An understanding of semantics with reference to subjectification, intersubjectification and grammaticalisation.
2.2.1. The how of (inter)subjectification
The concept of semanticisation is crucial to an understanding of the mechanisms at work in (inter)subjectification, because it discerns between two types of (inter)subjective meaning and delineates the research domain. Subjectivity is embodied in language: the very use of language by an individual makes it a subjective endeavour and because communication always involves one or more participants, language is also inherently intersubjective. This is called ‘pragmatic subjectivity’: it is not coded in any linguistic form, but is the very nature of the context in which language is used. However, linguistic forms may come to code speaker- (and/or addressee-) relatedness as a form-meaning pairing. For instance, the epistemic use of must expresses a factuality judgement by the speaker, not by anyone else. This type of subjectivity is semantic, and arises from the pragmatic kind (Traugott 2010: 32, Traugott & Dasher 2002: 35). For Traugott, only semantic subjectivity is involved in subjectification: My main concern is not with this [ambient pragmatic] context, but with linguistic MARKERS and EXPRESSIONS that index subjectivity and intersubjectivity and how they arise’ (Traugott 2010: 32, emphasis original).
De Smet and Verstraete (2006: 384-387) made an additional distinction in the domain of semantic subjectivity. Besides pragmatic subjectivity, which is inherent in language use and independent of semantics, there are two kinds of semantic subjectivity: ideational and interpersonal. The first kind is proposition-internal and works on the level of ‘representation of the extra-linguistic and extra-discursive world’ (De Smet & Verstraete 2006: 385). Calling someone an ass entails a subjective evaluation and is used to describe someone in the situation under discussion. The second kind works above the proposition, and codes a speaker position with respect to the description of the extra-linguistic world. An example is again the epistemic must, which denotes a subjective evaluation of the entire proposition, i.e. of the state of affairs in the world that is being discussed.
The mechanisms that code (pragmatic) subjectivity into linguistic forms are the same as those at work in the coding of grammatical meaning, i.e. pragmatic strengthening of an utterance’s informational content based on the metonymic process of invited inferencing. Traugott and Dasher (2002: 16–17) distinguish three types of meaning that are relevant for lexemes: coded meanings, utterance-type meanings and utterance-token meanings. The coded meanings are the basic, conventional readings of linguistic elements; i.e. their denotation. The utterance-token meanings are additional, pragmatically-enriched meanings which the interlocutor is invited to infer, i.e. invited inferences. Crucially, these meanings are not yet coded, but appear ‘on the fly’ in a specific situation; they present possible ‘extensions’ a coded meaning may give rise to. The utterance-type meanings are a further step; they represent generalised invited inferences, which are pragmatic readings that conventionally (that is, frequently) accompany a form, but may still be overridden. They are not yet coded meanings, but crystallized invited inferences associated with certain lexemes or constructions [that] can be exploited to imply/insinuate certain meanings’.
Pragmatic strengthening of the coded meanings leads to utterance-token meanings that, if adopted frequently by the linguistic community, become utterance-type meanings. As long as these meanings can be overridden the form is only pragmatically polysemous, but once ‘the original meaning becomes merely a trace in certain contexts, or disappears, […] the GIIN [Generalized Invited Inference] can be considered to have become semanticised’ (Traugott & Dasher 34–35). The form is then semantically polysemous. This process of gradual pragmatic strengthening is metonymical, because there is contiguity between what is expressed in the discourse and something outside of the discourse, in casu a subjective stance of the speaker as a participant in the speech event. (Inter)Subjectification, then, is a type of metonymy (Traugott & Dasher 2002: 54, 81); an instance of synecdoche (cf. Traugott & König 1991: 210–211).
The working of metonymy in subjectification naturally takes place within the normal context of pragmatic (inter)subjectivity, in the speaker–interlocutor ‘dyad’. We have seen that the speech community regards the completion of the semanticisation as imperative and in the case of intersubjectification the dyad becomes even more important, as the coding of explicit attention to the interlocutor requires a firm grasp of the socially determined set of conventions that hold in communication (Narrog 2010: 420–421). As we will see in the next section, the dyad becomes still more important in that it is on this level that we find an answer to the question why (inter)subjectification takes place.
