Vol. 13/ Núm. 3 2026 pág. 233
https://doi.org/10.69639/arandu.v13i3.2392
Gamification as a didactic strategy for developing speaking
skills in ninth-grade students at the Sagrado Corazón de Jesús
public school
La gamificación como estrategia didáctica para el desarrollo de la habilidad oral en
estudiantes de noveno grado de la Unidad Educativa Fiscomisional Sagrado Corazón
de Jesús
Elssie Mabel Narváez Cárdenas
mabelnarvaezcar@gmail.com
https://orcid.org/0009-0001-8819-7009
Universidad Politécnica Estatal del Carchi
Ecuador – Tulcán
Martha Aracelly Viveros Almeida
martha.viveros@upec.edu.ec
https://orcid.org/0000-0002-2441-6954
Universidad Politécnica Estatal del Carchi
Ecuador - Tulcán
ArtÃculo recibido: 10 junio 2026- Aceptado para publicación:16 julio 2026
Conflictos de intereses: Ninguno que declarar.
RESUMEN
Más de la mitad de los estudiantes de noveno grado en la escuela pública Sagrado Corazón de
Jesús llegaron al inicio del año con puntajes de expresión oral por debajo del nivel de aprobación
institucional. Este estudio evaluó seis componentes del habla por separado, fluidez,
pronunciación, vocabulario, gramática, cohesión e interacción, en 63 estudiantes, identificó
cohesión y vocabulario como los más débiles, y construyó a partir de ese diagnóstico una
estrategia gamificada sin tecnologÃa llamada Oral Story Circle. Tras cuatro semanas de
intervención, la prueba t pareada rechazó la hipótesis nula (p < .001, d_z = 3.58). Todos los
participantes mejoraron; la tasa de reprobación cayó del 57.1% al 4.8%. Las ganancias más
grandes se registraron en cohesión y vocabulario, los mismos componentes que la intervención
apuntó explÃcitamente. El Oral Story Circle ofrece una respuesta replicable y de bajo costo a los
déficits de expresión oral documentados en aulas de EFL de secundaria en Ecuador.
Palabras clave: gamificación, expresión oral en inglés, CEFR A2, educación básica
general, cohesión discursiva
Vol. 13/ Núm. 3 2026 pág. 234
ABSTRACT
Over half the ninth-grade students at the Sagrado Corazón de Jesús public school began the year
scoring below the institutional pass mark in spoken English. This study scored six speaking
components separately —fluency, pronunciation, vocabulary, grammar, cohesion, and
interaction— in 63 students, identified cohesion and vocabulary as the weakest, and built a low-
tech gamified strategy, the Oral Story Circle, around those specific deficits. After four weeks, a
paired-samples t-test rejected the null hypothesis (p < .001, d_z = 3.58): every participant
improved, the failure rate dropped from 57.1% to 4.8%, and the largest gains were in the two
components the design had targeted. The Oral Story Circle offers a replicable, infrastructure-
independent response to the speaking deficits documented in Ecuadorian secondary EFL
classrooms.
Keywords: gamification, EFL oral proficiency, CEFR A2, secondary education,
discourse cohesion
Todo el contenido de la Revista CientÃfica Internacional Arandu UTIC publicado en este sitio está disponible bajo
licencia Creative Commons Atribution 4.0 International.
ABSTRACT
Over half the ninth-grade students at the Sagrado Corazón de Jesús public school began the year
scoring below the institutional pass mark in spoken English. This study scored six speaking
components separately —fluency, pronunciation, vocabulary, grammar, cohesion, and
interaction— in 63 students, identified cohesion and vocabulary as the weakest, and built a low-
tech gamified strategy, the Oral Story Circle, around those specific deficits. After four weeks, a
paired-samples t-test rejected the null hypothesis (p < .001, d_z = 3.58): every participant
improved, the failure rate dropped from 57.1% to 4.8%, and the largest gains were in the two
components the design had targeted. The Oral Story Circle offers a replicable, infrastructure-
independent response to the speaking deficits documented in Ecuadorian secondary EFL
classrooms.
Keywords: gamification, EFL oral proficiency, CEFR A2, secondary education,
discourse cohesion
Todo el contenido de la Revista CientÃfica Internacional Arandu UTIC publicado en este sitio está disponible bajo
licencia Creative Commons Atribution 4.0 International.
