Exploring the interplay of trait-like L2 willingness to communicate, international posture, language anxiety and fluency in monologue L2 speech

2.1 L2 willingness to communicate and its antecedents

L2 WTC has been defined as “a readiness to enter into discourse at a particular time with a specific person or persons, using an L2” (MacIntyre et al. 1998: 547). In literature, L2 WTC is described as a behavioural intention determining L2 use (MacIntyre and Wang 2021). According to the classic conceptualisation of L2 WTC (MacIntyre et al. 1998; see Table 1) and its recent reconceptualization adapted to multilingual contexts (Henry and MacIntyre 2023), L2 WTC (Layer II in Table 1) is an L2 learner’s desire that precedes actual communicative behaviour (Layer I). However, there are numerous mutually interacting factors, presented in MacIntyre et al.’s (1998) heuristic model of L2 WTC in the form of more or less proximal layers that underlie the behavioural intention.

The immediate precursors of L2 WTC involve situated, individual and contextual aspects (Layer III), which trigger an L2 learner’s transient desires to initiate communication with a particular person, on a specific topic and in a defined context. Zhang et al. (2018) classified these antecedents into situational cues, such as the task, teacher and classmates, situational characteristics, including support, cooperation and objectives, and psychologically determined characteristics, like duty, positivity and negativity. These precursors of L2 WTC have frequently been investigated following a dynamic research design, whose findings generally support the role of these aspects in L2 WTC fluctuations (Zhang et al. 2022).

Further down in Layer IV, an L2 learner’s own view of their linguistic abilities and beliefs in communicative capabilities create a relatively stable set of propensities. This self-assessment, addressed in research as linguistic self-confidence, has been reported to be positively related to L2 WTC (Piechurska-Kuciel 2018). An examination into linguistic self-confidence and L2 communication confidence with the same participant group as reported in this study (Szyszka et al. under review) revealed how linguistic self-confidence is related to such measures of L2 speech fluency as speech rate (SR; the number of syllables per production time unit including pauses) and mean length of run (MLR; the average number of syllables between pauses), and L2 communication confidence, in turn, correlates positively with such measures of fluency as SR and articulation rate (AR; the number of syllables per production time unit excluding pauses), demonstrating the important role that these relatively stable characteristics may have in L2 speech fluency.

Affective-cognitive contexts, categorised as Layer V, encompass enduring factors such as intergroup attitudes and social situations. Among these factors, LA has been found to be one of the consistent predictors of L2 WTC (Elahi Shirvan et al. 2019), in the sense that higher levels of LA reduce L2 WTC. This is one of the reasons why we have decided to include LA in the present study. This negative emotion is often defined as “a distinct complex of self-perceptions, beliefs, feelings, and behaviours related to classroom language learning and arising from the uniqueness of the language learning process” (Horwitz et al. 1986: 128). LA has been reported not only to be inversely related to L2 WTC, but also to in-class participation (Das 2020) and L2 output (Cheng et al. 1999). MacIntyre and Charos (1996) confirmed that high LA was strongly associated with avoidance behaviour. These findings imply a link between LA levels and L2 speech fluency resulting from consistent engagement in L2 communicative practice (Lyster and Sato 2013).

Due to this potential connection, LA has recently attracted the attention of researchers aiming to investigate its relationship with L2 speech fluency (Aubrey 2022; Bielak 2022; Szyszka and Lintunen 2023; Szyszka et al. 2024; Zuniga and Simard 2022). For instance, with a subset of participants (n = 59) from the present study, Szyszka et al. (2024) revealed how LA, measured with the Input, Processing and Output Anxiety Scale and a task-specific post-session survey, was linked with speech fluency: participants with higher levels of internal processing and output anxiety used more filled pauses (FPs; such as ‘erm’), and high output and task-specific anxiety seemed to lead to lower SR and AR. A closer extreme case sampling analysis of six participants (Szyszka and Lintunen 2023), chosen based on the Foreign Language Classroom Anxiety Scale (FLCAS; Horwitz et al. 1986), showed how highly anxious individuals have slower SR and AR and more FPs than the least anxious participants. In Bielak’s (2022) study, task specific LA levels were inversely related to MLR (r = −0.5) and AR (r = −0.67), whereas Aubrey’s (2022) results revealed moderate to strong positive correlations between momentary LA changes and the mean length of pauses per second. Generally, further research is necessary due to inconclusive results from studies that have applied different methodologies and measuring tools. Compared to the above-mentioned research on LA and L2 speech fluency, the present study provides evidence on LA as a trait-like characteristic measured with the FLCAS (Horwitz et al. 1986) from a large sample (n = 102) to complement previous findings.