2.2.2. The why of (inter)subjectification
Traugott and Dasher (2002: 6-7) call the speaker–addressee dyad the link between grammar and use; it is a negotiation of meaning where the speaker makes inferences that the addressee has to pick up on and interpret. Moreover, the speaker translates into his expressions a certain estimation of the knowledge, linguistic competence and expectations of the addressee. In this way, both participants mould language and impose their perspective on it, but the catch is that this dyad is not symmetric: the speaker exerts the greatest influence during his/her turn in choosing what is said, how it is said and how it should be interpreted. Moreover, we know that even attention to the addressee is essentially a function of the speaker’s perspective. It is therefore no surprise that linguistic expressions will eventually come to code some of this subjectivity, given enough repetition of certain inferences in the same or similar contexts.
The inherent subjectivity of the speech event is taken to be the ultimate motivation of subjectification. This is a poignant illustration of Lehmann’s (1985: 10) observation that ‘languages change because speakers want to change them’ — not consciously, but simply by using language to say what they have to say. As with grammaticalisation, a speaker’s need to be creative without losing comprehension is what promotes the process (Finegan 1995: 9). Furthermore, again like grammaticalisation, the motivation is purely pragmatic (i.e., motivated extra-linguistically) thus proving that (inter)subjectification may be one of the most pervasive types of language change, but nonetheless, it is just one type.
Because Haspelmath’s (1999) explanation in Section 1.4.2. is essentially one of language change in general, it should come as no surprise that Traugott and Dasher (2002: 19) offer a strikingly similar account for (inter)subjectification. They postulate three heuristics (reminiscent of Horn 1990: 463) that govern subjectification and language change. The first of these, the Q(uantity)-heuristic, postulates that you should make your contribution as informative as required, while implying more thereby. The second, the R(elevance)-heuristic (called Relation by Horn (ibid)), implies that you should not write or say more than you must, but mean more thereby. Finally, the M(anner)-heuristic tells you to avoid prolixity, i.e. unnecessary wordiness. Breaking it results in a marked expression. The Q-heuristic is the most likely to change; it represents standardisation, which is necessarily conservative, and tends to avoid ambiguity. The R-heuristic represents the phenomenon of pragmatic strengthening, and make allowance for enrichment of meaning. Finally, the M-heuristic draws attention to a new use of an old form, and marks it as something special; something that will attract the addressee, thus enabling R.
This pragmatic characterisation of (inter)subjectification has one major downside: it makes it hard to search for structural correlates of the process. For instance, Traugott (2010: 56) speaks of a need to move from the admittedly rather vague notion of ambient subjectivity and interlocutor interaction as a motivating force in subjectification, toward identifying the types of linguistic context in which one might expect to find evidence for subjectification’. Traugott is only concerned with those meanings that are coded and that involve participants on the level of the speech event. This requires that the participants should not be coded in the proposition itself and that even an increase in, for instance, first person subjects cannot be taken to signal subjectification.
The main problem is that though subjectification can be shown to take place in a diachronic evolution, there is no objective set of parameters or principles to distinguish synchronic degrees of it in the way Lehmann and Hopper did for grammaticalisation. This is related to the fact that subjectification, unlike grammaticalisation, is a single process with no component processes to pick up on. De Smet and Verstraete (2006: 370–371) explicitly thematise this problem, and find that it is possible to distinguish ideational from interpersonal subjectivity by means of syntactic criteria. Nevertheless, it remains to be seen how their findings are to be expanded on a broader scale. The same observation holds for the research by Traugott (2010: 56–58) and Vanderbiesen’s study (in preparation). These are construction- and language-specific.
- The directionality of (inter)subjectification
Traugott (1989: 34–35) describes three tendencies evident in semantic change (both grammatical and lexical):
- Meanings based in the externally described situation > meanings based in the internally (evaluative/perceptual/cognitive) described situation.
