Imperfective aspect in Chinese conversation: do speakers imitate one another’s constructions?

1 Introduction

Speakers normally adapt to one another in conversation and often reuse grammatical structures recently produced by their interlocutors. Similar grammatical structure across speakers is often taken as an implicit phenomenon and referred to as syntactic priming (Branigan 2007; Chang et al. 2000). Priming has been widely studied in experimental settings (cf. Van Gompel and Arai 2018), including recent works on grammatical aspect (e.g., Magliano and Schleich 2000; Golshaie and Incera 2021). It is closely linked to the broader concept of cognitive alignment – understood as the dynamic coordination of mental representations between interlocutors during interaction (Pickering and Ferreira 2008; Pickering and Garrod 2022; Rasenberg et al. 2020). This study examines priming as it occurs in dialogue (e.g., Gries and Kootstra 2017) and focuses on how grammatical structure and meaning jointly influence the way speakers engage with and re-use each other’s constructions (cf. Goldberg 2006; Ziegler et al. 2019) – understood as holistic pairings of form and meaning (see Tantucci and Wang 2022a, on constructional priming). More specifically, we explore how speakers re-use each other’s imperfective constructions, with a focus on dialogic resonance (cf. Du Bois 2014), a form of structural and functional imitation where speakers re-use and re-adapt the constructions of their interlocutors.

The priming literature on grammatical aspect has mostly centred on the distinction between perfectivity and imperfectivity (e.g., Ferretti et al. 2007; Golshaie and Incera 2021). However, little attention has been paid to how different ways to construe an event aspectually may affect an interlocutor’s linguistic behaviour in natural conversation. We investigate imperfectives, examining whether focality – the degree of aspectual salience on an ongoing event (cf. Johanson 2000; Chen 2003) – influences the way speakers (creatively) re-use one another’s constructions. We frame our analysis in Dialogic Syntax (Du Bois 2014; Zima and Brône 2015; Tantucci and Wang 2002b) and statistically model speakers’ degree of (creative) resonance of in conversation (cf. the Dialogic Categorisation Model, Tantucci 2023a).

The present analysis is centred on Mandarin conversation and whether focal imperfectivity is retained more vividly in working memory than durative imperfectivity. Our assumption is that, as the conversation unfolds and progresses, dialogic resonance of focal imperfectives would decrease less than for durative imperfectives. The reason for this can be found in the cognitive and semantic characteristics of focal imperfectivity. When a speaker uses a construction centred on a distinct moment in which something is happening, that is likely to catch more attention than durative events that are temporally vaguer and more ambiguous. This means that an interlocutor is more likely to be primed for longer by an expression such as she was just entering the apartment rather than she lived in that apartment for years. Something similar also applies to the presence of a transitive or a locative object instead of events that would not include that. We assumed that this would also influence the extent to which an imperfective priming construction affects interlocutors’ ability or attitude towards re-using it in conversation. This may depend on the amount of salient information encoded by the construction. One thing is to say I was reading, another is saying I was reading Zizek’s latest essay that you told me about.

The paper is structured as follows: Section 2 introduces the notion of aspectual imperfectivity and the way it is expressed in Mandarin Chinese. Section 3 discusses dialogic resonance as an important mechanism of alignment in naturalistic conversation. In Section 4 we describe our data retrieval and annotation scheme. Section 5 is devoted to the statistical analysis of our data. We discuss the relevance and the implications of our findings in Section 6, before concluding in Section 7.

