Abstract
How do we determine whether individuals accept the actual consistency of a political argument instead of just its rhetorical good looks? This article answers this question by proposing an interpretation of political argument within the constraints of political liberalism. It utilises modern developments in the philosophy of logic and language to reclaim ‘meaningless nonsense’ from use as a partisan war cry and to build up political argument as something more than a power struggle between competing conceptions of the good. Standard solutions for ‘clarifying’ meaning through descriptive definition encounter difficulties with the biases of status quo idioms (long noted by theorists like William Connolly and Quentin Skinner), as well as partisan translations and circularity. Collectively called linguistic gerrymandering, these difficulties threaten political liberalism’s underlying coherency. The proposed interpretation of political argument overcomes this with a new brand of conceptual analysis that can falsifiably determine whether rhetoric has hijacked political argument.
Keywords
How do we determine whether individuals accept the actual consistency of a political argument instead of just its rhetorical good looks? To answer this question we first need to interpret what exactly is presupposed by an assertion in political argument that makes it part of the political argument. We then need the interpretation to render these assertions false if they have been hijacked by rhetoric.
An interpretation of the presupposition of political argument isn't difficult in an unrefined sense. In an unrefined sense, the problem of rhetoric doesn't even come up. We can say political argument is simply the process of giving reasons for thinking some solution to the problem of politics is acceptable (Barry, 2011: 2–12). Difficulties arise, however, when we try and refine this description further by elaborating what we mean by the political problem, which is to specify what makes the argument distinctly political. Political liberalism considers the problem as that of coordinating a peaceful coexistence within a plural society (Barry 1995; Rawls, 2005). From the liberal perspective, then, political arguments are attempts to render solutions to this coordination problem discursively acceptable to those affected by it. Such acceptability is a tenet of political liberalism to the extent that difficulties distinguishing between substantive political argument and non-substantive rhetoric threaten its underlying coherency. This article shows that modern developments in the philosophy of language have now provided political liberalism with a way around these difficulties.
While thin, the liberal perspective introduces a rather large constraint for the interpretation of political argument. Given plural societies are comprised of individuals with different conceptions of the good and corresponding comprehensive doctrines, no conception of the good can be presumed by the presupposition. That is,
THE LIBERAL CONSTRAINT No assumption is made concerning the truth of conceptions of the good or the truth of substantive beliefs of comprehensive doctrines in the presupposition of political argument.
So construed, political argument does not establish which claims about the good are true. That is what moral argument is for. Nor is it concerned with establishing the truth of cultural, religious, or literary platitudes. Political argument is only concerned with determining which solutions to the problem of politics are unreasonable. Reasonable solutions are solutions justifiable to everybody affected (including concerned onlookers) no matter their conception of the good. For example, an Islamic Caliphate imposed through force and brainwashing may be a solution to the coordination problem, but it presumably could not be justified to somebody who currently holds a non-Islamic conception of the good. That is to say, it does not comprise a ‘reasonable comprehensive doctrine’ (Rawls, 2005: xxx) and it is a position that would be ‘reasonably rejected’ in the contracting position (as specified by Barry, 1995) because it would not hold up in a political argument. It might turn out that secular conceptions of the good are the only acceptable conceptions for informing policy in a plural society, but this must only be assumed after political argument has established that religious conceptions promote unreasonable solutions to the problem of politics, not before.
Yet arguers often denounce arguments as ‘unreasonable’ merely for defending rival beliefs and intuitions different to their own. Even political theorists have been accused of packing their own cultural and religious prejudices into the concept of reasonableness (Parekh, 2000; Smith, 2010). Given this flouts the liberal constraint, the tendency for political liberalism to skirt the issue has been read as another nagging sign of its incoherence (Macintyre, 1988). I return that acceptable political argument can be distinguished independently of partisan beliefs and cultural background by way of conversational consistency and an openness to proving it. Political argument can be accordingly defined as a conversation where the conversation’s explicit assertions (S’s) tacitly presuppose. THE PRESUPPOSITION OF POLITICAL ARGUMENT S is consistent given the context of prior conversation, where S is directed towards rendering acceptable some X and X is a description of a solution to the coordination problem of peaceful coexistence in a plural society.
