second language vocabulary learning: the role of context versus

17
Second Language Vocabulary Learning: The Role of Context versus Translations as a Function of Proficiency Author(s): Peter Prince Source: The Modern Language Journal, Vol. 80, No. 4 (Winter, 1996), pp. 478-493 Published by: Blackwell Publishing on behalf of the National Federation of Modern Language Teachers Associations Stable URL: http://www.jstor.org/stable/329727 . Accessed: 04/09/2011 05:46 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Blackwell Publishing and National Federation of Modern Language Teachers Associations are collaborating with JSTOR to digitize, preserve and extend access to The Modern Language Journal.

Upload: others

Post on 19-Feb-2022

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Second Language Vocabulary Learning: The Role of Context versus

Second Language Vocabulary Learning: The Role of Context versus Translations as a Functionof ProficiencyAuthor(s): Peter PrinceSource: The Modern Language Journal, Vol. 80, No. 4 (Winter, 1996), pp. 478-493Published by: Blackwell Publishing on behalf of the National Federation of Modern Language TeachersAssociationsStable URL: http://www.jstor.org/stable/329727 .Accessed: 04/09/2011 05:46

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

Blackwell Publishing and National Federation of Modern Language Teachers Associations are collaboratingwith JSTOR to digitize, preserve and extend access to The Modern Language Journal.

http://www.jstor.org

Page 2: Second Language Vocabulary Learning: The Role of Context versus

Second Language Vocabulary Learning: The Role of Context

versus Translations as a Function

of Proficiency PETER PRINCE Centre de Recherche en Psychologie Cognitive Universiti de Provence 29 bld Robert Schuman 13100 Aix en Provence France

A widespread view of vocabulary learning is that it is advisable to make the shift away from

learning words with their translations and to rely on second language (L2) context as soon as possible. Such faith in context learning has not always received experimental support, however, nor is it commonly shared by L2 learners. An experiment in which subjects were tested on their recall of newly learned words was conducted to determine the relative

advantages and disadvantages of both context learning and translation learning as a func- tion of learner proficiency. Results reveal a superiority of translation learning in terms of

quantity, but an inability on the part of weaker learners to transfer their knowledge into L2 contexts. The possible reasons for this are discussed, and it is suggested that alternative

learning strategies that combine the advantages of the two techniques should be explored.

ONE ASPECT OF LANGUAGE LEARNING that in the past received little attention, but now has become a focus of much research, is the

learning of vocabulary. Increasingly, it is consid- ered that effective communication relies less

upon the mastery of grammatical rules than on the possession of an adequate and appropriate vocabulary (Vermeer, 1992). The size of the vo-

cabulary needed to achieve general communi- cation skills in English has been estimated at 5,000 words, with critical importance being at- tached to the learning of the most frequent 2,000 to 3,000 words as quickly as possible (Na- tion, 1993).

Within this area of study, one question that has yet to receive any precise form of answer is whether the learning of a new word best occurs when the word is met in context or in a "paired associate" condition, with the word being

linked to its native language equivalent. Many teachers work on the assumption that however effective this form of paired associate learning (henceforward termed translation learning) might be, it tends to reduce possibilities for sec- ond language (L2) autonomy by linking words too exclusively to first language (Li) equiva- lents; there is the added disadvantage that learners are likely to suppose or look for equiva- lence of meaning even when the contexts in which those words are used in different lan-

guages clearly diverge (McCarthy, 1990). As a result, most course books and teachers present L2 contextualized material right from the outset. Compelling evidence that such an ap- proach can be of benefit as far as receptive vo-

cabulary is concerned emerges from a study by Saragi, Nation, and Meisler (1978), in which the

reading of a novel led to successful recognition of 76% of new words tested. Similar conclusions about incidental learning of vocabulary are pre- sented by Nagy, Herman, and Anderson (1985).

It is to be noted, however, that when it comes to recall, much of the research conducted in

The Modern Language Journal, 80, iv (1996) 0026-7902/96/478-493 $1.50/0 ?1996 The Modern Language Journal

Page 3: Second Language Vocabulary Learning: The Role of Context versus

Peter Prince 479

this field has failed to demonstrate any clear

advantage of learning in context over transla- tion learning (Nation, 1982). Indeed, results

provided by Seibert (1930) indicate that learn-

ing words in pairs is consistently more effective than learning them in context. Addressing the same question, Pickering (1982) found learning in context slightly more advantageous, but not

enough for the difference to be considered in

any way conclusive. In their review of the ques- tion, Carter and McCarthy (1988) conclude that "it is difficult to draw precise lines to suggest when a move from key-word techniques, or translation in pairs, or from using a mono-

lingual or bilingual dictionary to context-based inferential strategies, is best instituted" (p. 15). They suggest that there are "no clearly marked

stages of transition" in the learning process, and that therefore "a mixture of approaches should be adopted" (p. 15). One reason for this recommendation of a mixture of approaches is that both methods have advantages and disad-

vantages, which will be discussed in the conclu- sion to the present paper.

Set against the uncertainty highlighted by Carter and McCarthy (1988), the sheer ability of learners to assimilate large quantities of words with their translations is impressive, although there are wide disparities between individual learners (Webb, 1962). This ability may corre-

spond to, and reinforce, a preferential strategy that learners adopt in the early stages of learn-

ing and subsequently find difficult to relin-

quish (Sautermeister, 1989). The obvious short-

coming of such a strategy is that it favours the creation of a single "trace"'1 in memory over the

multiple traces provided by contextual encoun- ters. Thus, although the information may be well established in memory, it is not necessarily available for use in appropriate contexts (Wise- man & Tulving, 1976).

One difficulty in this area of research, pointed out by Nation (1982), is that there is no standard definition of what is meant by context. In its broadest sense, of course, context may be said to comprise all the perceived phenomena that accompany the processing of a given stim- ulus, including the physical surroundings in which learning takes place. In this sense, a na- tive word that serves as the basis for the learning of a foreign word is a major element of the con- text. Other possible contexts are pictures or L2

synonyms (R6hr, 1993). Although a comparison between these different types of context would

undoubtedly be of interest, the present paper concentrates on the two most widely used con-

texts, namely the L1 translation equivalent of the word to be learned and an L2 sentence in which the word to be learned appears. Thus, the word "context" will be used in a narrow sense to refer specifically to the latter condition, as op- posed to translation learning in which the con- text (in the wider sense) is provided by a single L1 equivalent.

It is also important to specify the use to which the context is put: Words presented in transla- tion pairs are typically the focus of intentional

learning, whereas an unknown word seen em- bedded in an L2 text may frequently receive

only such attention as prevents it from imped- ing comprehension-that is, the reader might choose to ignore the word if he or she considers that it does not affect general understanding of the text. Hulstijn (1992) notes that the chance of readers remembering a word after encoun-

tering it in a text that is read for its content is

very small indeed. In the experiment described in this article, the context was to be used specif- ically for the learning of new vocabulary for the

purposes of production. Another element that appears to be relevant

is the degree of proficiency of the learner. Evi- dence provided by Kroll and Curley (1988) sug- gests that in the initial stages of learning, new words are strongly linked to their L1 equiva- lents, and a shift occurs after about 30 months of study, such that a network of links within the L2 begins to become effective. The data of Kroll and Curley come from a picture-naming experiment in which response times in the L2 decreased when subjects had more than 30 months of study behind them. The authors con- cluded that subjects had therefore formed a di- rect link between the L2 words and the underly- ing concept, and were no longer passing by way of their L1 in order to name the picture in their L2. Although the work of Kroll and Curley con- cerned lexical access as opposed to vocabulary learning, it seems likely that the patterns of ac- cess follow those established during the learn- ing phase; in other words, in the first months of learning, L2 words are more effectively stored in memory when they are linked to their L1 equivalents.