Finally, personality and intergroup climate, entailing, for instance, IP (Yashima 2002; Yashima et al. 2004) form the most distal Layer VI. In the present study, we investigate IP understood as “the general attitude toward international community and foreign language learning” (Yashima 2002: 54). This psychological tendency comprises interest in international affairs, unfolding how far a learner is interested in international events and news (IP-interest), as well as openness to other cultures and intercultural communication, which usually involves the use of L2 (IP-openness). IP impacts, among others, L2 WTC (Yashima et al. 2004). Thus, both of the dimensions of IP may also indirectly relate to fluency, but, to our knowledge, these relatively stable L2 learner characteristics have not been examined in connection with L2 speech fluency prior to the present study.

Generally, the heuristic model accounts for both transient and trait-like precursors of L2 WTC (Henry and MacIntyre 2023; MacIntyre et al. 1998). While situated factors determine fluctuating desires to engage in L2 communication on a moment-to-moment basis (Kang 2005), trait-like factors contribute to a consistent propensity to communicate across different situations, help learners actively seek opportunities for L2 interaction (Yashima et al. 2004), and predict the trait-level of L2 WTC (Elahi Shirvan et al. 2019). This consistent inclination for communication in an L2 increases chances for activation of automatization processes, which are indispensable for fluent speech (Segalowitz 2010). Therefore, the trait-like L2 WTC perspective is followed in the current study with the aim to explore the connection between “a stable tendency for some people to speak more than others over a long period of time” (MacIntyre and Ayers-Glassey 2020: 187) and selected aspects of utterance fluency, which echo, to a degree, the efficiency of underlying cognitive processes (Segalowitz 2010).

In the present study, L2 WTC has been operationalised following Mystkowska-Wiertelak and Pawlak’s (2017) and Peng and Woodrow’s (2010) models. The former is a multi-factor model including eight trait-like facets of L2 WTC: 1. communication confidence, 2. ought-to self, 3. classroom environment, 4. planned in-class WTC, 5. unplanned in-class WTC, 6. practice-seeking WTC, 7. IP-openness and 8. IP-interest. The latter focuses on L2 WTC experienced outside the classroom. The focus of the present study is on individuals’ readiness to initiate communication inside (facet 4 and 5) and outside the classroom (facet 6 and Peng and Woodrow’s (2010) outside the class L2 WTC), and their propensity for intercultural encounters (facet 7 and 8), which might be limited by trait-like LA. Motivational (facet 2) and contextual (facet 3) factors are scrutinised in detail in Szyszka (in preparation), whereas for more on the link between communication confidence (facet 1) and speech fluency, see Szyszka et al. (under review).

In our study, being willing to engage in more and less controlled communication in the classroom is labelled as planned in-class L2 WTC and unplanned in-class L2 WTC, respectively. The motivation behind including them in the present research is related to the assumption that readiness to engage in more spontaneous classroom interactions may be associated with greater reliance on implicit or highly automatised explicit L2 knowledge, whereas propensity to initiate planned in-class communication may not necessarily draw on this kind of knowledge (Pawlak 2012). Practice-seeking L2 WTC refers to the L2 learners’ pursuit of participating in various forms of verbal practice both within and outside the classroom, including practising “either with their friends and peers (e.g. during breaks), also falling back on modern technologies (e.g. sending text messages), or modifying their utterances if someone makes them aware that an error has been made” (Mystkowska-Wiertelak and Pawlak 2017: 95). Outside the class L2 WTC concentrates on being ready to interact with interlocutors in out-of-class settings without any obligations frequently imposed on learners in classroom situations (Peng and Woodrow 2010), such as mandatory classroom performance for teacher evaluation. Although classroom factors might partially explain context-dependent differences in L2 WTC, research reports that in-class L2 WTC levels are related (Zarrinabadi and Abdi 2011) or even predict learners’ readiness to communicate outside the classroom (Denies et al. 2015). IP (Yashima 2002), which determines an L2 learner’s readiness to be involved in intercultural communication, has been approached in this study from two perspectives: IP-openness and IP-interest. The first is understood as “willingness to gain new experiences through contacts with foreigners” (Mystkowska-Wiertelak and Pawlak 2017: 5) and the second is an L2 learner’s drive to expand knowledge regarding the target language culture in order to engage in intercultural communication. A higher level of IP was reported to be related to an increased communication frequency in L2 (Yashima et al. 2004). However, little is known whether this is reflected in L2 speech fluency.

