Pre-emptive interaction in language change and ontogeny: the case of [there is no NP]

1 Introduction

This paper aims at establishing a new pragmatic and semantic category of pre-emptive interaction. Pre-emptive interaction intersects with intersubjectivity, as it marks the speaker’s attention to ‘[the addresse]’s image needs’ (i.a. Schwenter and Waltereit 2010; Tantucci 2017a, 2020; Tantucci and Wang 2018, 2020a, 2020b; Traugott 2003; Traugott and Dasher 2002: 177) and underpins overt attempts to tackle how an interlocutor is expected to react as a result of an on-going speech event. Pre-emptive interaction occurs as a linguistically overt strategy to address a projected turn taking of a specific or generic interlocutor as a result of what is being currently said.

We ground our discussion in enactment cognition (cf. Engel et al. 2013), and we suggest that human ability to pre-emptively address an interlocutor’s potential reaction to what is being said is a form of prediction error minimisation (cf. Clark 2013; Friston 2010; Friston et al. 2012). We propose that, similar to the way we constantly rely on senso-motorial ability to predict possible problems arising from physical actions, we also monitor and predict possible issues that might arise through dialogues and conversations. In this sense, pre-emptive interaction constitutes an important human ability to predict and foresee the interlocutor’s reaction to what we are currently saying. We argue that semantic and pragmatics shifts from ongoing to pre-emptive interaction are important triggers of intersubjectification. We substantiate this claim by providing a case-study about the constructional change of [there is no NP], shifting from an original existential meaning and assertive illocutionary force (i.a Kissine 2013; Tantucci 2016a) towards a new intersubjectified usage. This new reanalysis leads to new pre-emptive refusal usages, whereby the speaker/writer rejects what the interlocutor or a generic social persona may potentially say about a current utterance. The new illocutionary force of the construct significantly intersects with new formal features such as presence of a complementing clause, discourse-new status of the NP and the generic reference to the ‘locus of existence’ of the NP. In the final part of the paper, we look at ontogenetic data from the CHILDES database, and verify whether similar pre-emptive polysemies of [there is no NP] arise before or after the 4th year of age of the child, viz. what is considered to be the critical period in which a Theory of Mind mechanism starts to fully develop ontogenetically (cf. Kovacs et al. 2010; Onishi and Baillargeon 2005; Surian et al. 2007; Tantucci 2020).

This paper is structured as follows: Section 2 provides an introductory discussion about enacted theory in cognitive psychology. A special focus is placed on the notion of prediction-error-minimisation (PEM) in connection with a conceptualiser’s epistemic reasoning. Section 3 tackles PEM in connection with intersubjectivity in cognitive linguistics and pragmatics. We argue that a fundamental element of intersubjectified constructions is the one of pre-emptively addressing how the speaker presumes the hearer may react as a result of his/her on-going speech. Section 4 is centred on the [there is no NP] construction and provides a corpus-based study of the constructional change of the chunk in the Corpus of English Dialogues (CED). From a multiple correspondence analysis (i.a. Nenadic and Greenacre 2007) and a mixed effects logistic regression of pragmatic and grammatical change of the chunk, will emerge that [there is no NP] started to acquire a new pre-emptive refusal usage out of an original assertive one throughout the 17th and the 18th centuries. Section 5 is finally focused on the spontaneous employment of [there is no NP] in first language acquisition. A large-scale corpus-based analysis from the CHILDES shows that pre-emptive refusal usages of [there is no NP] are acquired and mastered significantly later than assertive usages of the same construct. In particular, the child shows capacity to spontaneously master the construction only around the 4th year of age, viz. when a theory of mind ability tends to be extensively developed.

2 Prediction–error–minimisation in interaction

Over the last two decades, a new ‘pragmatic turn’ has emerged in cognitive science. This novel paradigm has been progressively drifting away from the traditional representation-centred model (Schilbach et al. 2013) of cognitive processes and started tackling cognition as being inherently ‘enactive’, that is as subserving action and sensorimotor skills (Clark 1998; Engel et al. 2014; Noe 2004; Varela et al. 1992). From this angle, neural activity patterns are studied with respect to their functional role in action generation. Cognitive states are then addressed as foregrounding and prescribing possible actions rather than statically representing states of the outside world.

