Decoding the dynamics of L2 interaction – A processability lens on the acquisition of turn-taking skills 

Anke Lenzing, University of Innsbruck 

In my talk, I focus on some of the key challenges L2 learners face in the acquisition of turn-taking. The bulk of human language use is conversational, and turn-taking is regarded as part of the “universal infrastructure for language” (Levinson, 2016). However, despite its ubiquitous occurrence in communication, turn-taking poses a huge cognitive challenge to interactants in a conversation. It forms part of an intricate system characterised by the rapid exchange of turns between interlocutors, aiming to minimise gaps and avoid overlaps (Sacks et al., 1974; Levinson & Torreira, 2016). 

To achieve this fine-tuned coordination between listening and speaking, interlocutors need to prepare their turns in advance, which includes making predictions about the current speaker’s turn and aligning with their conversation partner (e.g., Garrod & Pickering, 2015, 2021). 

How do L2 learners cope with this cognitive challenge, and which processes are involved in their acquisition of turn-taking skills? 

Widening the scope of Processability Theory (e.g., Pienemann, 1998; Pienemann & Lenzing, 2025) to these cognitive aspects of L2 interaction, I discuss theoretical considerations regarding the question as to how the psycholinguistic constraints that learners face in L2 acquisition affect their ability to make predictions in communicative interaction. I then present empirical insights into the turn-taking abilities of L2 learners of English at different stages of acquisition. This includes timing in turn-taking, learners’ alignment behaviour, as well as the learners’ abilities to make predictions. The analyses are based on audio and video data, as well as mobile eye-tracking data obtained from L2 learners’ communicative interactions (e.g., Lenzing, 2024; Lenzing, Schmiderer & Egger, 2025; Lenzing & Schmiderer, submitted/accepted). 

The results of the analyses shed light on some of the challenges involved in the acquisition of L2 turn-taking and point to relations between the L2 learners’ processing constraints and their turn-taking abilities. 

References: 

Garrod, S., & Pickering, M. J. (2015). The use of content and timing to predict turn transitions. Frontiers in Psychology, 6, 751. 

Garrod, S., & Pickering, M. J. (2021). Understanding dialogue. Language use and social interaction. Cambridge: Cambridge University Press. 

Levinson, S. C., & Torreira, F. (2015). Timing in turn-taking and its implications for processing models of language. Frontiers in Psychology, 6, 731. 

Sacks, H., Schegloff, E. A., & Jefferson, G. (1974). A simplest systematics for the organization of turn-taking for conversation. Language, 50(4), 696-735. 

Lenzing, A. & Schmiderer, K. (submitted/accepted). Processing in the Receptive, Productive and Interactive Use of an L2. In: Wright, C., Peltonen, P., Piske, T. & Steinlen, A. (eds.), Context Matters. Bristol: Multilingual Matters. 

Lenzing, A. (2024). Timing turn-taking: the art of making predictions in L2 communicative interaction. Inaugural lecture, University of Innsbruck, 29.11.2024. 

Lenzing, Schmiderer, Egger (2025). Investigating L2 multi-person interaction – what gaze shifts reveal about predictions in turn-taking. Paper to be presented at the EuroSLA conference, Tromsø, 25.-28.06. 2025. 

Pienemann, M. (1998) Language processing and second language development: Processability Theory. Amsterdam: Benjamins. 

Pienemann, M. & Lenzing, A. (2025). Processability Theory. Cambridge: Cambridge University Press.

Simulating language and learning in the context of PT

Manfred Pienemann 

Abstract

PT was conceptualised as a formal and thus testable contribution to the debate about the learnability of (second) languages. Its key objective was and still is to add a set of constraints to language learnability theory that is based on an aspect of the human mind, more specifically on language processing.

Recently, the stunning performance of so-called (Large) Language Models (LLMs) gained a great deal of public attention, and some researchers appear to imply that computational language models are now fully capable of replicating the linguistic abilities of human speakers and learners of language through statistical learning.   

To this long-term observer of the modern debate about language learnability it appears that with the above LLM-based position the pendulum swings back yet again to that end of the continuum in the learnability debate that was once occupied by S-R theory.

In this presentation, I advocate a position that balances statistical learning with human-specific constraints on language learning, thus embracing a limited role of computational language models in learnability theory. I will show that – despite their impressive performance - the role computational language models is limited, due inter alia to their lack of human capacities such as morality or reasoning that are based on the intentionality of the human mind.

I will look at two examples of the constructive role of SLA simulations in the context of PT-based SLA research. The first example is based on a collaborative research project conducted by Frank Lanze, Anke Lenzing, Howard Nicholas and myself. In this project, we simulate statistical L2 learning based on input and a set of constraining factors. Our software is designed to simulate the internal dynamics of the learning process as a dynamical system. This computational tool enables a comparison of natural learning data with simulated data – thus uncovering computational parameters that determine differential L2 dynamics. This computational tool permits us to test hypotheses such as the following: “L2 developmental and variational dynamics are fundamentally different”.

The second example (suggested by Frank Lanze) utilises a publicly available LLM as a research tool. I will show that the LLM “llama-3.2-3b-instruct“ can be used on a personal computer to generate a program that can analyse a narrow range of ESL samples according to PT principles. 

The above computational tools do have the capacity to promote the sophistication of SLA research. However, using or developing such tools does not imply that one needs to assume the now-popular position that LLMs are fully capable of replicating the linguistic abilities of human speakers and learners of language.