterization of admissibility:
I would say that, when we talk about sentences with indeterminate truth values, we are talking about whether utterances of them should be introduced into the microlanguage. When we are saying the truth value is indeterminate, we are saying that it is not admissible – it cannot be deployed to make an assertion. (Ludlow Reference Ludlow2014, 113)
Both participants agree that ‘Pat is unmarried or it is not the case that Pat is a bachelor’ is true, and the sentence can clearly be used to make an assertion. And this holds even if it has a disjunct that is unassertable. So, if it is inadmissible nonetheless, Ludlow owes an additional explanation. It seems to us that Ludlow would have to go in for an error theory regarding which sentences are legitimate parts of the microlanguage.
Second, we suppose one could deny P4 and hold that collective acceptance of a disjunction brings about tacit agreement among discourse participants that the disjuncts are determinably true or false. However, someone endorsing this would need to accept that genuinely underdetermined ranges can be made suitably determinate in strikingly easy fashion. Consider what this would mean for disjuncts involving vague predicates. To illustrate: let George be a man who is borderline bald. Then the admissibility of ‘It’s not the case that George is bald or George does not have a full head of hair’ would guarantee that George is no longer borderline bald, and this despite the fact that discourse participants may continue to disagree over whether George is bald.
Third, one could deny P2 or P3 – but we don’t see how this option is realistically open to Ludlow. We don’t understand how you could form a coherent disjunction out of two sentences, one of which isn’t even part of the micro language. The construction rules don’t seem to allow this in any case (Ludlow Reference Ludlow2014, 105).
We conclude that Ludlow’s attempt to ameliorate the difficulty raised by Braun is unsatisfactory.Footnote 11
3.3. Modulation and litigation
Ludlow rejects the view that the words we use have a core meaning, and that we come to better appreciate these meanings through investigation. Instead, he holds that processes of semantic investigation are typically processes involving changes of meaning. An obvious question, then, is: is any modulation we settle on as good as another? In Chapter 2, Ludlow argues that empirical discoveries trigger meaning litigation and that there are better and worse modulations, which are evaluable relative to internal norms of modulation and which bear on substantial normative issues. He also suggests that many of our key moral disputes are metalinguistic (Ludlow Reference Ludlow2014, 63). After a bit of exposition, our goal in this subsection is to make two critical points:
(1) The evidence adduced by Ludlow is highly equivocal. As such, it doesn’t make a strong case for his claim that important disputes in the social and political realm are often directed at meaning litigation (Ludlow Reference Ludlow2014, 55, 62). And even if these cases do involve (attempts at) metalinguistic litigation, it is not clear that the meaning of any one term is the target.
(2) Both of the cases considered by Ludlow that are described by him as concerning ‘more contentious and substantial issues’ – the cases of ‘rape’ and ‘person’ – share a feature that we think makes them poor general representatives of our disputes about substantial normative issues. This shared feature is that they are disputes in the legal realm, a realm where disputes often appear to concern the meaning of key terms in explicit legislative decrees in a way that is not typical. This makes it seem inappropriate to rely on such cases to support any general theses about the role of meaning litigation in disputes concerning substantial normative issues.
Ludlow begins this look at some purported cases of meaning litigation by looking at the recent discussion among astrophysicists about the meaning of ‘planet’. Ludlow describes his strategy, in part, as follows.
I’ll … take up an example case that is relatively less controversial – the definition of ‘planet’ – and use it to construct a model for how meaning litigation works. I’ll then turn to more contentious and substantial issues – the definition of ‘rape’ and the definition of ‘person’ and begin exploring how disputes about the meanings of those terms can be normative and fail to be normative. (Ludlow Reference Ludlow2014, 39)
So, the recent litigation concerning the meaning of ‘planet’ among astrophysicists is used as a guide for creating a model of how word meanings are and should be litigated. Ludlow is not as clear about what his model amounts to as we would like, but we will describe what we take the model to be below. After the ‘Pluto’ case is used to guide the creation of Ludlow’s model, the model is put to work in an effort to shed light on cases involving the modulation of more emotionally charged and normatively loaded words – in particular, the words ‘rape’ and ‘person’. Our exposition will follow this order of presentation.
