Study
Methods

STUDY DESIGN

This quasi-experimental mixed-methods study examined NPDT from three perspectives: whether students in the NPDT condition showed greater gains in overall Foreign Language Enjoyment (FLE), how particular NPDT behaviours were related to moment-to-moment changes in FLE, and how students themselves experienced these behaviours and explained their feelings. Three sources of data were used: pre- and post-intervention FLE questionnaires, classroom audio recordings, and post-intervention semi-structured interviews.

CONTEXT AND PARTICIPANTS

The study took place in a rural lower-secondary school in Hebei Province, China, where English was a compulsory subject in the senior high school entrance examination. Limited educational resources and strong exam demands meant that regular English lessons were largely teacher-fronted and exam-oriented. This setting therefore provided a useful context for examining NPDT as a different way of organizing classroom interaction.

A total of 811 students from 16 intact classes participated in the study, including eight Grade 7 classes and eight Grade 8 classes. Each student was assigned a unique anonymous identification code so that pretest and posttest questionnaire responses could be matched. After data screening, the final analytic sample consisted of 791 students, with 396 in the NPDT condition and 395 in the comparison condition.

The lessons were taught by four undergraduate students from Beihang University’s School of Foreign Languages and China Foreign Affairs University’s Department of English and International Studies. They were 19 to 20 years old and all had previous experience in volunteer English teaching. Each near-peer teacher taught both NPDT and comparison classes and worked with students in both Grade 7 and Grade 8.

THE INTERVENTION

Before the intervention, written consent was obtained from the school, students, and their guardians. Across two weeks, each class received five 90-minute lessons on alternate school days.

Both conditions received intensive reading lessons with the same teaching objectives and the same reading materials. The main difference was in teachers' instructional and interactional practices. The NPDT classes used the NPDT approach. The comparison classes followed conventional volunteer English instruction, with greater emphasis on explaining the reading texts and related language points and with classroom interaction consisting mainly of teacher questions and brief student responses.

Classes in both conditions used two Penguin Readers texts: The Last Photo and Marcel and the Shakespeare Letters. Using the same materials and learning goals helped keep the comparison focused on how the lessons were taught.

TEACHER PREPARATION

Before the intervention, the four near-peer teachers completed eight training sessions of approximately four hours each. In the first two sessions, the theoretical basis of NPDT and the classroom use of the approach were introduced. The teachers then discussed and refined the framework together.

From the third session onward, the teachers worked with an experienced lower-secondary school teacher to develop shared teaching materials and detailed lesson procedures for both the NPDT and comparison classes. The training concluded with lesson rehearsals and feedback.

MEASURING FOREIGN LANGUAGE ENJOYMENT

Students' FLE was assessed with the Chinese Version of the Foreign Language Enjoyment Scale, an 11-item measure developed for the Chinese EFL context. The questionnaire was administered before and after the intervention. The original item wording and response format were retained, while the instructions were adapted to fit the near-peer teaching context.

The questionnaire was completed in paper-and-pencil format because mobile phone use was restricted at the school. Students used their anonymous identification codes so that the two questionnaires could be matched. Internal consistency was satisfactory at both measurement points.

CLASSROOM PROCESS DATA

NPDT class audio recordings were imported into ELAN 6.7 and coded for NPDT behaviours and six audio-based FLE-related vocal signals derived from prior work on FLE dynamicity and vocal expressions of emotions. Before formal coding, two coders were calibrated with pilot-test recordings, and the coding manuals were finalized accordingly.

This is a key area for methodological refinement in the next phase of the NPDT project. We plan to use the idiodynamic method to examine moment-to-moment fluctuations in FLE in greater detail. For more information, see What’s Next.

STUDENT INTERVIEWS

Twenty-four students from the NPDT classes took part in semi-structured interviews. Within each NPDT class, students were ranked according to their change in FLE from pretest to posttest and grouped into the upper 30%, middle 40%, and lower 30%. One student was selected from each group. Pretest FLE scores were also considered so that the interview sample included students with different starting levels of enjoyment.

During the interviews, students were asked to recall concrete episodes from the two-week English lessons and describe the activities or tasks involved, the most noticeable teacher or peer behaviours, the feelings they experienced, and the reasons they believed they felt that way. All interviews were conducted in Mandarin and audio-recorded.

DATA ANALYSIS

Questionnaire data were analyzed using a linear mixed-effects model in R. Students' FLE scores served as the dependent variable. Time, Condition, and the Time × Condition interaction were included as fixed effects, with Grade and Teacher included as additional fixed effects. A random intercept for Student accounted for the repeated pretest and posttest observations from the same students. The model was specified as follows:

Estimated marginal means and simple-effects comparisons were used to interpret the Time × Condition interaction. Because students were nested within classes, standard errors were adjusted for class-level clustering using the Bell-McCaffrey CR2 adjustment with Satterthwaite degrees of freedom.

During audio coding, each NPDT event was first identified in the audio recordings and assigned a behaviour code. The event and the following 15 seconds were then checked for FLE-related vocal signals. NPDT behaviour categories were mutually exclusive, so only one behaviour code was assigned to each event. Vocal-signal categories were not mutually exclusive, allowing more than one signal type to be coded for the same event. To assess inter-coder reliability, 20% of the recordings were independently coded by two coders, yielding high agreement for NPDT behaviours (Cohen’s κ = .866) and FLE-related vocal signals (Cohen’s κ = .783–.820). The remaining recordings were then coded by the primary coder.

For analysis, a binary co-occurrence variable was created. NPDT events containing at least one FLE-related signal were coded as 1, and those without as 0. This variable was analyzed using a mixed-effects logistic regression model in R (version 4.5.1). The binary NPDT–FLE Signal Co-occurrence variable was the dependent variable. NPDT Behavior Type was included as the fixed effect, Teacher as an additional fixed effect, and Class as a random intercept to account for the clustering of events within classes. The model was specified as follows:

Estimated marginal probabilities of NPDT–FLE signal co-occurrence were calculated for each behaviour, followed by Tukey-adjusted pairwise comparisons.

Interview recordings were automatically transcribed, checked for accuracy, and imported into NVivo. Classroom episodes in the transcripts were first identified and coded according to the NPDT behaviour involved. These episodes were then coded inductively for students' feelings and perceived reasons. Related codes were compared across episodes and grouped into broader categories, which were refined throughout the analysis. Two coders jointly coded all transcripts and discussed coding decisions throughout the process.

IMPLEMENTATION FIDELITY

Implementation fidelity was assessed using classroom audio recordings from both conditions. The same coding framework was used to check whether NPDT behaviours were delivered as intended in the NPDT classes and whether they appeared in the comparison classes. This allowed the study to examine both intervention adherence and condition differentiation.