- Meanings based in the externally or internally described situation > meanings based in the textual and metalinguistic [metatextual] situation.
- Meanings tend to become increasingly based in the speaker’s subjective belief state/attitude towards the proposition.
The first tendency subsumes various changes from concrete to abstract notions (such as space to time); that is, from what can be observed and described of the outside world to the perception of this world by a sentient being. The second tendency encompasses changes of elements towards a text-structuring function (e.g. as temporal or causal connectives) or a metatextual function. A prime example of this is a discourse marker like anyway used as a means for the speaker to return to the topic (Traugott & Dasher 2002: 95). This tendency thus serves both for content (propositional) and procedural (pragmatic) functions. During this stage, forms evolve from describing the world being talked about to indexing the speaker’s organization of that world in the act of speaking (Traugott 1980: 47). The third tendency amounts to subjectification. At this stage, the speaker starts to use language to index his/her cognitive processes or sensitivity to social relations (ibid.: 96).
Crucially, Traugott (1989: 31) claims that all three tendencies follow a unidirectional sequence: the first tendency feeds into the second, and both may feed into the third. This unidirectionality becomes clear when the tendencies are disassembled to a number of correlating paths of directionality and mapped onto a cline:
Because speakers in communication naturally impose their perspective on the discourse, subjectification usually tends to be unidirectional. It is not normal for speakers to mask their input into the discourse (except in marked situations where this is required, but even then it seems likely that some subjective colouring will still occur), so the development will (almost) always be towards more subjectivity. Traugott (2003: 124, 134-136) argues further that the fact that intersubjectification cannot occur without subjectification (see infra) is further proof for the unidirectionality of change.
There is a natural affinity between this drive towards subjectivity and the other paths given in Fig. 1. Speakers will attempt to give their perspective on an increasingly larger scale, which is driven forward by their need for self-expression, primarily through pragmatically enriching the expressions used. Thus, elements that function in the proposition are recruited by them to encode their opinion about the proposition, and eventually they will attempt to pull the addressee(s) into their subjective perspective; for instance, by steering the discourse in a certain way. For them to do this, elements need to develop away from their pure content function and become procedural elements; i.e., forms that can take scope over and structure the discourse. Finally, as a result of the fact that the forms have been recruited to express a purely subjective meaning, they are no longer necessarily true or untrue nor do they conform the actual state of affairs. As with other types of change (for instance grammaticalisation), this cline is layered, meaning that it does not have to be run through all the way, and that older, less subjective forms can coexist with newer ones.
2.4. (Inter)Subjectification and grammaticalisation
In the course of this section, a number of times we have referred to parallels that exist between (inter)subjectification and grammaticalisation. In the most general sense, this is not surprising, because both are types of language change, and the motivations for one type of language change will by and large, be similar to those of another type. However, between the two processes under discussion there is a closer relationship.
As Traugott (2010: 38-42) points out: ‘Neither subjectification nor intersubjectification entails grammaticalization. Nevertheless, there is a strong correlation between grammaticalization and subjectification, and a weaker one between grammaticalization and intersubjectification’, but this not a one-to-one correlation (Narrog 2010: 386). Traugott (ibid.) continues to say that ‘subjectification is more likely to occur in grammaticalization than in lexicalization or in semantic change in general’, because many grammatical elements encode the speakers’ perspectives (e.g. tense relates to their present, mood to their beliefs, pronouns to their position or to the way they structure discourse, topic and focus to how they present their information,…). Unsurprisingly, Traugott (1995: 46) argues that subjectification is characteristic of all domains of grammaticalisation. This might be related to the above observation that speakers impose their perspectives on ever larger parts of the discourse, culminating in scope over the discourse itself. For an element to take this scope it must coincide with a bleaching of its original concrete, objective meaning, which in turn, may be accompanied by erosion … Still, subjectification in grammatical change is no different from that in lexical change, except that ‘in grammaticalisation the interplay between morphosyntactic and pragmatic/semantic factors leads to more complex trajectories of change than are usual in lexical change’ (Traugott 1995: 32). This might be related to the epiphenomenal nature of grammaticalisation, which means many (formal and functional) processes intersect with subjectification. Lexical change is a more straightforward alteration of lexical semantics.