3.3.1. ‘Pluto’ and norms of meaning litigation
At the end of the twentieth century, scientists and the rest of us alike counted Pluto among the planets of our solar system. As of roughly 2010, the consensus view among astrophysicists is that Pluto is not a planet. Instead, it is classified as a dwarf planet and a plutoid, where a dwarf planet is a sufficiently massive object that orbits the Sun but has not cleared other objects from its path, and a plutoid is a dwarf planet farther out in space than Neptune (IAU.org 2015).Footnote 12
Happily for Ludlow’s purposes, much of the recent discussion among astrophysicists about Pluto’s status as a planet and the popular media reports on that discussion characterized the issue as being about the definition of ‘planet’.Footnote 13 For example, Ludlow quotes Phil Plait in deGrasse Tyson (Reference deGrasse Tyson2009) as follows:
At the heart of the debate is our very definition of the word ‘planet’. Currently, there isn’t one. The International Astronomical Union … has no strict definition of planet, but has decreed that there are nine major planets, including Pluto … If the IAU doesn’t really know what a planet is, how can it know there are nine? (Ludlow Reference Ludlow2014, 43; Quoted in deGrasse Tyson Reference deGrasse Tyson2009, 104)
Furthermore, much of the discussion concerned the criteria an object must satisfy in order to count as a planet. For example:
Distinguish between a planet and a star by deuterium burning. There will be borderline cases, but so what? Some day, when we understand formulation of these bodies much better, finer distinctions can be made. (Quoted in Weintraub Reference Weintraub2007, 229)
The International Astronomical Union even formed a Planetary Definition Committee tasked with arriving at a definition of ‘planet’ in order to settle the matter of Pluto’s status (Ludlow Reference Ludlow2014, 43; IAU.org 2015; our emphasis). This is all prima facie evidence that the controversy was about the meaning of the term ‘planet’. The case seems to clearly exhibit the sorts of dynamic changes to word meaning that Ludlow thinks take place all the time, though often more covertly in less public and more messy contexts. All this suggests that the ‘Pluto’ case is a good candidate for guiding the construction of a model of meaning litigation.
Ludlow’s model seems to amount to this: discourse carries on for a time until a problem arises. In some cases the problem will hinge on non-semantic matters, but very often the problem will concern how the meaning of a key term should be modulated. In cases of this latter sort, discourse participants begin to adduce arguments for their favoured modulation(s). Ludlow identifies a number of strategies that such arguments appear to often rely on. Some look like good principles, others less so.
Ludlow tentatively endorses four “norms of meaning litigation”. We reproduce them here:
(i) Take undisputed cases and argue analogically for new cases (or against familiar cases).
(ii) Modulations should respect the bulk of canonical cases.
(iii) Modulations should track (not cross-cut) important properties.
(iv) Modulations should not be too taxonomically disruptive (Ludlow Reference Ludlow2014, 45–48).
To give an idea of how Ludlow sees these principles functioning, consider the following excerpt from Weintraub (2007):
Can we assert that a planet must have a moon? … This ‘must have a moon’ requirement would drop Mercury and Venus from the list of planets and make Mars questionable. Did Mars become a planet only after it captured its moons from the asteroid belt a few hundred million years ago? (Quoted in Ludlow Reference Ludlow2014, 46)
Ludlow reads Weintraub here as suggesting that modulations should preserve canonical examples and should not be “subject to contingencies like the comings and goings of their moons” – a clear enough appeal to (ii) and (iii), and perhaps (iv). (Ludlow Reference Ludlow2014, 46).Footnote 14
Among the lessons that Ludlow takes from the ‘Pluto’ case is that “empirical discoveries can serve as triggers” for re-evaluation of word meanings (Ludlow Reference Ludlow2014, 51). Indeed, the ‘Pluto’ case highlights this in an interesting way. For a long time, all agreed that Pluto is a planet. So, Pluto was clearly in the range of ‘planet’. Despite this, the discovery of thousands of trans-Neptunian objects and, later, the discovery that some of them are larger than Pluto, called into question the meaning of ‘planet’. If Ludlow is right, empirical discoveries can trigger meaning litigation even when previously there was no disagreement.