The intentional and goal-oriented dimension of enacted cognition is relevant for research in cognitive linguistics and pragmatics, as skills of sensory motor contingencies allow conceptualisers to develop higher cognitive abilities and social interaction (Di Paolo and De Jaeger 2012). In fact, learning and mastery of action–effect contingencies are at play through interaction between the subject and his/her surroundings. As enacted cognition inherently hinges on the process of reaching and maintaining an expected state, action and interaction both occur as a result of predicting how being in a desired condition would change the received sensory input (Hommel 2013). This clearly underpins the necessity to anticipate possible actions and linguistic acts from others and thus to enable effective coupling of agents in social contexts. This pre-emptive dimension of enacted interaction is based on perceptual and active inference geared to prediction–error–minimisation (PEM) (Clark 2013; Friston 2010; Friston et al. 2012). Subjective goals are pursued through sensory-motor skills and problematisation of contingent obstacles and possible errors that may occur through actions and interactions. An example of this is when one conceptualiser actively explores a pipe in his/her hand by turning it around and looking at it from different angles, then increasing the confidence that s/he is looking at a pipe and not merely at an image of a pipe (Hohwy 2016) or another object that simply resembles a pipe. From an interactional perspective, prediction–error–minimisation combines with Clark’s (1996) idea of joint projects and event-reaction pairs, whereby an interlocutor’s instigating event (generally at the illocutionary level) is inherently expected to lead to an addressee’s reaction (Clark 1996: 194). When interactional mismatches arise between instigating event and ‘expected reaction’, interlocutors encounter what Clark defines as ‘joint project problems’, which may lead to various forms of communication breakdown, hinging on (im-)politeness, face-threats or misunderstanding.

Most crucially, interlocutors often pre-emptively make attempts to tackle potential joint project problems, and try to pre-emptively minimise ‘errors’ that may arise through interaction. This is connected with the mechanism of vigilance (i.a. Mercier 2020; Mercier and Sperber 2017; Reboul 2017; Sperber et al. 2010) suggesting that evolution of human reasoning has happened in a context in which social and interactional pressures were of fundamental importance, in particular involving the need to win an argument and, at the same time, to carefully evaluate others’ arguments and objections. The anticipatory dimension of pre-emptive interaction is thus a crucial mechanism allowing speakers to monitor and foresee co-actional and interactional issues that may arise as a result of their turn takings. In this study, we argue that prediction–error–minimisation (PEM) overtly comes into play in language, as a pragmatic attempt to pre-emptively addresses forthcoming turn-takings that are likely to occur in conversation. Consider the following example of a so-called joint project as ‘a joint action projected by one of his participants and taken out by the others’ (Clarks 1996: 191):

The joint project above begins with Arthur projecting a joint task for Beth and himself to carry out. She is expected to tell him what modern poets she has been reading. It continues with Beth agreeing to the project as she answers Arthur’s question. Even more importantly, what is of interest here is the fact that Beth not only replies to Arthur’s question at the propositional level, but additionally addresses what she thinks is likely to be a further turn-taking on Arthur’s behalf, namely do you like them? / Which ones you like? Pragmatically, Beth infers the way the conversation is likely to continue and pre-emptively engages with a foreseeable topic.

3 Intersubjectivity and pre-emptive interaction

Pre-emptive interaction often intersects with intersubjectivity. Intersubjectivity in the literature studies is addressed from a number of different perspectives, e.g. as a grounding mechanism hinging on spatial/epistemic perspective-taking (i.a. Dancygier and Sweetser 2012; Langacker 1991; Verhagen 2005), as a shared dimension underpinning evidential reasoning as opposed to subjective opinions (i.a. Nuyts 2001, 2012; Tantucci 2013, 2016b), or as function to establish joint attention among interlocutors (e.g. Breban 2010; Diessel 2006; Ghesquière et al. 2012). Intersubjectivity has been treated and re-interpreted in a multitude of different ways (i.a. Guesquière and Van de Velde 2011; Langacker 1991; Nuyts 2001, 2012; Traugott 2003; Tantucci 2020). Spatial deictics have been discussed as markers of joint attention and intersubjectivity (Breban 2010; Diessel 2006; Guesquière et al. 2012; Langacker 1987, 1990, 1991). Demonstratives and determining elements like such and zulk have been similarly considered intersubjective, as they serve to create a ‘joint focus of attention’ (Diessel 2006: 465) by which the speaker negotiates discourse referent tracking for the hearer (Guesquière 2009; Guesquière and Van de Velde 2011).

A rather different approach to the notion of intersubjectivity is discussed by Traugott (cf. 2003, 2012). In her diachronic model of intersubjectification, speakers markedly code attention to the social self of the hearer (e.g. Traugott and Dasher 2002). She observes that less-intersubjective (e.g. literal) constructions show a clear tendency to progressively develop new polysemies with novel intersubjectified functions. For instance, pragmatic marking (PM) of adverbials such as actually or in fact are intersubjective as they function to hedge or mitigate what has been said and acknowledge the addressee’s actual or possible objections.

Verhagen (2005, 2007 also developed an influential account of intersubjectivity, that is centred on cognitive coordination between speaker and hearer. In his view, intersubjective constructions foreground the hearer as active interpreter and conceptualiser of utterances of the speaker. He remarks that the hearers’ reaction to what is said is often part of the semantics of intersubjectified expressions. A case in point are connectives such as but or moreover, which are often used to accommodate the addressee/reader’s projected expectations through an interaction.