The problem of rhetoric
Returning to our initial point then, arguers might use rhetoric to come across as seemingly consistent even when their arguments are substantively inconsistent or incomplete. It would be convenient if it were obvious when an arguer has asserted A after previously affirming not-A or when they have not said anything meaningful at all; however, individuals often concoct diversionary linguistic tricks to avoid exposing inconsistent or incomplete arguments. Insofar as the problem hijacks political argument, victory will rarely be with the consistent argument and will instead be with those individuals who can speak with the strongest ‘accents of infallibility’ (Keynes, 1995: 87). Such accents can be donned unintentionally when speakers pick up bad conversational habits from politicians and ideologues trying to ‘defend the indefensible’ (Orwell, 1950: 87) with linguistic trickery or sloppiness. More precisely this is THE PROBLEM OF RHETORIC Speakers can intentionally or unintentionally misuse language, which has a persuasive effect on their audience, but which makes consistency ascriptions of the speaker’s assertions indeterminate.
Some of liberalism’s more trenchant detractors have deemed this impossible. 3 It is technically impossible, they say, because of the way language works. The conventional discriminations propagated by natural languages impose certain values, beliefs, and intuitions onto political argument without justification. As Connolly (1983: 1) puts it, certain ideologies are propped up by the ‘institutionalized structure of meanings that channels political thought and action in certain directions’. A narrow definition of social power, for instance, supposedly aids those who would otherwise be described as powerful because it allows them to skirt the responsibilities and reactions analytically connected to such descriptions (Connolly, 1983: 97). Similar considerations apply to political terms such as ‘freedom’, ‘equality’, ‘rights’, and ‘justice’ and terms typically used in their definitions like ‘institutions’, ‘law’, ‘person’, and so on. Quentin Skinner (1998: xi) suggests that no matter how progressive a thinker, they will always be constrained by the political vocabulary available in their time, which will usually be determined by the ideological makeup of their age. Young (1990) has further argued that cultural minorities accommodate the biases of majority culture when they adopt their national language. The liberal constraint will be flouted from the get-go if individuals have to buy into comprehensive beliefs tacit in the conventional language in order to communicate their argument to their audience.
A simple translation of the conventional idiom into one’s own idiolect is often taken as the obvious solution. ‘Clarifying’ the meaning of an argument by defining one’s terms is a ubiquitous dogma in both political theory and political science. The demand for definition is thought to be the key to rooting out accents of infallibility. ‘Clarification’ by way of definition is even considered powerful enough to overcome the biases of the status quo vocabulary. The idea is that speakers can posit definitions of salient terms in the conventional idiom with descriptions of what they meant in their more nuanced idiom. ‘Freedom’ in the conventional idiom might translate into something like ‘not being exposed to the possibility of arbitrary interference’ in a republican’s nuanced idiom.
The danger here is that the conventional idiom can be re-defined into an ideological straitjacket in order to render any position acceptable by definitional fiat. The classic example of this is in Orwell’s Nineteen Eighty-Four where the term ‘Freedom’ is re-defined by Big Brother as ‘Slavery’, ‘Ignorance’ as ‘Strength’, and terms like ‘Revolution’ were said to have no definition whatsoever. This definitional practice was designed to make it linguistically and eventually cognitively impossible to make arguments against Big Brother.
Such linguistic manoeuvers can be used to avoid ever revealing that an argument has been hijacked by rhetoric. Individuals can claim that their argument only seems like it is fudging over inconsistencies because their position is so strongly unconventional and difficult to express in the conventional vocabulary. If the argument is rejected they can claim it only demonstrates that the audience had unfairly built a conception of the good or comprehensive belief into the presupposition of political argument. It seems like political liberalism must therefore tolerate the practice of definition/translation-giving to give such arguments a chance, despite the practice licensing ideologues to render their position acceptable by definitional fiat. The definitional gambit can be played over and over if the initial suspicions of rhetorical cunning do not recede, with the claim of unconventionality being applied to the language used in each definition/translation, then the language of those definitions/translations, and so on. If this definitional filibustering continues long enough, the definitions/translations will become circular, with certain definitions containing terms that were defined earlier in the ‘clarification’ process. Those terms will accordingly define themselves.
It would be wrong though to infer that exposing circularity will also expose those arguments hijacked by rhetoric. Most dictionaries contain cycles since words used in each definition should be themselves words that the dictionary defines. There is nothing overtly incriminating about circularity in political argument given it is an inevitable feature of natural language conversations (Quine, 1951). We cannot then use circularity to call out those ideologues who skirt the charge of inconsistency by definitional filibustering forever and a day.
I call this kind of dodge-work and straitjacketing linguistic gerrymandering. In the same way a politician can gerrymander electoral boundaries to win an election without necessarily changing how voters vote, an arguer can gerrymander the meaning of words in an argument to render their argument seemingly acceptable despite it being inconsistent or incomplete. Given cyclical definition-paths, no definition of a term is technically unacceptable. The corollary of this is that there are no technical standards for a definition’s acceptability either. An audience can reject any definition that cuts against their interests given there is no technical reason for them not to. An arguer must either buy into the conventional idiom and comprehensive doctrines it props up or depart from it where departure risks accusations of linguistic gerrymandering. The extension of linguistic gerrymandering is therefore likely to turn on the comprehensive doctrines of speaker and audience.
