A 32-Point Verbal-Nonverbal Gap in Grade 4
A Grade 4 team meets in March to decide whether a student, Mira, should enter the Grade 5 high-verbal-ability cluster. Mira has been in an English-medium school for two years and four months; her home language is Ukrainian and her ACCESS level is 3.2. On CogAT Form 8 she scores Verbal 92 (Verbal Analogies 94, Sentence Completion 86, Verbal Classification 97), Quantitative 118, Nonverbal 124, Composite 116. Each battery's score has a standard error of about five points. On a curriculum reading measure she is at grade level for decoding and fluency. Her science teacher describes her oral explanations as precise and well organised, and her writing has accurate content with noticeably limited vocabulary.
The verbal coordinator recommends excluding her because 'the cluster is for verbal strength and 92 is average.' Another teacher wants to re-test next autumn and decide then. The team has 20 minutes and needs a decision it can defend to parents.
1. Is the 32-point gap statistically meaningful, and is it evidence of a verbal reasoning weakness? Keep the two questions separate.
Read the worked answer
Is the gap real? Yes. With a standard error near 5, each 95% band is roughly ±10 points: Verbal 82–102, Nonverbal 114–134. The bands do not overlap, and the standard error of the difference is about √(25 + 25) ≈ 7, so 32 points is roughly 4.5 standard errors. Measurement error is not a plausible account of the gap.
Is it evidence of weak verbal reasoning? Not on this evidence. At Grade 4 the verbal battery is delivered in English and its items turn on knowing words and sentence patterns. For a student with two years and four months of English, the verbal score carries construct-irrelevant variance, and the size of that variance is unknown from the score alone.
Note the internal disagreement as well: 86 to 97 is an 11-point spread across subtests whose standard errors are larger than the battery's. The verbal profile is unstable as well as depressed. Treat the gap as a fact about the scores and a hypothesis about the student.
2. Rank the three verbal subtests by how much their scores are likely to be depressed by English proficiency, and justify the ranking from item demands.
Read the worked answer
All three are English-vocabulary dependent at Grade 4, so this is a hypothesis, not a finding. Expected order of language load, highest first:
- Sentence Completion (86) - the heaviest load. The student must supply the one word that completes a sentence frame, which demands precise vocabulary and often idiom. There is no route around an unknown word.
- Verbal Analogies (94) - a double demand: knowledge of four words plus the relation between them. Strong relational reasoning cannot compensate for an unknown word, but partial knowledge sometimes allows elimination, which usually places it in the middle.
- Verbal Classification (97) - the lightest of the three, because all the words are visible at once and partial meanings plus elimination can carry the item. It is not load-free: every word must still be recognised.
The observed order matches the predicted order, which is weak corroboration only. It would be tested by administering comparable reasoning tasks in Ukrainian or with picture support.
3. Write the placement recommendation the team can defend, and name the evidence that would change it.
Read the worked answer
Decision: admit Mira provisionally to the cluster now, with a written review in ten weeks, and do not record 92 as an average-ability finding.
Reasoning. 'The cluster is for verbal strength and 92 is average' assumes the verbal battery measured verbal reasoning for this student. At ACCESS 3.2 it measured English plus reasoning in unknown proportions, so exclusion on that score penalises her for a language she is still acquiring. But admitting her on Nonverbal 124 alone is equally unsupported: nonverbal scores are not language-free (directions, pacing, practice items), and they say nothing specific about verbal reasoning. Provisional admission with a scheduled review keeps the decision reversible, which is what the evidence supports.
Evidence to gather: reasoning tasks in Ukrainian administered by a bilingual colleague; a learning-rate measure showing how quickly she takes up a new verbal pattern with support; writing samples produced with vocabulary scaffolding.
What would change the decision: weak home-language reasoning, or verbal scores still flat after a year of intensive English while Quantitative and Nonverbal stay high. That pattern would point to a genuine verbal weakness rather than a language artefact.