2 Imperfective aspect

The imperfective aspect has been traditionally defined as a way to present a situation without its boundaries, focusing on its internal constituency independently from its beginning or end (Janda and Fábregas 2019: 690). In contrast to perfective aspect, which presents an event as a complete whole (e.g., Dahl 1985; Croft 2012), imperfective aspect focuses on the internal structure of an event, highlighting its ongoing nature, habitual recurrence, or extended duration. Because of this internal perspective, imperfective aspect can provide crucial cues about how speakers conceptualise time and action in interaction. Research on imperfective aspect generally follows two approaches:

1. Comrie (1976) views aspect as a categorical concept: it can be either perfective or imperfective. In his view, the perfective aspect refers to the viewing of a situation as a finished or complete whole, without reference to its internal structure. It presents an action or event as finished or complete. Imperfective aspect, on the other hand, involves viewing the event internally, focusing on its duration or repetition without emphasising its completion. Comrie initially distinguishes two types of imperfective:

   1. Habitual aspect: Refers to actions or events that occur repeatedly or habitually over time, but without specifying whether they have ended. It focuses on regular patterns of activity.

   2. Continuous aspect: Refers to actions or events that are ongoing or in progress.

      He further subdivides the continuous aspect into:

      1. Progressive aspect: This denotes actions that are currently in progress or actively unfolding at the moment being described.

      2. Nonprogressive aspect: This includes ongoing states or activities but without the dynamic sense of progress found in the progressive, who distinguishes imperfectives into habitual and continuous aspects and then divides the continuous aspect into progressive and nonprogressive).

2. Bybee et al. (1994) regard imperfectives as a highly grammaticalised category, representing a further development of the progressive aspect and encompassing various usages.

   1. Progressive: The action is currently unfolding or in progress, similar to Comrie’s progressive.

   2. Habitual: Actions that happen regularly or repeatedly, akin to Comrie’s habitual aspect. Stative: This refers to the expression of states or conditions that are ongoing but not actions (e.g., She knows the answer).

   3. Gnomic: Describes general truths or timeless statements (e.g., Water boils at 100 °C).

According to Bybee et al. (1994) model, the imperfective includes progressive, habitual, stative, and gnomic meanings, which form different combinatory subsets in specific languages. However, they also note that no coherent grammatical morpheme types for the continuous aspect were found in their cross-linguistic survey (1994: 139). This somewhat challenged Comrie’s aspectual taxonomy, as he did not clarify what exactly is the synchronic and diachronic status of the so-called nonprogressive aspect (cf. Comrie 1976: 25). Bybee et al. (1994) introduced the concept of the resultative aspect, limiting their definition to a continuous state after the completion of an action. According to Chen 2021 the nonprogressive aspect in Chinese corresponds to a so-called broad resultative aspect (Nedjalkov 1983: 6, 7). This, in turn, may develop into a progressive aspect, and further into other usages, becoming a more schematic, non-progressive imperfective. This is illustrated in the Chinese examples (1)–(3), showing the three functions where the actual use of the construction [V 着 zhe] is still primarily resultative (cf. Hopper 1991 on persistence) but showing a diachronic trend towards the latter two functions.

In (3) [V 着 zhe] is less compositional and more procedural (cf. Terkourafi 2011; Tantucci 2023b) than in (2), as it no longer simply describes the on-going phase of an event, but more broadly instructs the hearer to view the event as an unbounded state that is persisting at some time of reference. This could arguably be categorised as a durative non-progressive function in Comrie’s framework, or as a highly grammaticalised imperfective morpheme (originating from a resultative lexical source) in Bybee et al.’s model. Both views of imperfectives are equally valid. This paper is centred on synchronic data and will thus primarily refer to Comrie’s terminology. This is primarily a matter of analytical focus: Comrie’s framework more clearly foregrounds aspectual distinctions relevant to continuity, whereas Bybee et al. focus on degrees of grammaticalisation, more central to diachronic analyses.