Gekoski (1980) has described this phenome- non by stating that at lower levels of L2 profi- ciency, learners use L1 mediation in order to translate their "thoughts" into the L2. Because most of the words encountered in the initial stages are of a concrete, imageable nature, it may be that as soon as learners have ascertained the meaning of the word, either by accessing an

Page 4: Second Language Vocabulary Learning: The Role of Context versus

480 The Modern Language Journal 80 (1996)

imagen store (Paivio & Desrochers, 1980) or with the help of the teacher's explicit use of a

picture, they neglect the L2 context in order to attach the new word form to the L1 expression of the meaning. Thus, in effect, the learning occurs in a translation condition.

It should be noted, however, that although the Kroll and Curley (1988) data corroborate intuitions about vocabulary learning in the be-

ginning stage, the shift away from exclusive use of the translation link after about 30 months of

study does not imply that learning does not re- main more effective in the translation condi- tion. There are two reasons for this. First, as has

already been mentioned, the study of lexical ac- cess does not necessarily inform us about the most effective conditions for learning. Second, Kroll and Curley examined access to individual words, whereas a context typically comprises a sentence, with all the syntactic constraints that it entails. It is possible that the extra processing of syntax required to understand a sentence

prevents learners from efficiently learning vo-

cabulary in context for a period considerably longer than 30 months. Indeed, Sautermeister (1989) reports vocabulary learning behavior

among university, nonspecialist learners of

English, who are consistently presented with new words in context, but who are not satisfied until they have found an L1 equivalent to assist their learning. This behavior is similar to that described for beginners.

In summary, it appears that learning vocabu-

lary in context is widely perceived by the teach-

ing profession as desirable, but that students either actively resist it, believing the translation condition to be superior, or fail to elaborate the

strategies that might make it possible. This does not mean, however, that context learning is nec-

essarily ineffective because it can be argued that learners are unlikely to adopt appropriate strategies unless they are given guidance. The

question of strategies is thus germane to the issue of vocabulary learning, as indeed it is to L2 learning in general (O'Malley & Chamot, 1990). Proposals for improving learners' ability to use contextual clues to ascertain meanings of unknown words have been put forward by R6hr (1993) in the form of a methodical training pro- gram. Given the right guidance, therefore, it is

likely that even beginning learners can use

metacognitive strategies and metalinguistic knowledge to enhance their comprehension skills when a text contains unknown words. De-

spite the benefits of such training, however, there are two limits to its effectiveness. First, as

pointed out by Hulstijn (1992), learners fre-

quently make incorrect guesses about meanings because L2 contexts do not always provide suffi- cient information to make a correct guess possi- ble even when a learner has been trained to use all the semantic and syntactic clues available. Second, training is profitable only when put to effective use. Over and above the specific tech-

niques for noticing linguistic clues, R6hr points to the importance of fostering pleasure in learners when they search for the meanings of words. This is arguably the most important as-

pect of any language teacher's work, yet one which, in France at least, receives very little at- tention in teacher training programs (except those directed at teaching children as opposed to adolescents). Unfortunately, as Krashen (1987) remarks, most nonspecialist L2 learners are not dedicated linguists, and when faced with a choice between a high-effort strategy such as inferencing and a low-effort shortcut such as translation learning, they will tend to choose the latter.

Although training learners to elaborate meta-

cognitive strategies is of undeniable impor- tance, the experiment described in this study does not examine the effects of such explicit training, but concentrates instead on the

information-processing strategies developed by learners who have followed their natural incli- nations in vocabulary learning. In other words, the subject population under scrutiny consists of nonspecialist learners of English who have received only minimal training in metacogni- tive strategies and whose performance is of in- terest for that very reason. How efficient are such learners in guessing the meaning of un- known words, and to what extent are they able to learn words presented in L2 sentence con- texts? If they continue to show a preference for translation learning, is this indeed a handicap when it comes to accessing their knowledge for use in different contexts? The answers to such

questions will hopefully be useful in making teachers more keenly aware of the processing strategies commonly adopted in inferencing, so that either alternative or complementary strate-

gies may be proposed. The method adopted in the experiment de-

scribed in the present study was to compare per- formance in vocabulary learning as a function of proficiency of the learner, condition of learn-

ing (translation vs. context), and condition of recall (again, translation vs. context). The aim was to assess the ability of learners of different levels (a) to exploit sentence context informa-

Page 5: Second Language Vocabulary Learning: The Role of Context versus

Peter Prince 481

tion both in learning and recall and (b) to transfer what they have learned from one con- text to another. The first point concerns the level of proficiency at which an L2 text starts to

provide a useful basis from which to learn new words. It attempts to shed light on the question raised by Carter and McCarthy (1988) about when it is beneficial to move from one strategy to another. The second point is also important because effective learning for the purposes of

production can be said to have taken place only when the information stored as a result of learn-

ing in one context can be retrieved and used in another. Presumably, teachers encourage learn- ers to learn new words directly in L2 sentence contexts because they feel that learning new vo-

cabulary in the restrictive context of translation links is not conducive to the retrieval and use of that vocabulary in the richer contexts of L2 sen- tences. However, such a belief, although intu-

itively plausible, remains to be demonstrated

empirically. One limitation of the experimental studies

conducted by Seibert (1930) and by Pickering (1982) was that the effect of proficiency upon learners' performance was not examined. Pick-

ering gives no indication of level other than to

say that the subjects were in secondary school.

Although these subjects found context learning no easier than translation learning, we cannot know if this result would still be obtained with more advanced learners.

Cohen and Aphek (1980) examined the effect of proficiency level in their study of Hebrew vocabulary learning by native English speakers. They found that their more advanced learners

displayed clearly superior performance in re- call when the recall task involved L2 contexts.

Specifically, subjects were required to provide the translations of certain words, presented ei- ther in Hebrew or in English, within a Hebrew text. In a pure translation task, on the other hand, no difference appeared between the two

groups of learners. Cohen and Aphek (1980) conclude that context only provides a useful basis once learners have reached a level where

they are not "over-stimulated" by an L2 context. This explanation is plausible if one considers

that subjects' controlled attentional resources are limited (Shiffrin & Schneider, 1977) and that the comprehension of a text, less automatic in a weaker learner, diminishes the resources normally available for the recall of the word. In theory, however, the recall task used allowed subjects enough time to proceed serially, first understanding the text and then retrieving the

translation. Alternatively, given that the task re-

quired retrieving the translations of words that were actually presented, a possible strategy could have been to ignore the context alto-

gether and concentrate on the recall aspect of the task. Clearly, however, the weaker subjects were unable to put either of these strategies into effect and suffered from over-stimulation. The effort required to understand the text appar- ently prevented them from using strategies available to them.

In view of the preceding remarks, it was pre- dicted that in the present experiment, advanced learners would make more efficient use of con- text than weaker learners, both during the

study phase and during recall, but that little difference would emerge as regards the transla- tion condition. Although this result would con- form both to the intuitions and observations of teachers and to the data provided by Cohen and

Aphek (1980), it is important to analyse the pos- sible causes as a first step towards establishing a model of the way L2 learners compensate for a lack of lexical knowledge when processing sentences.

Cohen and Aphek (1980) did not examine different conditions of learning, only of recall; however, it is to be expected that the effort re-

quired to process text would make learning new words in an L2 context considerably more diffi- cult than learning with translations. Although many types of processing are possible in transla- tion learning-such as using keyword tech-

niques (Pressley, Levin, Hall, Miller, & Berry, 1980; Roediger, 1980)-there is only one man-

datory step, that is, associating the L2 word with a familiar, easily retrievable word in the L1.