2.2 L2 speech fluency and L2 WTC

Fluency is a complex phenomenon understood by Lennon (2000: 26) as “the rapid, smooth, accurate, lucid, and efficient translation of thought or communicative intention under the temporal constraints of on-line processing”. A more recent approach to defining L2 fluency, proposed by Tavakoli and Hunter (2018), involves four levels of understanding the concept. First, in a very broad meaning, L2 fluency functions as a general language proficiency entailing all skills. Second, L2 fluency in a broad sense refers to an overall speaking proficiency. Next, in a narrow sense, it is associated with the flow and continuity of L2 performance, where fluency is often assessed alongside linguistic complexity and accuracy. Finally, in a very narrow perspective, adopted for the purposes of the present study, fluency is conceptualized in terms of objectively measurable temporal characteristics, such as speed, breakdown and repair features of L2 speech (Skehan 2009; see 3.2.2). According to Segalowitz’s (2010) fluency framework, these observable manifestations of the smoothness of speech refer to only the utterance fluency dimension, which partially reflects the underlying cognitive processes of an L2 speaker, known as cognitive fluency. Additionally, cognitive fluency forms the foundation for how listeners perceive the fluidity of L2 speech, that is, perceived fluency. Generally, the efficiency and speed of cognitive processes play a role in both measurable manifestation of speech fluency and impressions of fluency from a listener’s standpoint. The present study takes the perspective of utterance fluency reflecting, to an extent, the underlying cognitive processes (Segalowitz 2010).

The conceptualisation of parallel and serial processing in the L2 speech production model provides a theoretical framework for understanding the role of trait-like L2 WTC in building L2 speech fluency. Briefly, in Levelt’s (1989) model, later adapted to L2 speech by De Bot (1992) and Kormos (2006), speech production begins at the conceptualisation stage where micro- and macro-planning are activated at a conceptual level. Next, at the formulation stage, a pre-verbal message is formulated and encoded immediately before the third stage – articulation. Finally, speech can be monitored and modified at all of the aforementioned stages. The speed of these processes is determined by the level of L2 speech automatisation (Segalowitz 2010). Therefore, high levels of readiness to communicate that lead to increased speaking practice (e.g., Kim et al. 2022), necessary for the automatization of speech production processes, can be, theoretically, related to L2 speech fluency. In other words, L2 learners with trait-like propensity to engage in L2 speech constantly and systematically increase their chances to automatize speech processes and augment efficiency of the underlying cognitive processes related to L2 speech fluency. As a result of automatization, for more proficient speakers, speech production processes may operate in parallel at conceptualisation, formulation and articulation stages, and are manifested in fluent speech with minimal undue repairs, such as repetitions (Duran-Karaoz and Tavakoli 2020). For instance, high values of SR and AR, as well as low frequencies of FPs and repetitions are expected when parallel processes are activated. However, if the processes are less automatic, serial or controlled processing takes place, which is typically marked by slower SR and signs of disfluency (Kormos 2006). Theoretically, high levels of trait-like L2 WTC create conditions for accelerated automatization of speech processes associated with L2 fluency. However, so far, results of scarce research investigating trait-like L2 WTC and L2 speech fluency have not been entirely consistent.