Vol. 13/ Núm. 3 2026 pág. 235
INTRODUCTION
Speaking a foreign language is one of the harder skills to teach in secondary school, and
English as a Foreign Language (EFL) classrooms in Ecuador are a clear example. The national
curriculum sets A2.2 on the CEFR scale as the target for ninth grade (Ministerio de Educación,
2024). Most students do not reach it. In a typical class you see the same pattern: one-word
answers, long pauses, and a quiet majority who would rather not speak at all.
Low motivation and limited exposure explain part of this, but not most of it. Work done
in comparable Latin American settings (Morales, 2025; Zambrano & Méndez, 2025) keeps
pointing at the same three things: a thin active vocabulary, almost no use of discourse connectors,
and the fluency breakdown that happens when a student cannot retrieve a word fast enough or
link two ideas. Each of these makes the others worse. A student with too few words runs out
quickly; without connectors, what comes out is closer to a list than a turn; and after a couple of
stalled attempts in front of the class, staying quiet starts to feel like the rational choice.
Gamification has been one of the responses studied in this kind of situation. Meta-
analyses (Hamari, Koivisto, & Sarsa, 2016; Sailer & Homner, 2020) report that points, levels,
badges, and quick feedback can raise engagement and lower anxiety, and there is more specific
evidence in EFL speaking suggesting gains in willingness to communicate and oral output
(Dehghanzadeh et al., 2021; Zainuddin, Chu, Shujahat, & Perera, 2020). The literature has two
gaps that matter here, though. Most of the better-known studies rely on digital platforms that
assume stable internet, which is not given in Ecuadorian public schools. And very few of them
start by diagnosing which speaking sub-skills are weakest in the current group of students and
then build the mechanics around those weaknesses, instead of dropping a generic gamified app
into the class.
The starting point for this study was less theoretical than that. After several years of
teaching ninth grade, it had become obvious that conventional methods were not reaching a
sizable group of students, especially those who had already decided early on that English speaking
was not for them. What this intervention tries to narrow is the everyday gap between recognizing
vocabulary on the page and producing it out loud while peers are listening.
Three objectives organize the work. The first is a skill-based baseline of speaking
proficiency in 63 ninth-grade students at the Sagrado Corazón de Jesús public school, scored on
six components separately (fluency, pronunciation, vocabulary, grammar, cohesion, interaction)
rather than collapsed into a single mark, so the weakest sub-skills can be identified and prioritized.
The second is the design and implementation of a gamified strategy, Oral Story Circle, which
runs offline, uses small student circles, and requires participants to use specific connectors and
vocabulary items under mild time pressure. The third is the evaluation: pretest-posttest
INTRODUCTION
Speaking a foreign language is one of the harder skills to teach in secondary school, and
English as a Foreign Language (EFL) classrooms in Ecuador are a clear example. The national
curriculum sets A2.2 on the CEFR scale as the target for ninth grade (Ministerio de Educación,
2024). Most students do not reach it. In a typical class you see the same pattern: one-word
answers, long pauses, and a quiet majority who would rather not speak at all.
Low motivation and limited exposure explain part of this, but not most of it. Work done
in comparable Latin American settings (Morales, 2025; Zambrano & Méndez, 2025) keeps
pointing at the same three things: a thin active vocabulary, almost no use of discourse connectors,
and the fluency breakdown that happens when a student cannot retrieve a word fast enough or
link two ideas. Each of these makes the others worse. A student with too few words runs out
quickly; without connectors, what comes out is closer to a list than a turn; and after a couple of
stalled attempts in front of the class, staying quiet starts to feel like the rational choice.
Gamification has been one of the responses studied in this kind of situation. Meta-
analyses (Hamari, Koivisto, & Sarsa, 2016; Sailer & Homner, 2020) report that points, levels,
badges, and quick feedback can raise engagement and lower anxiety, and there is more specific
evidence in EFL speaking suggesting gains in willingness to communicate and oral output
(Dehghanzadeh et al., 2021; Zainuddin, Chu, Shujahat, & Perera, 2020). The literature has two
gaps that matter here, though. Most of the better-known studies rely on digital platforms that
assume stable internet, which is not given in Ecuadorian public schools. And very few of them
start by diagnosing which speaking sub-skills are weakest in the current group of students and
then build the mechanics around those weaknesses, instead of dropping a generic gamified app
into the class.
The starting point for this study was less theoretical than that. After several years of
teaching ninth grade, it had become obvious that conventional methods were not reaching a
sizable group of students, especially those who had already decided early on that English speaking
was not for them. What this intervention tries to narrow is the everyday gap between recognizing
vocabulary on the page and producing it out loud while peers are listening.
