7.6 Check Consistency and Avoid Overfitting

Consistency beats cleverness

A tempting wrong rule often explains one example perfectly. A good rule explains all examples in the item. Before choosing an answer, test your rule against each given sentence and make sure it produces the stated meaning every time.

This is especially important when the practice set contains repeated pieces. You may decide that a final syllable marks past tense, then notice another example where the same syllable appears with a present meaning. That does not mean the item is unfair. It means the first rule was too quick.

Practice-style example, not official DLAB content:

A first rule may be subject, object, verb. That fits the first three examples. The fourth example changes only lim to lum, so the final word carries the action plus negation contrast. A target meaning like "The student does not draw the flag" should preserve subject and object while using the negative action form: veko sar lum.

Now avoid overfitting. The examples do not prove that every word ending in m is a verb. They do not prove a general sound rule for all negative verbs. They prove only that lim and lum contrast as draw and does not draw inside this practice item.

Overfitting wastes time. It also creates confident wrong answers. A candidate may invent a rule such as "words with a are objects" because two examples happen to contain object words with a. If another example breaks that idea, drop it.

Use a consistency check that is fast enough for timed practice. State the rule in one line, then run it across the examples. If the rule requires exceptions, choose the simpler rule that fits all evidence. If two rules fit equally, prefer the one that directly maps the changed meaning.

Consistency also protects ethics. Public DLAB detail is limited, and protected tests should not be reconstructed from real questions. Practice examples should be original and clearly labeled as practice-style. The value is the reasoning process, not the fantasy that a drill is a copy of the exam.

When reviewing misses, separate two problems. Did you infer the wrong rule from the evidence, or did you infer the right rule and apply it inconsistently? The first calls for better comparison. The second calls for slower final checking or a clearer mental checklist.

A good final check asks three questions. Does my rule explain every given example? Does my answer include every required meaning feature? Did I add anything unsupported? If the answer is yes, yes, and no, move on.

This skill becomes more useful as items become denser. Long constructed examples can create noise, but the same principle holds. Find the rule that accounts for all evidence, apply it once, and avoid clever additions that the item did not earn.