What is more, the accusation is not just confined to crass ideologues. It routinely pops up in political theory and political science. Brian Barry (1975), for instance, thought Nozick’s argument for the minimal state resurrected cheap and nasty sentiments that had been dormant since the Elizabethan Poor Law simply by re-defining the term ‘coercion’ and using it as an argumentative trump. Critics of the libertarian position have run up against circularity when trying to call this supposed linguistic straitjacketing out (e.g. Cohen, 1995: 60). Libertarians also have similar complaints about proponents of redistribution (Nozick, 1974: 160).
A theory of language goes a long way
To accommodate for this, we need a theory that decisively identifies such gerrymandering as a misuse of language in the problem of rhetoric. Again,
THE PROBLEM OF RHETORIC Speakers can intentionally or unintentionally misuse language, which has a persuasive effect on their audience, but which makes consistency ascriptions of the speaker’s utterances indeterminate.
Theories of language typically appealed to in political theory fail to meet these demands. 4 Grice’s (1957) account of meaning, for example, cannot interpret the phrase ‘unintentional misuse of language’ without assuming a particular conception of the good or set of substantive beliefs exogenous to the theory. Meaningful communication for Grice is the assertion of a meaningful sentence S, where S is meaningful when the audience’s belief that x is produced by the audience’s recognition of the speaker’s intention for them to believe that x by the utterance of S. This theory would interpret an unintentional misuse of language, then, as the unintentional misuse of an utterance by a speaker where the effect of that utterance was as the speaker intended. The only possible way to explain how such an utterance might be a ‘misuse’ would be to appeal either to a conception of the good and condemn the speaker’s assertion as (unintentionally) immoral or a comprehensive belief and condemn the assertion as (unintentionally) false. Nothing endogenous to the theory of language will do it. Similar problems occur with Wittgenstein’s (2001) popular concept of ‘meaning as use’ for interpreting the problem. If the meanings of words are to be determined by their use, then the notion of misusing language is dubious at best. If we use our words to obscure inconsistent reasoning, this use forms part of their very meaning. There is therefore no way to conceptualise the problem of rhetoric with Wittgenstein’s theory of language without supplementing it with a conception of the good or comprehensive belief to explain why some assertions are specifically a ‘misuse of language’.
Robert Stalnaker’s (1999, 2004) two-dimensional theory of language is one of the few (if not only) theories to meet the demands described above. His theory is a version of what is known as possible worlds semantics. Possible worlds semantics interprets truth-conditions as possible worlds. Truth-conditions here are relevant to rhetoric since they are intimately connected to meaning: For S [where S is a sentence] to be meaningful is for it to represent the world as being a certain way, which is to impose conditions that the world must satisfy, if it is to be the way S represents it. (Soames, 2010: 1)
Stalnaker’s version of this theory of meaning is distinctive for his pragmatic accommodation of the problem the discovery of the necessity of identity (∀x,y(x=y⊃□x=y)) posed for possible worlds semantics. The accommodation requires a crucial pragmatic distinction, which I suggest is also the distinction pertinent to conceptualising the difference between rhetoric and non-rhetoric without presupposing a substantive belief or conception of the good. Using Saul Kripke’s (1974) lead, the problem is demonstrable with sentences like ‘Water is H2O’, ‘Bob Dylan is Robert Zimmerman’, and ‘the evening star is the morning star’. These sentences are unquestionably informative but they do not seem to partition possible worlds. Since the evening star and the morning star refer to the same object (Venus), the sentence ‘the morning star is the evening star’ can be interpreted as representing the world in exactly the same way as ‘the morning star is the morning star’. The sentence ‘A is B’ implies A and B can be substituted for one another in a way that preserves the truth-conditions of the sentences in which they are used, so ‘A is B’ shares the same truth-conditions as ‘A is A’ and ‘B is B’, which are obviously necessary truths. Given any informative sentence (and the assertion ‘the evening star is the morning star’ would be very informative to an Ancient Greek astronomer) must surely be meaningful, these sentences cast doubt on possible worlds semantics since there are some informative sentences that do not seem to partition the set of possible worlds at all because they are necessary and so true in all possible worlds.