There is also a well-known epistemic strand in the literature about intersubjectivity. Nuyts (2001a, 2001b, 2012 argues that intersubjective constructions hinge on modal meanings ‘presented as being shared between the assessor and a wider group of people, possibly (but not necessarily) including the hearer’ (Nuyts 2012: 58) as for constructions such as it is likely, unfortunately and so on. In a similar manner, Tantucci (2013, 2017a, 2017b, 2020 distinguishes meanings that are specifically centred on the hearer’s potential reactions to what is said, from meanings that include a specific or generic social persona, who is assumed as the social bearer of the utterance. Intersubjectivity is also an important topic in interactional linguistics. In this case, it is often brought to the fore as ‘shared understanding’ (Linell and Lindstrom 2016) which is realised in the form of turn-takings of repair, reaction or expansion to achieve a socially shared cognition (Schlegoff 1991, 1992). Equally influential are qualitative discussions about the intersection of counterfactual reasoning and intersubjectivity as in the case of Mental Space Grammars (i.a. Dancygier and Sweetser 2012).

In this paper, we take the stance that intersubjectively marked expressions are inherently geared towards projected reactions of an interlocutor to a current utterance. Our argument is that prediction–error–minimisation (PEM) is a crucial element of intersubjectively marked interaction. Strategies hinging on (im-)politeness and meta-discursive functions such as turn-giving, agreement-seeking or elicitation of response (Traugott 2012: 10) are all cases in point whereby an interlocutor foresees and pre-emptively tackles possible issues arising from an ongoing exchange. Sentence-peripheral usages of constructions such as clearly, actually, believe it or not, no doubt and so on, all encode speaker’s attitude towards the addressee, rather than directly modifying the verb (Tantucci 2017a; Traugott and Dasher 2002). An instance of this can be the imperative construction Look! when it is used as an attention-giving device (e.g. Trillo 1997):

In (3), look is redundant at the propositional level. However, it expresses the intention to overtly account for the potentially negative impact of the utterance towards addressee’s ‘positive face’ (cf. Brown and Levinson 1987). It is reasonable to expect a child younger than 4 being able to utter the former expression (2), viz. in the form of a directive. On the other hand, it is not farfetched to suggest that the latter usage in (3) will be ‘cognitively’ mastered at a comparatively later stage of ontogenetic development. In fact, with the overt employment of Look as pragmatic marker (PM) in (3), the speaker predicts that the addressee/reader may somehow react negatively to his/her utterance and tries to minimise potential joint project problems that may arise from the current utterance. In this sense, intersubjective forms often occur as PEM operators and are overtly codified as a ‘surplus of meaning’ over mere propositional or co-actional meaning (Tantucci 2020).

From a usage-based perspective, intersubjectivity can be operationally identified as surplus of meaning that interlocutors codify in addition to ego-centric goals and intentions that could be expressed merely in the form of a co-action (Tantucci 2020; Tantucci in press). The surplus-approach (cf. Tantucci et al. 2018) is centred on the premise that new intersubjectifed polysemies of a construction arise and become conventional as a response to interactional needs. Speakers re-cycle an existing form to express a new meaning centred on the addressee (or a third party) and thus improve the chances to achieve some sort of perlocutionary effects. This may unveil a gradient continuum from ego-centric co-actional engagement to more complex awareness of social cognition. The latter is reflected by the progressive ability of the child to spontaneously employ increasingly intersubjectified polysemies or functions of the same construct (cf. Tantucci 2020). In some cases, the intersubjective ‘surplus’ of peripheral PMs may be characterised by a ‘division of labour’ between right and left peripheries. Downing (2001) observes that left peripheral employment of surely in evaluations intersects with seeking agreement or corroboration:

The clause-peripheral employment of surely acts as an overt ‘surplus’ of meaning pre-emptively codifying the speaker/writer’s awareness of addressee/reader’s potential reactions to his/her evaluation. In fact, when surely is used as a PM, it should not be logically compatible with rhetorical interrogative speech acts geared towards agreement seeking. Even in this case, we argue that this is due to the pre-emptive dimension of the intersubjective function of surely, through which the speaker/writer overtly foresees and engages with his/her interlocutor’s agreement with what is said.

