Measuring L2 Lexical Growth Using Hypernymic Relationships
Total Page：16
File Type：pdf, Size：1020Kb
lang_508 langxml-als.cls (1994/07/13 v1.2u Standard LaTeX document class) 3-4-2009 :626 LANG lang_508 Dispatch: 3-4-2009 CE: N/A Journal MSP No. No. of pages: 28 PE: Matthew 1 Language Learning ISSN 0023-8333 2 3 4 5 Measuring L2 Lexical Growth Using 6 7 Hypernymic Relationships 8 9 Scott Crossley 10 Mississippi State University 11 12 Tom Salsbury 13 Washington State University 14 15 Danielle McNamara 16 University of Memphis 17 18 This study investigated second language (L2) lexical development in the spontaneous 19 speech of six adult, L2 English learners in a 1-year longitudinal study. One important 20 aspect of lexical development is lexical organization and depth of knowledge. Hyper- 21 nymic relations, the hierarchical relationships among related words that vary in relation 22 to their semantic specificity (e.g., Golden Retriever vs. dog vs. animal), are an important 23 indicator of both lexical organization and depth of knowledge. Thus, this study used 24 hypernymy values from the WordNet database and a lexical diversity measure to analyze 25 lexical development. Statistical analyses in this study indicated that both hypernymic 26 relations and lexical diversity in L2 learners increase over time. Additionally, lexical 27 diversity and hypernymic values correlated significantly, suggesting that as learners’ lex- 28 icons grow, learners have access to a wider range of hypernymy levels. These findings 29 are discussed in relation to developing abstractness in language, extending hypernymic knowledge, and the growth of lexical networks. 30 31 32 33 The research reported here was supported by the Institute of Education Sciences, US Department 34 of Education, through Grant IES R3056020018-02 to the University of Memphis. The opinions 35 expressed are those of the authors and do not represent views of the Institute or the US Department 36 of Education. The authors would also like to thank Dr. Philip McCarthy of the University of 37 Memphis for his statistical and methodological assistance. Additionally, the authors are indebted 38 to the three anonymous reviewers who provided critical assessments of the early and late versions of this article. 39 Correspondence concerning this article should be addressed to Scott A. Crossley, Department 40 of English, Mississippi State University, P.O. Box E, Starkville, MS 39759. Internet: sc544@ 41 msstate.edu Language Learning 59:2, June 2009, pp. 307–334 307 C 2009 Language Learning Research Club, University of Michigan lang_508 langxml-als.cls (1994/07/13 v1.2u Standard LaTeX document class) 3-4-2009 :626 1 Crossley, Salsbury, and McNamara Hypernymic Relationships 2 3 Keywords hypernymy; hyponymy; superordinates; subordinates; lexical networks; cor- 4 pus linguistics; lexical diversity; computational linguistics 5 6 7 Introduction 8 Lexical proficiency in second language (L2) learners is far detached from 9 simplistic performance assessments such as number of words used or com- 10 prehended or the ability to match dictionary definitions with strings of letters. 11 It is now more fully recognized that an L2 learner’s lexical proficiency in- 12 cludes knowledge of syntactic properties, conceptual levels, sense relations, 13 and complex lexical association models (Haastrup & Henriksen, 2000; Huckin 14 & Coady, 1999). The lexical proficiency of L2 learners is of special interest 15 because misinterpretations of lexical items are key elements in global errors 16 that inhibit communication (de la Fuente, 2002; Ellis, 1995; Ellis, Tanaka, & 17 Yamakazi, 1994). Lexical proficiency is also crucial because the understanding 18 of lexical acquisition in relation to its deeper, cognitive functions can lead to 19 increased awareness of how learners process and produce an L2. Additionally, 20 studies into the development of L2 lexical proficiency, until recently, have been 21 an often neglected area of study in L2 acquisition (Meara, 2002). Those stud- 22 ies that have looked at lexical development have examined broad measures of 23 growth, such as lexical accuracy, lexical frequency, and lexical diversity (Polio, 24 2001). Studies that consider lexical growth in L2 learners are important be- 25 cause lexical growth strongly correlates with academic achievement (Daller, 26 van Hout, & Treffers-Daller, 2003). 27 This study explores one aspect of lexical proficiency: sense relations. Of 28 interest is the development of sense relations in L2 lexical acquisition and 29 how these sense relations relate to the richness of word knowledge and the 30 depth of vocabulary knowledge exhibited by L2 learners (Wesche & Paribakht, 31 1996). There are multiple variables that can be studied when analyzing depth 32 of L2 lexical knowledge. These include measures of polysemy, hypernymy 33 (superordinate and subordinate terms), synonymy, semantic overlap, and other 34 variables related to the conceptual meaning of words such as concreteness and 35 imageability. Although studies of L2 depth of lexical knowledge have been 36 rare, those studies conducted have generally demonstrated that L2 learners 37 generally do not have as much depth of lexical knowledge as do first language 38 (L1) speakers of those languages (Ordonez, Carlo, Snow, & McLaughlin, 2002; 39 Verhallen & Schoonen, 1993, 1998). 