The effect of unknown word density on reading comprehension

Abstract

This study aims to find out the effect of unknown words in a text on reading comprehension. The assumption which lies behind the study is that vocabulary knowledge is a strong, though not the only, predictor of reading comprehension. A 633-word narrative text about a lady being tricked by a caller about an escape from a mental institution was used as the reading material in the study. Four versions of the text were prepared by making two kinds of adjustments. The first adjustment was to replace the low frequency words based on Francis & Kucera's frequency list by nonsense words. Then the second adjustment was to make sure all the words not being replaced were within the most frequent 2,000 words of English (West's (1953) General Service List). The four versions were 100% text coverage with all the words in the text from the most frequent 2,000 words, 95% text coverage with one unknown word in every twenty tokens, 90% text coverage with one unknown word in every ten tokens, and 80% text coverage with one unknown word in every five tokens. All the subjects participating in the experiment were non-native speakers of English who had vocabulary knowledge well beyond the most frequent 2,000 words of English. This was to ensure they would not have any vocabulary difficulties with the text except for the nonsense words. After reading the text, all of the subjects were required to sit two comprehension tests - a multiple choice test and a cued recall task. Before sitting the tests, they were also asked to predict how they thought they would perform on the tests. The results indicated that there was a predictable relationship between unknown vocabulary density and reading comprehension. The subjects leading the 100% version were mostly able to achieve adequate comprehension. The subjects reading the 95% and 90% versions showed quite a big disparity among their performance, with a minority achieving adequate comprehension. None of the subjects reading the 80% version achieved satisfactory comprehension. The subjects' prediction scores mostly agreed with their comprehension scores, that is, the density of unknown words strongly influences their impression of text difficulty. A level of 98% coverage of known words seemed to be a satisfactory level for unassisted extensive reading Teaching implications about choosing reading materials for both intensive and extensive reading were discussed and possibilities for further research were suggested.