78 Then 96: A Retest Gain That Has to Be Explained Before It Is Used
Candidate Kowalski takes the DLAB twice. First attempt: 78, at 0630 at the end of a 12-hour overnight shift, on a walk-in slot he accepted because it was the only opening before his reenlistment window closed. Second attempt: 96, eight weeks later, after he completed a self-paced online course that advertises "DLAB-style" practice with invented grammar rules and stress-pattern drills. His Russian course requires 90. The audio subtest results barely moved between the two sittings; the visual rule-application portion accounts for almost all of the gain. Unit policy allows a maximum of two attempts per career. Both scores are on his record, and the second one is now the score of record. His first sergeant wants him loaded into the next Russian class; the training office wants a written assessment before it commits the seat, because it has been burned before. The class starts in six weeks and there are two alternates already qualified.
1. Is the 96 usable? Walk through the considerations in the order that matters.
Read the worked answer
1. Is the score real? It was produced under standard conditions and is the score of record. Nobody gets to discard it because the gain is suspicious. The question is not authenticity but meaning.
2. How much of the gain has an innocent explanation? Two candidates for the 18-point jump. First, the initial sitting was compromised by fatigue and an unscheduled walk-in - a 0630 test after a night shift is a well-known deflator, and the low first score may understate him. Second, eight weeks of practice on the specific item formats can raise a score through familiarity rather than aptitude.
3. Does the profile discriminate? Yes, and this is the useful part. The audio subtests barely moved; the visual rule-application portion carried the gain. Practice on written invented-language rules transfers readily to a written rule-application section. It transfers poorly to paced listening. So the gain is concentrated precisely where format familiarity pays, which is the weaker of the two explanations.
4. What does that mean for Russian? Russian is a Category III course with a heavy listening and speaking load. The portion of his profile that improved is not the portion the course stresses.
Verdict: usable as a threshold pass, but not as a strong predictor. Score him in, and treat the seat decision on other evidence.
2. What should the training office actually put in writing, and what should it do with the two qualified alternates?
Read the worked answer
The memorandum should do three things.
Record the facts without editorializing. Two sittings, dates, conditions, subtest-level change, and the self-study course completed between them. Note the first-sitting conditions as a possible deflator rather than as an excuse.
State the interpretation. The candidate meets the 90 threshold on the score of record. Confidence in the prediction is moderate, not high, because the gain is concentrated in the most practice-sensitive portion and the listening profile is unchanged. Recommend load, with early academic monitoring - first unit examination, and a check-in at the mid-course point - rather than a delayed load.
Avoid the wrong remedy. Do not delay him a cycle to "get a cleaner score." He is at his attempt limit, he meets the threshold, and a delayed third sitting is not available. Delaying costs a year for no additional information.
On the alternates: they are qualified now. Do not hold them as insurance past the load date in the expectation that the primary will fail; that is how a unit loses both. If the alternates can be seated in a later class, hold them; if not, release them to other programs.
3. The commander's takeaway is "so we should send everyone to the online prep course." Argue against that inference using this case.
Read the worked answer
This case is evidence that coaching moves scores. It is not evidence that coaching improves outcomes, and the two are being conflated.
Look at what the coaching taught. The described course teaches invented grammar rules, stress-pattern drills, and - in the visual section - item formats. Score gains produced by familiarity with item formats measure familiarity. The training pipeline does not care how well a soldier handles the DLAB's visual rule section; it cares whether he can do Russian for 63 weeks.
Notice where the gain landed. If genuine aptitude had improved, we would expect some movement in the paced listening subtests, which are the harder thing to drill. They did not move. A gain that appears only in one section, in the direction format practice predicts, is best explained by format practice.
Notice the sample. n=1. One candidate, one retest, one course, and a first sitting with a plausible deflator. That is not a base for a policy.
What would justify expansion: a pre/post design with a comparison group, and - the decisive outcome - comparison of language-school graduation and washback rates between coached and uncoached soldiers. If coached soldiers score higher and graduate at the same rate, the course has not improved selection; it has degraded the screen. That is worse than doing nothing, because it consumes seats and instructor time to make the decision input noisier.