Semantic category effects in second language word learning

The present study addresses a long-standing assumption in the field of applied linguistics: that presenting new second language (L2) vocabulary in semantically grouped sets is an effective method of teaching. Participants learned 32 new L2 labels for concepts from four different semantic categories in either a related or unrelated condition. At test, participants translated words in both translation directions. We found a semantic interference effect both during the encoding of information into memory and during the retrieval of information in translation. We discuss these findings in terms of theoretical models of L2 lexical representation and development, as well as in more practical terms of L2 curriculum design and vocabulary instruction. A long-standing question within the field of second language acquisition (SLA) has been how second language (L2) vocabulary should be taught: in semantic groups or not? Several researchers and textbook authors assume the efficacy of presenting learners with new L2 vocabulary in semantically related sets. For example, Gairns and Redman (1986), in their well-known book Working with Words, argue that grouping words by meaning “can provide greater precision in guiding students towards meaning, and in helping them to define the boundaries that separate lexical items” (p. 32). Similarly, Seal (1991) explains that when words are learned in semantic sets, “the learning of one item can reinforce the learning of another,” as well as facilitate understanding because “items that are similar in meaning can be differentiated” (p. 300). These authors assume at least two benefits associated with learning words in semantic sets: the similarity between items (a) serves to facilitate the learning task and (b) causes the learner to notice fine-grained distinctions between words, which leads to a better understanding of the words that are learned. A brief examination of several of the current English as a second language textbooks reveals that the practice of introducing new vocabulary in semantic sets is a popular one. For example, in one widely used textbook, Vistas (Brown,  2003 Cambridge University Press 0142-7164/03 $12.00 Applied Psycholinguistics 24:3 370 Finkbeiner & Nicol: L2 word learning 1991), the “target vocabulary” for chapter 1 consists of items necessary in a classroom (e.g., paper, pen, pencil, chalk, blackboard, eraser). In another equally popular textbook, ExpressWays (Molinsky & Bliss, 1996), family members (e.g., husband, wife, mother, father, brother, sister) and places in the community (e.g., airport, bank, post office, mall, park, library, museum) serve as the target vocabulary in the first two lessons. Although many SLA theorists and practitioners endorse (implicitly or explicitly) the seemingly sensible position that teaching new L2 vocabulary in semantically grouped sets is an effective method of teaching, there is actually very little empirical evidence to support this position. The body of literature often cited in support of presenting learners with semantically grouped words includes (monolingual) memory studies that involve the following: first there is a study phase, in which subjects are given a series of words (all of which are well known to the subjects) and told to memorize them; then there is a test phase, which requires subjects to either recall all the words from the study phase or to recognize which words from a list of words had appeared in the study phase. Such studies have found that semantically grouping the study words facilitates later recall or recognition (Bousfield, 1953; Cofer, 1966; Cohen, 1963). In a closely related set of studies, it has been shown that the use of category labels during the study phase and/or during later recall of word lists has improved performance (Lewis, 1971; Tulving & Pearlstone, 1966; Tulving & Psotka, 1971). Some SLA researchers have interpreted these findings as providing support for the idea of using semantic sets in L2 vocabulary teaching. For instance, Schmitt (1997) argues that, just as grouping works to enhance recall of words in list-recall tasks in native speakers, “there is no reason to believe it does not do the same for L2 learners” (p. 213). It is not entirely clear, however, how a task that requires subjects to recall already well-known words that appeared on a study list is similar to the task of learning brand new L2 labels for already established concepts that have well-established first language (L1) labels. In a recall task, subjects could benefit from semantic grouping because it provides an organizational schema. If the subject notices that the study items all belong to a particular semantic category, at test the participant simply has to generate a list of appropriate exemplars for that category and check each one off against a record of items established during the study phase. If a generated item matches one of the subject’s record of items, then she or he can “recall” it for the experimenter. This effectively turns the recall task into a recognition