The pre-emptive dimension is similarly present in [and don’t you VP] imperatives. This is a chunked construction, whereby the connective and does not express some kind of logic coordination among two sentences, but is rather used to introduce a directive speech act of ‘pre-emptive prohibition’. In fact, with this construction, the speaker is not referring back to some discourse-given topic that was introduced previously. Rather, s/he pro-actively hypothesises a possible reaction of the speaker as a result of what is being currently said and pre-emptively prohibits him/her to perform it. Consider examples (5–6) from the spoken section of the BNC:

Examples (5) and (6) are both demographically sampled. In the first case, the speaker (Cassie) informs her interlocutor (Bonnie) that she does not trust a common friend (Catherine) anymore. During her own turn taking, Cassie makes use of the [and don’t you VP] construction to pre-emptively forbid her to let Catherine know about what is she currently saying. The [and don’t you VP] is uttered by Cassie to pre-emptively address the way she imagines or foresees Bonnie may react as a result of what she is currently saying. Something similar is at stake in (6), whereby the speaker (Martine) foresees that her interlocutor (Marielle) may burst out laughing as a result of her utterance. She similarly employs the [and don’t you VP] construction to pre-emptively admonish her not to do so. In both cases, the speaker opts for a marked surplus of meaning that would not be necessary in a simple imperative construction [don’t VP]. The latter, in fact, simply expresses an order, yet it is not marked as a construction that is inherently geared towards the anticipation of what the speaker expects the hearer might do as a result of his/her turn taking.

4 There existentials and the [there is no NP] construction

In this section, we introduce the so-called there existential construction [there is NP] and then specifically discuss pre-emptive usages of its negative form [there is no NP].

There is extensive literature on there existentials in English (i.a. Abbott 1993; Biber et al. 1999; Huddleston and Pullum 2002; Jenkins 1975; Lakoff 1987). Most accounts deal with the semantics of definiteness and specificity, particularly in connection with the declarative form of the construction (i.a. Abbot 1997; Clark and Marshall 1981; Prince 1981, 1992). The bulk of the literature centred on existential constructions has touched upon the common feature of ‘attributing a location’’ to the central entities (e.g. Anderson 1971; Bolinger 1977; Fawcett 1987; Fillmore 1968; Kahn 1966; Lyons 1975). Existential there is therefore often claimed to designate the abstract location of some central entity, which may be further specified by the locative phrase found in many existential clauses. Enumerative vs cardinal distinctions are also discussed in relation to there existentials (cf. Davidse 1999; Milsark 1977). These challenge the locative interpretation of English canonical existentials—according to which the existential there is analysed as an adverb designating an abstract location, as in Lyons (1975)—in favour of a quantified instantiation of the relevant-type specifications conveyed by the existent NP.

That being said, our enquiry is based on the negative form of there existentials. In this case, the definiteness status of the NP is not a central issue in our discussion, which in negatives is by default generic. The present account is rather centred on the pre-emptive usages of the chunk such as the one in (7):

In the House of Commons’ debate reported above, the MP is referring back to a decision that has been made about the ‘Gibraltar case’. Throughout his monologue, he foresees possible objections to this act and even potential attempts to subvert it in Parliament. The inferred turn that other MPs may take at the end of his speech is pre-emptively tackled by the same speaker with the negative existential [there is no way that VP]. From an enactive perspective, the MP is here trying to minimise a possible ‘error’ (which interactionally can be intended as a form of conflict, misunderstanding or face threat) that he predicts might arise as a result of his ongoing speech. At the same time, the illocutionary force of p is the one of refusal in that s/he markedly signals that s/he is not willing to address that issue further than what the real addressee or a generic third party in society may be expected to do (cf. Tantucci 2013, 2020 on extended intersubjectivity; see also Roberts 2012 on coercion in information structure). Simply put, the speaker pre-emptively forces the resolution (cf. Stommel 2011: 78; Lohnstein 2016 on so-called verum focus) of a fictive polar question (cf. Pasqual on fictive interaction, but see also the notion of QUD, e.g. Roberts 2012) that a generic social interlocutor may be expected to ask under the same preparatory conditions: Can we change those without the agreement of all other members community? (cf. Searle 1976 on preparatory conditions of speech acts).

Something similar is at stake in (8) below:

The politician B in (8) is being interviewed about gender equality in Parliament. While he makes his point against a numerically equal partition of seats, he predicts how his interlocutor—or a generic social persona—may refer to the idea of reaching gender equality and pre-emptively rejects it by resorting to the [there is no NP] construction. This strategy again, is interactionally triggered by the enactive awareness of what is likely to be the next turn taken by A. This is also exemplified in (9) below, where the speaker pre-emptively engages with what his/her interlocutor might suppose as a result of what s/he is currently saying.