Three objectives organize the work. The first is a skill-based baseline of speaking
proficiency in 63 ninth-grade students at the Sagrado Corazón de Jesús public school, scored on
six components separately (fluency, pronunciation, vocabulary, grammar, cohesion, interaction)
rather than collapsed into a single mark, so the weakest sub-skills can be identified and prioritized.
The second is the design and implementation of a gamified strategy, Oral Story Circle, which
runs offline, uses small student circles, and requires participants to use specific connectors and
vocabulary items under mild time pressure. The third is the evaluation: pretest-posttest
Vol. 13/ Núm. 3 2026 pág. 236
comparison, gains broken down by skill and by group, and effect sizes reported next to p-values
so practical and statistical significance can be read together.
The aim is not to add another general claim that gamification helps speaking. It is to test
whether a diagnostic-driven, low-tech design that fits the actual constraints of an Ecuadorian
public classroom can move both the scores and, eventually, how students relate to speaking tasks.
Given that English proficiency still works as a filter for university admission and for a fair share
of jobs in the country, the question is worth asking carefully. The methodology, results, and
discussion below follow the three objectives in order.
MATERIALS AND METHODS
Research design
The study adopted a quantitative approach and used a pre-experimental pretest–
intervention–posttest design (O¹ → X → O²) with two intact ninth-grade groups from the Sagrado
Corazón de Jesús public school. The scope is correlational-descriptive: the analysis examines
associations between the intervention and changes in speaking sub-skills without random
assignment or a control group. Group A comprised 31 students and Group B comprised 32, Figure
1.
Participants
All 63 participants were ninth-grade students at the same institution who had received at
least three years of English instruction at three hours per week. Students with diagnosed speech
or language disorders and those absent from more than 20% of sessions were excluded. Written
informed consent was obtained from parents or guardians, and student assent was collected before
data gathering began. The school is identified by a placeholder to protect confidentiality.
comparison, gains broken down by skill and by group, and effect sizes reported next to p-values
so practical and statistical significance can be read together.
The aim is not to add another general claim that gamification helps speaking. It is to test
whether a diagnostic-driven, low-tech design that fits the actual constraints of an Ecuadorian
public classroom can move both the scores and, eventually, how students relate to speaking tasks.
Given that English proficiency still works as a filter for university admission and for a fair share
of jobs in the country, the question is worth asking carefully. The methodology, results, and
discussion below follow the three objectives in order.
MATERIALS AND METHODS
Research design
The study adopted a quantitative approach and used a pre-experimental pretest–
intervention–posttest design (O¹ → X → O²) with two intact ninth-grade groups from the Sagrado
Corazón de Jesús public school. The scope is correlational-descriptive: the analysis examines
associations between the intervention and changes in speaking sub-skills without random
assignment or a control group. Group A comprised 31 students and Group B comprised 32, Figure
1.
Participants
All 63 participants were ninth-grade students at the same institution who had received at
least three years of English instruction at three hours per week. Students with diagnosed speech
or language disorders and those absent from more than 20% of sessions were excluded. Written
informed consent was obtained from parents or guardians, and student assent was collected before
data gathering began. The school is identified by a placeholder to protect confidentiality.
Vol. 13/ Núm. 3 2026 pág. 237
Figure 1
Pre-Experimental Research Design and Outcome Summary (n = 63)
Instrument and scoring rubric
Both pretest and posttest used the same individually administered oral production test,
lasting five to seven minutes per student and consisting of four tasks: a guided monologue
(describe your best friend), a picture description of an everyday scene, a role-play (borrow an
object from a classmate), and oral reading of two sentences targeting /θ/, /ð/, and /ʃ/.
Six analytical components were scored on a 0–10 scale: fluency, pronunciation,
vocabulary, grammar, cohesion, and interaction. The global score was the mean of these six
components. Descriptors were adapted from Ministerio de Educación del Ecuador (2024)
guidelines and the CEFR A2.2 level.
Two trained raters scored 20% of the tests independently at both time points. Inter-rater
reliability was calculated as weighted Cohen’s kappa with quadratic weights. Intra-rater reliability
for the primary researcher was estimated on a random 10% of the sample scored twice at a two-
week interval.
Diagnostic use of pretest results
Pretest scores were examined component by component to identify which sub-skills were
weakest before the intervention was built. Cohesion and vocabulary scored lowest across both
groups. The game mechanics were therefore designed to require connector use on every turn and
to reward productive vocabulary deployment, rather than addressing speaking generically. Full
pretest findings are reported in the Results section.