Learning in context, however, requires that the subject (a) process the sentence to gain suf- ficient understanding of it, (b) use that under-

standing to infer the meaning of the unknown word, and (c) associate the meaning with the form of the unknown word in such a way that a

representation is formed that is available for future use. The third step may involve accessing the translation equivalent in the L1, thus in ef- fect reverting to translation learning, or form-

ing a representation that fits directly into an L2 network. These two processes are not mutually exclusive and may indeed occur in parallel. Whatever the exact nature of the links that are formed, it is clear that the effort required for context learning is greater than that for learn- ing with translations, and such an effort may put a limit on the quantity of material that can be learned in this way.

Page 6: Second Language Vocabulary Learning: The Role of Context versus

482 The Modern Language Journal 80 (1996)

It follows that the purpose of this experiment was not to make an overall comparison between context learning and translation learning, but to look at the two conditions in terms of learners' performance in accessing and using the learned material. As far as the learning itself was concerned, there were three reasons to ex-

pect that the translation condition would pro- vide better results overall. First, according to the encoding specificity principle (Tulving, 1983; Wiseman & Tulving, 1976), probability of recall depends on the similarity between the in- formation provided by contextual cues at recall and the memory trace formed from contextual information during learning. It follows that per- formance should be better when subjects meet the same conditions in the recall phase as in the

study phase. In this experiment, however, this

principle should affect the different combina- tions of conditions unequally. In the translation condition, the contextual cue (i.e., the transla- tion equivalent) did indeed remain the same between the study phase and the recall phase, but in the context condition, the sentences given in the recall phase were not the same as those seen in the study phase. Clearly, therefore, when subjects were placed in the same condi- tion for the study and recall phases, the transla- tion condition gave rise to a lighter cognitive load than the context condition because the cue provided at recall was the same as that seen

during study. The change in the sentences between study

and recall constitutes a major difference from the method used by Pickering (1982), who pre- sented the same sentences during the two

phases. Although such a method makes the task easier for participants, one drawback is that the

experimenter cannot be sure to what extent the unknown word has been learned, or simply been recalled because the context sentences were identical. This distinction is important: The ability to learn a word in an L2, whether it be in context or with a translation, is only of value if the representation thus formed can be transferred to other contexts.

A second difference between the present study and the Pickering (1982) experiment is that his sentences were presented in the sub-

jects' L1, the only L2 item being the word to be learned. Although this provides a guarantee that the sentences have been understood, it does not constitute a natural condition of learn-

ing, and it renders invalid the mechanisms nor-

mally used in L2 reading. The extent to which an L2 context is a help or a hindrance in learn-

ing new words is therefore impossible to detect with this method. For these reasons, it was

judged preferable to present sentences in the L2, having first tested them with a sample of learners of the same level as the experimental subjects in order to make sure that all the sen- tences were easily understandable.

A further element of difficulty in the context condition concerns the time available to sub-

jects during the two phases. In the translation condition, the time available was 30 seconds per word to be learned, and although subjects in the context condition were allowed to work un- til they judged that they had completed the task, it was thought likely that the time spent learning the unknown word, as opposed to de-

coding the sentence, would be less than in the translation condition.

These differences between the two condi- tions mean that a direct comparison between them is of less interest than a comparison be- tween how the two groups of learners make use of the two conditions. A direct comparison be- tween the conditions would necessitate an ex-

perimental design involving not only a cogni- tive load that was equal in each condition, but also two strategies that were opposed. It ap- pears nearly impossible, however, to affirm that this opposition exists because there is nothing to prevent subjects in the context condition from using a translation strategy as well. There- fore, rather than attempt to make the context condition in some way equal to the translation condition, this study kept the two conditions as close as possible to those that are actually met and used by learners and considered the two

approaches from a developmental standpoint. One final remark concerns another aspect of

vocabulary learning examined in this experi- ment, namely the finding that such learning is

asymmetrical, with recall performance being better in the L2 to L1 direction than in the L1 to L2 direction. In other words, having learned, for example, that the French word rideau is a translation equivalent of the English curtain, it would be easier for an English speaker to re- trieve curtain from memory when given rideau than to retrieve rideau when given curtain. This

finding is common in the literature (Ellis & Beaton, 1993), and it was expected that it would be replicated here.

METHOD

The experiment included a factor L (learning condition), with two modalities (context and

Page 7: Second Language Vocabulary Learning: The Role of Context versus

Peter Prince 483

translation), and a factor R (recall condition), also with two modalities (context and transla- tion). In the context learning (CL) condition, the L2 words to be learned appeared in a series of L2 sentences, one unknown word per sen- tence. In the translation learning (TL) condi- tion, subjects received 44 L2 (English) words to learn, accompanied by their L1 (French) equiv- alents. In the context recall (CR) condition, subjects received a series of sentences in which the words studied were replaced with blanks that they were requested to fill. The translation recall (TR) condition involved receiving a list of the words studied and writing their transla- tions. For half of the subjects, the list was given in L2, and for the other half, in L1. The TR con- dition therefore comprised a factor D (direc- tion of translation) which did not appear in the CR condition.

The two conditions, translation and context, were combined with the two phases, learning and recall, in such a way that each subject car- ried out half of the recall phase in the same conditions as the learning phase (TL-TR and CL-CR) and the other half in a different condi- tion (TL-CR and CL-TR).

As well as the factors L (learning condition) and R (recall condition), there was a factor G, corresponding to the two groups of subjects of

differing levels (24 in each group). The test sheets given out in the recall phase were com-

posed in such a way that the words studied were counterbalanced across the two recall condi- tions (translation and context) and, within the translation recall condition, across the two mo- dalities of D (L1 to L2 and L2 to Li).

MATERIAL

The material was composed of 44 English words with an average frequency of 25 per mil- lion (Thorndike & Lorge, 1944). Each word had one principal translation equivalent in French, which was clear and unequivocal. All the words were concrete, referring either to actions or to

objects. Apart from the necessity of using words with one principal translation, the main crite- rion for choosing words was that they should have a high probability of being unknown to the

experimental subjects. This meant that certain words had a very low frequency (less than one per million, for example, for squid) and there- fore might not be considered useful for the av- erage student to learn. However, usefulness was not a criterion adopted in the selection of mate- rial, which had to conform to the constraints

imposed by the experimental task. It may none- theless be supposed that the processing strate-

gies adopted with more frequent words of simi- lar imageability and translatability are akin to those used with the present material.

PROCEDURE

In order to constitute a list of words unknown to the subjects taking part in the experiment, a

pretest was administered comprising three

phases. Subjects first received a sheet with 100 words in the L1 (French) to be translated into the L2 (English). Forty four of the words were assumed unknown to them, and the other words were fillers. The sheets were collected, and sub-

jects received another sheet with 100 English words to be translated into French. The transla- tions of the 44 preselected words appeared on the second sheet, the other words being a differ- ent set of fillers from those used in the previous phase. This procedure controlled for the possi- bility that subjects might fail to find the L2 translation of a word presented in L1, yet re-

spond correctly when the word was subse-

quently presented in L2. Finally, subjects re- ceived a list of 44 sentences in English in which the words presumed unknown were replaced by a blank. Subjects were instructed to fill the blank with a single English word, and if they found no word that they considered suitable, to put a French word instead. This was to verify that the sentence had indeed been understood and that subjects had accessed the meaning of the missing word. On a few rare occasions, a correct English word was found in this contex- tual condition without having been found in the translation conditions.

Among the 44 words assumed to be un- known, the number of words produced in one of the conditions of the pretest was minimal for the weaker group and slightly higher for the advanced group. On account of the diversity of

knowledge, however, there was a wide range of words that were known by only one or two sub-

jects; it was therefore decided to retain all the words for the second part of the experiment, but to take account of subjects' pretest knowl-

edge when analysing the results. A few subjects who knew more than 5 words out of 44 (i.e., 11.4% of the words presented) were nonetheless eliminated from the subject pool.

