📄 Pronunciation of English Words with /th/ Sounds among Senior High School Learners

Summary

The study examines how Filipino senior high school learners pronounce English words containing the two dental fricatives: voiceless /θ/ (as in thing) and voiced /ð/ (as in this). Neither sound exists in Filipino or other Philippine languages, and mispronunciations can change meaning, for example they heard as day. The author surveyed an entire Grade 11 HUMSS class of 32 students at a small rural school in Zamboanga City. The students spoke three home languages: Chavacano, Tausug and Visayan. The design was cross-sectional with total enumeration. Students practised with a word sheet and then took minimal-pair read-aloud tests. A teacher scored them with a rubric, and phone audio recordings were used to recheck scores. Analysis used percentage frequency distributions. Most learners handled /θ/ well but struggled with /ð/. Tausug and Visayan speakers did better on /θ/, while Chavacano speakers did better on /ð/. The author explains the errors as interlingual transfer from the mother tongue (Richards, 1974), worsened by little everyday exposure to English. The work matters for pronunciation teaching in multilingual rural schools.

Key points

• Participants were 32 Grade 11 learners (14 male, 18 female): 13 Chavacano, 13 Tausug and 6 Visayan speakers.
• Instruments:
  • a modified School Form 1 recording each learner's background and L1;
  • a practice sheet drawn from the Phoenix Language Package textbook;
  • minimal-pair read-aloud tests (Set A and Set B);
  • a scoring rubric, with mobile-phone recordings used for rechecking.
• Ratings were banded by error count: excellent at roughly 0–3 errors, average at 4–6, poor at 7–9.
• Results for /ð/ were 31.3% excellent, 28.1% average and 40.6% poor. For /θ/, most learners were rated excellent, but 11 of 32 (34.3%) were poor.
• The author attributes errors to mother-tongue transfer and to a local environment where Chavacano, Tausug, Visayan or Filipino dominate daily talk.

Interesting ideas

• I1. Minimal pairs isolate one contrast so that any error points to a single cause, which makes the test diagnostic rather than just a total score.
• I2. Scores are broken down by the learner's L1 group rather than reported only as one overall figure, which exposes subgroup-specific weaknesses that an average would hide.
• I3. Raw recordings were kept so that the rubric judgment could be rechecked before analysis, which makes the scoring auditable.
• I4. Raw error counts are mapped to coarse bands (excellent, average, poor), giving an interpretable summary while the counts remain available.
• I5. The paper traces a systematic substitution (/ð/ produced as /d/, they → day) to L1 transfer rather than to random noise.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (learn): Build minimal-pair sentence sets that differ in one feature to act as negative controls. A parser or frame step that cannot tell the pair apart has a precisely located weakness.
• I2 (learn): Report evaluation per source or text type (fables, maths problems, school tests), so a regression on one slice is not masked by gains elsewhere.
• I3 (answer): Store the evidence behind each graded answer, such as the tree, frames and matched facts, so a human or second pass can re-verify the verdict.
• I4 (answer): Grade student-style responses on school-course tests with explicit error-count bands, keeping the band explainable by listing the counted errors.
• I5 (tokens): Treat predictable L2-English substitutions such as th→d spellings as known variant patterns in normalization. This lets learner-written text be lemmatized correctly without guessing.