Figure 1
Pre-Experimental Research Design and Outcome Summary (n = 63)
Instrument and scoring rubric
Both pretest and posttest used the same individually administered oral production test,
lasting five to seven minutes per student and consisting of four tasks: a guided monologue
(describe your best friend), a picture description of an everyday scene, a role-play (borrow an
object from a classmate), and oral reading of two sentences targeting /θ/, /ð/, and /ʃ/.
Six analytical components were scored on a 0–10 scale: fluency, pronunciation,
vocabulary, grammar, cohesion, and interaction. The global score was the mean of these six
components. Descriptors were adapted from Ministerio de Educación del Ecuador (2024)
guidelines and the CEFR A2.2 level.
Two trained raters scored 20% of the tests independently at both time points. Inter-rater
reliability was calculated as weighted Cohen’s kappa with quadratic weights. Intra-rater reliability
for the primary researcher was estimated on a random 10% of the sample scored twice at a two-
week interval.
Diagnostic use of pretest results
Pretest scores were examined component by component to identify which sub-skills were
weakest before the intervention was built. Cohesion and vocabulary scored lowest across both
groups. The game mechanics were therefore designed to require connector use on every turn and
to reward productive vocabulary deployment, rather than addressing speaking generically. Full
pretest findings are reported in the Results section.
Vol. 13/ Núm. 3 2026 pág. 238
Gamified intervention: Oral Story Circle
The Oral Story Circle was designed as a structured oral production task with an embedded
experience-point (XP) system. The core logic was deliberate: every mandatory constraint in the
game corresponded to a diagnosed deficit, and every reward was tied to productive language
behavior rather than to participation alone. The design was also low-tech by choice, requiring no
internet access and no devices beyond printed cards and an XP tracking sheet, in direct response
to the infrastructure constraints documented in Ecuadorian public schools (Bae et al., 2023).
Students worked in circles of eight to ten, passing a soft ball at random. The teacher
provided the opening sentence for each round. On each turn, the student holding the ball was
required to: (a) repeat the previous sentence (+5 XP), (b) use a connector (+10 XP), (c) include a
vocabulary item from the weekly theme (+10 XP), and (d) complete the turn within five seconds,
shown on a visible timer (+5 XP). The four requirements were sequential and non-negotiable; XP
was awarded only for requirements actually met.
Two bonus criteria were available per turn. Using a connector different from either of the
previous two turns earned +15 XP, directly targeting the narrow-repertoire problem documented
in EFL learners’ discourse-marker use. Using a vocabulary item not pre-taught in class earned
+20 XP, rewarding lexical initiative beyond the taught syllabus. The bonus values were
deliberately set above the mandatory ones to make variety and autonomy the most rational
strategy once a student had the basics.
After each round, accumulated XP placed students at one of five levels: Novice (0–199),
Explorer (200–399), Ninja (400–599), Epic (600–799), and Legend (800+). Three badges were
awarded for specific milestones: Connector Ninja (five different connectors in one round), Master
Builder (collaborative story reaching eight sentences), and Linguistic Legend (Level 5). Badges
carried no academic grade but granted small classroom privileges. The combination of points,
levels, and badges follows the gamification elements with the strongest empirical support in EFL
speaking contexts (Zhang & Hasim, 2023).
The five-second window was a deliberate fluency constraint: forced rapid retrieval pushes
learners toward proceduralization of formulaic chunks rather than effortful sentence construction
(De Jong, 2018). The ball-passing protocol distributes production risk across all students rather
than concentrating exposure in those who would otherwise volunteer. The collaborative
storytelling format provides contextualized, low-anxiety oral practice consistent with the
affective-filter reduction mechanisms described in structured oral task research (Maher & King,
2023; Tajik, 2025).
Materials consisted of one soft ball per circle, five printed thematic vocabulary card sets
(20 words each: school, family, food, city, free time), printed picture-sequence cards, individual
XP tracking sheets, and a classroom progress chart. Digital backup tools (Genially and Wordwall)
were used only when stable internet was available.
Gamified intervention: Oral Story Circle
The Oral Story Circle was designed as a structured oral production task with an embedded
experience-point (XP) system. The core logic was deliberate: every mandatory constraint in the
game corresponded to a diagnosed deficit, and every reward was tied to productive language
behavior rather than to participation alone. The design was also low-tech by choice, requiring no
internet access and no devices beyond printed cards and an XP tracking sheet, in direct response
to the infrastructure constraints documented in Ecuadorian public schools (Bae et al., 2023).