The study phase and the recall phase took place a week later. In order to reduce the al- ready unlikely probability that subjects might encounter the words in the experiment during

Page 8: Second Language Vocabulary Learning: The Role of Context versus

484 The Modern Language Journal 80 (1996)

the week, they received homework in which none of the experimental words appeared.

During the study phase, subjects were divided into two groups, corresponding to the two mo- dalities of the factor L, the learning condition. Half of them received a sheet with the 44 words and their translations and were instructed to learn them (TL condition). The other half re- ceived the list of 44 sentences that had been used in the pretest, but this time with the target words appearing in place of the blanks. Subjects were instructed to read the sentences, to guess the meaning of the unknown words, and to learn them (CL condition). During this phase, subjects were not allowed to write.

The study phase was of roughly the same

length for the two groups, in other words 20 to 25 minutes, or about 30 seconds per word to be learned. Thirty seconds per word was the time chosen by Pickering (1982), whose method was different in so far as the material to be learned was projected onto a screen, so that the time each item was attended to could be controlled.

Despite being a more rigorous experimental pro- cedure, this method was considered problematic in that learners were unable to use the strategies they normally adopt, such as forming associa- tions between items. The time constraint im-

posed in the present experiment was nonetheless relaxed to allow certain subjects to complete, or consider that they had completed, the task.

After the study phase, which took place at the

beginning of a normal language class, subjects moved on to a different activity for about 40 minutes. In the course of this activity, they en- countered none of the 44 words they had stud- ied. Then the recall phase took place, lasting about the same time as the study phase. Again, the slower subjects were given a few extra min- utes to complete the test. For this part of the

experiment, each subject received a sheet con-

taining 22 words to translate (TR condition) and 22 English sentences that included a blank to be filled with one of the target words (CR condition). In the TR condition, half the sub-

jects received French words to translate into

English, the other half English words to trans- late into French. The sentences used in the CR condition, different from those seen in the

study phase, had been tested previously on a

group of eight learners of the same level as the

subjects belonging to the weaker group in this

experiment. In this test, learners indicated in French the word that was missing in the sen- tence, thus demonstrating that the contexts

presented did indeed allow them to access the

meanings of the missing words. In order to en- sure that this was also the case during the recall

phase of the experiment itself, subjects were asked to fill in with a word in French the blanks

they had not filled in during the test. This in- struction was only given at the end of the exper- iment in order to prevent subjects from resort-

ing to such a strategy too rapidly.

SUBJECTS

Forty eight students, enrolled at the Phar-

macy Faculty of the University of Montpellier, took part in the experiment on a voluntary basis. All the subjects had been studying English for between 5 and 8 years; English was a

compulsory subject in their university studies. For the purposes of the experiment, they were divided into two groups, weak and advanced, on the basis of a placement test taken shortly be- fore the experiment. The test administered was the TOEFL, on which the weak group scored an

average of 397 and the advanced group 480. The placement test was their first encounter with the TOEFL. Subsequent tests revealed that both groups were able to improve their scores

substantially as soon as they became familiar with the specificities of the examination. The main skill not tested by the TOEFL, namely oral

expression, was also assessed separately on a subsection of four students from each group, who were asked to tell a story as depicted by a comic strip without captions. On a number of criteria, ranging from ease of expression (as measured by the number of words produced per minute) to grammatical accuracy, the gap between the two groups was equally as wide as that suggested by the TOEFL scores. Although both groups had studied English for roughly the same amount of time, a questionnaire re- vealed that the weaker group had rarely found the subject motivating or participated actively in class. Their general L2 performance was characterised by a lack of confidence, a fear of

taking risks, and a marked tendency to revert to their L1 whenever possible.

As regards the vocabulary learning pro- cedures previously developed by subjects, the

highly centralized and tightly controlled nature of the French educational system means that these may be regarded as broadly similar for all learners; that is to say that vocabulary and gram- mar are L2 contextualized from the outset, but

specific training in metacognitive strategies is at present sporadic or minimal. Within this gen- eral framework, however, there is obviously

Page 9: Second Language Vocabulary Learning: The Role of Context versus

Peter Prince 485

room for individual teaching practices to vary considerably on the specific matter of vocabu-

lary learning and for learners themselves to

adopt a broad array of strategies.

RESULTS

The averages of correct answers, by learning condition, recall condition, and group, and ex- pressed as a percentage of words to be learned, appear in Table 1.

There was a main effect of group (F [1/23] = 16.57, p < .0002), whereby the more advanced learners had a higher rate of correct answers than the weaker group (67.33% vs. 48.73%). This situation prevailed whatever the learning condition, in context (CL) or with translations (TL). There was no interaction between learn-

ing condition and group. Both groups performed better in the TL than

in the CL condition, and this factor was also significant (F [1/23] = 14.35, p < .0005). This result is hardly surprising, especially as far as the weaker learners are concerned; for reasons already stated in the introduction, the CL con- dition constituted a more arduous task than the TL condition, and this difference in difficulty was also felt by the more advanced learners.

The factor R, condition of recall, was also highly significant (F [1/23] = 35.9, p < .0001). Subjects found it easier to recall words in a translation con- dition than in the context condition (64.43%

vs. 50.64%). This effect is to be interpreted in con- junction with a significant interaction between re- call condition and group (F [1/23] = 53.61, p < .0001), whereby the weaker group performed con- siderably less well in context recall than in transla- tion recall (32.3% vs. 65.2%), whereas the ad- vanced group performed slightly better in the context recall condition (69% vs. 65.7%).

A second aspect of the experiment concerned the possibility that in translation recall, perfor- mance would be better from L2 to L1 than from

L1 to L2. The results in terms of percentage of material recalled are shown in Table 2.

Again, there was a significant effect of the fac- tor L (F [1/23] = 28.09, p < .0001), whereby the translation condition gave rise to better perfor- mance than the context condition (80.2% vs. 50.7%). This factor interacted with the group factor (F [1/23] = 4.49, p < .04), the difference between the two learning conditions being sig- nificantly higher for the weaker group.

Contrary to expectations, the direction of translation did not prove to be a significant fac- tor. Although it did not reach significance, the difference in performance for the weaker group, with 94.6% of correct responses for the L2 to L1 condition as opposed to 77.1% when translating from L1 to L2, is in agreement with previous findings (Ellis & Beaton, 1993). It is to be noted that the advanced group displayed no such difference, their performance being simi- lar irrespective of direction of translation.

TABLE 1 Percentage of Correct Responses by Group, Learning Condition, and Recall Condition, Plus Percentage of Total Correct Responses by Learning Condition and Recall Condition

TL CL M Weak Learners TR 85.81 44.52 65.17

(Group 1) CR 32.37 32.22 32.30 M 59.09 38.37 48.73

Advanced Learners TR 74.54 56.84 65.69 (Group 2) CR 74.03 63.92 68.97

M 74.29 60.38 67.33 TR 80.18 50.68 65.43

Total CR 53.20 48.07 50.64 M 66.69 49.38 58.03

Note. TL = translation learning; CL = context learning; TR = translation recall; CR = context recall.

TABLE 2 Percentage of Correct Answers in Translation Recall, as a Function of Direction of Translation

Weak Learners Advanced Learners (Group 1) (Group 2)

L2-L1 L1-L2 L2-L1 L1-L2 TL 94.55 77.07 TL 73.82 75.27 CL 42.76 46.28 CL 62.90 50.78 Note. TL = translation learning; CL = context learning.

Page 10: Second Language Vocabulary Learning: The Role of Context versus

486 The Modern Language Journal 80 (1996)

DISCUSSION

Given that the second aspect of the experi- ment (i.e., the possibility that vocabulary would be recalled easier when translating from L2 to

L1 than vice versa) did not give rise to a signifi- cant effect of asymmetry, discussion will be lim- ited to the main question under consideration (for a thorough review of the possible reasons for asymmetrical performance, see Ellis & Beaton, 1993; DE Groot, Dannenburg, & van Hell, 1994; Keatley, Spinks, & DE Gelder, 1994; Kroll & Stewart, 1994; Snodgrass, 1993).