The principles that Ludlow identifies serve as guides and criteria for the evaluation of proposed meaning changes. Interpretation of these principles, Ludlow points out, is interest sensitive. So, for example, in the ‘Pluto’ case what counts as an important property per (iii) will be sensitive to the values and goals of astronomy. On the other hand, when meaning modulation plays out in a context not so closely tied to scientific inquiry, importance will typically be sensitive to a different set of interests.
The case of ‘Pluto’ involved the needs and interests of the scientific community. The next two cases, those of ‘rape’ and ‘person’, are cases concerning terms that connect more immediately with substantial moral issues.
3.3.2. ‘rape’
Ludlow tells us that he will follow the “discussion of the evolution of the meaning of ‘rape’ in Schiappa (Reference Schiappa2003), particularly as it relates to the question of marital rape” (Ludlow Reference Ludlow2014, 52). Ludlow judges that in many past legal cases, the possibility of marital rape was excluded as a matter of the meaning of ‘rape’. He offers up as evidence arguments made in various legal contexts. Here is one such argument, attributed to Lord Matthew Hale (writing in the 1600s):
the husband cannot be guilty of rape committed by himself upon his lawful wife, for by their mutual matrimonial consent and contract the wife hath given up herself in this kind unto her husband, which she cannot retract. (Ludlow Reference Ludlow2014, 52; Quoted in Schiappa Reference Schiappa2003)
And another, from Sir William Blackstone’s Commentaries published in Reference Blackstone[1765] 1859:
[B]y marriage, the husband and wife are one person in law: that is, the very being or legal existence of the woman is superseded during the marriage, or at least is incorporated into that of the husband. (Ludlow Reference Ludlow2014, 52; Quoted in ‘Rape and Battery’ 1954, 723n)
On Ludlow’s analysis, each of the above quotes corresponds to a different argument that held sway in many legal contexts at least into the 1970s (Ludlow Reference Ludlow2014, 52–53). The first is an instance of what Ludlow calls the ‘mutual consent justification’ (Ludlow Reference Ludlow2014, 52). The idea was that in marrying, a woman permanently gives consent to her husband. Presumably, then, this argument against the possibility of marital rape was to be filled out along the following lines: because a married woman cannot withhold consent from her husband and rape is forced sexual contact without consent, a woman cannot be raped by her husband. The second quote is illustrative, according to Ludlow, of what he calls the ‘marriage as property’ argument (Ludlow Reference Ludlow2014, 52). Presumably, moving from the premises of the marriage-as-property argument to its conclusion goes as follows. A person cannot freely choose to act upon him/herself without having consented to be acted upon by him/herself. Rape is forced sexual contact without consent. In the eyes of the law, the married man and woman are one person. So, in the eyes of the law, if a married man freely chooses to have intercourse with his wife, consent is secured. So, such acts cannot count as rape in the eyes of the law.
Next, Ludlow cites a countervailing argument given by Brownmiller (Ludlow Reference Ludlow2014, 53).
[C]ompulsory sexual intercourse is not a husband’s right in marriage, for such a ‘right’ gives the lie to any concept of equality and human dignity...A sexual assault is an invasion of bodily integrity and a violation of freedom and self-determination wherever it happens to take place, in or out of the marriage bed. (1975, 381; quoted in Ludlow Reference Ludlow2014, 53)
Ludlow suggests that Brownmiller is guided by some of the principles of meaning litigation: she is trying to establish that cases of marital rape are like other cases of rape with respect to the important properties of rape (Ludlow Reference Ludlow2014, 53). With his remarks to this effect, we take Ludlow to be offering evidence that the ‘rape’ case involves discourse fundamentally aimed at the litigation of ‘rape’. The argument seems to be that since we can see discourse participants using exactly the strategies used in cases of litigation according to Ludlow’s model, we have reason to accept that this is a case fundamentally involving meaning modulation and litigation.