3 Resonance and dialogic imitation

This study adopts a usage-based perspective on natural speech as composed of constructions – that is, as holistic pairings of form and meaning (Goldberg 1995, 2006; Kay and Fillmore 1999; Langacker 1987; Tomasello 2003; Traugott and Trousdale 2013). The usage-based model highlights individuals’ capacity to identify and categorise constructions based on natural exposure to language use. Recent years have seen renewed emphasis on co-construction of meaning within conversational exchanges, exploring how speakers jointly conceptualise them during dialogue (e.g., Haugh 2007; Arundale 2010; Weigand 2018). Dialogic Syntax (Du Bois 2014; Tantucci and Wang 2021; Zima and Brône 2015) extends this line of enquiry with a strong focus on grammatical structure, that is on constructions that result from speakers’ joint efforts to categorise meaning through interaction. For instance: a salutation such as [A: How’re you doing? | B: Not too bad, how about yourself?] is constructed by two individuals as a unit with structural properties and semantic meaning, performing the joint project of a greeting. In this way, constructions emerge dynamically (Hopper 2011) because of interlocutors’ dialogic engagement. This often involves creative re-elaboration of forms and meanings throughout an interaction, e.g., a telephone exchange such as [A: I’m now taking the bus ] B: [Ok, I’m just cooking dinner .], with B re-using the [I’m ADV_Temp ^[6] V_Prog ^[7] Obj] structure in her response. Linguistic processing is inherently ‘recombinant’ (Tantucci and Lepadat 2024; Tantucci and Wang 2024), as speakers constantly and quickly adjust structure and meaning to new dialogic stimuli. Such re-use and re-combination of an interlocutor’s constructions is often creative (Tantucci 2023) and is defined as dialogic resonance. Resonance is a distinctive form of alignment (Pickering and Garrod 2021), as it more specific than mere cognitive and communicative coordination. It represents a form of complex imitation (Arbib 2012), which may occur either implicitly or explicitly, as speaker B recognises and builds upon what speaker A has said, treating it as a structural and functional substrate for subsequent turns (Goodwin 2013):

Since you (A) said that X, I (B) recombine X in the new form of X’.

Dialogic resonance has been found to be comparatively more impeded in Autism Spectrum Disorder (e.g., Tantucci and Wang 2023), partly because it plays a crucial role in guiding verbal engagement – allowing speaker B to formally acknowledge the relevance of what speaker A has said in order to sustain the interaction. This constructional approach to dialogue enables the quantitative analysis of verbal engagement and joint categorisation of form and meaning among speakers, as in the example below retrieved from an Air UK sales meeting in the British National Corpus^[8] (BNC1994):

In this interaction, resonance occurs with a Stuart (B) chatting with a colleague (A). A’s original construction [I could hear her thinking it] is re-elaborated by B in the form of [he could hear you thinking that]. There are semantic, pragmatic, and morphosyntactic analogies between the two forms. There is a verbatim repetition of specific words, could, hear and think. The subject and the object of each construction before and after hear are all Personal Pronouns (Pp). This allows the categorisation of the more schematic construction [Subj_Pp ^[9] could hear Obj_Pp ^[10] thinking Obj] as a pairing of form and meaning. It expresses the idiomatic meaning of someone (in the Subj_Pp position) ‘hearing’ the thoughts of someone else (in the Obj_Pp position). Rhetorical effects are also present: B re-combines A’s expression to engage with them, as s/he empathises with A’s ‘presumed’ state of mind. The emergence of a dialogic construction is referred to as a diagraph (Du Bois and Giora 2014: 354) as given in Table 2. When the original ad hoc construction is modified, that is marked as underlined text (in case of replacement) and in brackets (in case of (addition)):

Resonance may also involve cognitive alignment in the way both speakers verbally construe events. More specifically, [Subj_Pp could hear Obj_Pp thinking Obj] is a construction of high focality (Chen 2003) as the evoked scenario, i.e., the thinking activity – despite being abstract – is processed as ongoing at a specific point of reference. This would be different to, say, [He often imagined you thinking it], which would extend to a longer and more blurred stretch of time.