Stalnaker thinks it possible to accommodate this problem by making a demarcation between the facts that determine the content of what is said and the facts that determine the truth or falsity of that content. While the truth of ‘Water is H2O’ depends on the empirical world (the observations of gas when water is heated) and certainly is not true by linguistic convention alone, it has the same conversational function as a purely linguistic definition. 5 It updates the facts in the conversational context that determine what is said by ‘this glass is full of water’ as much as ‘feminism means promoting equal-opportunity between women and men’ updates the facts that determine what is said by ‘I subscribe to feminism’. ‘Water is H2O’ does not ultimately determine the truth-value of the content of ‘this glass is full of H2O’. Facts that do this (e.g. that the only glass in the room is filled with something) are contingent and part of what Stalnaker calls the conversational context’s second dimension, whereas facts that determine the content of what is said are part of its first dimension. This demarcation accommodates for identity statements being both informative and necessary: being necessary, they do not partition possibilities in the second dimension but being informative, they partition the conversational context’s first dimension.
Before showing how these two dimensions uniquely capture the problem of rhetoric, it is worth unpacking this concept of conversational context, for it paints the picture of political argument I wish to advocate by filling out ‘the context of the prior conversation’ in the presupposition of political argument. The conversational context is the features of our world that give assertions their interpretations. This, Stalnaker suggests, is the common ground between conversers, It is common ground that
The context-set gives assertions their interpretation and in turn is updated by those very assertions. Take
Political argument involves giving reasons to defend some solution to the political problem. These reasons are assertions. By Stalnaker’s (2014) interpretation of discourse then, the goal for each arguer is to refine the initial common ground between individuals, through a chain of accepted assertions, to a new common ground in which a solution they think is good is uniquely acceptable. An audience does not have to think the possible solution true by their own comprehensive doctrines, only that the conversation has not yet discarded it, even if it might in future. Any ‘overlapping consensus’ (Rawls, 2005: 150) is not over what is considered true in comprehensive doctrines, but over what is false in the context of conversation. This bridges the gap for liberalism (see Nagel, 1991: 163) between propositions held true in one’s own comprehensive doctrine and propositions held provisionally acceptable for the sake of political argument and peaceful coexistence.
The question remains though as to what determines an assertion’s acceptability such that we might coherently discard possibilities and narrow the context-set towards this end. Stalnaker’s demarcation between the first and second dimensions of discourse gives our answer.
Both dimensions constitute the common ground that evaluates the input of an assertion. Yet they play two different functions in this evaluation to the extent that we might think of the context-set as being two distinct decks of cards, one for each dimension. The second dimension determines whether what is said is acceptable in the sense that what is said
If I were then to wriggle around their unacceptability by asserting, ‘Only marriage in accordance with religious scripture is true marriage’, and ‘By feminism I don’t mean equality, I mean the liberation of women from patriarchal institutions’, the assertions function as definitions and update the accepted facts that determine the content of assertions. The second dimension of the context-set does not update, however, since no facts independent of facts about the conversation’s language are accepted. Assuming there is no suspicion of linguistic gerrymandering or defective communication, these assertions update the first dimension of conversation by discarding all the possibilities where the words ‘marriage’ and ‘feminism’ partition secular and equality-related content, respectively. We might think of the first dimension as another deck of cards separate to the second dimension, where its cards (think of them as the game’s rule cards) represent the rules for translating assertions into particular partitions of the second-dimensional deck. Updates to the first dimension can discard rules that would otherwise prescribe multiple and conflicting second-dimensional partitions to assertions. For a second-dimensional assertion ‘A’ to be acceptable, then, there must be a first-dimensional interpretation of ‘A’ into a partition
Note the assignment of second-dimensional truth-values to the content of what I previously said might subsequently change if the first dimension updates. You might believe following scripture is the polar opposite of slavery, for instance, rendering my initial assertion, ‘The institution of marriage is domestic slavery’, now false in the context of the conversation.
That the facts in a context-set determine the acceptability of assertions in political argument seemingly conflicts with a popular view that facts are only relevant to normative justification by virtue of ‘a principle that is not a response to a fact’ (Cohen, 2008: 232). The second-dimensional claim ‘Income tax stifles economic productivity and economic productivity is key to raising living standards through real income’, for instance, might support the principle, ‘We should decrease income tax’. Yet the facts only ground the principle by virtue of a deeper principle, ‘We should raise standards of living’, supported in turn by ‘Standards of living are important for well-being’, explained with ‘We should promote well-being’, and so on. G.A. Cohen (2008) thought this regress inevitably continues until a basic fact-insensitive principle is revealed. I suggest this principle, in liberal terms, is a basic conception of the good that grounds out facts in our comprehensive doctrines. 6
Given the assumption of pluralism then, these basic principles cannot be expected to be part of our shared context-sets. It is tempting, in light of the regress, to infer from this that neither can the facts the principles support. Yet the temptation is mistaken. Context-sets contain every conceivable fact excepting only those that have already been discarded by accepted assertions or common knowledge. There will be more facts in the context-set the less arguers have argued with one another. The acceptability of a political argument turns on what detractors can make with the facts remaining in the context-set. This acceptability does not depend on the way the argument’s assertions are expressed, only on the second-dimensional partitions they make. If an audience hearing an assertion cannot work out a rejoinder that discards all the ‘true’ facts from the partition the assertion makes in the second dimension, they should accept the assertion for the sake of political argument until someone can, even if it cuts against their own comprehensive doctrines.
