![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
ABSTRACT
Talk of the functions of language or concepts plays a central role in developing an appealing pragmatic approach to conceptual engineering. But some have expressed skepticism that we can make any good sense of the idea of function as applied to concepts or language, or argued that the most we can say is that the function of ‘F’ is to refer to the Fs. In this paper, however, I argue that identifying linguistic functions is not hopeless, and that we can make progress by working at the level of system functions, and drawing on work in systemic functional linguistics. For that enables us to develop a better framework for thinking about the functions that language serves, and the ways its subsystems contribute to those functions. This approach to understanding linguistic functions enables us to develop a pragmatic approach to conceptual engineering that provides standards for conceptual engineering without metaphysical mysteries. It also enables us to make progress in figuring out what functions certain philosophically interesting parts of language serve, and how they are introduced into language, in ways that may disentangle us from a range of old philosophical problems.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 The reasons for this will become clearer below, once we get a better view of the diverse functions different parts of language serve. For discussion of the problems with assuming that our logical terms track the 'logical structure of the world', see Thomasson (Citation2015, Chapter 11).
2 For development of this idea that language should be thought of as a cultural artifact which, like other cultural artifacts, should be seen as serving functions, see my (2021), drawing on work by Irmak (Citation2019) and Kaplan (Citation1990).
3 Michael Williams here is summarizing Richard Rorty's views.
4 Where 'Represent' is here taken in the sense of mirroring, standing for, or tracking worldly features (Price Citation2011, 5). This does not preclude the idea that most mature language does serve an ideational macrofunction, in the sense to be described below.
5 Some have also appealed to sameness of function to address what is often thought of as Strawson’s challenge (Citation1963)—to identify sameness of topic, across conceptual revisions, or, as it is sometimes put, to address ‘the limits of revision’ (Cappelen Citation2018, 180–82). This won’t be my central concern. As I’ve discussed elsewhere (Citation2020a), I don’t think that we should take the question ‘when is a concept or topic the same?’ with deep metaphysical seriousness (as if there were some boundaries to be discovered); and when we ask whether we should make certain revisions in our conceptual scheme, the way to evaluate that is not on the whole to ask if it involves a replacement (rather than a modification) of a concept. There is again a parallel with civil engineering here: questions about what ought to be modified, torn down, restructured, etc. in civil engineering should not be guided at all by metaphysical questions about whether the same walls, screws, boards, and buildings would remain, or if they’d be ‘replaced’. Instead, the questions resolve more directly into questions about our goals and how best to fulfill them, regardless of these ‘metaphysical’ identity issues.
6 This focus on language also harmonizes with Macarthur and Price (Citation2007, 94), who (with some reservations) similarly speak primarily about 'linguistic' functions, in part to avoid introducing the inapt term 'representations' as the standard way of covering both linguistic and conceptual items.
7 For further discussion of whether to think in terms of engineering language or concepts (and arguments for a psychological version of a conceptual approach), see Isaac (Citation2021a and Citation2021b).
8 It is also crucial to distinguish a stable function from variable uses if we are to avoid the notorious Frege/Geach problem in proposing non-descriptive approaches to an area of discourse. That is, if we can identify a stable function, and identify meaning with the rules a term follows that enable it to fulfill those functions, then we may be able to identify a stable sense of meaning that remains even in embedded contexts where the term is not used to perform the characteristic speech acts with which it is associated. (For work on this approach, for the case of modal terms, see my Citation2020b, Chapters 2 and 3, drawing on work by Michael Williams (Citation2011)).
9 It is important to note that these functions needn’t be beneficial to all or benign; there is room here for critique. More on this below.
10 Later we will see additional reasons that it is helpful to begin by asking about the functions of language and linguistic formations rather than concepts. Nonetheless, it would also be extremely interesting to see such functional analyses done in psychological/conceptual rather than linguistic terms--I simply will not be undertaking that approach here. For some additional considerations in favor of starting at the linguistic level, see also my (2021).
11 On the lack of rivalry, see also (Halliday Citation2014, 56). And as Halliday also puts it "different coexisting models in linguistics may best be regarded as appropriate to different aims, rather than as competing contenders for the same goal" (Citation1964, 13). For further discussion of the place of systemic functional linguistics in the history of linguistics, see also Bateman (Citation2017).
