The paper reports the results of a pilot study investigating the efficiency of on-line glosses in L2 reading. This study represents the initial stage in a series of pilot studies to be conducted with the purpose of establishing the optimum format and amount of information an L2 learner should have access to in the process of computer-assisted reading.

On-line glossing has been shown to be an effective means of teaching reading in a foreign language: it makes reading more efficient, enhances comprehension, improves incidental vocabulary learning.

Given the capacity of computers, software designers and language instructors seek to provide the learners with access to a variety of information—grammar explanations, L2 definitions and thesaurus definitions, sound, pictures, video etc. Evidence from research, however (Davis and Lyman-Hager), indicates that when learners have various informational options available, they tend to rely almost exclusively upon English definitions and largely ignore other options. The obvious implication of these data is that there is a certain ceiling on the effectiveness of on-line glosses beyond which technology ceases to be effective. Identifying this ceiling of actually needed on-line support becomes possible only after a careful examination of what specific functions are performed by computers in the process of L2 reading by humans.

The purpose of the study is twofold: 1) to conceptualize the role of on-line glossing in L2 reading and examine how it actually aids L2 reading; 2) to develop a psychologically sound on-line support system consistent with the actual needs of the learner in reading a foreign language text.

Reading is a multi-level process and is subserved by a number of cognitive processes described as bottom-up or top-down. Bottom-up processes include perception of words, activation of semantic representations, encoding representation into meaning and encoding of basic sentence level propositions. Top-down processes involve higher order knowledge structures (schemata) and metacognitive abilities. The role of computerized glosses in L2 reading may be defined as the one that aids L2 reading by performing most bottom-up functions and thus allowing more of the readers' attention to top-down processes.

If glossing does mediate reading comprehension by speeding bottom-up functions, then the next step in software design is to identify the format of glossing that would be consistent with the characteristics of lexical processing in reading a foreign language text. Psycholinguistic studies indicate, that lexical items are activated at two different levels—the level of syntactic and semantic information, and the level of the morphological form. This paper will examine two hyper text glossing strategies when the reader is provided with: (a) a standard L2 (Russian) dictionary form of the word and its L1 (English) equivalent (the level of syntactic and semantic information); and (b) an L1 contextual translation of L2 words (the level of the morphological form).

The hypothesis of the study is based on the supposition that strategy (b) would be more beneficial for comprehension, retention and incidental vocabulary learning since it establishes an immediate match between the form and the meaning part of the word. To test this hypothesis, the authors use the following methods: a multiple-choice comprehension test, a cloze-test, recall protocols, a vocabulary test, and subjective satisfaction questionnaires and interviews.