6.3 Multi-Feature Clues and Distractor Control
Key Takeaways
- Multi-feature items require deciding which visual changes are meaningful and which are incidental.
- Distractors often feel obvious because color, size, or shape attracts attention before evidence supports it.
- A good rule explains all examples and transfers to a new arrangement with minimal extra assumptions.
- When two rules seem possible, candidates should prefer the one supported by more contrasts.
Seeing past the loudest feature
A visual pattern can carry many clues at once. A picture might show three small red circles inside a box with an arrow pointing left. The label could encode number, size, color, shape, containment, direction, or some combination. Multi-feature reasoning is the art of finding which features the system actually uses, and it is where the DLAB separates a 95-level candidate from a 110-plus candidate.
The loudest feature is not always the relevant feature. Bright red catches your eye, but if red appears with several unrelated labels, it is weak evidence. A tiny position change may matter more if the label changes exactly when the position changes. In timed practice this is the chief error source: attention chases salience instead of correlation.
Build a contrast grid
List the features across examples and mark which label pieces change. If "ka" appears with every plural image even when object and color change, "ka" is a strong number marker. If "ka" appears with both singular and plural images, number is weaker.
| Image description | Label | Useful clue |
|---|---|---|
| one red circle above box | lom pa | above may be "pa" |
| one blue square above box | tir pa | above still "pa" |
| one blue square below box | tir vo | below may be "vo" |
| three blue squares below box | ka tir vo | "ka" may mark plural |
This original practice-style grid suggests color, number, and relation each have label pieces. The strongest clue is the one that repeats across different contexts: "above" stays "pa" even when object and color change; "below" stays "vo" when number changes; plural surfaces as "ka" only with three items.
Distractor control
Distractor control means asking what an answer choice would require. If a choice claims "pa means circle," it fails because "pa" also appears with a square. If a choice claims "tir means below," it fails because "tir" appears in an above example. Wrong answers usually pin a label to a feature present in one example but not all relevant examples. Three high-frequency distractor families:
- One-example anchors: a feature that appears once and is generalized too far.
- Surface-match lures: a choice that reuses a symbol from the first example because it looks familiar.
- English-syntax lures: a choice ordered the way an English speaker would say it, ignoring the invented system's order.
When two hypotheses survive, choose the narrower, better-supported one. If "ka" appears only once with three objects, it might be plural or it might mean blue if that image is also blue. Look for another blue image or another plural image to break the tie. If none exists, lean on the answer choices and avoid overclaiming.
The two-pass scan
A reliable timing method is the two-pass scan. On pass one, name every visible feature quickly (object, property, number, relation, direction). On pass two, discard features that do not track label changes. If a feature never changes across examples, it cannot explain a label contrast. If a feature changes but the label does not, it is probably incidental for that rule. This discipline aligns with the DLAB's purpose: it measures potential to learn an unfamiliar language, so the test rewards separating meaningful contrasts from noise rather than recalling any real vocabulary.
A worked multi-feature item
Work a harder case. Four examples: (a) one red circle inside box = "sol fa lom," (b) one blue circle inside box = "sol fa tir," (c) two blue circles inside box = "sol fa-fa tir," (d) two blue squares beside box = "nem fa-fa tir." Compare systematically. The relation word changes between "sol" (inside) and "nem" (beside), so position is encoded first. The object word stays "fa" for circle and is absent for square (it becomes part of "nem"? no, the object piece is the constant middle token), so isolate carefully: "fa" appears in a, b, c (all circles) and the plural shows as "fa-fa." Color sits last: "lom" for red, "tir" for blue.
Now a new target, two red circles beside box, predicts "nem fa-fa lom." Notice that the only way to reach that answer is to hold four features apart simultaneously; collapsing any two would scramble the order.
Reading the distractor set
In a real DLAB-style multiple-choice set the distractors are not random. Typically one choice reverses the relation (inside vs. beside), one drops the plural marker, one swaps the color, and one is correct. If you have built the contrast grid, you can knock out the relation-reversal and the dropped-plural choices in a single glance, then decide between the remaining two on color alone. The grid does double duty: it finds the rule and it pre-sorts the distractors.
Where multi-feature skill pays off
The heaviest multi-feature items cluster near the difficult end of the test, where the marginal points decide whether you clear 105 (difficult languages such as Russian or Hebrew) or 110 (the hardest Category IV languages). Candidates who freeze on these items because they try to hold the whole picture in mind at once are the ones who stall around the mid-90s. The contrast grid is the cheap, repeatable habit that converts that freeze into a methodical elimination.
Practice-style: The label piece "pa" appears with a red circle above a box and a blue square above a box. It changes when the object moves below the box. What is "pa" most likely marking?
When two visual rules both fit the examples, which should you prefer?
What is the main risk of choosing the most visually prominent feature in a multi-feature item?