12 See Aurora (Citation2015). Roman Jakobsen ‘explicitly considers Husserl’s phenomenology [especially the third Logical Investigation] as one of the main sources underlying Prague structuralism and Russian formalism’ (Aurora Citation2015, 14). Husserl also gave a lecture to the Prague Circle (invited by Jakobsen) on November 18, 1935.
13 This Husserlian work also influenced Ryle’s later work on linguistic categories and category mistakes.
14 For a brief overview, see Malmkjaer (Citation1991), 141-6.
15 Interested readers are instead referred to Eggins (Citation2004) and Halliday (Citation2009).
16 These are really characteristic ‘uses’ of language—unlike the macrofunctions that come later.
17 In saying this, I do not mean to endorse a merely local form of expressivism (endorsed, e.g. by Simon Blackburn (Citation2013)) as opposed to the global expressivism defended by Price (Citation2013). For the distinction between the introduction rules and functions governing words like 'dog' or 'pig' on the one hand, and those governing terms for properties and numbers (for example) on the other hand can also be captured by Price's distinction between terms that are (and are not) 'e-representational'. Those congruent terms that are introduced observationally to serve a heuristic function seem to be 'e-representational' in the sense that their "job … is to covary with something else--typically some external factor or environmental condition" (Price Citation2011, 20). For discussion of the global versus local expressivist debate, see the full range of essays in Price (Citation2013). For a defense of a form of global pragmatism against concerns that it can't properly address discourse about ordinary objects, see my (Citation2019).
18 The relevant notion of meaning used is of ‘meaning potential’—what a speaker can do, linguistically (Halliday Citation1973, 44), where this is a subset of general behavioral options (1973, 47). So the thought is that grammar enables us to (linguistically) do the same thing, in different ways—and often by combining the primary goal with other functions. ‘So a category like that of ‘threat’ … will be realized in the language system through a number of different grammatical options’ (1973, 49)—each of which may realize ‘more delicate options in the meaning potential’,(1973, 50). For example, a gangster might say "I'll kill your dog if you don't pay your debts", or (adding subtlety and plausible deniability), "You sure owe me a lot of money. It'd be a shame if something was to happen to your dog".
19 It also seems to coincide with the small-r, 'internal' sense of representation identified by Price (Citation2011), on which something 'counts as a representation, in this sense, in virtue of its … role, in some sort of cognitive or inferential architecture' (2011, 20). For it is plausibly the ability to carry propositional content that enables pieces of language to be used in inferences. Having an ideational/representational macrofunction in this sense must be distinguished from the idea that the relevant language is congruent language, observationally introduced (roughly, from the idea that it is 'e-representational' in something like Price's sense (2011, 20)).
20 For a further exemplification of such options, see Eggins (Citation2004, 118–9).
21 It is the ability of terms such as modal terms to enter language with a primarily non-descriptive (regulative) function, but go on to (serve ideational functions as well and) be used in expressing propositional content that can be reasoned with, that leads to the notorious Frege-Geach problem. I will have to leave discussion of this problem for another occasion, but for preliminary work in that direction, see my (Citation2020b, Chapters 2 and 3).
22 Unfortunately there is not space here to do the relevant reverse engineering work on moral or modal language. I have aimed to make a start at work on metaphysical modal language (as serving a regulative function of mandating, conveying, or renegotiating semantic rules) elsewhere (Citation2020b), though I had not yet found the material on systemic functional linguistics to work from. For an initial approach to moral discourse through this framework, see Warren and Thomasson (Citationforthcoming).
23 Though they may still be expressed in propositional form and come to serve an ideational macro-function, the crucial point here is that moral or modal claims (even expressed propositionally) originate from forms of speech introduced to serve inter-personal macro-functions, and thus have introduction rules that differ in crucial ways from those for congruent, observationally acquired language introduced to serve an ideational macrofunction.
24 Those familiar with my other work will see why: for I think that it is a mistake to think of nominalizations to speak, say, of properties as merely pretending, rather than as giving rules for what it takes to speak of properties, in the only sense that has sense. For details of the argument for a deflationist rather than fictionalist interpretation of such nominalizing discourse, see my (Citation2013). Macarthur and Price similarly endorse a pragmatist quietism about metaphysical issues--which means they are normally "happy to stand with the folk and affirm the first-order truths of the domain in question", contrasting this with an anti-realist fictionalist approach (Citation2007, 99). For a similar reaction to fictionalism, see Blackburn's (Citation2005) response to David Lewis.