It is important to Ludlow to point out that empirical discoveries played a part in motivating the controversy over whether a woman can be raped by her spouse. As Ludlow puts it, part of the reason for denying that marital rape counts as rape was belief in “the myth that [marital rape] is not as damaging or harmful as an attack by a stranger” (Ludlow Reference Ludlow2014, 53). But this is a myth that empirical research has disconfirmed.
We think it unclear that these legal theorists were primarily engaged in meaning litigation. Moreover, it would not be appropriate for Ludlow to take this for granted in this context. Among the points he is trying to establish is that disputes in the social and political arena about the proper modulation of a word’s meaning are sensitive to empirical discoveries and bear substantially on moral issues. But all will agree, for example, that whether a man can be prosecuted for marital rape bears substantially on moral issues. So it would be dialectically inappropriate to simply assume that the legal cases at issue are fundamentally about the meaning of ‘rape’. Though Ludlow is not explicit on this point, for these reasons, we read Ludlow as offering an argument that the ‘rape’ case is one fundamentally involving modulation and litigation.
However, we find it puzzling that Ludlow thinks citing the arguments he does supports his diagnosis of the case as fundamentally involving modulation of ‘rape’. Let’s grant for the moment that what was at issue was the (legal) meaning of some key term. Then the quoted arguments equally support the claim that all parties understood ‘rape’ to mean something to the effect of forced sexual intercourse without consent but that what was at issue was the meaning of some other word, perhaps ‘marriage’ (or ‘consent’). Notice too that our (we believe charitable) reconstructions of the mutual-consent and the marriage-as-property arguments both move from a premise articulating the operative meaning of ‘rape’ as forced sexual contact without consent. If our reconstructions are approximately correct, then it seems all parties associated a relatively stable meaning with ‘rape’.Footnote 15 So, it looks at least as plausible that the dispute concerned the meaning of ‘marriage’ and whether this meaning guaranteed that a woman has no right to withhold consent from her husband. And this is so even granting that the dispute fundamentally concerned the meaning of a word. We think this illustrates that it is often unclear exactly which word-meaning is at the centre of a metalinguistic dispute, even when holding fixed the assumption that the dispute is metalinguistic.
With this last point in mind, notice that the participants in a dispute that all agree is metalinguistic in character could each take a different word meaning to be at the centre of dispute. We can imagine, for example, a case in which all parties involved seek to settle whether a person can be prosecuted for marital rape, but some parties are working to modulate the meaning of ‘rape’ while others are working to modulate the meaning of ‘marriage’. And still others work to modulate the meaning of ‘consent’. Each of the arguments of Hale, Blackstone and Brownmiller might be given in the course of such a dispute, and such a dispute might end with changes to the law having the effect that people can be prosecuted for marital rape. Such imagined cases pose a problem for Ludlow’s picture of meaning modulation because it seems that at the end of such a dispute no single term will have been narrowed (or broadened) or sharpened (or unsharpened) by the lights of the users of the microlanguage. If that is right, then no single term in the microlanguage will have been modulated despite there being real changes to the legal consequences of certain actions. What this shows, we think, is that even with respect to some cases that are agreed to be metalinguistic in character, it is not clear how Ludlow’s modulation story can be applied.
So should we think that the legal episodes discussed by Ludlow concerning the possibility of marital rape are fundamentally metalinguistic in character? An alternative, and we think very attractive, characterization is that the important issue in these legal disputes was something like: should married women maintain the legal right to withhold consent? But that kind of dispute needn’t be metalinguistic. That is, to make sense of such a dispute we needn’t posit any important disagreement about the meaning of words such as ‘rape’, ‘marriage’, etc. Moreover, it seems to us that the fact that people were employing some of the strategies that Ludlow identifies as meaning-litigation strategies doesn’t provide much evidence one way or the other. This is because, for example, a person who argues about whether Pluto is a planet or Secretariat is an athlete may accept appeals to strategies like (i)–(iv) to settle such disputes, while denying that the disputes are metalinguistic. Such a person might simply judge that these principles are being used to sort things into kinds according to their properties, and that has nothing to do with language. Consequently, we find his discussion of the purported evolution of the meaning of ‘rape’ problematic.