Turning to the question of the patterns and

strategies of learning, it emerges clearly that the more advanced group performed far better than the weaker group when the conditions called for a transfer of information learned. Of the four combinations possible between the

learning conditions and recall conditions, three

required transfer of the information learned: (a) TL-CR and (b) CL-TR, but also (c) CL-CR because the sentences changed between the study and recall phases. In the two conditions where sentence contexts were used for recall, the advanced group scored 69%, as opposed to 32.3% for the weaker group. The advantage was smaller in the CL-TR condition: 56.8% versus 44.5%. It is interesting to note that the weaker

group performed better in the sole condition in which no transfer of knowledge was required- where translation learning was followed by translation recall, the scores were 85.8% and 74.5% for the weak and advanced groups, respectively.

The discussion that follows is divided into three broad sections. In the first section, learners' use of translation "links"'2 is exam- ined. The second section looks at learners' use of context, and finally, conclusions will be drawn regarding teaching practice.

Translation Links

It emerges clearly that the weaker group per- formed better when recall was by translation, irrespective of the learning condition. This re- veals that students in the weaker group are far more comfortable with the limited operations involved in forming and retrieving a one-to-one

correspondence than with the more complex and less salient links formed when processing sentences. It is possible that in the CL condition, the weaker subjects used the context to arrive at the meaning of the word, which was then linked to its translation equivalent. The score of 44.5%

in the CL-TR condition suggests that such a strat- egy was at least partially used. The poor score in both the CR conditions indicates an inability to operate outside the one-to-one link. Note that performance is also higher in the TL conditions for the more advanced group: The critical differ- ence between the two groups lies in the ability of the advanced group to use a word learned with its translation in an appropriate L2 context.

The inability of the weaker learners to trans- fer knowledge represents a clear limit to the usefulness of their otherwise impressive ability to learn words with their translations. It may indeed be plausibly suggested that a highly de-

veloped ability to learn words via translation links may in some cases be detrimental to the establishment of the skills and strategies re-

quired to handle discourse. It must be kept in mind that the weak learners who participated in the experiment were by no means at the begin- ning stage of learning English, but had several years of study behind them; it seems likely, therefore, that they are overdependent upon translation links and so have failed to develop certain processing strategies crucial to the ef- fective use of context.

Two remarks may be ventured in connection with this dependence on translation links. First, although the processing required by the experi- mental task involved several aspects that are not

easily teased apart, it may be suggested that on the continuum between controlled and auto- matic processes, the performance of the weaker

subjects reflected a preferential strategy that results at least in part from automatic process- ing. That is to say that linking a new word to its translation equivalent is certainly the most

rapid way of ascertaining its meaning, and until such time as an L2 network is sufficiently organized, it may well be automatic, notwith-

standing teachers' efforts to use pictures or L2 context to convey meaning. Second, the type of

processing adopted depends to a large extent on past habits and attitudes, therefore the re- sults reported here should not be taken as sug- gesting that weak learners are weak because they link L2 words to their translations; after all, had

they received appropriate training in dealing with context, they might well have performed considerably better. Despite this caveat, it seems

likely that persistent reliance on L1 is one of a

complex cluster of factors that lead to ineffec- tive L2 learning and that this reliance stems

largely from a desire to understand quickly. The results reported here suggest that the weaker

group belong to a population of learners who

Page 11: Second Language Vocabulary Learning: The Role of Context versus

Peter Prince 487

use low-effort strategies that lead to rapid learn-

ing but are ultimately ineffective in using broader skills. In view of this, it seems impor- tant, as R6hr (1993) stresses, to foster the pleas- ure of exploring L2 lexical relationships at the earliest possible stage of learning.

The Use of Context

Accounting for the difference between the two groups in the experiment requires a theory of the task under consideration. One important element of context theory is the semantic gap that is created by the presence of an unknown word. Each gap may be said to have a certain

shape, the contours of which are fixed by the clues provided by the other words in the sen- tence. Take an example from the experimental material: "Let's make dinner together, shall we? If you peel the potatoes, I'll cut them and make the chips." The shape of the gap is determined

mainly by the word "potato," which, when com- bined with the reader's previous knowledge of

chip-making, serves to create a mental model of the situation (Johnson-Laird, 1983). Before cor-

rectly ascribing a meaning to the word "peel," however, the learner must eliminate other possi- bilities by noticing, for example, that the cut-

ting involved in chip-making is already ac- counted for, so peeling is probably a different

activity from cutting. The most plausible re-

maining candidate for the meaning of "peel" is "to remove the skin." In uncontrived sentences, of course, the contours of the gap are rarely so well defined, so in practice many sentences may be required before the features that make up the semantic domain of a word are fully ascer- tained. Indeed, it could be objected that even in the sentences used in the present experiment, meanings could not be guessed unequivocally: It is also typical to wash potatoes before cooking them, so "peel" could be construed as a syn- onym for "wash." The fact that subjects were able to fill in the blanks with the correct L1I word during the pretest argues against this pos- sibility, but they may well have felt less confident than the stronger group that their guess was indeed correct. This lack of confidence may be caused by the elaboration of a mental model that is less precise and in which the gap created by the unknown word is correspondingly fuzzy.

Why should this be so when the sentence con- tains only one unknown word? The explanation offered here is an elaboration of what Cohen and Aphek (1980) presumably mean by the "over-stimulation" of textual processing, namely

the "cost" involved in arriving at an under-

standing of the sentence. There are likely to be two causes of this cost. First, learners might process texts more slowly because their L2 net- works are not as richly developed as those of more advanced learners. For example, in the weaker learner's mental lexicon, the word "cut," although a well-known, high-frequency word, does not have as many links with other L2 words as for an advanced learner, who may have formed links with less common words, such as "slice," "chop," "scissors," or "shortcut." The basic premise here is that the richer the net- work in which a word is integrated, the more

automatically the word is accessed, both in

meaning and in form (Keatley, Spinks, & DE

Gelder, 1994). Thus it has been consistently shown that word recognition is slower in an L2 than in one's L1 (e.g., Favreau & Segalowitz, 1983), and, although the difference may only consist in tenths of a second, it is increased as task load increases.

Second, the weaker learner will be slower to use syntactic information. Indeed, research by Barnett (1986) indicates that a poor level of syn- tactic knowledge diminishes reading profi- ciency to the point where understanding is not

improved even by increased lexical knowledge. One implication of this could be that when readers stumble over syntactic analysis, they cannot make the necessary inferences to arrive at the meaning of an unknown word. Thus, in our example, the words "let's" and "together" presuppose that the situation involves more than one person, and the words "you" and "I" in the second half specify the number as two. The word "them" must also be understood as

referring to "potatoes." Even where a sentence contains no potentially ambiguous words, our weaker subjects may simply have been less sensi- tive to these anaphoric elements, thus ham-

pered in their ability to guess the meaning of unknown words. It should be remembered, however, that the material in the present experi- ment was tested on learners whose level was sim- ilar to that of the experimental subjects, who found no difficulty in providing the correct missing word in their L1, and that at the end of the experiment, the subjects themselves were also able to provide the appropriate L1 word. Thus, it cannot be stated that the weaker sub- jects were prevented from making correct infer- ences, only that they were slower to do so. A broader claim that has been made is that where syntactic processing is costly, L2 reading tends to be more conceptually guided (Ulijn, 1980).

Page 12: Second Language Vocabulary Learning: The Role of Context versus

488 The Modern Language Journal 80 (1996)

However, postulating qualitative differences be- tween L1 and L2 processing may not be entirely founded (Frenck-Mestre & Pynte, 1995), and the present analysis serves only to underline that reading in the L2 is undeniably slower (Mack, 1986).