Explicit assertions of basic conceptions of the good are the one exception. The liberal constraint means we cannot expect audiences to accept our conception in the argument, we can only expect them to accept our conception as our conception (see the final section for a technical justification of this point). De facto affirmation of a comprehensive doctrine’s policy recommendation like ‘Income tax should not be decreased’, which is not a basic principle, 7 will nevertheless amount to de facto affirmation of the conception of the good that grounds it. Provisional support is given in political argument when all the rival policy recommendations of the audience’s comprehensive doctrines have been discarded. I might assert ‘Higher standards of living are not actually correlated with well-being’, using methodology so tight that those who believe otherwise are at a loss for what to say to go about discarding the possibilities the method assumes. This methodological argument would discard the argument against income tax sketched above and give de facto support for those comprehensive doctrines recommending the contrary. If n > 1 policy recommendations remain, argument must then shift to discarding possibilities for choosing between them (with considerations of fairness, equality, and justice presumably important fixtures).
Consistency and defective context-sets
The problem of rhetoric is tripped when the first dimension of the context-set is defective but where the second dimension is still being updated to accommodate assertions. In more precise terms, THE PROBLEM OF RHETORIC REDESCRIBED Assertions (perfectly heard/read/brought to mind) can update the second dimension of the context-set relevant to the argument’s subject matter, despite the relevant first dimensional facts being defective, thereby rendering consistency ascriptions for the assertions indeterminate.
The above interpretation of the problem relies on the term ‘consistency’ which has not yet been specified. Dropping the term forces us to describe the particular ways in which the context-set is defective when the problem of rhetoric is tripped, instead of leaving it unspecific (as whichever ways happen to make ‘consistency ascriptions for the assertions indeterminate’). Consistent assertions are simply assertions that have not been accepted as false in the conversational context and are not the product of a context-set defective in the two ways specified here.
The first is when conversers update their beliefs about the accepted facts in conversation differently to one another, leading them to believe a different
The second situation is where there are insufficient facts shared by the conversers to interpret the assertion into the second dimension at all. Say I cannot interpret the assertion ‘All democracies are just’ in a way that divides up truth-conditions in the context-set’s second dimension and yet nevertheless assert it (if I am the speaker) or tacitly accept it (if I am the audience). Given the assertion has been accepted into the conversational score, there is a shared belief that the assertion updates the second dimension, but I am unable to do so at present. Conversers bank the fact of the assertion’s utterance into the first dimension of the context-set as a fact that will (hopefully) in time determine the content of what was said, but needs further facts to do so. The assertion then plays the role of a wildcard in conversation. It functions a bit like the blank tile in Scrabble that slots in for any letter the player needs to make a word. The uninterpreted sentence slots in to discard any possibilities that need to be discarded for the truth (in the context of the conversation) of a follow-up assertion, say ‘The recent elections in Russia are a victory for justice’. If this follow-up reason accords with some weak first-dimensional beliefs about what sorts of reasons follow from the words used in the wildcard assertion (like the word ‘democracy’ having some conventional connection to elections), the follow-up reason is accepted without scrutinising the way such reasoning represents the world to be and any questions the representation might beg.
This is what the literature calls presupposition for accommodation. Lewis (1979: 340) describes it accordingly, ‘If at time t something is said that requires presupposition P to be acceptable, and if P is not presupposed just before t, then – ceteris paribus and within certain limits – presupposition P comes into existence at t’. Say I were to utter to somebody ‘I hope my sister will be able to proofread this article’. If they did not know that I had a sister, they should accommodate this fact into the presupposition of the conversation by discarding those possibilities where I do not, or else a perfectly informative sentence would be rendered meaningless. While accommodation is usually a benign phenomenon in sentences designed to inform, it is more insidious in sentences designed as reasons in a political argument (where agents stand to gain materially from certain outcomes). The question then is where the ‘certain limits’ in Lewis’ description of accommodation fall.