40 This study will concentrate on the growth of hypernymic relations 41 in L2 learners’ lexicon. Hypernymic relations are semantic links between Language Learning 59:2, June 2009, pp. 307–334 308 lang_508 langxml-als.cls (1994/07/13 v1.2u Standard LaTeX document class) 3-4-2009 :626 1 Crossley, Salsbury, and McNamara Hypernymic Relationships 2 3 conceptually related words such as animal and dog. In this example, ani- 4 mal is semantically linked to dog but functions as a superordinate term, as it 5 is more abstract than the concrete, subordinate term dog. Although there have 6 been a few studies that have analyzed the hypernymic knowledge of bilingual 7 speakers (Defour & Kroll, 1995; Ordonez et al., 2002; Sharifian, 2002) and 8 L2 learners (Ijaz, 1986; Levenston & Blum, 1977), to our knowledge no study 9 has investigated the idea that as L2 learners acquire language; they also extend 10 and further develop lexical hypernymic relations. The possibility of investigat- 11 ing hypernymic relations in relationship to the growth of lexical networks was 12 raised by Haastrup and Henriksen (2000), but they did not directly investigate 13 hypernymy and instead measured synonymy, lexical gradation, and antonymy. 14 They argued that these were more crucial aspects of adjectives, which was the 15 linguistic category that was the focus of their study. This study, however, will 16 analyze verb and noun use and is thus well suited for investigating hypernymic 17 relationships. An analysis such as this could provide crucial information about 18 L2 lexical growth in three ways. First, it could provide evidence for the de- 19 velopment of lexical proficiency based on the use of extended sense relations. 20 Second, it could provide supporting evidence for the development of lexical net- 21 works in L2 language systems. Finally, it could provide evidence of abstraction 22 in L2 language use. 23 24 25 26 Hypernymy Defined 27 As stated earlier, this study will examine the growth of lexical hypernymic rela- 28 tions in L2 learners. It will not analyze the growth of L2 hypernymic concepts 29 because the data examined in this study come from adult L2 learners, who, 30 ostensibly, have fully developed conceptual knowledge of the world. Lexical 31 hypernymy is considered a fundamental semantic relationship that is founded 32 on the connection between general and specific lexical items (Chaffin & Glass, 33 1990; Haastrup & Henriksen, 2000). Hypernymic relations are hierarchical 34 associations between hypernyms (superordinate words) and hyponyms (subor- 35 dinate words). A hypernym is defined as a word that is more general than a 36 related word (animal compared to dog) and a hyponym is more specific than 37 a related word (dog as compared to animal). Another example of hypernymy 38 is the association between car and vehicle. In this case, car is the hyponym of 39 the hypernym vehicle because car has a narrower and more specific denotative 40 scope than vehicle, which would also include trucks, go-carts, golf carts, and 41 hearses. 309 Language Learning 59:2, June 2009, pp. 307–334 lang_508 langxml-als.cls (1994/07/13 v1.2u Standard LaTeX document class) 3-4-2009 :626 1 Crossley, Salsbury, and McNamara Hypernymic Relationships 2 3 An important aspect of hypernymy is the notion of basic level categories. 4 The theory of basic level categorization avers that an object at a specific level of 5 categorization has a superior status and is the concept that is used most often to 6 discuss an object (Brown, 1958). An important element that identifies whether 7 an object is a basic-level category is cue validity. Cue validity is premised 8 on the collection of features that distinguish an object from other objects. 9 The strongest cues are those that automatically place an object in a category 10 (e.g., gills for fish). Cue validity is important to basic categories because basic 11 categories are the level at which the largest number of attributes are contained. 12 All other members of a category (subordinates and superordinates) belong to 13 that category based on the number of features they share with the basic category. 14 Thus, car is a basic category, as it contains the most features that allow it to 15 be distinguished from other objects at a similar level (e.g., motorcycles, trucks, 16 and golf carts). Its subordinate, sedan, has lower cue validity because most of 17 its cues are shared across the category and few features distinguish it from a 18 car. In comparison, the superordinate category vehicle contains fewer shared 19 attributes of the basic category car (Rosch, Mervis, Gray, Johnson, & Boyes- 20 Braem, 1976). In summary, a basic category word is preferred for concept 21 labeling because subordinates are less distinctive but more informative, whereas 22 superordinates are more distinctive but not more informative.