‘Philosophy must plough over the whole of language’, as Wittgenstein famously stated. But which language? Singularising the noun allows a deceptive slippage between some language whose premises we take for granted (‘The limits of my language are the limits of my world’ was another great, and corrective, line of his) and ‘language’ in some dangerously, presumptively general sense. One of the great what-if questions for linguistics, philosophy and cognitive science is how different the last two millennia of western thought would be if we had built our disciplines on the foundations of languages radically different in what their grammars prioritise from Greek, Latin, Hebrew – or, more recently, German, French or English.

If we trace the development of language studies in the west, we find an early bisection between logic – which came to be associated with the study of meaning and reasoning, which should be shorn of all context – and rhetoric, where the relations between speaker and hearer are all-important but which does not primarily focus on grammar. The echoes of this continue today as, for example, formal semanticists continue to wrestle with how to define definiteness (the noisy neighbour) in ways that are based on set theory and mathematics rather than embodying representations of the speech setting or of ‘theory of mind’. But the recent ‘subjective’ turn among cognitive linguists, and the even more recent ‘intersubjective turn’, continue to throw up evidence about how deeply woven into language we find the search for common ground and its constant renegotiation, rooted in the speaker’s attention to their addressee’s attentional, emotional and belief states.

I’ve long been interested in ‘multiple perspective’ (Evans 2006) – originally sparked by special ‘triangular’ kin terms in Australian languages with meanings like ‘the one who is your mother and my daughter, me being your grandmother’. The conversation that was the genesis of this article took place with Jon Landaburu in 2004, on a bus trip from Mexico to Teotihuacan. In discussing the general issue of multiple perspective, I asked him whether he thought that similar biperspectival constructions ever get grammaticalised with modality. He then told me about his work on ‘engagement’ in the Colombian language Andoke, which later came out in the seminal article we refer to in ours. I was particularly struck by his suggestion that ‘it is the traditional influence on grammar of logic, whether scholastic or formal, which has led led to the separation of these two dimensions (i.e. speaker knowledge and intersubjective assessment) which are necessarily present and interwoven in any act of communication’ (Landaburu 2007:30-1, my translation). The crucial part of his article is that it allows for a bidimensional distribution of attention/knowledge or its absence: both speaker and hearer (assertion based on common ground/attention), neither (question for which no informed answer is expected), speaker-only (authoritative statement by speaker, who assumes the hearer doesn’t know or attend), addressee-only (inquiry of presumably better-informed addressee).

Although I loved the article, I was frustrated by the paucity of examples, and the lack of an interrogable corpus. I also thought that the full significance of his discovery wouldn’t be clear until comparable phenomena were found in other languages, and a proper analytic framework developed based on cross-linguistic comparison. In the succeeding years I gradually accumulated case studies that could form the basis of a typology. Perhaps more importantly, I was joined in this research by my two co-authors in this article, Henrik Bergqvist and Lila San Roque, each of whom, unlike me, has deep, firsthand experience working on languages in which engagement is a central part of the grammar – respectively Kogi in Colombia (close to, but unrelated to, Andoke) and Duna in Papua New Guinea. Their detailed fieldwork-based insights allowed a deeper probing of how this phenomenon really works in those languages. Even so, however, our understanding of how the grammar of engagement works is at an early stage and a major priority in the next stage of research is to gain more detailed field data, through a combination of recording naturalistic conversation and developing protocols for what Nikolaus Himmelmann calls ‘staged communicative events’ which allow us to vary, with more confidence, the factors at work.

In talks I’ve given on this phenomenon, I’m often asked: ‘isn’t this just the same as saying “so” or ‘hey!” or “actually” in English’. Well, there are clear overlaps in communicative function, and these words are the closest we can get to translating sentences from languages like Andoke in a reasonably natural way. But they are not systematically integrated into the grammar in the same way, forming tightly bound paradigms that require constant scanning of the addressee’s attention and knowledge states by the speaker.
Linguists have long believed that it makes a difference whether something is part of the grammar or simply an optional add-on – just as it makes a difference whether we need to obligatorily specify tense and definiteness, as in English, or do so in more roundabout ways, as in Chinese. Viewed another way, from real-time processing, psycholinguist Paula Rubio-Fernandez asks: is attributing mental states to others .. too costly to be used as a basis for online interaction? As an inflectional category, engagement definitely needs to be processed online, and it definitely involves attributing mental states to others – so our emerging typology is directly relevant to this central question for the psychology of social cognition.

Had the intellectual tradition of linguistics begun with Foe, Duna, or Kogi rather than Greek and Latin our notions of what is basic might have looked very different – and intersubjective categories would have demanded treatment from the very beginning. Ironically, what I have presented as a third, ‘intersubjective’ stage in linguistic approaches to the grammar of meaning – collaborative communication between speakers attending to each other, altering their common ground through time, and representing what each other knows – is arguably the precondition for the evolution of language in the first place.

One language’s grammar is another language’s discourse. The fact that we have had to travel to far-flung lands to find the ‘pure’ examples of engagement discussed in our article does not imply that their functional equivalents won’t turn up in interesting ways in all languages. At the same time, the interesting question arises of what difference does it make for a category like engagement to be elevated to a central grammatical position. For this we need sensitive cross-linguistic studies, preferably involving parallel or semi-parallel corpora, as well as psycholinguistic studies comparing the learning and processing of intersubjectively relevant categories, and taking as the independent variable the gquestion of whether the language has clearly grammaticalised engagement categories.

This piece is inspired by the Language and Cognition article “The grammar of engagement I: framework and initial exemplification” by Nicholas Evans, Henrik Bergqvist, and Lila San Roque. Read it online now.


Evans, N. (2006). View with a view: Towards a typology of multiple perspective. Berkeley Linguistics Society (BLS) 32, 93–120.
Landaburu, J. (2007). La modalisation du savoir en langue andoke (Amazonie colombienne). In Z. Guentchéva & J. Landaburu (eds.), L’énonciation médiatisée II: Le traitement épistémologique de l’information: Illustrations amérindiennes et caucasiennes (pp. 23–47). Leuven: Peeters.

Leave a reply

Your email address will not be published. Required fields are marked *