Despite the fact that sentence processing is more laborious for less proficient learners, their weaker performance in the present experiment still calls for a fuller explanation because al- though they had less time to devote to learning the words, it is improbable that lack of time was the sole cause of the difference in performance. Additional factors that could account for the difference must, therefore, have come into play after the moment when subjects had guessed the meaning of the word and should have been

engaged in learning it. The task employed in this experiment was not sensitive to the finer

aspects of learning strategy, but it may be specu- lated that one or more of three different ap- proaches were used. As already mentioned, sub-

jects could have linked the L2 word form to its

L1 equivalent, especially when they were confi- dent that their guess was correct. This would lead to a certain success rate in the TR condi- tion but would not make transfer of knowledge to other contexts easy. Alternatively, they could have relied on context notjust in order to guess the word's meaning but also to provide the main network support for the learning of its

form-using, in other words, the L2 network at their disposal in order to assimilate new word forms. This is possible not only because the CL condition explicitly encouraged it, but also be- cause the weaker learners, despite possessing a less richly developed lexicon, nonetheless have sufficient L2 lexical knowledge to be able to use it as a basis for further learning. Indeed, re- search by Frenck-Mestre and Prince (1995) sug- gests that these learners are able to access lexi- cal knowledge automatically for the purposes of

recognition, so there is no reason why they should not be able to extend the links in their L2 network when conditions are right. Such a result would in fact correspond to what teachers

presume and hope will happen when words are

presented in context. However, the present ex-

periment did not call for the recognition of

newly learned word forms but for their produc- tion, and there is reason to suppose that pro- duction taps into specific mechanisms that may not be available when a network has been devel-

oped mainly for recognition (Green, 1993). A final reason that words were not easily recalled could simply be that once subjects felt that they

had understood the sentence during the study phase, they did not attend to the new word forms enough to be able to retain them. That is, learners failed at recall to retrieve either a translation link or a within-L2 link, where such a link was not established during the study phase. They behaved as if the instruction had been to understand the sentence rather than to learn the unknown word. If this third inter-

pretation is correct, one implication is that the

problem may well be one of attitude and method rather than an inability to exploit con- textual cues per se.

As one might gather from the speculative na- ture of the preceding discussion, ascertaining precisely what processing mechanisms are in- volved in learning vocabulary from context is difficult, given the complexity of such an activ- ity, in which top-down and bottom-up processes are highly interwoven. Although it must be stressed that the interpretations that are given here of these finer aspects are indeed only ten- tative and that further research is required in order to tease them apart, the results of the

experiment do, in fact, appear to be unequivo- cal as regards the main conclusions, namely (a) L2 words are easily learned when presented with their translations and (b) this is no guaran- tee that they will be successfully accessed for use in an L2 context.

Implications for Teaching Practice

The final part of the discussion turns to the

question of how the findings might be taken into account in teaching practice. The sugges- tions put forward are not, of course, to be con- sidered as necessary corollaries of the experi- ment itself because they lie outside its scope. They are nonetheless advanced in the hope they will make a useful contribution to the issue of

vocabulary learning. The picture that emerges is one in which a combination of factors leads to a certain difficulty in transferring knowledge acquired in one context to another context. This difficulty mainly affects learners who have been studying for a number of years but who nonetheless remain ineffective in their use of L2. We suggest that these learners form a spe- cific population with at least the following characteristics:

1. A belief that L2 words are best and most

rapidly understood by linking them to L1 equiv- alents. The presupposition adopted by these learners here is that L2 words can in fact be

adequately linked, from a semantic point of

Page 13: Second Language Vocabulary Learning: The Role of Context versus

Peter Prince 489

view, in a one-to-one correspondence to L1 words.

2. An inability to exploit the L2 lexicon effec-

tively for the purposes of production. 3. Relatively effortful processing of L2 sen-

tences due to slower word identification and

syntactic processing. These three characteristics do not preclude

the existence of a fourth, namely the develop- ment of automatic links within the L2 network, along semantic and associative lines, for the

purposes of recognition. This fourth charac- teristic is not in contradiction with the third, which concerns the recognition of word forms as opposed to word relationships.

In light of these comments, what conclusions

may be drawn for teaching practice? One gen- eral point is that the development of metacogni- tive strategies and attitudes is of prime impor- tance in this domain (O'Malley & Chamot, 1990; Oxford, 1990). Notably, initial expectan- cies as to the effort involved in language learn- ing should be probed and discussed so that the

pitfalls of low-effort strategies like translation learning are well understood from the begin- ning. In the absence of such discussion, some learners are liable to focus on the development of rapid strategies for understanding, to the detriment of the more effortful processes in- volved in production. These observations cor- roborate and extend the output hypothesis as elaborated by Swain (1985).

Turning to the more specific question of learning words in context or with translations, it is noteworthy that when the correct translation link receives sufficient attention during the study phase, subjects' ability to retrieve it is not in doubt. The weakness lies less in the retrieval process itself than in the use to which the infor- mation is put once retrieved. In other words, weaker subjects appear unwilling to envisage the use of an L2 other than in a translation situation. Here again, the problem is one of atti- tude rather than of processing skill itself, indi- cating a need to remedy the situation not only by practice at developing the appropriate skills but also by working upon learners' perceptions of long-term aims and of the ways the task at hand contributes to them.

A further aspect, pointed out by Nation (1982) and Hulstijn (1992), is that a distinction should be drawn between reading for compre- hension and reading as a means to increase one's vocabulary. Although reading for compre- hension no doubt gives rise to incidental learn- ing (Herman, Anderson, Pearson, & Nagy,

1987; Nagy, Herman, & Anderson, 1985), thus contributing to an increase in receptive vocab-

ulary, it is probable that the formation of a rep- resentation that remains accessible for the pur- poses of production requires more elaborate processing than that provided by inferencing. One form of processing would be to pay atten- tion specifically to the form of the word, both orthographic and phonological, by, for exam- ple, repeating it aloud (Ellis & Beaton, 1993); another would be to associate the new word consciously with words already known. A fur- ther stage would involve the retrieval of newly learned words in a variety of production tasks in which the clues provided draw upon all the semantic and associative links that contribute to the meaning of the word to be learned. Where the aim is to master the most common words in a language rapidly (Nation, 1993), such an approach is especially important be- cause it cannot be assumed that even frequently encountered words are readily available for ex- pression. All these strategies for learning vocab-

ulary involve isolating the word from the con- text, so that context provides the means to identify the meaning of the new word and not necessarily the means to learn it. Although a sufficiently imageable context may do that also, the value of context lies above all in its authen- ticity, the benefits of which are of three differ- ent sorts.

First, assessing the meaning of a word in context obliges the learner to develop strategies, such as anticipating and inferencing, which become in- creasingly profitable as learning progresses be- cause they instill an attitude of self-reliance that is the hallmark of proficiency. Second, system- atically meeting new words in context under- lines the fact that words are indeed used in dis- course for purposes of communication. Finally, context provides an indication of the way the words are used. All these factors may be said to contribute to a learner's L2 autonomy and to facilitate the transfer of knowledge that accom- panies it. They also point to the fact that the mental representation of a word's meaning de- velops with successive encounters in different contexts, so it may in principle always undergo modification as new and finer semantic distinc- tions are added (Monsell, 1985).

