7.6 Check Consistency and Avoid Overfitting

Consistency beats cleverness

A tempting wrong rule explains one example perfectly. A valid rule explains every example in the item. Before you commit, run your proposed rule across each given sentence and confirm it produces the stated meaning every single time. This is the single discipline that separates a high constructed-grammar score from a frustrating one, because the DLAB rewards rules that generalize across the evidence, not clever guesses that fit one line.

When a quick rule breaks

Suppose you decide a final syllable marks past tense, then a later example shows the same syllable with a present meaning. That is not an unfair item — it is a signal your first rule was too quick. The correct move is to revise the rule so it covers all the data, not to force the data to fit the rule.

This self-correction loop is itself a measured aptitude. Successful language learners constantly form a hypothesis, test it against new input, and revise when it breaks — and the DLAB construction items reward exactly that flexibility. A rigid test taker latches onto the first rule and defends it against contradicting examples; a strong one treats the first rule as provisional and lets the evidence overrule it. Building the habit of cheerfully discarding a broken rule, rather than rationalizing it, is one of the highest-leverage things you can practice before test day.

Worked example (practice-style, not official DLAB content)

A first-pass rule — subject, object, verb — fits the first three lines. The fourth line changes only lim to lum, so the final word carries action plus negation: lim = draws, lum = does not draw. Target "The student does not draw the flag" preserves subject (veko) and object (sar) while using the negative action form → veko sar lum.

Naming overfitting precisely

Notice what these examples do not prove. They do not prove every word ending in m is a verb. They do not establish a general sound rule for all negatives. They prove only that lim and lum contrast as draw versus does-not-draw inside this item. Overfitting looks like this:

Overfit rules waste time and manufacture confident wrong answers. The fix is the Occam test: if two rules both fit, choose the simpler one; if a rule needs exceptions to survive, drop it for one that does not.

A fast final check

State your rule in one line, then ask three questions before moving on:

1. Does my rule explain every given example?
2. Does my answer include every required meaning feature?
3. Did I add anything the examples did not earn?

Yes, yes, and no — then commit. When reviewing misses, separate two distinct failures: did you infer the wrong rule (a comparison problem, fixed by better contrast reading), or infer the right rule and apply it inconsistently (an attention problem, fixed by slower final checking)? The diagnosis changes the remedy.

Ethics and honest practice

Public DLAB detail is deliberately limited, and the live test is protected material. Practice drills must be original and clearly labeled as practice-style; reconstructing real questions from memory is both against testing rules and useless, because the value lies in the reasoning process, not in any fantasy that a drill copies the exam.

Competing-rule tie-breaks

Sometimes two rules each explain all the examples. When that happens, prefer the rule that most directly maps the changed meaning to the changed form, and avoid the rule that requires a hidden extra step:

The simpler rule is not just elegant; on the DLAB it is statistically the safer bet, because the test is engineered around clean, learnable patterns rather than baroque exceptions.

The closing discipline

As constructed examples grow denser, the same principle holds: find the one rule that accounts for all evidence, apply it once, and resist clever additions the item never justified. Run the three-question final check — does the rule explain every example, does the answer include every required feature, did I add anything unsupported — and commit on yes, yes, no. This single loop, repeated across the construction section, is what converts raw pattern-spotting into a reliable qualifying score, and it is the habit most worth rehearsing in the final days before test day.