The first dimension of discourse will set these limits for the ‘banked’ wildcard sentences that trip the problem of rhetoric. The facts that determine what is said include facts that determine what kinds of assertions count as follow-up assertions from these wildcard sentences. While we may not be able to interpret these sentences we accept that they have nothing to do with representing certain possible worlds. For instance, the sentence ‘All democracies are socially just’ will presumably not accommodate for ‘there is more annual rainfall in the north than in the south’ since we know the two sentences have nothing to do with one another via linguistic convention alone. Sentences utilising the words ‘democracy’ and ‘justice’ are almost never used to prove a point about rainfall. Yet when two sentences are commonly used in the same argument (like sentences using the words ‘democracy’ and ‘justice’) accommodation will often occur. As Lewis (1979: 340) wrote, ‘conversational score does tend to evolve in such a way as is required in order to make whatever occurs count as correct play’. In a political argument, however, this sort of wild card accommodation, where sentences like ‘The recent elections in Russia are a victory for justice’ are rendered acceptable without due scrutiny, is not ‘correct play’, it is cheating.
The two sorts of defection listed above are exhaustive of cases that trip the problem of rhetoric. We can accordingly follow Stalnaker’s (1984: 82) standard of consistency, as three conditionals made up of propositions ( C1. If C2. If C3. If
The phrase ‘set of accepted propositions’ here constitutes the second dimension of the context-set. When these conditions are upheld, we have a paradigm case of non-defective communication, where there is no confusion over what is said and no disagreement over the truth-value of what is said in the context of conversation.
The above is an interpretation of both consistency and the problem of rhetoric that is endogenous to a theory of language and also adheres to the liberal constraint. Nevertheless, we still have no practical method to overcome the problem of rhetoric. We need conditions that, when adhered to, prove that the first dimension of argument is non-defective. If I were to argue consistently in a language my audience did not know and refused to submit myself to translating that language into the audience’s idiom, it does not matter whether or not I was consistent, I have not justified my position acceptably to my audience.
Stalnaker’s suggestions are not helpful to this end. He suggests that we ‘repair’ defective context-sets in the first dimension by defining our terms (e.g. Stalnaker, 2007: 258). When somebody is under the mistaken impression that the word ‘optician’ refers to a specialist for diseases of the eye, for instance, they might get worried if a family member told them ‘I am going to the optician this afternoon’. Defining what an optician means (i.e. someone who fits glasses) will repair it adequately. The problem though is that political argument is not merely informative communication. There are incentives in-built to having a seemingly acceptable political argument and therefore to hijacking conversation with linguistic gerrymandering in the first dimension to ensure reasons given are never unacceptable. In terms of the two-dimensional apparatus, the technical problem is that definitions do not discard possibilities from the context-set’s first dimension or refine it in any strict way because of the ever-present possibility that the definition-giving will be circular. Given there is always room to manoeuvre around past definitions with a first-dimensional house of cards of indeterminateness, caveats, and circularity, definitional/translation exercises will never refine the first dimension of a political argument’s context-set because no definition will render subsequent definitions unacceptable.
The solution is elimination, not definition
Arguers must be able to demonstrate that they can reduce the opportunities available to them to gerrymander in the first dimension of conversation while retaining their ability to partition the second dimension as they had supposedly done before to prove their argument does not exploit the problem of rhetoric. We must adopt a method that hotwires the first dimension and forces it to shrink, unlike conventional conceptual analysis, without triggering a parallel reduction in what we can do with the second dimension (i.e. without reducing what we can actually say but just what we have to say it with).
Instead of substituting a political term T for a description or synonym (i.e. definition), then, we should eliminate T and substitute the whole sentence S in which T was used. A similar strategy has been adopted by David Chalmers (2011; 2013) to identify a set of pointless disputes in philosophy and more informally by Richard Feynman (1969) to identify instances of failed scientific communication. We rephrase the sentence in a way that preserves its acceptability in the context of conversation, but does not re-use the term T or any of T’s cognates. Crucially, the term T is then barred from re-use in the remainder of our argument. As we repeat this process, our vocabulary will slowly be diminished and we will start speaking in increasingly austere ways. If we cannot rephrase sentences framed by the presupposition of political argument without using T in a way that preserves the initial distribution of acceptance and/or non-acceptance of the assertion, then this is evidence that the acceptance and/or non-acceptance was not substantive. It is proof that either the speaker is linguistically incompetent or their argument depends on the problem of rhetoric for its (un)acceptability.
We impose constraints on the first dimension of discourse by restricting the resources agents have to determine what is said, but without restricting what can be said. If agreement or disagreement in political argument turns on some trivial linguistic fact, the solution is to shrink the first dimension piecemeal in a way that will eventually remove that fact from the context-set. Arguers can still meaningfully talk about abstract principles and ideals without having a unique name for them. If we suspect the problem of rhetoric has been tripped, we assume inconsistency and place the burden on the speaker to prove otherwise. They do this by eliminating terms from their assertions and rephrasing the assertions in a way they think preserves the consistency conditions (C1, C2, and C3). The more they can subject their argument to the pressures of elimination, the more proof there is that their position does not depend on empty rhetoric. Yet if we assume competency and a rephrase is not forthcoming, this is evidence in favour of the initial hypothesis of inconsistency and the argument should be provisionally rejected.
