Dağarcık Nasıl Ölçülür — B2–C1 İngilizce Okuma Parçası

How many words does a person know? The question sounds simple, and every answer to it is an estimate. There is no procedure by which a researcher can open someone’s memory and count what is stored there. What can be done instead is to test a small selection and reason carefully from the result.

This is why vocabulary size is sampled rather than counted. An educated adult may know tens of thousands of word families, and no test can present them all. Researchers therefore divide the language into frequency bands — the most common thousand words, the second thousand, and so on — and select a fixed number of items from each. If a participant identifies six of eight items drawn from the second thousand, the assumption is that roughly three quarters of that entire band is available to them. The proportion observed within the sample is projected onto the whole.

The method is only as reliable as the sampling behind it. Items must be selected at random from the band rather than chosen by the researcher, since any human selection tends to favour words that feel typical. The number of items per band must also be sufficient; with only two or three questions, a single lucky guess shifts the estimate by hundreds of words. Most published instruments therefore use a minimum of eight to ten items per band and report their results as a range rather than a single figure.

A less obvious requirement concerns where the test items come from. If they are drawn from the same material the learner has been studying, the instrument no longer measures vocabulary at all; it measures how much of that specific material was retained. The estimate is inflated, and the error cannot be seen, because the score looks entirely normal. Sound practice keeps a reserve of items outside the teaching material — a straightforward application of the distinction between training data and test data.

What makes any of this worth the trouble is the relationship between coverage and understanding. Comprehension does not improve smoothly as vocabulary grows; it changes character at particular thresholds. Around ninety-five per cent coverage, a reader can follow a text with a dictionary at hand and considerable effort. Around ninety-eight per cent, the same reader can follow it without help. Three percentage points separate assisted reading from independent reading, and that is precisely why the exact figure matters.

An estimate, then, is not a weaker form of a count. It is a different kind of statement, one that carries its own margin of error and its own conditions. Reported honestly — with the bands named, the sample size given and the reserve described — it tells a learner something a count never could: not merely how many words are known, but which part of the language remains closed.

Measuring What Someone Knows — Kelime Dağarcığı Nasıl Ölçülür

Sözlükçe

Anlama soruları

Bu parçanın kelime tabanı

Metinde çalışan gramer