task, which is much easier to perform. Recall is much more difficult when participants are unable to make use of an organizational schema (Bower, Karlin, & Dueck, 1975). In contrast, in an L2 vocabulary–learning task, it is unclear how awareness of the category label would be beneficial. The learning of new L2 labels for already established concepts requires the creation of new form–meaning connections. A strong, stable, one-to-one connection between a concept and its L1 form needs to give way to a one-to-two connection between concept and both the L1 form and L2 form (c.f. Kroll & Stewart, 1994; Potter, So, Von EcApplied Psycholinguistics 24:3 371 Finkbeiner & Nicol: L2 word learning kardt, & Feldman, 1984). The precise nature of these connections has been the subject of much debate since the early work of Kolers (1963, 1966), but it is reasonable to suppose that the connection between concepts and L2 forms is quite weak in the initial stages of acquisition (Talamas, Kroll, & Dufour, 1999). The process of L2 vocabulary learning could thus be characterized as one of first establishing and then strengthening the new L2 form–meaning connections. This is clearly quite different from remembering a series of L1 words. Thus, it is not surprising that studies that focus on the learning of new labels show a different effect of semantic grouping. In studies that record the number of learning trials needed to reach a predetermined learning criterion, it has been shown that participants take longer to learn new labels for sets of semantically related items than for sets of semantically dissimilar items (Higa, 1963; Kintsch & Kintsch, 1969; Nation, 2000; Tinkham, 1993, 1997; Underwood, Ekstrand, & Keppel, 1965; Waring, 1997). Note, however, that although learning semantically related words appears to take longer, it is possible that words learned under these conditions are learned better for the purpose of actual language use (e.g., the retrieval of vocabulary during production and comprehension). That is, the very difficulty associated with learning the new labels may make them easier to process once they are learned (c.f. Craik & Lockhart, 1972; Craik & Tulving, 1975, for a discussion of the effect of “depth of processing” on retrieval). Conversely, it is possible that whatever makes semantically grouped words more difficult to learn also makes them more difficult to retrieve during later language use. There is, in fact, some evidence that L2 learners are slower to process words presented in semantic clusters. Kroll and Stewart (1994), for example, found that their Dutch–English bilingual subjects (for whom Dutch was the L1) were slower to translate Dutch words into English when the words were grouped by semantic category than when they were not. This result only held for L1–L2 translation, not for L2–L1 translation. Kroll and Stewart argued that this asymmetry is due to the existence of a direct link between L2 words and L1 words and that translation from L2 to L1 simply does not invoke conceptual representations. On Kroll and Stewart’s view, no conceptual effects (such as semantic grouping) are expected for L2–L1 translation. This view has not been without challenge. Both Altarriba and Mathis (1997) and La Heij, Hooglander, Kerling, and van der Velden (1996) have reported semantic interference effects in L2–L1 translation tasks. In the Altarriba and Mathis (1997) study, subjects learned a small number of L2 words in association with L1 translation equivalent forms and were then asked to perform a translation verification task, in which pairs of L1 and L2 words were presented. The subjects were to decide whether the pairs were translation equivalents. Among the “no” cases were pairs that were semantically related but not equivalent. Reaction times to these semantically related foils were longer than for unrelated words, indicating the involvement of conceptual representations. This suggests that direct L2 form–meaning connections had been established even after one training session. Overall, then, there is evidence that even novice L2 learners may exhibit semantic interference effects (presumably the result of direct associations between newly learned L2 Applied Psycholinguistics 24:3 372 Finkbeiner & Nicol: L2 word learning forms and meaning representations). Furthermore, it appears that presenting new labels by semantic category makes the labels harder to learn and, once learned, harder to retrieve. In the present study, we tested these ideas directly by having a training phase in which participants learned new labels for familiar concepts and then a test phase in which they were required to retrieve those labels in a translation task. We manipulated semantic grouping both during training and during test. In addition, our test phase included both L1–L2 translation and L2–L1 translation, thereby providing an opportunity to extend the findings of Altarriba and Mathis (1997) to a language production task and to test whether Kroll and Stewart’s (1994) asymmetrical pattern of translation results extends to novice language learners.