25 Children typically process grammatical metaphor only after age 8 or 9 (Halliday Citation2009, 46).
26 This also seems like a way of making good on the idea that there are optional additional frameworks that can be added onto the ‘thing’ language, in Carnap’s (Citation1950/1956) terms.
27 Whether the grammatical metaphor is introduced from a congruent verb or adjective, or from another part of speech, such as a modal term, may make important differences in the ways we articulate their introduction rules, in what epistemological story is appropriate, etc. But such investigations will have to be left for another occasion.
28 Halliday also shows that parallel nominalization constructions appear in Chinese to similar effect (Citation2009, 135).
29 Non-content carrying words include prepositions, conjunctions, auxiliary verbs and pronouns.
30 ‘Grammar has always had this potential for ‘cross-coupling’. But it came well to the fore in the classical languages of the iron age, such as Chinese, Sanskrit and Greek, where it became the resource for creating abstract, technical objects … hundreds of verbs were nominalized as technical terms, and these nouns, together with their associated nominal group constructions, formed the core of a new, typically written, mode of discourse’ (Halliday Citation2009, 119). We might even speculate that it may be no coincidence that these coincide with the origins of much of philosophy as we know it, across traditions, as many philosophical questions cannot be formulated without a rich vocabulary of grammatical metaphors.
31 Martin (1990: ‘Literacy in science: learning to handle text as technology’, cited in Halliday Citation2009, 125) has shown that specialized technical discourse cannot be created without deploying grammatical metaphor.
32 One might even think that this is essential for moving from a situation of individualized power and commands (in the 'rule of man'), to the rule of law.
33 The above work enables us to justify the decision noted earlier, to think first in terms of linguistic functions rather than in terms of the functions of concepts. Asking first about the functions of individual concepts can lead us astray, since asking about the concept of F inevitably nominalizes the F. That is, we can ask about the concept of redness, but this elides the functional differences we might discover between the function of having congruent adjectives such as ‘red’ versus nominalized forms (redness) which we must use in naming concepts. (This of course does not prevent us from drawing similar conclusions about the functions of phonetically distinct terms that would be good translations or synonyms).
34 To use Gilbert Ryle's (Citation1932) phrase.
35 This should bring Ryle (Citation1949) back to mind—one of the earliest philosophers I know of who was onto the idea that there are important functional differences in different modes of speech, which show up in the paradoxes and puzzles that arise with category mistakes. P. F. Strawson seems to be onto a similar point, where he talks about how the subject/predicate form is applied in the basic case to spatio-temporal particulars, but by ‘imaginative extension’ this logical form is also carried over to ‘higher levels’, enabling us to make predications of colors, numbers, etc. The basic case is a model for the other cases, but ‘From this fact spring both the delusions of Platonism and the delusions of anti-Platonism. They are indeed, but two sides of the same delusion’ (Citation1974/2016, 30).
36 A full disentangling from various philosophical problems would require extensive separate work, including understanding the different introduction rules that govern congruent observationally acquired nouns, versus nouns that enter as grammatical metaphors from various other forms of speech. There is not space here to do that detailed work, but only to gesture at how various inappropriate expectations may arise and lead to puzzles.
37 For discussion and defense of such 'easy' ontological inferences, see my (Citation2015), especially Chapter 3, which draws on and builds from work by Hale and Wright (Citation2001) and Schiffer (Citation2003).
38 See, for example, Hale and Wright (Citation2001).
39 We can also give due credit to the intuitions often raised against these functionally alternative proposals: that moral, modal, or mathematical discourse has propositional content that enables us to state truths 'about' numbers, moral qualities or possible worlds. For we can notice the ideational functions that are simultaneously served.
40 For a start on this, see Warren and Thomasson (Citationforthcoming).
41 For insightful comments on earlier versions of this work, I wish to thank Manuel Gustavo Isaac, Katherine Ritchie, Herman Cappelen, and David Woodruff Smith, and the editors and an anonymous referee from this journal, as well as audiences at (virtual) talks at Arche, Hong Kong University's Concept Lab, Temple University, and the University of California, Irvine.