Students worked in circles of eight to ten, passing a soft ball at random. The teacher
provided the opening sentence for each round. On each turn, the student holding the ball was
required to: (a) repeat the previous sentence (+5 XP), (b) use a connector (+10 XP), (c) include a
vocabulary item from the weekly theme (+10 XP), and (d) complete the turn within five seconds,
shown on a visible timer (+5 XP). The four requirements were sequential and non-negotiable; XP
was awarded only for requirements actually met.
Two bonus criteria were available per turn. Using a connector different from either of the
previous two turns earned +15 XP, directly targeting the narrow-repertoire problem documented
in EFL learners’ discourse-marker use. Using a vocabulary item not pre-taught in class earned
+20 XP, rewarding lexical initiative beyond the taught syllabus. The bonus values were
deliberately set above the mandatory ones to make variety and autonomy the most rational
strategy once a student had the basics.
After each round, accumulated XP placed students at one of five levels: Novice (0–199),
Explorer (200–399), Ninja (400–599), Epic (600–799), and Legend (800+). Three badges were
awarded for specific milestones: Connector Ninja (five different connectors in one round), Master
Builder (collaborative story reaching eight sentences), and Linguistic Legend (Level 5). Badges
carried no academic grade but granted small classroom privileges. The combination of points,
levels, and badges follows the gamification elements with the strongest empirical support in EFL
speaking contexts (Zhang & Hasim, 2023).
The five-second window was a deliberate fluency constraint: forced rapid retrieval pushes
learners toward proceduralization of formulaic chunks rather than effortful sentence construction
(De Jong, 2018). The ball-passing protocol distributes production risk across all students rather
than concentrating exposure in those who would otherwise volunteer. The collaborative
storytelling format provides contextualized, low-anxiety oral practice consistent with the
affective-filter reduction mechanisms described in structured oral task research (Maher & King,
2023; Tajik, 2025).
Materials consisted of one soft ball per circle, five printed thematic vocabulary card sets
(20 words each: school, family, food, city, free time), printed picture-sequence cards, individual
XP tracking sheets, and a classroom progress chart. Digital backup tools (Genially and Wordwall)
were used only when stable internet was available.
Vol. 13/ Núm. 3 2026 pág. 239
The intervention ran over four weeks at two 45-minute sessions per week, totaling 12
contact hours. Week 1 used vocabulary games (Bingo, Password) to establish the lexical base.
Week 2 introduced connectors through sequencing activities. Week 3 combined both components
without time pressure. Week 4 added the five-second rule, activated all bonus criteria, and closed
with a competitive round (Boss Battle). The researcher facilitated all sessions; the regular English
teacher co-facilitated throughout.
Procedure
In Week 1, the pretest was administered individually in a quiet room adjacent to the regular
classroom. All responses were audio-recorded. Weeks 2 through 5 comprised the intervention.
Each session followed a fixed structure: five-minute warm-up, five minutes of instruction and
modeling, 25–30 minutes of game play with the researcher rotating between circles at
approximately three minutes per circle, five to ten minutes for XP updating and feedback, and a
five-minute cool-down. In Week 6, the posttest was administered under conditions identical to
the pretest, by the same researcher, in the same room and time window. The researcher remained
blind to pretest scores during posttest administration and scoring.
Data Analysis
All analyses were conducted in R (R Core Team, 2024). Normality was assessed with the
Shapiro–Wilk test and visual inspection of Q–Q plots. Because raw pretest scores departed from
normality, nonparametric tests were used for between-group baseline comparisons (Mann–
Whitney U, coin package) and within-group pre–post comparisons (Wilcoxon signed-rank, stats
package). Effect sizes for paired comparisons were calculated as dₓ (Cohen’s d for paired samples;
Lakens, 2013) using the effect size package.
The primary inferential model was a linear mixed-effects model (LMM) fitted with lme4
(Bates et al., 2015): Global ~ Time × Group + (1 | student_id), with Pretest and Group A as
reference levels. p-values were obtained via Satterthwaite degrees of freedom using lmerTest
(Kuznetsova et al., 2017). Estimated marginal means and pairwise contrasts were computed with
emmeans (Lenth, 2024). Because multiple paired comparisons were performed simultaneously
(six components × two groups), p-values were adjusted with the Benjamini–Hochberg false
discovery rate correction (Benjamini & Hochberg, 1995).
For practical educational interpretation, individual scores were mapped onto the four bands
of the Ecuadorian grading scale: Failing (<7.0), Passing (7.0–7.9), Good (8.0–8.9), and Excellent
(≥9.0). Percentage-point shifts across bands are reported as descriptive measures of redistribution.
