Grade-5 Screener: Odd Word Out or Odd Rule Out?
Ms. Okafor is reviewing a 20-item verbal classification screener completed by Devon, a grade-5 student reading at grade level on a curriculum-based measure. Each item lists four words and asks which one does not belong with the other three. Devon scored 13/20 in 11 minutes, inside the 15-minute limit. Item-level review shows a clean split. He is correct on 9 of 9 items where the three related words share a part of speech and the odd word is a different category, for example hammer, saw, drill, lettuce. He is correct on only 4 of 11 items where all four words are nouns and the odd word is either a broader category, as in rose, tulip, daisy, plant, or differs by a subtler attribute, as in lake, river, stream, puddle. He changed two answers, both from correct to incorrect, and left no items blank. His vocabulary screening score sits at the 62nd percentile. Ms. Okafor must decide whether the pattern reflects a classification-rule deficit, a vocabulary gap, or a test-taking style, and what to recommend to the classroom teacher.
1. Which hypothesis - classification-rule deficit, vocabulary gap, or test-taking style - best fits the item-level data? Give your reasoning and name the single piece of evidence that would most weaken your conclusion.
Read the worked answer
Vocabulary gap is the weakest fit. A vocabulary explanation predicts errors scattered across item types in proportion to word difficulty. Instead the errors sort perfectly by item structure: 9/9 where part of speech separates the odd word, 4/11 where it does not. Several words in the easy set (drill, lettuce) are no more frequent than words in the hard set (tulip, puddle), and a 62nd-percentile vocabulary score is adequate for these words.
Test-taking style explains little. He left nothing blank, used 11 of 15 minutes, and worked in order. The two correct-to-incorrect answer changes are a mild self-monitoring flag, but that accounts for at most 2 of the 7 errors.
Classification-rule deficit fits best. The discriminating variable is whether a surface cue (part of speech) is available. When it is, he can respond; when all four words are the same class, he must select the dimension of comparison himself - category level in the plant item, size or scale in the puddle item - and he does not. That is a rule-selection problem, not a word-knowledge problem.
Evidence that would weaken this: if all 7 misses were low-frequency words and all 4 hits were high-frequency, vocabulary becomes viable again. Check word frequency for the missed items before finalizing.
2. The classroom teacher proposes pre-teaching the 20 vocabulary words from the missed items. Evaluate the proposal and replace it with an activity you can justify from the data.
Read the worked answer
The proposal targets a mechanism the data do not support. Devon is a grade-level reader at the 62nd percentile on vocabulary; the missed sets use common words (rose, tulip, plant, lake, puddle). It is very likely he knows all four meanings in each set and still cannot choose. Pre-teaching those words would raise scores on exactly those items and transfer nowhere - item-specific learning dressed up as intervention.
Replace it with a justification routine. Present four words and require two things: circle the odd word and write the rule in one sentence. Then ask the follow-up that matters: could a different word be the odd one out under a different rule, and what would that rule be? For boot, sandal, slipper, sock, the odd word is sock if the rule is outer footwear, but a student can defend boot if the rule is ankle coverage. Forcing a stated rule makes the dimension of comparison explicit and public, which is precisely the step Devon skips.
Sequence the work: teacher models two sets, then pairs justify to each other, then independent written justification. Progress-monitor on novel word sets, never on the same 20 words, and score the justification as well as the choice.
3. Write three probe items in the same four-word format that would separate rule-based responding from associative responding, and state what response pattern would confirm each hypothesis.
Read the worked answer
Probe set
- shark, tuna, cod, dolphin - intended answer dolphin (mammal, not a fish).
- hammer, wrench, screwdriver, nail - intended answer nail (a fastener, not a tool).
- boot, sandal, slipper, sock - intended answer sock (not outer footwear).
How to read the responses
- Rule-based: he selects the intended word and states a categorical rule for at least two of the three. Some choice variation is fine as long as a rule is articulated and applied.
- Associative: he selects the word that is most strongly linked in use or context rather than by category - for example nail because you hammer it, or dolphin because it swims with the others - and justifies with a scene rather than a rule. That is the same failure seen on the lake/river/stream/puddle item.
- Unstable / no dimension: choices are inconsistent across the three and he cannot state a reason, or his reason changes after he hears a peer's. This points to dimension selection not yet being established rather than to a preferred wrong strategy.
Run the three probes individually, ask for the reason before showing any feedback, and record both the choice and the stated rule. The choice alone will not discriminate; the justification is the data.