An important diagnostic to disentangle existential and pre-emptive refusal usages of [there is no NP], is that only the latter are compatible with subsequent reinforcement of what has just been said by the speaker/writer in the form of this (indeed) cannot be done (or this is not possible/true in case of stative predicates). Conversely, mere existential usages of the chunk cannot be referred back in the same way. Crucially, in the spoken section of the BNC nominal collocates in the NP slot of the constructions displaying the highest log-likelihood value are doubt (236.3), reason (221.9), way (214.3), need (172.6), evidence (83.9), guarantee (59.6), point (54.7), difference (35.6) and problem (33.8). None of those refer to concrete objects of entities that construe a prototypical existential meaning. Rather, they all tend to be geared towards the pre-emptive rejection of what a specific or generic interlocutor may wonder (there is no doubt that p, there is no reason that p), plan (there is no way that p) consider to be necessary (there is no need that p) and so on. In this sense, the working definition of rejection is not necessarily of a coercive kind, but rather underpins the speaker’s procedural dismissal of some real or epistemic project that s/he imagines a specific or generic interlocutor may propose under the same preparatory conditions. This interactional strategy is grounded in collective intentionality (cf. Tomasello 2019) and extended intersubjectivity (Tantucci 2017, 2020; Williams 2018), as it is based on the understanding of the conventional behaviour and foreseeable reactions of the generic social persona of a socio-cultural community of practise. This is reflected linguistically by the reanalysis of constructions from a literal meaning to pre-emptive usages aimed at fictively capturing what the speaker expects the addressee—as well as anyone else in his/her socio-cultural community of practise—may have to say about his/her ongoing interaction, of which [don’t you VP] and [there is no NP] are cases in point.

In the following section, we provide a corpus-based account of the diachronic formation of pre-emptive polysemies of the negative existential [there is no NP] in British English dialogues and we shed light on the contextual, pragmatic and formal variables that significantly intersect with its development. An important fact that will emerge from this survey is the intersection between new intersubjectified polysemies and their characteristic of being pre-emptively geared towards the interlocutor’s potential reactions to the speaker’s current utterance.

4.4 A constructional analysis of [there is no NP] in the CED

This section is centred on the constructional change (cf. Traugott and Trousdale 2013) of [there is no NP] from the 16th up to the 18th century in the CED corpus. The main focus of the present analysis lies on the assumption that pre-emptive interaction is a process that is most prominently activated in dialogic interaction, whereby the interlocutor explores his/her own turn taking in order to engage with what they expect the addressee’s turn may be like. We plotted a multiple correspondence analysis (i.a. Nenadic and Greenacre 2007) looking at the holistic interaction of time in the CED with the significant co-variants among the ones we described in Section 4.4 on a two-dimensional plane. In multiple correspondence analysis modelling (MCA), associations among variables are shown by calculating the chi-square distance between different categories of the variables and between observations. These associations are thus represented graphically in the form of a map, which eases the interpretation of the structures in the data, e.g. the closer the distance between variables, the stronger the statistical correspondence. More specifically, we looked at the interaction between illocutionary force and reference with time in the CED, as illustrated in Figure 1 (R package: FactoMineR, Lê et al. 2008). We fitted this model to specifically visualise the relationship between the semasiological shift from assertions to pre-emptive refusals and reference. This allows to capture the diachronic shift of the [there is no NP] in the direction to generic expressions and extended intersubjectivity (cf. Formato and Tantucci 2020; Tantucci 2013, 2017a, 2017b, 2020):

Figure 1:

Multiple correspondence analysis of illocutionary force and reference of [there is no NP] in the CED.

The two dimensions on the x and the y axis in Figure 1 above do not refer to a specific variable. They exclusively bear statistical significance, viz. they determine spatially the attraction versus repulsion of between the three variables of this model, i.e. the century in which [there is no NP] occurred, the reference type (specific vs generic) and the illocutionary force (that is whether the construction was used as a bare assertion, or whether it involved a pre-emptive refusal). From the plot, it is possible to note that in the CED, there is a strong correspondence of a pre-emptive refusal force with 1700, in combination with generic reference. In fact, they appear quite close to one another, at the right-hand side of the map (cf. Tantucci and Wang 2018, 2020a, 2020b for specific applications of MCA in cross-cultural, diachronic and developmental pragmatics). As we argued in Section 3.1, generic reference intersects with intersubjectivity, as the speaker is not propositionally referring to an object being present in some physical or abstract space, but rather metalinguistically referring to a forthcoming turn taking on the addressee’s behalf. On the left-hand side of the plot, original assertive usages (labeled as asser) of [there is no NP] are closely associated with specific reference (labelled as spec) usage during the 1600. The above model is based on two dimensions, which in some cases may not be enough to capture the overall variation among the variables considered. A general rule of thumb to assess the degree of accuracy of MCA plotting is to calculate the percentage of variation that is captured on a two-dimensional plain and to see whether it is superior to at least 70% (cf. Levshina 2015: 382). This can be verified with the mjca() function in the R ca package (cf. Greenacre et al. 2016). The two dimensions above can capture more than 82% of the overall variation among century, illocutionary force and reference, which makes the plot in figure 2 a reliable visualisation of the correspondence among the three variables. Some examples from the present dataset are given below:

(13)

It is more pitie that there is no more disciplin among men.