In this paper, we look at whether the imperfective constructions used by speaker A influence speaker B's aspectual choices in response . Similarly, we are interested in the dialogic effects of different aspectual construing processes on working memory, i.e., whether high focusing imperfectivity (i.e., imperfectivity that represents a specific moment in time) is more likely to stay activated and re-enacted in dialogue than continuous imperfectivity (underpinning an unspecific period in which some state persists). The usage-based assumption behind this study is that verbal representation of event structure is directly affected by dialogic stimuli in naturalistic interaction. This may depend on implicit and explicit mechanisms. Alignment (Pickering and Garrod 2021) is argued to result from the automatic ability to adjust to others’ behaviour. At the same time, it has also been found that verbal imitation (i.e., dialogic resonance) often involves creativity and correlates with explicit intersubjective marking at the sentence periphery and politeness reciprocity, which are, in turn, explicit interactional phenomena (Culpeper and Tantucci 2021; Culpeper et al. 2025; Tantucci et al. 2022). Priming persistence is found to arise from both implicit learning and explicit memory, with the latter contributing to immediate but transient effects (Bernolet et al. 2016) which are mostly resource-limited and attention-dependent (Zhang et al. 2020). Whether the activation of specific aspectual representations across interlocutors is an implicit or explicit mechanism goes beyond the scope of this paper. We acknowledge the theoretical importance of this distinction, but disentangling these mechanisms would require experimental methods or longitudinal data beyond what is available in a naturalistic corpus-based study.

What remains of interest for our study is whether aspectual representation is significantly affected by the dialogic input by another interlocutor. Put simply, if speaker A opts to represent linguistically some event in a particular way, is speaker B going to opt for a similar way to represent the same event? This study will aim to answer the following research questions:

1. Do highly focal imperfectives have a stronger priming effect in naturalistic conversation than durative ones?

2. What is the priming effect of complexity? Do larger constructions lead to greater resonance in conversation?

3. What is the role of information structure? Does the presence of transitive/locative objects affect dialogic resonance of imperfectives?

4 Data retrieval and annotation

The data of this research were retrieved from the recorded audio and transcripts of Mandarin Chinese speakers’ telephone conversations in the CallHome (among family members) and CallFriend corpora (among friends). Each corpus consists of 120 and 60 unscripted telephone conversations, respectively, each totaling approximately 250,000 words (Liu et al. 2006). Callhome and Callfriend speakers were aware they were being recorded but were not given any guidelines regarding the content of their conversations. The situated nature of this context (e.g., absence of proxemics, kinesthetic cues) provides an opportunity to focus primarily on the textual dimension of verbal interaction. We retrieved in total 236 imperfective aspect constructions, including 174 from CallHome corpus and 62 from the CallFriend corpus.^[11]

For the annotation, we focused on several dimensions: whether the imperfective was ‘dialogically’ primed by a previous construction (resonating vs independent); the source of resonance, if any (self vs other); the degree of dialogic resonance; the distance from the dialogic prime to the point of resonance, the imperfective function (durative vs focalised), and the speaker’s accents (Northern or Southern Mandarin) were all annotated. Accent was assessed aurally during the review of the recordings and included as a variable to explore whether structural differences in the use of Mandarin imperfectives might also be influenced by dialectal variation – an intriguing possibility tentatively raised by Liu (2022: 8). Finally, the form of imperfective aspect expressions could be analysed as a construction occupying nine syntactic slots at most (irrespective of the core verb in counting), based on the typological features of Chinese. These are given in Table 1, Section 2.2.

5 Results and analysis

A key goal of this study was to assess how imperfectivity is resonated across interlocutors in Mandarin conversation. To do so, we calculated the degree of resonance in each of B’s turn where A had just used an imperfective.

5.3 Findings and interpretations

The first important finding of this regression is that increasing distance (in intonation units, IUs) from the imperfective leads to a resonance decrease. In other words, the more that is said after speaker A uses an imperfective, the lower the degree of resonance of that imperfective by speaker B. This is what we expected: speakers’ working memory does a better job at alignment with constructions they have just been heard in conversation. In contrast, their ability to resonate and re-use the linguistic material they heard will unavoidably decrease with time and the progression of the conversation. This is particularly evident in Figure 1, where increasing IUs of speaker A’s turn (given as Distance on the x-axis) predicts a decrease of resonance of the imperfective in the following speaker B’s turn (β = −0.18, 95 % CrI: −0.25, −0.11).

