📄 Competency Tracking for English as a Second or Foreign Language Learners

Summary

The paper tackles a practical gap in English language teaching. Instructors rarely know which skills their students already learned elsewhere, so classes often repeat material or skip over hidden gaps. Bishop proposes a competency-tracking setup built only from stock Moodle features. Every CEFR grammar point and language function for English becomes a Moodle "outcome." Each outcome is rated on a five-step scale adapted from the Interagency Language Roundtable descriptors. One Moodle "course" per CEFR level serves as a gradebook rather than a real class. Students, teachers, parents and supervisors can all see the same record. Learners are first placed with an external test or an in-house test, then probed with per-grammar-point question banks. Functions are judged in person against CEFR rubrics. A backup-and-restore workaround lets an individual record be exported, carried to another school and merged into a new class gradebook. The paper is a system description with a single ongoing pilot and no quantitative evaluation. Its value is a cheap, portable and standardized way to make prior learning visible for curriculum planning.

Key points

• Competencies come from the CEFR for English, grouped into six levels (A1, A2, B1, B2, C1, C2). Each level has roughly 30 grammar points that are deliberately left unordered within the level.
• The rating scale runs from 5 (mastery, native-like ease) through 4 (acceptable), 3 (working), and 2 (limited) to 1 (minimal). A hyphen in Moodle marks a skill that was never taught; it should be treated like a 1 or tested.
• Placement can use IELTS, TOEIC or TOEFL scores mapped to CEFR levels. Question banks for the level below (for example A2 items for a B1 learner) are then used to find gaps.
• The class-level view uses an illustrative cohort of five famous writers:
  • On "should have / might have," three learners had prior exposure and two had none.
  • On cause-and-effect connectors, three scored 3 and two had no record, so the topic was safe to include in the class.
• A single-student record exports as a Moodle backup (.mbz) of about half a megabyte. In Moodle 2.3 it can be merged into an existing course. No plugins are used, and a pilot is running at one small institution where class membership changes weekly.

Interesting ideas

• I1. The scale keeps "never assessed" separate from "assessed and weak," and unassessed skills default to needing a test rather than being assumed known.
• I2. Skills are partially ordered: levels follow a fixed sequence, but items inside a level can be tackled in any order, with completion of the whole level required to advance.
• I3. Diagnosis deliberately probes one level below the placement level to catch gaps that a single overall score hides.
• I4. A shared, standardized inventory of named grammar points (past modals, cause-effect connectors, embedded questions) serves as the common currency for describing what a learner can handle.
• I5. Records are portable and mergeable because everyone uses the same skill list; local changes go into a separate copy so the shared core stays uniform.

For Mova

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

• I1 (learn): Evaluation reports could keep an explicit "untested" status per construction or question type, so missing coverage is never mistaken for a pass and directs where to build the next test.
• I2 (learn): Self-play and training curricula could group phenomena into levels with free order inside each, unlocking harder text types only when every item of the previous level reaches an acceptable score.
• I3 (answer): When a reader passes school-course tests at some grade, it should also be quizzed on lower-grade items, since an aggregate score can mask specific failures underneath.
• I4 (tree): These CEFR grammar points could be written as explainable UD tree patterns, letting Mova tag which constructions a text uses and estimate its difficulty level.
• I5 (learn): Per-skill scorecards for different parser and reader versions could share one fixed schema, with experimental skills kept in a side layer, so results from different runs can be merged and compared directly.