(13)

*a.

This cannot be done.

CED / D2HOCHUR / 1600-1639

(14)

In an indictment of Rape, there is no time of prosecution necessary, for nullum tempus occurrit regi.

(14)

a.

This cannot be done.

CED / D2WMERVI / 1601

(15)

the King is concerned for His life: you are to speak the truth and the whole truth; for there is no reason in the world that you should adde any one thing that is false.

(15)

a.

This cannot be done.

CED / D3TCOLEM / 1678

The illocutionary force of (13) is assertive, as what is stated is essentially an existential meaning, e.g. the mere absence of something real or abstract with reference to some state of affairs. The locus of existence in (13) is specific, namely among men and the proposition cannot be referred to with this is not to be (cannot) done (*13a). From our data, assertive employment of [there is no NP] shows a tendency to intersect with absence of a complementing clause, while the opposite is true for pre-emptive refusals. Both (13–14) above are cases of pre-emptive refusal, as in both cases, the speaker/writer could anaphorically reinforce his/her statement with this is not to (cannot) be done (14a–15a). Both (14–15) are also inherently discourse-new, as the speaker/writer negatively foresees an addressee or a generic third party’s stance associated with some future or hypothetical state of affairs.

At this point, it was important to shed light on the variables that are significantly at play in the reanalysis form assertive to pre-emptive refusal usages of [there is no NP]. To do so, we modelled a mixed effects logistic regression (cf. Baayen et al. 2008). The latter can help to assess whether refusal usages of [there is no NP] would indeed be favoured at comparatively later time-spans of the CED. Mixed effects logistic regression helps to model binary outcome variables (Baayen 2008; Jaeger 2008) with two types of factors that are mixed in this kind of analysis: fixed factors and random factors. Random factors are used to systematically exclude variation that can be deemed as ‘random’ or unpredictable, and thus only indirectly affecting the response variable. For instance, factors that could arguably be considered to be random are the number of speakers in a given corpus or the number of verbs that are found in a given construction. The separation of the effects of random factors allows the analyst to assess more reliably the effects of the remaining fixed factors, and the findings can be generalised beyond the current data set with greater confidence (cf. Hilpert 2013: 52–53).

Our model thus included Illocutionary force as a binomial response variable (i.e. Assertive vs Refusal), Century as independent variable, and Genre as a random effect, other fixed effects were discarded due to multicollinearity. The results of this analysis are reported below (cf. Baayen et al. 2008; Hilpert 2013 for mixed effect models with diachronic linguistic data):

In the random effects section, the column at the right-hand side of the table is called Std. Dev. (Standard Deviation) and indicates the variability from the predicted values due to the random effects added to the model (viz. the textual Genre in which the construction occurs). It thus reflects the fact that every utterance has some unexpected factors that affect usage in addition to the fixed effects. In the fixed effects section, under the Estimate column, the coefficients Intercept, Century 1600 (1.33) and Century 1700 (0.97) indicate the slope of the categorical effects of polysemic usages of [there is no NP]. These are the dimensions are assumed to correlate with a prevalence of pre-emptive refusals vs assertive usages of [there is no NP]. From the Estimate column, we can see that refusal usages have positive values both for 1600 and 1700. What this means is that in the comparison between 1500 and 1600, there is a significantly positive value (Z = 3.39, p < 0.0005), indicating that the latter shows an increase of pre-emptive refusal usages at the expenses of assertive ones. The same comparison also leads to significant results in relation to 1700 (Z = 2.6, p < 0.05).

Put simply, the model indicates that [there is no NP] across 1500 and 1600 underwent a process of intersubjectification. A literal meaning of expressing that some object was located in some space shifted towards a new function of pre-emptive refusal, literally expressing I am pre-emptively declaring that p cannot be done. With the latter, the speaker would go beyond mere co-actional interaction and encode a surplus of meaning aimed at responding to projected turn-taking of a specific or generic addressee.

5 Pre-emptive polysemies of [there is no NP] throughout ontogeny

It has been acknowledged that phonological reduction and morphosyntactic change are not useful diagnostics for comparing diachronic and ontogenetic change. Nonetheless, important similarities have been noted in terms of unidirectional patterns of semantic–pragmatic reanalysis (Diessel 2011; Tantucci 2018) (viz. where the form of construct remains constant) as they indeed seem to hinge on similar mechanisms of change (Ziegeler 1997) or similar adaptive behaviours (Givon 2009). Drawing on that, it has been proposed that research on intersubjectification can be informed by experimental findings about the ontogeny of ToM and help formulate a gradient, more elaborate redefinition of the intersubjective paradigm (Tantucci 2020).