Figure 1:

Main effect of dialogic distance on resonance.

This figure illustrates the baseline relationship between dialogic distance (in intonation units) and resonance in the following speaker’s turn. It provides a foundational reference point for interpreting the more detailed interaction patterns explored in Figures 3 and 4. A second important finding of this study is that the larger the size of the priming imperfective (its complexity, i.e., the number of words it is made of), the higher the degree of resonance in the subsequent turn. As Figure 2 shows, roughly every four words used in the imperfective construction by speaker A leads to an increase of one resonating constituent in the imperfective subsequently uttered by speaker B (β = 0.29, 95 % CrI: 0.15, 0.44).

Figure 2:

Size of imperfectives (constructional complexity) as a predictor of resonance.

Figure 3:

Interaction between distance and focality as a predictor of resonance.

Figure 4:

Interaction between distance and objecthood as a predictor of resonance.

Example (11) illustrates a case of shorter size (often called low complexity in construction grammar cf. Bybee 2010; Tantucci 2021) leading to low resonance.

(11)

[low complexity, low resonance]:

A:

还在做, 还没写呢, 哎呀。

hái zài zuò, hái méi xiě ne, àiya

still prog do, still not yet write sfp Asp, sfp Mod

‘I am still doing (experiments). I haven’t written the paper, oh.’

B:

啊, 在做实验, 你这反正关键靠实验呃啊?

a zài zuò shíyàn, nǐ zhè fǎnzhèng guānjiàn kào shíyàn e a

back 15 prog do experiment, you this anyway crucially rely on experiment sfp Mod

‘Ah, you are doing an experiment, does your paper rely on the experiments crucially?’

CallHome 0782

1. 15

   Backchannel.

In (11) A’s imperfective [在做 Obj] zài zuò ‘be doing’ is resonated by B as [在做实验] zài zuò shíyàn ‘be doing an experiment’, giving rise to the schematic construction [在做 Obj], with a resonance value of 3.

In (12) below we have a case of larger size (higher complexity), leading to increasing resonance in the subsequent turn:

(12)

[high complexity, high resonance]

A:

现在正在办签证呢, 欸。
xiànzài	zhèng	zài	bàn	qiānzhèng	ne	ei
now	just	prog	apply	visa	sfp Asp	sfp Mod
I am just applying for a visa now.’

B:

正在办签证, 是吧?
Zhèng	zài	bàn	qiānzhèng,	shì	ba
Just	asp	apply	visa,	be	sfp Mod
‘You are just applying for a visa, isn’t it?’
CallHome zho 0766

In this case, A’s form [正在办签证呢, 欸] zhèng zài bàn qiānzhèng ne ‘I am just applying for a visa now’ is resonated by B as [正在办签证, 是吧?] zhèng zài bàn qiānzhèng, shìba ‘just applying for a visa isn’t it?’, giving rise the dialogic construction [正在办签证 SFP], with a resonance value of 5: 正 + 在 + 办 + 签证 + SFP_Mod.

The model also showed important interactions involving the type of aspect used across speakers. We have already seen that increasing distance from the priming imperfective leads to a decrease in resonance. What is more important is that such a decrease is sharper for durative imperfectives than focal imperfectives (β = 0.11, 95 % CrI: 0.03, 0.19), as shown in Figure 3. This is the tendency that we expected: the more specific the construal of an ongoing event, the more vivid the impression on the hearer in the subsequent turn.^[16]

Another important interaction is between imperfectives that include a transitive or locative object and ones without any. We found that even in this case, the priming effects of imperfectives that include an object are stronger across speakers’ turns, as resonance tends to decrease more slowly than for objectless constructions (β = 0.11, 95 % CrI: 0.04, 0.19).^[17] No other construction type that we analysed had a similar effect on resonance. This indicates that [Subj V Obj_Pat/Loc] constructions tend to prime hearers more than any other, presumably because they are more informative on the one hand, and also easier to process than, say, obliques, on the other. This is particularly evident in Figure 4.