3.3.3. ‘person’
Moving on to Ludlow’s discussion of ‘person’, Ludlow tells us that his aim is to “get clear on the nature of the debate and to outline the form that a productive debate would take” (Ludlow Reference Ludlow2014, 55). The legal episodes Ludlow discusses here, we are told, largely concern whether fetuses are in the range of ‘person’ in legal contexts. Ludlow also makes a radical claim about the character of many “key moral disputes”. He claims that:
[O]ur debates about what is a person are at bottom disputes about how to modulate ‘person’, not about some concept of person in Plato’s heaven. Of course the dispute is not merely definitional since there are profound consequences to our choice of modulation and our choice is not arbitrary but founded by important norms for the litigation of word meanings.
This claim may sound tendentious. Is it feasible to think that all moral disputes (or at least many of the key disputes) are metalinguistic? I believe the answer to this is yes. Many (perhaps all) moral disputes are fundamentally metalinguistic disputes (depending on what we count as a moral dispute; I am not counting disputes that all parties agree turn on empirical facts – for example on what is the more effective social policy). (Ludlow Reference Ludlow2014, 63)
Let us set aside for the moment whether when meaning litigation happens it can have substantial repercussions for normative issues. Our concern is that it is not clearly appropriate to generalize from disputes in legal contexts with normative repercussions to disputes over key moral issues in other contexts. Legal contexts are peculiar in that they often concern canonical texts – those laying out laws or the like – so, it seems that the fundamental questions in legal cases often hinge upon some sentence in a text and its correct interpretation in an atypical way.
Consider, for example, an exchange considered by Ludlow in which a judge points out that if it were established that a fetus is a person, the correct legal resolution of the case at issue would be decided, because of the 14th amendment’s use of the term ‘person’ and the protections it lays out for persons (Ludlow Reference Ludlow2014, 57). This suggests that the legal meaning of the word ‘person’ is an important part of the discussion at issue, which at first looks good for the case Ludlow is making. But it is hard to think of many clearly moral disputes outside the legal realm that will hinge in an analogous way upon the meaning of some explicit decree.
In any case, it looks like we can procure examples of cases that won’t be sensitive in this way to any such decree. Think, for example, of a dispute among parents, or ethicists for that matter, over whether it is morally permissible to spank a child. All parties could be clear enough on the meanings of the key terms, or agree to put aside issues stemming from such unclarities for the sake of argument, yet still argue quite coherently. This makes it appear that such disputes are often not fundamentally metalinguistic. Or take for example a question that thousands of undergraduates are told each semester corresponds to a key moral dispute among philosophers: are wrong actions wrong because of their consequences or for some other reason? Disputes over the right answer to this question often begin with the offering up of a number of action types that all parties agree are wrong. Next the parties try to hash out why they are wrong. But there is nothing obviously metalinguistic about this sort of inquiry. We can discuss the properties had by tokens of a kind without our discussion going metalinguistic. Our point is that the way that disputes in legal contexts often concern the meanings of key terms is special. This raises the worry that they are unsuitable for drawing lessons about the relations between meaning litigation and normative disputes in general.
These cases reinforce the worries regarding the transparency of litigation and modulation. It seems that on many issues, disputants will be in error about what they are doing if their dispute is analyzed as metalinguistic. There is nothing wrong with an error theory, per se – people can be wrong. But it’s a cost that should be paid for in persuasive evidence for the analysis.
