A Depth of Individual Word Knowledge Model for the Mental Lexicon

Brent Wolter

Hokkaido University

  Introduction

Word association tests investigating the structure of the L2 mental lexicon have indicated that it is in some ways structurally different than that of native speakers (Channell 1990; Meara, 1982, 1984). The support for this position comes mostly from the results of word association tests, which have shown that non-native speakers produce high proportions of syntagmatic and so-called eclangf responses to prompt words. Native speakers, on the other hand, have been shown to produce primarily paradigmatic responses.[1] This lead researchers to conclude that (a) the connections between words in the second language learnerfs mental lexicon are less stable, (b) phonology appears to play a much more prominent role in organizing the L2 mental lexicon, and (c) the semantic links between words tend to differ in a systematic way from those of native speakers. However, most of the comparable data for native speakers have been generated from a relatively small number of common prompt words that tend to elicit a similarly limited set of stable and predictable responses. When less common prompt words have been used, however, native speakers have been shown to produce a substantial number of what could be described as gnonnative-likeh responses (Postman, 1970; Stolz & Tiffany, 1972).

When viewed as a whole, these results are enough to pose the question of whether or not the deviations observed in responses of nonnative speakers are in fact the result of a structurally different mental lexicon, or perhaps a structurally similar, although often smaller, mental lexicon. Therefore, the purpose of this study was to test two hypotheses:

1) The L1 and L2 mental lexicon are structurally similar; and

2) Depth of individual word knowledge (i.e. how well a particular word is known) can account for how words are stored in the mental lexicon.

Method

Two lists of 45 prompt words were constructed using word frequency information from the Bank of English corpus created by COBUILD at Birmingham University. The first word list (PWL 1) was administered to both a group of nonnative speakers and a group of native speakers. The second list (PWL 2) was administered only to the group of native speakers. The idea behind creating two lists was to include prompt words that both groups of subjects would be familiar with to various degrees. Responses were categorized as being paradigmatic, syntagmatic, clang/other , or no response. Depth of word knowledge was assessed for each prompt word using the Vocabulary Knowledge Scale (VKS; Wesche & Paribakht, 1996). The VKS produces a score of 1-5 for each word tested. A score of 1 (the lowest score) indicates that the word is not familiar, while a score of 5 (the highest score) shows knowledge of the wordfs meaning and the ability to use it correctly in a sentence. Categories were then created using each of the scores generated by the VKS and the response pattern within each category for both groups was compared. Thus native speaker patterns for all words which were assigned a VKS score of 1 were compared to nonnative speaker VKS 1 responses and so forth. The underlying assumption here was that the two groups would not differ significantly in their patterns of responses within each category. Finally, in order to test the second hypothesis as stated above, patterns of response were compared based on the VKS categories for each group.

Results

The results offer mixed support for the first hypothesis of a structurally similar L1 and L2 mental lexicon (see Table 1). No significant difference was detected for response patterns in the VKS 1 and VKS 2 categories, but there was a significant difference in the VKS 3 and VKS 5 categories. Comparisons in the VKS 4 category were not possible, as there were no native speaker responses in this category.

 

Table 1. Comparison of Mean Ranks Between NNS and NS Groups According to VKS Categories as Determined by the Mann-Whitney Test (from Wolter, 2001).

 

 

 

Number of

responses

 

 

Mean rank

 

 

 

 

 

 

VKS

 

NNS

NS

 

NNS

NS

 

Mann-Whitney

 

Significance

1

 

71

24

 

46.99

51.00

 

      780

 

ns*

2

 

162

64

 

114.12

111.92

 

     5083

 

ns*

3

 

42

32

 

32.88

43.56

 

      478

 

.025

4

 

36

0

 

18.50

-

 

        -

 

-

5

 

268

284

 

260.09

291.99

 

    33658

 

.01

p > .05

There was better support for the second hypothesis, however, as both groups showed a trend towards a greater proportion of syntagmatic and later paradigmatic responses as familiarity with the prompt words increased (see Table 2).

Table 2. Mean Ranks, Number of Responses According to VKS Category, df, and c2 Values as Determined by the Kruskall-Wallis Test (from Wolter, 2001).

 

 

 

VKS

Group

df

c2

1

2

3

4

5

Nonnative speakers

mean rank

 

4

 

275.89*

 

122.56

 

180.59

 

286

 

347.85

 

393.35

     n

 

 

71

162

42

36

268

Native speakers  

     mean rank

 

3

 

153.01*

 

54.58

 

83.48

 

208.31

 

-

 

241.17

n

 

 

24

64

32

0

284

*p < .001


Discussion

Although the results of this study do not completely support the first hypothesis, they do at least indicate the L2 mental lexicon is highly structured, and that depth of word knowledge plays a role in the development of associations between words. Of particular interest is the difference between the native speaker group and non-native speaker group in respect to patterns of responses for words that were rated as well known (words in the VKS 5 category). In this category, both groups produced about the same proportion of phonological responses (around 3%), indicating that phonology plays no more significant of a role in the L2 mental lexicon once words become well known. However, nonnative speakers still demonstrated a preference for syntagmatic over paradigmatic responses, while the opposite was true for the native speakers. Some possible explanations for this will be discussed during the colloquium.

References

Channell, J. (1990). Vocabulary acquisition and the mental lexicon. In J. Tomasczyk & B. Lewandowska-Tomasczyk (Eds.), Meaning and lexicography (pp. 21-31). Amsterdam: Benjamins.

Meara, P. (1982). Word association in a foreign language: A report on the Birkbeck Vocabulary Project. Nottingham Linguistic Circular, 11 (2), 29-37.

Meara, P. (1984). The study of lexis in interlanguage. In A. Davies, A. Howart, & C. Criper (Eds.), Interlanguage (pp. 225-235). Edinburgh: Edinburgh University Press.

Postman, L. (1970). The California norms: Association as a function of word frequency. In L. Postman & G. Keppel (Eds.), Norms of word association (pp. 241-320). New York: Academic Press.

Stolz, W.S. & Tiffany, J. (1972). The production of gchild likeh word associations by adults to unfamiliar adjectives. Journal of Verbal Learning and Verbal Behavior, 11, 38-46.

Wesche, M., & Paribakht, T.M. (1996). Assessing vocabulary knowledge: Depth vs. breadth. Canadian Modern Language Review, 53, 13-40.

Wolter, B. (2001). Comparing the L1 and L2 mental lexicon: A depth of individual word knowledge model. Studies in Second Language Acquisition, 23, 41-69.


[1] Paradigmatic responses are from the same word class as the prompt word, and as such can presumably perform the same function as the prompt word in a sentence (e.g. dog animal, canine). Syntagmatic responses bear more of a collocational relationship to the prompt word (e.g. dog bite, bark). Clang responses resemble the prompt word only phonologically, with no overtly clear semantic connection (e.g. dog bog).