A fundamental condition for the gradient cross-validation of intersubjectivity is polysemy resulting from semasiological change (i.a. Traugott and Dasher 2002). Simply put, a construct (e.g. [there is no NP]) needs to remain morphologically the same, yet develop multiple meanings through time, some of which tend to be more intersubjectified than others. The same construct can then be searched in a corpus of first language acquisition. Polysemies that are hypothesised to be more intersubjectified should then be spontaneously mastered at stages of ontogenetic development, that is when a ToM should be increasingly matured. In this respect, the 4th year of age is widely acknowledged to be as a critical stage of development towards ToM (Apperly 2010; Tantucci 2020, forthcoming). The aim of this approach is to provide usage-based evidence for empirically validating whether the meaning of a construct is more or less intersubjectified, which is a diagnostic that is crucially missing in current linguistic research in intersubjectivity. From this perspective, children are clearly not to be considered as the ‘drivers of language change’. Rather, spontaneous mastery of intersubjectified polysemies of the same linguistic form throughout ontogeny can help to shed light on different degrees of intersubjective complexity and therefore, overt linguistic mastery of ToM.

In this section, we put this gradient approach into play and we compare the semantic and illocutionary change of [there is no NP] that we observed diachronically with children’s process of first language acquisition of the polysemic usages of the same construction. To do so, we looked holistically at all the there is no/there’s no forms in all the CHILDES^[3] corpora of British and American English, respectively, in the 0–3, 4–6 and 7–9 age spans. The CHILDES dataset is composed of different subcorpora (one for each child), covering more than 30 languages and 230 children and adolescents ranging from 0 to 18+ years of age. For our analysis, we looked at a collection of 92 subcorpora amounting to a total of 29, 480, 736 tokens, uttered by roughly by 1,507 children. More specifically, we focused on transcriptions of children aged between 0 and 9 years, divided as follows: 0–3, 4–6 and 7–9. It is important to notice that each subcorpus contains utterances by both the child and other people, such as the interviewer(s) and/or carers, the child’s parents, grandparents, relatives and so on. Our analysis only took into account those utterances spoken by the child in contexts of spontaneous interaction (e.g. not being elicited for a specific task), therefore limiting the scope of the queries to a specific set of utterances out of the total available. This allowed us to control the settings of usage of the [there is no NP] forms and satisfy the independence assumption. Table 3 contains the details for each selection, listing the total number of tokens including the non-child ones (tokens); the total number of tokens for only the utterances by the children (child-only tokens) and the number of occurrences of [there is no NP] (including the variant [there’s no NP]).

We first normalised our annotated occurrences based on the larger subcorpus, (the one including data of the 0–3 age span, 4,580,207 tokens, entailing a per-3.2-milion-word normalisation of the child-only annotation strings from the two remaining subcorpora).^[4] We then fitted a binomial logistic regression of the developmental acquisition of the chunk and its assertive vs refusal usages. While in Section 4.5, our main effects were based on centuries, in this case, we used children’s age as main effects. The choice of the latter as the only predictor resulted from a forward stepwise selection, followed by ANOVA testing for possible interactions (cf. Levshina 2015:§12).

The model above is slightly different from the one in Section 4.5 as it does not include the genres of texts as a random effect. This is due to the fact that the context in this case is quite stable, viz. based on spontaneous spoken interaction between the child and mother/carers, with the main effects satisfying the independence assumption (cf. Winter 2013). One may wonder why a variable “speaker” was not included as a random factor in the model above. This is due to the disparity of the CHILDES datasets that were included in our query, some of which not comprising speakers’ identity. It is also important to remark that no significant interactions among other potential predictors were at play, which was functional to the interpretation of age as a predictor in isolation (cf. Winter 2020: §5). The main effects here taken into account are children’s ability to spontaneously produce pre-emptive refusal usages of the [there is no NP] construction, entailing extended intersubjectivity and generic reference of the locus of existence of the NP. From the table above, we can see positive coefficients associated both with children’s capacity to utter pre-emptive refusals in the 4–6 age span (Z = 2.82, p = 0.0047) and the 7–9 age span (Z = 3.59, p = 0.0003) in contrast with a negative value at the intercept, corresponding to the 0–3 age span.

All in all, from our data a clear developmental trajectory emerged among formal and illocutionary features of [there is no NP], indicating a progressive ontogenetic ability to engage with pre-emptive refusal usages out of mere assertive employment of the same form. This pattern is very similar to the diachronic development of the same chunk towards intersubjectivity and pre-emptive interaction observed in Section 4.5. In fact, children’s spontaneous usage of [there is no NP] is increasingly geared towards pre-emptive refusal strategies and intersubjectively marked interaction.

Examples (16–17) illustrate early usages of the chunk, which are merely aimed at assertively stating the (negative) existence of an entity in some location:

In both (16–17), [there is no NP] merely refers to discourse-given information, without pre-emptively addressing how the interlocutor or a third party might react as a result of the child’s utterance. This can operationally be tested with the possibility of dropping the NP after there is no, and simply replying with a pronominal form there is none, which distinctively hinges on discourse-given information. This indicates that PEM is here not an overtly coded element of the usages of [there is no NP].