Nevertheless, subjectification is more typical of primary (lexical to grammatical) than of secondary (less to more grammatical) grammaticalisation (Traugott 2010: 40-41, Traugott & König 1991: 189, Traugott 1995: 47), because primary grammaticalisation often relies on pragmatic strengthening of elements in specific contexts (cf. 1.4.1.). This also what underlies subjectification. Secondary grammaticalisation, however, usually entails the automatisation of structures and a reduction of the possibilities a speaker has to express a certain function with a certain form. In other words, the speaker no longer has conscious access to the form, which means it is harder for him/her to impress his/her perspective on it. The development of grammatical markers that explicitly code intersubjectivity is said to be cross-linguistically rare (Traugott 2010: 41).
Our discussion has revealed that grammaticalisation and subjectification share their motivation (expressivity/extravagance, guided by maxims), operation (metaphor and primarily metonymy through pragmatic strengthening) and unidirectionality (stemming from the need for expressivity). In discussing Haspelmath’s (1999) study on the motivations of grammaticalisation we noted that it is imperative for the linguistic community to understand why the process occurs. Therefore, we can safely say that grammaticalisation too is ultimately motivated in the speaker-addressee dyad. Moreover, the three tendencies in semantic change identified by Traugott (1989) obviously also pertain to grammaticalisation, and are explicitly adopted and integrated into a grammaticalisation framework by Diewald and Smirnova (2010: 158).
In conclusion, we may say that (inter)subjectification is a possible component process of grammaticalisation that is, as are other sub-processes, independent of it. Nevertheless, within the context of grammaticalisation, subjectification works in a special way, cooperating with syntactic, morphological and phonological factors (Traugott 1995: 45). Unlike other component processes, however, (inter)subjectification is in and of itself categorically unidirectional, which may result in cases where a reversion of grammaticalisation is accompanied by an increase in subjectivity (ibid.). Still, subjectification itself is not epiphenomenal – it is one semantic process that may accompany the semantic change in grammaticalisation, but need not (Cuyckens, Davidse & Vandelanotte 2010: 6). That they share motivations and mechanisms is due to the fact that neither grammaticalisation nor (inter)subjectification are distinct from other types of language change. However, what makes them special is that they are pervasive types of language change that give us insight into the human interaction with language and the language context. They function as substantial backing for an elimination of borders between pragmatics and semantics: langue and parole.
 An important caveat is that an exclusive focus on only coded meanings may eventually run into problems, or at least make one overlook important aspects. Boye and Harder (2009) have shown this in exemplary fashion for evidentiality (see Section 3), and Cuyckens, Davidse and Vandelanotte (2010: 6) also remark that the study of pragmatic (inter)subjectivity is not a priori uninteresting.
 The relevance of adoption by the community is not to be underestimated. In his explanation of language change, Haspelmath (1999) explicitly points to the restrictive nature of conformity, and Milroy (1992: 79, quoted in Traugott & Dasher 2002: 38, italics original) states that ‘a change in the output of a single speaker might be regarded as the locus of change in the system, whereas of course a change is not a change until it has been adopted by more than one speaker’.
 Cf. Lehmann (1985: 10), who notes that speakers apparently do not have a general predilection for litotes or understatement.
 It is important to note, however, that the unidirectionality of (inter)subjectification is independent of the linguistic community. There may be social norms governing the coding of (inter)subjectivity (e.g. honorifics in Japanese), but in principle the only thing an element needs to subjectify is the appropriate semantics (cf. Traugott 2003: 135). Conversely, subjectification is constrained by the semantics of the element and the inferences that can be drawn from it (Traugott 1995: 46).
 It will be noted that the very fact of subjectification, i.e. an increase in subjective meaning or speaker-relatedness, speaks against a view of grammaticalisation as pure loss.