Despite the benefits of contextual presenta- tion, it appears that effective learning of words requires a stage in which the word is in fact isolated from its context and submitted to elab- orative processing. The extent to which links between concepts and L2 word forms can be

Page 14: Second Language Vocabulary Learning: The Role of Context versus

490 The Modern Language Journal 80 (1996)

activated without some form of L1 activation remains a matter of debate (e.g., DE Groot, Dannenburg, & van Hell, 1994; Kroll & Stewart, 1994), but it is probably safe to assume that for most words that have an obvious translation

equivalent, ascertaining the meaning of a word is in many cases tantamount to activating an L1 representation, regardless of whether the L1 word is actually presented or not. However, given that learning with translations, despite its

efficiency in terms of quantity, includes none of the benefits just described for contextual pre- sentation, there appears to be a need for learn-

ing techniques and material that manage to combine both efficiency and learner autonomy. As Hulstijn (1992) rightly points out, the ques- tion is not so much whether learning in context is beneficial, but exactly what form it should take to be most effective.

One solution might be to train learners not

simply to ascertain meaning from context, but to select consciously from within that context the most pertinent word associated with the word to be learned, thus creating a paired- associate learning condition without resorting to translations. This initial link between two L2 word forms would serve mainly as a prop that would help to sustain the newly learned word form in memory, but in order for the meaning of that word to become more firmly integrated in a network, it is essential that the word be encountered at regular intervals and in a variety of contexts and tasks. In order to achieve maxi- mum effectiveness, such tasks would have to in- clude some form of elaborative processing (Schacter & Graf, 1986). If this does not hap- pen, the resulting representation would possi- bly be accessible for recognition purposes or for tasks involving implicit memory, but not for ex-

plicit recall. On the part of the teacher, such an

approach requires careful planning, both in se-

lecting the vocabulary items that are most suited to the learners' needs, and in setting up the tasks, such as the generation of synonyms, collocations, or whole sentences, which will lead to effective learning. These tasks should also be an object of occasional analysis themselves, such that the learning process becomes trans-

parent to the learner, who ideally should be able to perform such tasks autonomously, and more

automatically, as proficiency increases. In view of the ease with which L1 representa-

tions are accessed, it is to be expected, however, that in quantitative terms any technique that relies only upon the L2 network would still be less efficient than translation learning, espe-

cially in the early stages; however, this would be more than counterbalanced by the fact that the learner is operating to a greater degree within the L2. At the very least, such a technique pro- vides an approach to vocabulary learning in which the dichotomy between learning with translations and learning in context is replaced by a conscious awareness, on the part of the learner, that both learning strategies have their

disadvantages and that alternative approaches could usefully be explored.

NOTES

1 The word "trace" is used here in accord with the view that an encounter with any stimulus results in that stimulus and the conditions in which it was en- countered being stored in memory as a "trace" that will decay or be disrupted by other traces if not subse- quently reactivated (Baddeley, 1990, pp. 43-50).

2 The term "link" is used here to cover not only the objective linguistic and semantic relationships that exist between words, but also the manner in which such relationships are stored in the mental lexicon. Concerning the latter aspect, it is worth noting first, that outside of highly meaningful events, memory traces will need to be activated many times for a link to become firmly established, and second, that the links thus formed are not all of the same strength.

REFERENCES

Baddeley, A. (1990). Human Memory, Hove, England: Lawrence Erlbaum Associates.

Barnett, M. A. (1986). Syntactic and lexical/semantic skill in foreign language reading: Importance and interaction. The Modern Language Journal, 70, 343-349.

Carter, R., & McCarthy, M. (1988). Vocabulary and Lan- guage Teaching, New York: Longman.

Cohen, A. C., & Aphek, E. (1980). Retention of second-language vocabulary over time: Investi- gating the role of mnemonic associations, Sys- tem, 8, 221-235.

DE Groot, A. M. B., Dannenburg, L., & van Hell, J. G. (1994). Forward and backward word translation by bilinguals. Journal of Memory and Language, 33, 600-629.

Ellis, N., & Beaton, A. (1993). Factors affecting the learning of foreign language vocabulary: Imag- ery keyword mediators and phonological short- term memory, The QuarterlyJournal ofExperimental Psychology, 46A, 3, 533-558.

Favreau, M., & Segalowitz, N. (1983). Automatic and controlled processes in the first- and second-

Page 15: Second Language Vocabulary Learning: The Role of Context versus

Peter Prince 491

language reading of fluent bilinguals. Memory & Cognition, 11, 565-574.

Frenck-Mestre, C., & Prince, P. W. J. (1995). Second

language autonomy. Unpublished manuscript. Frenck-Mestre, C., & Pynte, J. (in press). Reading in

the native versus the second language. Quarterly Journal of Experimental Psychology.

Gekoski, W. L. (1980). Language acquisition context and language organization in bilinguals.Journal of Psycholinguistic Research, 9, 429-449.

Green, D. (1993). Towards a model of L2 comprehen- sion and production. In R. Schreuder & B. Weltens (Eds.), The bilingual lexicon (pp. 249- 278). Amsterdam/Philadelphia: John Benjamins.

Herman, P. A., Anderson, R. C., Pearson, P. D., &

Nagy, W.E. (1987). Incidental acquisition of word meaning. Reading Research Quarterly, 22, 263-284.

Hulstijn,J. H. (1992). Retention of inferred and given meanings: Experiments in incidental vocabu-

lary learning. In P. Arnaud & H. Bejoint (Eds.), Vocabulary and Applied Linguistics. (pp. 113-125). London: Macmillan.

Johnson-Laird, P. N. (1983). Mental models. Cambridge, England: Cambridge University Press.

Keatley, C. W., Spinks, J. A., & DE Gelder, B. (1994). Asymmetrical cross-language priming effects.

Memory and Cognition, 22, 70-84. Krashen, S. D. (1987). Principles and practice in second

language acquisition. London: Prentice Hall International

Kroll, J. E, & Curley, J. (1988). Lexical memory in novice bilinguals: The role of concepts in re-

trieving second language words. In M. M.

Gruneberg, P. E. Morris, & R. N. Sykes (Eds.), Practical aspects of memory, Vol 2 (pp. 389-395). London: John Wiley and Sons.

Kroll,J. F., & Stewart, E. (1994). Category interference in translation and picture naming: Evidence for asymmetric connections between bilingual memory representations. Journal of Memory and

Language, 33, 149-174. Mack, M. (1986). A study of semantic and syntactic proc-

essing in monolinguals and fluent bilinguals. Jour- nal of Psycholinguistic Research, 15, 463-489.

McCarthy, M. (1990). Vocabulary. Oxford, England: Ox- ford University Press.

Monsell, S. (1985). Repetition and the lexicon. In A.W. Ellis (Ed.), Progress in the psychology of lan-

guage, Vol. 2 (pp. 147-195). London: Lawrence Erlbaum Associates.

Nagy, W. E., Herman, P. A., & Anderson, R. C. (1985). Learning words from context. Reading Research Quarterly, 20, 233-253.

Nation, I. S. P. (1982). Beginning to learn foreign vo- cabulary: A review of the research, RELC Jour- nal, 13, 1, 14-36.

Nation, I. S. P. (1993). Vocabulary size, growth, and use. In R. Schreuder & B. Weltens (Eds.), The bilingual lexicon. (pp. 115-124). Amsterdam/ Philadelphia: John Benjamins.

O'Malley, J., & Chamot, A. (1990). Learning strategies in second language acquisition. Cambridge, England: Cambridge University Press.

Oxford, R. (1990). Language learning strategies: What ev-

ery teacher should know. Rowley, MA: Newbury House.

Paivio, A., & Desrochers, A. (1980). A dual-coding approach to bilingual memory. Canadian Journal of Psychology, 34, 388-399.

Pickering, M. (1982). Context-free and context-

dependent vocabulary learning: An experi- ment. System, 10, 1, 79-83.

Pressley, M., Levin, J. R., Hall. J. W., Miller, G. E., &

Berry, J. K. (1980). The keyword method and

foreign word acquisition. Journal of Experimental Psychology: Human Learning and Memory, 6, 163- 173.