RESULTS AND DISCUSSION
Baseline speaking proficiency
Sixty-three ninth-grade students at the Sagrado Corazón de Jesús public school were
assessed on the institutional 0–10 oral rubric, which scores six analytical dimensions
The intervention ran over four weeks at two 45-minute sessions per week, totaling 12
contact hours. Week 1 used vocabulary games (Bingo, Password) to establish the lexical base.
Week 2 introduced connectors through sequencing activities. Week 3 combined both components
without time pressure. Week 4 added the five-second rule, activated all bonus criteria, and closed
with a competitive round (Boss Battle). The researcher facilitated all sessions; the regular English
teacher co-facilitated throughout.
Procedure
In Week 1, the pretest was administered individually in a quiet room adjacent to the regular
classroom. All responses were audio-recorded. Weeks 2 through 5 comprised the intervention.
Each session followed a fixed structure: five-minute warm-up, five minutes of instruction and
modeling, 25–30 minutes of game play with the researcher rotating between circles at
approximately three minutes per circle, five to ten minutes for XP updating and feedback, and a
five-minute cool-down. In Week 6, the posttest was administered under conditions identical to
the pretest, by the same researcher, in the same room and time window. The researcher remained
blind to pretest scores during posttest administration and scoring.
Data Analysis
All analyses were conducted in R (R Core Team, 2024). Normality was assessed with the
Shapiro–Wilk test and visual inspection of Q–Q plots. Because raw pretest scores departed from
normality, nonparametric tests were used for between-group baseline comparisons (Mann–
Whitney U, coin package) and within-group pre–post comparisons (Wilcoxon signed-rank, stats
package). Effect sizes for paired comparisons were calculated as dₓ (Cohen’s d for paired samples;
Lakens, 2013) using the effect size package.
The primary inferential model was a linear mixed-effects model (LMM) fitted with lme4
(Bates et al., 2015): Global ~ Time × Group + (1 | student_id), with Pretest and Group A as
reference levels. p-values were obtained via Satterthwaite degrees of freedom using lmerTest
(Kuznetsova et al., 2017). Estimated marginal means and pairwise contrasts were computed with
emmeans (Lenth, 2024). Because multiple paired comparisons were performed simultaneously
(six components × two groups), p-values were adjusted with the Benjamini–Hochberg false
discovery rate correction (Benjamini & Hochberg, 1995).
For practical educational interpretation, individual scores were mapped onto the four bands
of the Ecuadorian grading scale: Failing (<7.0), Passing (7.0–7.9), Good (8.0–8.9), and Excellent
(≥9.0). Percentage-point shifts across bands are reported as descriptive measures of redistribution.
RESULTS AND DISCUSSION
Baseline speaking proficiency
Sixty-three ninth-grade students at the Sagrado Corazón de Jesús public school were
assessed on the institutional 0–10 oral rubric, which scores six analytical dimensions
Vol. 13/ Núm. 3 2026 pág. 240
independently: pronunciation, fluency, grammar, vocabulary, interaction, and cohesion. The
global score is the mean of these six components. Component-level scoring, rather than a single
composite, makes it possible to identify which sub-skills are weakest and to interpret post-
intervention gains at the same level of granularity (Saito & Plonsky, 2019). These baseline
numbers translate directly into classroom reality. At the start of the intervention, most students
produced short, disconnected utterances, relied on a narrow set of memorized phrases, and
avoided extended turns altogether. The low cohesion scores reflected the absence of linking
language between ideas; the low vocabulary scores reflected the gap between words students
could recognize on the page and words they could retrieve under the mild social pressure of
speaking in front of peers. Identifying these two dimensions as the priority targets was not only a
statistical decision but a pedagogical one: the intervention needed to give students a structured
reason to use connectors and new words repeatedly, in a low-stakes context, before those forms
could become reliable oral resources.
Between-group differences at baseline
Shapiro–Wilk tests indicated departures from normality in both groups (p < .05), and
group sizes were small and unequal. Between-group differences were therefore tested with the
Mann–Whitney U test, the standard nonparametric alternative to the independent-samples t-test
when normality cannot be assumed (Field, 2024; Lakens, 2022). Figure 2 displays the distribution
of baseline global scores. Table 1 reports the corresponding descriptive statistics.
Figure 2
Baseline Speaking Proficiency Distribution by Group (n = 63)
Note. Boxplots show median (center line), interquartile range (box), and minimum–maximum values (whiskers).
