S. Zhou, Gamifying Language Education 2024

Abstract

This mixed-methods study investigates the impact of digital game-based learning (DGBL) on enjoyment, ideal L2 self, and intrinsic motivation among Chinese English as a Foreign Language (EFL) learners. Seventy participants were divided into DGBL and control groups. The DGBL group engaged in Duolingo activities, while the control group received traditional EFL instruction. Data collection included pre-post self-determination theory (SDT) questionnaires, experience sampling method (ESM) to assess real-time enjoyment, and stimulated recall interviews. Quantitative analysis using paired-samples t-tests, one-way ANCOVAs, and multilevel modeling revealed that DGBL significantly enhanced enjoyment and ideal L2 self-perception, with pre-existing autonomy and ideal L2 self predicting greater enjoyment during gameplay. Qualitative findings highlighted increased engagement, perceived learning gains, and a sense of autonomy fostered by DGBL. Importantly, this study demonstrates that integrating DGBL into EFL classrooms can significantly boost enjoyment and cultivate a positive self-concept as language learners among Chinese students. These findings have practical implications for educators, suggesting that incorporating welldesigned game-based activities can create a more motivating and effective learning environment, addressing the specific challenges faced by Chinese EFL learners.

Study Design

Participants were recruited from a public university in Mainland China, focusing on second-year students enrolled in the university’s EFL program. To ensure homogeneity in their foreign language background, participants had to meet the following criteria: (1) English as their primary foreign language, (2) aged between 18 and 22, (3) regular access to a computer or tablet with a reliable internet connection, and (4) no prior experience with the specific DGBL platform used in the study. This last criterion aimed to isolate the effects of the DGBL intervention itself, minimizing any influence from pre-existing familiarity with the platform.

A total of 70 learners (32 male, 38 female) were recruited through convenience sampling from two intact EFL classes with similar proficiency levels (random assignment), as determined by a standardized English placement test administered at the beginning of the academic year. Recognizing the limitations of convenience sampling, this approach was chosen due to logistical constraints and the need for intact class settings (cluster randomization)

To control for potential class-based biases, the two classes were randomly assigned to either the DGBL group (n = 35) or the control group (n = 35) using a random number generator (e.g., through a software like Microsoft Excel or a similar tool). This randomization process ensured that any observed differences in outcomes between the groups could be attributed to the intervention rather than pre-existing differences in participant characteristics.

Following the two-week intervention, a subset of 15 participants from the DGBL group was purposefully selected for semistructured interviews. The selection criteria aimed to capture a diverse range of experiences and perspectives by considering gender, pre-test scores on the IL2SS and self determination theory subscales, initial English proficiency levels, and reported enjoyment levels during the intervention. The interviewed participants included 7 males and 8 females, with an average age of 19.5 years (SD = 0.8). Ten of them had prior gaming experience, while five did not. In terms of English proficiency, 4 were classified as low, 6 as medium, and 5 as high based on the placement test results. This purposive sampling approach allowed for a richer and more nuanced understanding of the impact of DGBL on learners with different characteristics and experiences.

Efforts were made to maintain demographic balance within the sample, considering gender and other relevant factors. Additionally, the study was conducted following ethical guidelines, with approval from the university’s Institutional Review Board (IRB). Informed consent was obtained from all participants, ensuring they were aware of the study’s purpose, procedures, and their right to withdraw at any time without penalty.

Study design analysis

→ Sampling
This study is a quasi-experimental design, utilizing a cluster randomized setting to randomly assign one group to an intervention (experimental group) and one to go without (control group). There are several threats to internal validity in using this design, however due to requirements of using an entire intact class (to prevent contamination) required the researcher to consider setting up convenience sampling rather than random assignment

There are a number of artifacts that might arise out of this set up. Due to differences in peer culture, teacher expectations, motivation levels and learning strategies as well as cohesion and classroom climate, artifacts may threaten internal validity.

Pre-test ANCOVA helps, but does not eliminate unmeasured baseline differences. (OpenAI, 2025). Some examples of artifacts according to Chatgpt:

Unequal Exposure Time Artifact
Because the experimental class used Duolingo for class time and was encouraged to use it outside of class, differences in:

• total English exposure
• time on task
• frequency of feedback
  may arise. This is not strictly due to “motivation” but to differential dosage, which is an artifact of having intact conditions.

Hawthorne Effect Intensified by Cluster Assignment
When entire classes know they are using a “special method,” the novelty may elevate motivation temporarily.

This is amplified when applied to a whole group, where group enthusiasm reinforces itself (“collective Hawthorne effect”).

→ Variables

Due to a lack of true randomization, there are a number of extraneous variables that have not been controlled. There are a series of confounding variables that might distort the relationship between the dependent and independent variables.