Contrasting findings have emerged from studies examining the relationship between L2 WTC, conceptualised as a trait-like characteristic, and L2 speech fluency. D’Amico’s (2012) study explored the relationship between L2 WTC and speech fluency in two groups of study abroad (SA) and at home (AH) intermediate level learners (9 in SA, 14 in AH) of L2 Spanish. Data were collected using pre- and post-programme interviews as well as pre- and post-programme WTC questionnaires. Although significant differences were found in SR, silent pauses (SPs), average length of fluent run and repairs between SA and AH after a six-week period, no significant correlations were identified between L2 WTC and L2 speech fluency markers in either group. Similarly, Kim et al. (2022) investigated this relationship in two groups of L2 learners who were short-term residents in Australia: students and workers. In their study, L2 speech fluency was measured with SR, phonation-time ratio (proportion of time spent speaking during the speech sample) and MLR. They collected speech samples based on an interview question about an unpleasant experience while communicating in English. Peng and Woodrow’s (2010) scale was used to calculate the levels of L2 WTC. The outcomes did not show any significant links between L2 WTC and L2 speech fluency markers. However, Zabihi et al. (2021) found evidence of a positive correlation between trait-like L2 WTC and perceived fluency. They applied a questionnaire to measure the levels of L2 WTC and invited 26 experienced teachers to evaluate the fluency of speech samples of 100 learners of L2 English, who performed a picture narration task similar to the one used in the present study. In this study trait-like L2 WTC was found to be a predictor of the subjective evaluation of oral fluency.

More consistent results have emerged from L2 WTC research following the idiodynamic design, marking changes in variables on a moment-to-moment basis. A moderate positive relationship between dynamic WTC and speaking time was found in Nematizadeh and Cao (2024). Nematizadeh (2021), Nematizadeh and Wood (2019) and Wood (2016) have also demonstrated that heightened L2 WTC states align with fluent speech moments. Wood (2016) pioneered the investigation of the dynamics between L2 WTC and speech fluency. In his study, the analysis of the performance of four participants led to the conclusion that both variables influenced each other. The aim of Nematizadeh’s (2021) study was to monitor per-second changes in L2 WTC and identify relatively stable states lasting longer than a few seconds, defined as attractors. These attractor states were later analysed along with speech fluency measures, such as SR, MLR, duration and number of SPs and FPs of 11 participants. The results showed that attractors emerged mostly when the participants reported having lexical resources, interest in the task content and confidence to continue speech. A lack of these led to extended pausing. In other words, high L2 WTC reflected in stable attractor states coincided with fluent speech. Similarly, Nematizadeh and Wood (2019) investigated dynamic interactions between L2 WTC and speech fluency measures, such as SR, AR, pausing patterns and MLR. They followed a stimulated recall research design to identify the L2 WTC levels of four participants, whose performance on a picture description task was also analysed from the perspective of fluency features. Higher than average MLR was found to be related to high L2 WTC, whereas frequent pausing was associated with lower levels of L2 WTC. These studies confirm that the dynamic nature of L2 WTC can shape and be shaped by real-time speech fluency. However, research investigating trait-like L2 WTC and speech fluency provides less conclusive evidence, highlighting the importance of the present study.

Further research is needed to understand the intricate relationship between L2 WTC and L2 speech fluency at different proficiency levels and contexts. In search of this interplay, the present study follows the perspective of trait-like characteristics of planned in-class L2 WTC, unplanned in-class L2 WTC, outside the class L2 WTC, practice-seeking L2 WTC and two strong predictors of L2 WTC – LA and IP. The aim of the study is to assess the levels of different types of L2 WTC and the strength of their interrelationship in L2 learners to obtain a detailed picture of the individuals under investigation. Obtaining a precise group profile regarding trait-like characteristics provides a foundation for mapping complex relationships between affect and L2 speech fluency. The present study, thus, seeks to examine the interplay of the aforementioned types of L2 WTC, two trait-like antecedents of L2 WTC and temporal indices of L2 utterance fluency, including SR, AR, frequency of SPs, frequency of FPs and repetitions (R). The following two research questions (RQs) are addressed:

1. What are the levels and strengths of relationships between trait-like L2 WTC (Planned and Unplanned in-class L2 WTC, Outside the class L2 WTC and Practice-seeking L2 WTC) and trait-like antecedents of L2 WTC (LA, IP-openness and IP-interest) of L2 learners?