5.3.1 Hierarchical clustering of imperfective construction types

At this stage, we examined the distribution of various imperfective constructions in Mandarin conversation and the sociolinguistic factors that may influence the preference for certain structural features over others. We thus fitted a hierarchical clustering model, an algorithm that groups data into hierarchically nested clusters based on their similarity (we used Euclidean distance with Ward.D2 amalgamation, cf. Levshina 2015). The process starts with each data point as its own cluster, then successively merges the most similar clusters together, forming a tree-like structure called a dendrogram (e.g., Gries 2010; Tantucci 2020; Tantucci and Wang 2022). Different levels of constructional similarity was obtained for all imperfective slots in our data (see Table 1) and their attraction to accent, resonance (coded as High vs Low, depending on the Median value), imperfective functions, objecthood, presence of Sentence Final Particles (SFP) and so on (see the corresponding dendrogram in the Appendix). We then reduced the clustering structure into a two-dimensional representation via multidimensional scaling (MDS). This maps the attraction among constructions with the highest degree of similarity, as given in Figure 5 (Bubble sizes indicate frequencies).

Figure 5:

Two-dimensional scaling of hierarchical clustering of imperfective construction types in Mandarin conversation.

The map reveals a distinctive division between two areas, one on the left-hand side of the map, mostly including the H slot (Aspectual 呢 ne, see Table 1), and one at the bottom right-hand side, clustering around constructions including the D slot (pre-verbal 在 zài). We then looked at the factors that most decisively determined this partition. Accent emerged as a key factor in the distribution of the imperfective use of either form in Mandarin conversation. This is easily captured in the snake plot in Figure 6.

Figure 6:

Snake plot of the factors determining constructional similarity among Chinese imperfectives.

Figure 6 indicates how variables tend to cluster together towards the left (red) versus the right-hand side (grey) of the plot. The x-axis represents the difference in mean proportion of each variable between Cluster 1 and Cluster 2. Variables are sorted in ascending order on the Y axis according to their relative importance in distinguishing the clusters. We can see that the D slot (D.Y, i.e., presence of preverbal 在 zài) seems to be the stronger factor ‘pulling’ Southern Accent towards the left. Similarly, we can see that the absence of D (D.N), is the stronger factor ‘attracting’ Northern accent, together with H (Aspectual 呢₁ ne) and E (着 zhe) slots. This pattern highlights a clear sociolinguistic divide: 在 zài is rarely used in the North but commonly found in Southern speech, a striking contrast that, to our knowledge, has not been previously articulated so explicitly.

5.3.2 Bayesian analysis of accent variation

To confirm this exploratory finding, we conducted a mixed-effects logistic Bayesian regression, with all slots as predictors of Accent, and Conversation ID as a random effect, see Table 8.^[18]

Similar to what the hierarchical clustering suggested, three predictors (slots) showed important regional tendencies: D (在 zài), H (呢₁ ne) and E (着 zhe). 在 zài is distinctively preferred by Southern speakers (β = 3.10, 95 % CrI: 1.35, 4.86) while H (呢₁ ne) and E (着 zhe) are predominantly used by Northerners (β = −6.59, 95 % CrI −8.79, −4.43; β = −3.55, 95 % CrI: −6.05, −1.08).

In Southern dialects, 在 zài is frequently used as a progressive and imperfective marker (Hashimoto 1985; Hu 2005; Wang 1999). In Northern dialects, the imperfective functions of 呢₁ ne are much more prominent, e.g., spoken Beijing dialect (Chen 2022: 35–37; Liu 2022: 8). This could also have to do with Mandarin Chinese (also named 普通话 pǔtōnghuà ‘Common Language’) as a lingua franca (cf. Li 2006; Ostler 2022: 424) across China, integrating diachronically various constructions from Southern to Northern dialects, now all compatible in the same synchronic system. We often found in our data that 呢₁ ne can be self-sufficient in describing highly focalised progressive events among northern speakers, as in (13) below:

(13)

[Northern speaker using 呢1 ne as a progressive]:

A:

哎周雷你先坐一会儿, 对不起啊, 我给我们家打电话呢。
Ai	Zhōu Léi	nǐ	xiān	zuò	yìhuìer,	duìbùqǐ	a,	wǒ
back	Zhou Lei	you	sit	for	a while,	sorry	sfp Mod,	I
gěI	wǒmén	jiā dǎ	diànhuà	ne
to	our	family	call	sfp Asp
‘Zhoulei, please sit for a while, I am talking with my family on the phone’.
CallHome/1525

On the contrary, 在 zài is used frequently by Southern speakers (see also Hu 2003 on 呢 ne being more likely interpreted modally – rather than aspectually – by Southerners).

(14)

[Southern speaker using 在 zài as progressive]

A:

我现在因为很多朋友都在学电脑, 他们都说这个好。
wǒ	xiànzài	yīnwéi	hěn	duō	péngyǒu	dōu	zài	xué
I	now	because	very	many	friend	all	PROG	learn
diànnǎo,	tāmén	dōu	shuō	zhègè	hǎo
computer,	they	all	say	this	good
‘Because many friends are all learning IT now, they all say that this major is good’.
CallFriend/4227

6 Discussion

Research on grammatical aspect priming has largely focused on the distinction between perfectivity and imperfectivity, yet little attention has been given to how varying construals of imperfectivity influence an interlocutor’s linguistic behavior. We addressed this from a new angle, as we looked at whether focality affects priming and alignment across speakers. Additionally, while most research on aspectual priming is based on controlled experimental settings, this study examined spontaneously produced dialogic data from telephone call interactions. A way to control for alignment across speakers in conversation is via dialogic resonance, the way speakers (often creatively) re-use one another’s constructions. We provided a replicable annotation framework called the Dialogic Categorisation Model (DCM) (Tantucci 2023a), which we implemented across all interactions in the CallHome and CallFriend corpora.

What we found is that resonance naturally decreases as the dialogue continues to unfold: the longer the stretch of conversation from A’s imperfective, the lower B’s resonance with that form. However, it increases when the imperfective is made of a larger construction, e.g., 她目前正在睡觉呢 tā zhèngzài shuìjiào ne ‘she is just sleeping at the moment’ rather than 她在睡觉 tā zài shuìjiào ‘she is sleeping’. This effect may be both cognitive and socio-pragmatic. An increasing size of the original imperfective provides more information about the event and inhibits entrenched use of that construction (cf. Tantucci and Di Cristofaro 2020): highly conventionalised constructions often undergo chunking (Bybee 2010; Bybee and Moder 2024) as they tend to be shorter and semantically more bleached (e.g., He’s coming) than longer forms that are construed ad-hoc and which include richer information about what is happening (e.g., he is now just coming down the stairs). Larger constructions are more likely to be processed compositionally (cf. Kay and Michaelis 2019), and thus leave a stronger impression on speakers’ memory during an exchange. This suggests that alignment in conversation is responsive to both the salience and specificity of grammatical constructions. While disentangling automatic alignment from implicit learning lies beyond the scope of this study, the resonance patterns we observe – especially for focal and high-complexity imperfectives – are consistent with a mixed view, as they reflect sensitivity to structural detail, recent input but also semantic vividness. From a cognitive-pragmatic angle, politeness reciprocity (Culpeper and Tantucci 2021; Tantucci et al. 2022) may also be at work here. Speaker A’s effort in producing a more detailed construction of the event may be conversationally ‘rewarded’ (i.e., reciprocated) by stronger engagement by speaker B, hence with a higher degree of dialogic resonance in B’s turn. This is particularly evident for the constructions we tackled in this analysis, with length varying from 2 to 6 internal constituents. Things may differ for even larger ones, as that may come with a cognitive ‘cost’. Excessive amounts of information may have a negative effect on relevance and hearer’s attention, as predicted by Relevance Theory (cf. Sperber and Wilson 2012).