Roediger, H. L. (1980). The effectiveness of four mne- monics in ordering recall. Journal ofExperimental Psychology: Human Learning and Memory, 6, 558- 567.

R6hr, G. (1993). Erschliessen aus dem Kontext. Berlin, Ger-

many: Langenscheidt. Saragi, T., Nation, I. S. P., & Meister, G. F. (1978).

Vocabulary learning and reading. System, 6, 72- 78.

Sautermeister, C. (1989). Pour une meilleure compe- tence lexicale. In A. H. Ibrahim (Ed.), Lexiques (pp. 121-133). Paris : Hachette C1e Internationale.

Schacter, D. L., & Graf, P. (1986). Effects of elabora- tive processing on implicit and explicit mem-

ory for new associations. Journal of Experimental Psychology: Learning, Memory and Cognition, 12, 432-444.

Seibert, L. C. (1930). An experiment on the relative

efficiency of studying French vocabulary in as- sociated pairs versus studying French vocabu-

lary in context. Journal of Educational Psychology, 21, 297-314.

Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human information processing: II. Perceptual learning, automatic attending, and a general theory. Psychological Review, 84, 2, 127-190.

Snodgrass, J. G. (1993). Translating versus picture naming: Similarities and differences. In R. Schreuder & B. Weltens (Eds.), The bilingual lexi- con (pp. 83-114). Amsterdam/Philadelphia: John Benjamins.

Swain, M. (1985). Communicative competence: Some roles of comprehensible input and comprehen- sible output in its development. In S. Gass and C. Madden (Eds.), Input in second language acquisi- tion (pp. 235-253). Rowley, MA: Newbury House.

Thorndike E. L., & Lorge, I. (1944). The Teacher's Word- book of 30,000 words. New York: Bureau of Publi- cations, Teacher's College, Columbia Univer- sity.

Tulving, E. (1983). Elements of Episodic Memory. Oxford, England: Oxford University Press.

Page 16: Second Language Vocabulary Learning: The Role of Context versus

492 The Modern Language Journal 80 (1996)

Ulijn, J. (1980). Foreign language reading research: Recent trends and future projects. Journal of Re- search in Reading, 3, 17-37.

Vermeer, A. (1992). Exploring the second language learner lexicon. In L. Verhoeven & J. De Jong (Eds.), The construct of language proficiency. Applica- tions of psychological models to language assessment

(pp. 147-162). Amsterdam/Philadelphia: John Benjamins.

Webb, W. B. (1962). The effects of prolonged learning on learning vocabulary. Journal of Verbal Learning and Verbal

Behavior, 1, 173-182. Wiseman, S., & Tulving, E. (1976). Encoding specific-

ity: Relations between recall superiority and

recognition failure. Journal of Experimental Psy- chology: Human Learning and Memory, 2, 349-361.

APPENDIX

Material Used

Representative sentence contexts are given for 10 items. Remaining items are given with their French translations.

(a) = context condition, study phase (b) = context condition, recall phase (c) = translation condition

1. (a) -Why are those eggs all broken?

-Well, someone put the eggbox on the chair, and I didn't see it, and I sat on it and crushed it.

(b) Wine is made by grapes and let-

ting the juice ferment. (c) to crush (icraser)

2. (a) The cowboy, chased by a group of ferocious

Indians, whipped his horse to make it go faster.

(b) In the 19th century the punishment for desert-

ing the army was to be across the back fifty or a hundred times.

(c) to whip (fouetter)

3. (a) When it was built, many people said that the

Titanic could not sink, but it went down on its

very first voyage. (b) Most people manage to float on the water but I

don't know how; I just to the bot- tom.

(c) to sink (couler)

4. (a) The owl has the reputation of being a very in-

telligent bird because it often looks profoundly pensive.

(b) The is a nocturnal bird that feeds mostly on mice.

(c) an owl (un hibou)

5. (a) She broke a leg skiing two weeks ago and now

she uses crutches to help her walk.

(b) are very useful because they al- low you to move about when you have a leg in

plaster.

(c) crutches (des bMquilles)

6. (a) That man's been drinking for hours; look-he

has to lean against the bar to stop himself from

falling over. (b) Don't against that wall! The

paint isn't dry-your pullover will be ruined! (c) to lean (s'appuyer)

7. (a) Don't come in here without shoes! I've just

broken a glass and I've got to sweep the floor.

(b) In October I have to the patio every day because of all the leaves that fall from the acacia.

(c) to sweep (balayer)

8. (a) -Don't spit your coffee out like that! It's

disgusting! -You would too if you put salt in it instead of

sugar! (b) A llama is a South American animal that has

the reputation of at people when it is angry.

(c) to spit (cracher)

9. (a) A butcher usually wears a white apron when he

is working, to stop blood getting on his clothes.

(b) I always wear a big blue when I do the cooking; if I don't I get my clothes all

dirty. (c) an apron (un tablier)

10. (a) Let's make the dinner together, shall we? If you

peel the potatoes, I'll cut them and make the

chips. (b) Before eating an orange you have to

it, as the skin is not nice to eat.

(c) to peel (iplucher)

Remaining items, with their translations: to dive (plonger) hedgehog (herisson) lobster (homard) witch (sorcidre) dwarf (nain) to faint (s'vanouir) toad (crapaud) to knit (tricoter) stilts (echasses) to stir (remuer)

Page 17: Second Language Vocabulary Learning: The Role of Context versus

Peter Prince 493

lightning (foudre) stool (tabouret) dove (colombe) to bounce (rebondir) to wrap (emballer) to wink (cligner)

to plough (labourer) to juggle (jongler) to grind (moudre) skittles (quilles) beaver (castor) to kneel (sagenouiller)

log (Iniche) dusk (cripuscule) lice (poux) squid (calamar) leek (poireau) to knead (petrir)

dew (rosie) cobbles (pavis) ladle (louche) crow (corbeau) jug (pichet) to stab (poignarder)

In Recognition of Excellence in Dissertation Research

THE MLJ IS PROUD TO ANNOUNCE THE WINNER OF THE 1996 EMMA Birkmaier Award for Doctoral Dissertation Research in Foreign Language Education, presented by the American Council on the Teaching of Foreign Languages (ACTFL). Sharon Wilkinson, West Virginia University Thesis title: "Foreign Language Conversation and the Study Abroad Transition: A Case Study." Directors: Jeannette Bragger

Rebecca Kline

Pennsylvania State University Professor Wilkinson was honored and awarded a $500 cash prize from the MLJat the ACTFL Annual

Meeting in November in Philadelphia.

ACTFL Seeks Candidates for the 1997 Birkmaier Dissertation Award

THE MODERN LANGUAGE JOURNAL SEEKS CANDIDATES FOR THE ACTFL EMMA MARIE Birkmaier Award for Doctoral Dissertation Research in Foreign Language Education. This award was established in 1980 to recognize the author of doctoral dissertation research in foreign language education that contributes to the advancement of the profession. 1. Dissertations are nominated either by the dissertation advisor, another faculty member, or the candidate. 2. The nomination must consist of an application form (available from ACTFL), an abstract of the thesis, and a letter of nomination from a faculty member. The letter is to describe the thesis, situate it in the field, and relate the significance of its contribution to the theory or practice, or both, of second or foreign language learning. 3. Committee members rank the nominations to determine a list of finalists. 4. Finalists submit full dissertations from which the award winner is selected.

For inquiries or nominations, contact: The chair of the 1997 Birkmaier Award Committee or ACTFL

Judith Liskin-Gasparro 6 Executive Plaza Dept. of Spanish & Portuguese Yonkers, NY 10701-6801 University of Iowa 914-963-8830 211 Schaeffer Hall FAX: 914-963-1275 Iowa City, IA Email: [email protected]

The recipient will receive a $500 cash award, provided by the MLJ. The award will be presented at the

ACTFL annual meeting.