2. To what extent do selected temporal aspects of L2 utterance (dis)fluency (SR, AR, SP, FP, R) correlate with Planned and Unplanned in-class L2 WTC, Outside the class L2 WTC, Practice-seeking L2 WTC, LA, IP-openness and IP-interest in the whole sample and in the groups of the participants at B2 and C1/C2 levels?

3.1 Participants

The participants were 102 L1 Finnish university students learning English as a foreign language, taking part in a larger project focusing on L2 speech fluency from different perspectives. Within this cohort, 66 were females, 28 males, four chose not to disclose their gender and four marked the option ‘other’. Their mean age was 22.5 years (SD = 5.25). On average, they declared a 10.3 year exposure to English prior university (SD = 1.46). Their English language proficiency level was assessed with the Lexical Test for Advanced Learners of English (LexTALE), a validated measure of overall proficiency instrument (Lemhöfer and Broersma 2012). The mean score the participants obtained was 85.4 (SD = 8.8; scale 0–100), which corresponds to a C1 level on the Common European Framework of Reference for Languages (CEFR, Council of Europe 2001). However, individual scores ranged from 62.5, indicating B2 level (60–80 %), to 98.75, matching the upper C1/C2 levels (80–100 %). There were 26 at B2 and 76 participants at C1/C2 levels.

3.2 Measurement

3.3 Procedure and materials

The data collection, taking place during the COVID-19 pandemic, was organised as part of the FDF2 project. The participants signed informed consents prior to the production of monologue speech samples in the language laboratory. A logically sequenced cartoon story was used as a stimulus material. The narrative was depicted in six linearly developed pictures without captions or subtitles. Picture descriptions have been widely used in fluency research (Kormos and Dénes 2004), mainly because they are a guarantee of comparability between samples and elicitation of free production (Peltonen 2018). Individuals were instructed in English to describe the story immediately after a two-minute preparation period given to familiarise themselves with the pictures and to plan before speaking. Participants were allowed to look at the pictures while they were being recorded to avoid memory overload. The average total time of speaking was 63.41 seconds (minimum 19.05, maximum 131.25, median 60.82). The speech samples were saved as MP3 files, later converted to .wav for further analysis. After the recordings, participants completed a battery of online questionnaires, including biographical information, the Affective Questionnaire (see Section 3.2.1) and the LexTALE (Lemhöfer and Broersma 2012). The Affective Questionnaire was translated into Finnish before its distribution. Back translation was applied to ensure accuracy of the content. As the original tool was designed to suit Polish context (Mystkowska-Wiertelak and Pawlak 2017), minor adaptations were introduced. For instance, in Practice-seeking L2 WTC scale the word Finnish was added, as in I am willing to use English to speak to/text my Finnish friend in my free time.

3.4 Analysis

All recorded monologues were transcribed and double-checked. Repetitions were calculated manually. Speech analysis software Praat (Boersma and Weenink 2020) was used to manually annotate FPs. SPs were identified with a script but manually checked. Lennes’s (2002) script was applied in Praat to calculate the total duration of labelled segments, while De Jong and Wempe’s (2009) script, albeit with manual adjustments, was used to identify the number of syllables. Fluency measures in the study – AR, SR, SP, FP, R – were standardized per minute of speaking time.

Statistical analyses were performed in SPSS (version 27). Since the sample size was relatively small, determining the distribution of data was important for the selection of statistical tests. The Shapiro-Wilk’s tests of normality indicated that the data for FPs (W = 0.87, p < 0.001), repetitions (W = 0.90, p < 0.001), Planned in-class L2 WTC (W = 0.95, p < 0.001), Practice-seeking L2 WTC (W = 0.93, p < 0.001) and IP-interest (W = 0.86, p < 0.001) were non-normally distributed, so non-parametric Spearman’s correlation coefficients were used. A groupwise approach to correlations was applied to identify the correlation coefficients for the whole group of the participants, as well as for those at B2 and C1/C2 levels. The interpretation of effect sizes followed Plonsky and Oswald’s (2014) recommendations – r = 0.25 small, r = 0.40 medium, r = 0.60 large effect size.