It seems correct for Ludlow to claim that the choices we make about meaning can bear substantially on moral and otherwise normative issues. It is easy to see, for example, that because of the protections and rights for persons described in the 14th amendment, how we litigate the meaning of ‘person’ will have a substantial impact on how agents and others in the moral sphere are treated. And we think Ludlow is correct to suggest that how scientists litigate the meaning of terms such as ‘planet’ makes a difference to, for example, the explanatory power of the results of their research and the fruitfulness of their investigations. Indeed, all this appears to have the consequence that metalinguistic disputes are often normatively important and are governed by internal norms. But Ludlow’s claim that many disputes are fundamentally metalinguistic does not seem well supported. This is because there is good reason to doubt that the case of ‘rape’ is fundamentally about meaning litigation and because both of his cases studies from the social and political sphere involve legal contexts, which often hinge on word meanings in a special way.
4. Conclusion
Living Words constitutes an interesting attempt at developing two philosophical starting points that are seemingly at odds. First, there’s the idea that linguistic communication requires linguistic stability for speakers and hearers to be able to interpret one another. Second, there is the idea that microlanguages exhibit quickly occurring semantic shifts based on conversants’ needs and purposes. Ludlow argues, persuasively we think, that respecting the second starting point is perfectly compatible with importing the constraints identified by the abstract study of language (i.e. the thematic structure of predicates, syntactic theory, generative linguistics, etc.) and discourse analysis. As Ludlow puts it:
Presumably, there must be stable elements … otherwise it is difficult to see how [linguistic interpretation and interaction] could be possible. It is a safe bet that most of the non-terminal semantic rules are stable across these shifts (they may well be stable across all human languages). Thus the real dynamic portion would be in constructing the terminal (lexical) rules on the fly. (Ludlow Reference Ludlow2014, 114)
In other words, language users’ freedom to dynamically alter the meaning of their terms is compatible with certain static linguistic constraints that are relatively immune to shift.
However, Ludlow also argues that the meanings of terms are not merely dynamic but (perhaps necessarily) underdetermined. We, however, are at some pains to work out what Ludlow’s theory gains by his positing of underdetermination, other than trouble with the T-schema. Suppose that the predicate ‘is a planet’ is fully determinate and fails to include Pluto. If the language is suitably dynamic, under this supposition, nothing prevents us from viewing disputes over Pluto’s status as a planet as metalinguistic proposals to change the meaning of ‘is a planet’. Furthermore, the extension of ‘is a planet’ may well be expected to shift in light of conversational needs and empirical discoveries. It strikes us that under such circumstances we would expect to see the same patterns of dialogue, meaning litigation and the like. That is, wouldn’t the phenomena surrounding language usage look the same if ranges were determinate but dynamic?
Much of the argumentation in LW feels underdeveloped, sometimes frustratingly so. For example, the idea that words have ‘skeletal’ meanings is floated, but the development is left to scattered suggestions regarding thematic structure of predicates, which are sketchy at best. One final example to illustrate the point: the book ends with the following idea regarding linguistic entrapment.
We cannot be imprisoned by something that does not exist (that is by a fixed language). On the other hand we do need to be cautious in when and how we defer to the linguistic practices of our discourse partners, and we need to insist that semantic deference be paid only when it is warranted. More to the point, when necessary we should not hesitate to modulate word meanings aggressively so as to express the ideas (and feelings) we want to express. (Ludlow Reference Ludlow2014, 163)
The idea is tantalizing both with respect to semantic theory and political thought, but not enough attention is paid to its development. We worry about all the ways in which conversational styles, gendered linguistic practices, power relations and the like can enhance or hamper one’s ability to modulate word meanings successfully. One can presumably be just as imprisoned in a dynamic microlanguage they lack the power to change as in a static one.Footnote 16
Living Words is best seen, in our view, as laying the foundations for a project rather than presenting a true fait accompli. It is bound to frustrate some readers who are searching for the devil in the frequently absent details. But it has the potential to stimulate new and exciting lines of research into the fraught relationship between the fixed and the free aspects of language and language use. It is rich with interesting philosophical insights and it illustrates its main points using real, messy case studies rather than anodyne, sterile and contrived examples. All of which makes LW an intellectually worthwhile and entertaining read.