Individual student scores are overlaid as points. Group A (n = 31, M = 6.44, SD = 0.65) scored significantly lower than
Group B (n = 32, M = 7.15, SD = 0.49) at baseline (Mann-Whitney U = 192, p < .001). Scores on a 0–10 scale.
independently: pronunciation, fluency, grammar, vocabulary, interaction, and cohesion. The
global score is the mean of these six components. Component-level scoring, rather than a single
composite, makes it possible to identify which sub-skills are weakest and to interpret post-
intervention gains at the same level of granularity (Saito & Plonsky, 2019). These baseline
numbers translate directly into classroom reality. At the start of the intervention, most students
produced short, disconnected utterances, relied on a narrow set of memorized phrases, and
avoided extended turns altogether. The low cohesion scores reflected the absence of linking
language between ideas; the low vocabulary scores reflected the gap between words students
could recognize on the page and words they could retrieve under the mild social pressure of
speaking in front of peers. Identifying these two dimensions as the priority targets was not only a
statistical decision but a pedagogical one: the intervention needed to give students a structured
reason to use connectors and new words repeatedly, in a low-stakes context, before those forms
could become reliable oral resources.
Between-group differences at baseline
Shapiro–Wilk tests indicated departures from normality in both groups (p < .05), and
group sizes were small and unequal. Between-group differences were therefore tested with the
Mann–Whitney U test, the standard nonparametric alternative to the independent-samples t-test
when normality cannot be assumed (Field, 2024; Lakens, 2022). Figure 2 displays the distribution
of baseline global scores. Table 1 reports the corresponding descriptive statistics.
Figure 2
Baseline Speaking Proficiency Distribution by Group (n = 63)
Note. Boxplots show median (center line), interquartile range (box), and minimum–maximum values (whiskers).
Individual student scores are overlaid as points. Group A (n = 31, M = 6.44, SD = 0.65) scored significantly lower than
Group B (n = 32, M = 7.15, SD = 0.49) at baseline (Mann-Whitney U = 192, p < .001). Scores on a 0–10 scale.
Vol. 13/ Núm. 3 2026 pág. 241
Table 1
Baseline speaking proficiency by group
Group n M (SD) Range Failing <7.0, n
(%)
Group A 31 6.44 (0.65) 5.2–7.8 20 (64.5%)
Group B 32 7.15 (0.49) 6.2–8.2 16 (50.0%)
Total 63 6.80 (0.67) 5.2–8.2 36 (57.1%)
Note. Scores on a 0–10 scale. Between-group difference: Mann-Whitney U = 192, p < .001. Failing = Global score <
7.0 (Ecuadorian grading scale).
Group A (M = 6.44, SD = 0.65) scored significantly lower than Group B (M = 7.15, SD
= 0.49), U = 192, p < .001. The 0.71-point gap was accompanied by a high concentration of failing
students (<7.0 on the Ecuadorian grading scale): 64.5% in Group A and 50.0% in Group B,
totaling 36 students (57.1%) of the sample.
Skill-level diagnosis
Component-level inspection identified cohesion and vocabulary as the two weakest
dimensions across both groups, with pronunciation showing the smallest baseline deficit (Figure
3). This pattern is consistent with regional reports placing Ecuadorian learners among the lowest-
ranked English speakers in Latin America (EF EPI, 2023) and with the broader Latin American
secondary EFL literature, where oral production lags behind receptive skills due to limited
speaking practice in large, mixed-ability classrooms (Cantos et al., 2024).
Treating speaking as a unitary construct would have masked these asymmetries. Saito
and Plonsky (2019) argue that sub-skills follow different acquisition trajectories and respond to
different instructional inputs, so component-level diagnosis is a precondition for targeted
intervention. The cohesion deficit specifically reflects a well-documented limitation in EFL
learners' use of discourse markers: Ebrahimi and Xodabande (2023) and Guba et al. (2023) report
that learners typically rely on a narrow set of basic connectors until explicit instruction broadens
their repertoire. Identifying cohesion and vocabulary as the priority targets directly shaped the
rule structure of the intervention described in the next section.
Figure 3
Baseline speaking proficiency by sub-skill and group (n = 63)
Note. Bars represent pretest means for each rubric sub-skill, ordered from lowest to highest total-sample mean. Error
bars indicate 95% confidence intervals. The shaded region marks the failing band (<7.0) on the Ecuadorian grading
scale. Group B scored consistently above Group A across all six sub-skills.
Table 1
Baseline speaking proficiency by group
Group n M (SD) Range Failing <7.0, n
(%)
Group A 31 6.44 (0.65) 5.2–7.8 20 (64.5%)
Group B 32 7.15 (0.49) 6.2–8.2 16 (50.0%)
Total 63 6.80 (0.67) 5.2–8.2 36 (57.1%)
Note. Scores on a 0–10 scale. Between-group difference: Mann-Whitney U = 192, p < .001. Failing = Global score <
7.0 (Ecuadorian grading scale).