A key aspect of this study is that aspectual focality remarkably influences the way speakers align with each other in conversation. Highly focalised imperfectives stay more vivid in interlocutors’ memory than durative ones. This also applies to imperfective constructions that include transitive or locative objects. This suggests that the more saliently an on-going event is construed, the stronger the priming effects on the hearer. This is especially relevant in Chinese, where the imperfective aspect is often not obligatory (Wu 2005). It comprises a distinctively diverse range of construction types with high degree of flexibility (Xiao and McEnery 2004). This makes speakers’ recombinant creativity (Tantucci 2023a; Tantucci forthcoming) a fundamental mechanism for the alignment of imperfective constructions in natural conversation: Chinese imperfectives are highly malleable and thus often creatively re-combined in natural conversation. It would be worth investigating whether similar degrees of aspectual priming emerge in languages with more constrained and morphologically obligatory aspectual systems, such as those found in the Germanic or Romance families. Related to this point, we found clear differences in the imperfective constructions used by Northern and Southern Mandarin Chinese speakers. The postverbal aspect marker 着 zhe and sentence-final aspect marker 呢 ne are preferred by Northern speakers, while the preverbal aspect marker 在 zài is preferred by Southern speakers, which reflects the influence of substrate dialects on the regional varieties of Mandarin Chinese and status of Mandarin Chinese as a lingua franca. Chinese imperfective constructions have been discussed both diachronically and synchronically (Hu 2003; Wang 1999).

7 Limitations

This study is not without limitations. For one, the size of CallHome and CallFriend corpora is limited to roughly 250,000 words each. This is reflected in some degree of uncertainty (Est. Error is around 1/3 of CI for main interactions in the first regression) of our prediction. Future development of highly controlled spoken corpora of Chinese could provide richer resources for analysing aspectual priming in Mandarin dialogue. Related to that, the dialogic use of imperfective constructions could be extended to other construction types that may be retrieved from larger corpora. Also, our study controlled for Accent as a binary variable. However, dialectal variation is gradient and with larger datasets informed by speakers’ demographics more fine-grained analyses of dialectal variation could be conducted. One example could be the preverbal aspect marker 有 yǒu, which only recently became a conventionalised Mandarin construction, as it was originally present only in Southern Dialects. Dialogic resonance may provide a powerful lens for investigating recent patterns of linguistic propagation (cf. Croft 2010) in recent years.

8 Conclusions

This paper has important implications for the theoretical understanding of Chinese grammar, for advances in cognitive research on priming and interactional alignment, and for conversational approaches to engagement in human interaction. It shed new light on the grammatical characteristics of Mandarin imperfective aspect in naturalistic interaction, the construction types it favors, the ones that are more likely to prime interlocutors’ turns in conversation and the use of different imperfective constructions by Northern and Southern Speakers. We contributed to research in dialogic priming and alignment as we provided new evidence showing that imperfective events that focus more vividly on specific time intervals (focalised) are more likely to affect speakers’ subsequent use of imperfectives in contrast to events that encode less detailed time spans (durative). We found that larger constructions, including ones with richer information structure, are also more prone to dialogic resonance and thus being re-used and re-combined by speakers in conversation. There is currently a paucity of research on aspect from a dialogic perspective and the way imperfective meanings are re-used and recalibrated across interactants in naturalistic interaction. This study also aimed to address this gap and provided a replicable framework for the quantification of dialogic resonance for the study of Construction Grammar in dialogue. Applied avenues where this may be explored are contexts of First and Second Language Acquisition, (Im)politeness and intercultural communication and neurodivergent interaction (e.g., on the autism spectrum or in contexts of neurocognitive degeneration). Additionally, priming in naturalistic interactions may be studied multimodally, by controlling for gestures, expressions and/or prosody. Future research could also examine how aspectual priming varies typologically across languages and dialects.