The Role of Illustrative Examples in Productive Dictionary Use Hilary Nesd Introduction One of the distinguishing features of learners' dictionaries is an abundance of illustrative examples, and lexicographers and reviewers write convincingly of the value of dictionary examples as aids to successful language production. Cowie (1989), for example, claims that a wellconstructed example in a learners' dictionary can show the learnerwriter acceptable complementation patterns, collocations, and native stylistic norms. Other lexicographers, such as Landau (1984, 166), Marello (1987, 226-27), Creamer (1987, 243) and Drysdale (1987) extol the value of examples in a similar vein, implying that, without the benefit of dictionary examples, learner-writers would make many more mistakes. Studies show, however, that learners' dictionaries often fail to provide the information that learners need to avoid producing errors (Huang [1985], Nesi [1987], Meara and English [1988]). Learners have also been found to seriously misinterpret the grammatical, collocational and semantic information learners' dictionaries do provide (Nesi and Meara [1993]). Perhaps the lexicographers' views derive from a consideration ofwhat should happen when a skilled dictionary user consults a skillfully chosen example. Such views may not take into account what does happen when ordinary dictionary users, with a tendency to misread dictionary entries, consult examples that do not adequately reflect all the lexical features that they need to know. The small amount of evidence available from testing and observing learner use ofdictionary examples is far less positive about their usefulness. Black (1986) found no significant difference between com- ____________Illustrative Examples in Productive Dictionary Use_________199 prehension test scores for words defined with examples and words defined without examples, and although Miller and Gildea (1985) found that native-speaker children produced more acceptable sentences when they had access to examples, they decided to draw no definite conclusions from their study (1987, 90). Miller and Gildea's research method had run into problems because, when their subjects were asked to write sentences with the aid of examples, many simply reproduced the examples they had been given. Black's study is also somewhat unreliable because subjects often seemed to have simply guessed the correct multiple choice answer. Both studies also ignored the possibility that subjects might already have known some of the look-up words. The task set in the study is a computer-based version of the productive task devised by Miller and Gildea (1985, 1987). Using a program written by Paul Meara at University College Swansea, I was able to regulate the quantity of dictionary information that appeared on the screen and record not only the sentences produced by each subject, but also the extent to which subjects accessed dictionary information and the length of time spent reading that information. Subjects were prevented from simply repeating example sentences from the dictionary entry, because they were required to include both the target word and a given high frequency word in the sentences they produced. They were also allowed to choose whether or not to access dictionary information before writing each sentence. This made it less likely that subjects would look up words they already knew well; instead, look up was likely to be motivated by inadequate productive knowledge of the target words. The study The best type of EFL dictionary entry is, presumably, one that can be quickly absorbed and put to effective practical use. The two research questions posed for this study therefore concerned these two attributes of the dictionary entry: 1.Do definitions with examples take longer to read than definitions without examples? 2.Are definitions with examples more helpful in productive dictionary use than definitions without examples? Eighteen target words were chosen for the experiment from the 200____ Hilary Nesi middle bands of Nation's "University Word List" (Nation 1990). The words were enlighten, err, gravity, incorporate, intersect, perpetrate, retard, rudimentary, symptom, version, agitate, civic, clarify, collide, compute, controversy, interact, and interlude. A pilot study suggested that the majority of subjects would be unlikely to know these words. Two versions of the test were prepared using entries taken from the second version of the Longman Dictionary of Contemporary Englüh (LDOCE). In each version, example sentences and phrases for half the target words were removed. In version A, examples for the first...