📄 Developing Cooperative Learning Based E-Module to Teach Basic English Grammar of the First Semester of English Study Program Students at FKIP – UIR

Summary

The paper reports the first, needs-analysis stage of an R&D project (Borg and Gall model) at FKIP-UIR, an Indonesian university. The goal is an e-module for the Basic English Grammar course that blends self-study material with cooperative learning (STAD). The author argues that the existing textbook-and-whiteboard teaching leaves first-semester English students unmotivated and gives them too little cooperative practice. To find out what the module should cover, 30 first-semester students wrote paragraphs, and their grammatical errors were counted and sorted into ten categories, each with finer subtypes. In total 386 errors were found. Verb tense was the largest category, followed by context of meaning, finite verbs and singular/plural agreement. These categories and subtypes became the module's table of contents, and the author gives sample exercises such as combining sentences while shifting tense. The contribution is mainly practical: course material is built directly from an error profile of the actual students rather than from a generic textbook syllabus.

Key points

• Participants were 30 first-semester students of the English Study Program at FKIP-UIR. Data came from a writing test, analysed as the needs-analysis step of the Borg and Gall R&D cycle.
• The 386 errors broke down as follows:
  • verb tense: 106 (27.46%)
  • context of meaning: 55 (14.25%)
  • finite verb: 51 (13.21%)
  • singular/plural agreement: 45 (11.66%)
  • pronoun: 28 (7.25%)
  • non-finite verb: 25 (printed as 2.48%, but 25/386 is about 6.48%)
  • word order: 25 (6.48%)
  • word choice: 22 (5.70%)
  • preposition: 20 (5.18%)
  • context of usage: 9 (2.33%)
• Verb-tense errors included wrong regular and irregular past forms, double past marking and incomplete past progressive. Non-finite errors included a stray -ed, a missing "to" and a wrong -es.
• The module's contents (Table 3) map one-to-one onto the ten error categories, with about four subtopics each.
• The paper has internal inconsistencies:
  • The text says the material has eight chapters, but the table lists ten topics.
  • The conclusion mentions "second semester" students.
  • The conclusion also refers to auxiliaries and passive voice, which are not among the ten counted categories.

Interesting ideas

• I1. The curriculum is derived from a frequency-ranked error taxonomy of real output, so effort goes first to the biggest error classes rather than to a fixed syllabus.
• I2. The taxonomy is two-level: a broad class (verb tense) splits into concrete, checkable subtypes such as double past marking, a missing infinitival "to" or a substituted infinitive for simple past.
• I3. One exercise tells the learner in advance how many errors a paragraph contains and asks them to find all of them.
• I4. Another exercise merges information from a second sentence into the first while converting present tense to past.

For Mova

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

• I1 (learn): rank Mova's parse and frame failures into named categories by frequency, and let that ranking decide which self-induced repair rules or training data to work on first.
• I2 (tokens): the morphology layer could flag these learner-style subtypes explicitly (e.g., "didn't went", missing "to" after a verb), so that noisy or non-native input yields an explainable diagnosis instead of a silent mis-tag.
• I3 (learn): inject a known number of controlled grammatical errors into clean text as a negative control, then measure whether Mova's checks recover exactly that count.
• I4 (back): use this as a generation test, where two clause graphs are merged, the tense is changed on the event node, and the output is checked for correct morphology and agreement.