Quite differently, children progressively learn to use the same chunk with a pre-emptive refusal illocutionary force, thus overtly problematising some hypothetical state of affairs that the addressee or a generic third party may virtually bring to the fore.

Both (18–19) represent a first step towards the pre-emptive reading of [there is no NP]. In fact, the child is not merely interested in stating the absence of some entity in some location. Despite an existential meaning is still expressed propositionally, at the procedural level s/he pre-emptively rejects the idea of some possible project that the addressee or a generic social persona may consider as valid or plausible in those specific contexts. Similar to what we argued in Section 4, both usages in (18–19) express a procedural meaning that can be referred back with the expression this cannot be done, which, in turn, is not felicitous in mere existential usages such as in (16–17). Similarly, they cannot be used pronominally in the form of there is none, as they are aimed at expressing new information, rather than stating the absence of some object that is already part of the ‘textual’ common ground among the interlocutors. This is because they are grounded in social cognition and the collective awareness of what a generic persona in his/her socio-cultural community of practise would be plausibly say under the same preparatory conditions. Usages of this kind are significantly absent before the 4th year of the child’s ontogenetic development, viz. when ToM abilities are shown to become increasingly sophisticated (cf. Kovacs et al. 2010; Onishi and Baillargeon 2005; Surian et al. 2007; Tantucci 2020). As suggested at the beginning of this section, despite evident morphosyntactic mismatches existing between language change and first language acquisition, however similar pathways of change are yet present for what concerns the semantic-pragmatic reanalysis of a linguistic form that remains formally the same (i.a. Diessel 2011; Givon 2009; Tantucci 2020; Tantucci and Wang 2020b; Ziegeler 1997). This phenomenon can be defined as onto-semasiology (cf. Tantucci in press) and underpins a matching trajectory in language change and child development, shifting from literal meanings to more idiomatic usages of the same form that are increasingly oriented towards intersubjectivity and collective intentionality (cf. Tomasello 2019). The implications of this method are the ones of exploiting phenomena of language change and intersubjectification for applied purposes, as increased semantic-pragmatic complexity that emerges diachronically matches higher complexity and (comparatively) late acquisition throughout the ontogenetic development of a Theory of Mind. The present study is case in point about this relationship and provides the methodological desiderata for addressing intersubjective complexity as an interactionally pre-emptive mechanism emerging diachronically. New polysemies of pre-emptive interaction that arise diachronically can become a powerful resource to shed new light on ontogenetic capacities to overtly express a Theory of Mind throughout naturalistic interaction.

6 Conclusions

This paper addressed the pragmatic relationship between pre-emptive interaction and intersubjective awareness in language change and ontogeny. Pre-emptive interaction occurs through dialogues as a form of prediction-error-minimisation (PEM), viz. when an interlocutor addresses potential reactions that may result from his/her ongoing speech. We argue that this form of awareness is an indicator of intersubjectification and may constitute a trigger of constructional change. To support this claim we provided a corpus-based analysis of the change of the [there is no NP] construction in Early Modern English. The chunk shifts from a mere existential construction, characterised by an assertive illocutionary force, in the direction of a new pre-emptive refusal usage. This reanalysis significantly correlates with a number of formal and semantic changes, such as decrease of compositionality, increase of procedurality and generic reference to the place or state of affairs connected with [there is no NP]. The semantic and pragmatic changes that we observed throughout the constructional change of the chunk also emerged from a large-scale analysis of the CHLDES data. This cross-validation revealed a higher degree of complexity and extended intersubjectification of pre-emptive usages of [there is no NP], as children would be able to spontaneously master them at comparatively later stages of ontogenetic and ToM development. Our analysis indicates that it is only around the fourth year of age that the child starts to spontaneously master pre-emptive polysemies of [there is no NP], often intersecting with generic reference, presence of a complementing clause and reduced compositionality.

This novel corpus-based method is centred on the cross-validation of intersubjective complexity arising diachronically and the ontogenetic capacity to express a ToM as linguistically overt mechanism throughout spontaneous interaction. The implications of this methodology aim at a new applied turn of studies in language change, serving as a powerful resource for the study of intersubjective complexity in FLA, autistic and neuro-typical interaction. Naturalistic exchanges can therefore be analysed as being characterised by overt intersubjectified expressions displaying a gradient spectrum of complexity, ranging from literal meanings to extended intersubjectified polysemies. Spontaneous mastery of the latter can provide data-driven diagnostics to assess interlocutors’ pragmatic and semantic ability to overtly express ToM awareness and social cognition (cf. Tantucci in press) as a by-product of naturalistic interaction.