Group A (M = 6.44, SD = 0.65) scored significantly lower than Group B (M = 7.15, SD
= 0.49), U = 192, p < .001. The 0.71-point gap was accompanied by a high concentration of failing
students (<7.0 on the Ecuadorian grading scale): 64.5% in Group A and 50.0% in Group B,
totaling 36 students (57.1%) of the sample.
Skill-level diagnosis
Component-level inspection identified cohesion and vocabulary as the two weakest
dimensions across both groups, with pronunciation showing the smallest baseline deficit (Figure
3). This pattern is consistent with regional reports placing Ecuadorian learners among the lowest-
ranked English speakers in Latin America (EF EPI, 2023) and with the broader Latin American
secondary EFL literature, where oral production lags behind receptive skills due to limited
speaking practice in large, mixed-ability classrooms (Cantos et al., 2024).
Treating speaking as a unitary construct would have masked these asymmetries. Saito
and Plonsky (2019) argue that sub-skills follow different acquisition trajectories and respond to
different instructional inputs, so component-level diagnosis is a precondition for targeted
intervention. The cohesion deficit specifically reflects a well-documented limitation in EFL
learners' use of discourse markers: Ebrahimi and Xodabande (2023) and Guba et al. (2023) report
that learners typically rely on a narrow set of basic connectors until explicit instruction broadens
their repertoire. Identifying cohesion and vocabulary as the priority targets directly shaped the
rule structure of the intervention described in the next section.
Figure 3
Baseline speaking proficiency by sub-skill and group (n = 63)
Note. Bars represent pretest means for each rubric sub-skill, ordered from lowest to highest total-sample mean. Error
bars indicate 95% confidence intervals. The shaded region marks the failing band (<7.0) on the Ecuadorian grading
scale. Group B scored consistently above Group A across all six sub-skills.
Vol. 13/ Núm. 3 2026 pág. 242
The baseline asymmetry between groups also constrained the analytical strategy: post hoc
matching on baseline proficiency was not possible, so within-group rather than purely between-
group comparisons were prioritized for the post-intervention analysis.
The oral story circle intervention. Game mechanics and scoring system
Figure 4 presents the complete game cycle, including the four mandatory rules per turn,
the bonus criteria, and the post-round XP system that maps onto five proficiency levels and four
achievement badges.
Each student turn requires four actions in sequence: repeating the previous sentence (+5
XP), using a connector (+10 XP), including a target vocabulary item (+10 XP), and producing all
of this within a 5-second window (+5 XP). Two bonus criteria reward variety (+15 XP for using
a connector different from the previous turn) and lexical autonomy (+20 XP for introducing a
vocabulary item not pre-taught). After each round, accumulated XP determines progression
through Levels 1–5 and unlocks the four achievement badges. A 45-minute session contains two
to three rounds, with the teacher rotating between circles of 8–10 students every three minutes.
The format is deliberately low-tech: it requires no internet access, no devices, and only
basic materials (a ball, vocabulary cards, a printed XP tracker). This responds directly to
documented infrastructure constraints in Ecuadorian public schools and avoids the technology
dependence that Bae et al. (2023) found correlates with smaller gamification effects.
The baseline asymmetry between groups also constrained the analytical strategy: post hoc
matching on baseline proficiency was not possible, so within-group rather than purely between-
group comparisons were prioritized for the post-intervention analysis.
The oral story circle intervention. Game mechanics and scoring system
Figure 4 presents the complete game cycle, including the four mandatory rules per turn,
the bonus criteria, and the post-round XP system that maps onto five proficiency levels and four
achievement badges.
Each student turn requires four actions in sequence: repeating the previous sentence (+5
XP), using a connector (+10 XP), including a target vocabulary item (+10 XP), and producing all
of this within a 5-second window (+5 XP). Two bonus criteria reward variety (+15 XP for using
a connector different from the previous turn) and lexical autonomy (+20 XP for introducing a
vocabulary item not pre-taught). After each round, accumulated XP determines progression
through Levels 1–5 and unlocks the four achievement badges. A 45-minute session contains two
to three rounds, with the teacher rotating between circles of 8–10 students every three minutes.
The format is deliberately low-tech: it requires no internet access, no devices, and only
basic materials (a ball, vocabulary cards, a printed XP tracker). This responds directly to
documented infrastructure constraints in Ecuadorian public schools and avoids the technology
dependence that Bae et al. (2023) found correlates with smaller gamification effects.