Had Tony Blair’s ambiguity in Northern Ireland been subjected to the pressures of elimination and the term ‘violence’ been eliminated, he might have been incapable of rephrasing without revealing his hand to the unionists. Eliminating the term ‘freedom’ and rephrasing explicitly in terms of the positive and negative descriptions might also have improved Cold War diplomacy between East and West. The rhetorical accommodations involved with the positive (socialist) rephrase, roughly ‘the ability to be one’s own master’, which Berlin bemoans for having given considerable leeway for socialist tyrants to justify state expansion in the name of self-mastery, would plausibly have been less malleable had the term ‘master’ been eliminated in turn. Of course it is unlikely politicians will ever subject themselves to such a pedantic process, but their political arguments can be taken over by political theorists who, we can assume, are linguistically competent and can devote considerable time to the elimination exercise.
This gives a methodological mandate to political theorists as competent specialists with respect to the rhetorical properties of political arguments. It amounts to a new brand of conceptual analysis for political theory. Laypeople can confidently appeal to the results of this analysis in political argument because elimination is both falsifiable and accumulative. A hypothesis that a certain argument is substantive will be vindicated the longer the argument survives the elimination process. The hypothesis is falsified if, at some point, a rephrase is not forthcoming. Somebody can in turn reject this falsification by demonstrating that a rephrase is possible.
We have noted suspicions (Barry, 1975) that the libertarian revival in political theory is more of an ideological coup-by-definition than substantive argument. We can test this with the method of elimination without imposing our own definitions (and the comprehensive doctrines that license them) on argument. The method is designed to unravel rhetorical arguments on their own terms and in their own language. A central assertion of most libertarians is roughly, ‘Taxation is coercive given it interferes with individual rights and freedoms and should therefore be avoided.’
Eliminating ‘taxation’, ‘coercive’, ‘freedoms’, and ‘rights’ would encourage a more austere and lengthy rephrase, but one that is entirely in the libertarian’s own court. We could take Nozick’s (1974: 164–166) own suggestion and rephrase with Sen’s proof of the impossibility of the Paretian Liberal, Interfering with an individual’s decisiveness over a social ordering of the alternative uses of their property mistakenly takes these alternatives to be an object of social choice. Individual decisiveness over property instead restricts the alternatives the social choice can range over. This is demonstrated by the impossibility of preserving a principle of social choice (Pareto optimality) come what may when the basic good of minimal individual decisiveness is held constant.
If we eliminate forever and a day we eventually exhaust our vocabularies. We have to set a limit such that this does not happen.
Chalmers (2011: 543) introduces the notion of bedrock terms. These terms cannot be eliminated because we would not be able to communicate basic information about the world if we did not have them. The bedrock terminology constitutes an austere language which political argument should ideally tend towards when confronted by the problem of rhetoric. It is, however, a piecemeal process towards the ideal and only rarely will political argument need to descend into such pedantic meanderings.
Chalmers (2011: 543–549) posits a list of bedrock expressions. They include, existential quantifiers ‘some x exists such that’, universal quantifiers ‘for all x such that’, ‘consciousness’, ‘I’, and ‘that’. The basic idea is that truth-conditions described with terminology not on the list could also be described with the bedrock terminology. Non-bedrock terminology on the other hand cannot capture all the distinctions bedrock terminology can.
‘The good’ as bedrock
‘The good’ is bedrock in political argument. Bedrock terms include all their cognates, so the bedrock status of ‘the good’ is extended to its cognates like ‘should’, ‘ought’, ‘x is better than y’ and so on. If we were to eliminate all these terms, it would be impossible to rephrase moral claims. For example, it would be impossible to rephrase ‘you should not eat meat’ once ‘should’, ‘good’, and cognates have been eliminated. This raises the question why we should not lump moral claims under the category of non-substantive rhetoric given moral claims do not stand up to the pressures of elimination.
Following our two-dimensional interpretation, however, the problem of rhetoric can only be tripped if the second dimension is being updated and I suggest basic moral claims are not second dimensional claims at all. They are instead a special class of necessary claims, partitioning the first but not the second dimension of conversation. They are special because given the assumption of pluralism (and in keeping with the liberal constraint), the reference of ‘is good’ will be context-sensitive in political argument, depending on the views of the speaker to determine the extension of each token utterance. Most informative first-dimensional truths like ‘Water is H2O’ are not context-sensitive and therefore not bedrock. If we were to eliminate ‘water’ or ‘H2O’, we could either rephrase in terms of the other or with context-sensitive terms like ‘that thing there’.
