The accuracy and confidence of Greek learners guessing English word meanings

Abstract

Lexical inferencing is a key feature of the strategic competence of learners. In particular, we need to know how successful different types of learners are in guessing meanings of words when different kinds and numbers of cues are available. We also need to study learners' ratings of their own success: authentic writing does not typically come with adequate cues built in to allow the meaning of every word unknown to a reader to be guessed accurately, so good readers need to be aware of when their guess is likely to be accurate, as against when there are insufficient or conflicting cues, or where their own relevant knowledge base (eg content schema or mastery of wordformation processes) is inadequate. Otherwise over-or under-use of dictionaries and/or learning of faulty information may occur.
In Study 1 three groups of subjects, Greek NS, Greek/English bilinguals, and English NS were tested for both accuracy of guesses and confidence in them, responding to stimuli consisting of pseudowords of Greek and Latin origin in verb form in a words-out-of-context test.
In Study 2 two groups of subjects, Greek NS and English NS, and in Study 3 four levels of Greek learners of English were tested for both accuracy of guesses and confidence in them, responding to stimuli, half in verb and half in noun form, consisting of (a) decomposable pseudowords of Greek and Latin origin; (b) Nondecomposable monosyllabic ones. Each pseudoword was put in context containing cues of three different sorts: (i) convergent, (ii) vague, and (iii) divergent.
Results include the following: British subjects are more accurate on Greek words than Latin, and as accurate as Greek subjects on Greek words; however, they are underconfident overall and their accuracy/ confidence correlations are not so good as the Greek subjects'. Greek subjects of all studies, now, are most successful guessing words of Greek origin; however, partly because of the disguise of English transliteration they cannot achieve accuracy close to 100%; they tend
to some overconfidence particularly on Latin and Nondecomposable. On the whole stems are easier to guess than prefixes and nouns than verbs. Contrary to expect at ions, there was not a steady increase in success over the four levels studied.

Date of Award	28 Jul 1992
Original language	English
Awarding Institution	University of Wales, Bangor