This does not mean moral claims are uninformative though. ‘T is good’ has the same conversational function as informative identity statements that use context-sensitive terms like ‘I am T’ and ‘that liquid is T’. Like ‘Water is H2O’, these statements are both informative and (being identities) have the same truth-value in all second dimensional possibilities, but unlike ‘Water is H2O’, the assertions do not fix the reference of ‘I’, ‘that’, or ‘is good’ for subsequent conversation.
While we can eliminate ‘I’ in contingent assertions like ‘I am free’ and rephrase by substituting ‘I’ for a proper name ‘T’, we cannot eliminate it in the necessary identity ‘I am T’. ‘I am T’ explains the acceptability of the initial rephrase, whereas the timeless ‘T is T’ does not. Even ‘the speaker is T’ fails to explain it without the additional qualifier ‘and I am the speaker!’ (Perry, 1979; Stalnaker, 2014: 108–126). Likewise, we can eliminate ‘the good’ in contingent assertions like ‘Australia’s inflation rate is good’ by substituting ‘good’ for something like ‘unlikely to fall into recession’, but we cannot eliminate it from assertions explaining the rephrase’s acceptability. Namely, something (T) inherent to recessions is essentially bad, which is expressed with the necessary ‘not-T is good’.
Given the reference of ‘good’ changes from speaker to speaker, we cannot replace ‘good’ in the first dimension with ‘not-T’ as another speaker with a different comprehensive doctrine might pick out T with their use of ‘the good’. If we eliminate ‘the good’ from the first dimension, the function it plays in giving something a first-shot justification is lost like the function of locating a speaker or audience in the set of possible worlds is lost with the elimination of ‘I’ and ‘you’. This basic ‘not-T is good’ claim functions like what Cohen (2008) calls an ultimate principle in his fact-insensitivity thesis (just tweaking fact-insensitivity here to stand for second-dimension-insensitivity). 8 A linguistic community without a simple term referring to freedom or social justice but with a robust natural language would still be able to meaningfully refer to the same property (with a conjunction or disjunction of other terms) as those communities that possessed the term. They would still be able to describe the property of freedom with a complex of other expressions, perhaps including some expression for the good. A linguistic community without an expression for the good might be able to talk about social justice and freedom but they would not be able to describe these properties as appraisive and so explain why they are to be respected in political argument.
Basic claims about the good will not trip the problem of rhetoric given they do not have second-dimensional content. Second-dimensional content is important for refining the context-set to discard or reject such claims in political argument, but not for the initial token moral claim itself. Moral claims will not update the second dimension and therefore will not trigger the problem of rhetoric since the problem can only be triggered when the second dimension actually updates.
Conclusion
With this interpretation, the presupposition of political argument does not assume any conception of the good or comprehensive belief. Their extension is determined by context. The interpretation does not accordingly flout the liberal constraint. Neither does it concede anything to relativism. While context-sensitive, these conceptions and beliefs will be refined (i.e. discarded if rendered false by accepted second-dimensional content) over the course of political argument even when it is the arguer's own cultural or ideological beliefs that are discarded. Views are discardable in political argument if they cannot be consistently defended. Relativists might counter that political positions are always defendable with rhetoric and linguistic gerrymandering, but this is a technical rather than philosophical point that can be overcome with the method of elimination.
As a technical fix, I suggest the method of elimination better serves political theory than orthodox (historical or definitional) descriptivism. While this article has been concerned with an interpretation of political argument that meets the demands of political liberalism specifically, I believe it can be generalised. Given the interpretation assumes no comprehensive doctrine, it is at least reconcilable with accounts of the political that do. 9 It even appears to describe the procedure for meeting the ‘basic legitimation demand’ of political realism without flouting what Williams (2005) calls the critical theory principle. 10 It is of course outside of this article’s scope to defend these strong claims. I make them only to raise the prospect of the method of elimination as a general alternative for conceptual analysis in political theory.
Footnotes
Acknowledgements
The author wishes to thank David Chalmers, Keith Dowding, Sophie Fehlberg, Philip Pettit, Brad Taylor, and Albert Weale for advice on earlier drafts and/or the argument in its abstract. Keith Dowding has been especially instrumental. Thanks are also due to the two anonymous reviewers and Enzo Rossi for their constructive comments. This paper was presented at the 2014 APSA Conference at Sydney University and the 2015 ECPR Conference in Montreal. A related argument was presented at the Moral, Social and Political Theory Seminar at the Australian National University.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.