TL;DR
This paper investigates how French-speaking Belgians learning Dutch (Francophone Written Dutch, FWD) differ from native Dutch speakers (Standard Written Dutch, SWD) in their use of French loanwords. They propose a computational model of lexical interference, suggesting that bilinguals use more French loanwords to maintain a "normal" level of vocabulary diversity, and that these words are ordered in a conceptual store.
What Problem Does This Paper Address?
The paper tackles the phenomenon of language interference, specifically how a bilingual's native language influences their acquisition and use of a second language. The focus is on the lexical level, examining the integration of French loanwords into Dutch by Belgian francophones. The authors aim to move beyond qualitative observations to a more quantitative, computational analysis of these interference patterns, seeking to understand the underlying mechanisms and provide numerical values for these linguistic phenomena. They are particularly interested in how this interference affects vocabulary growth and word choice in comparison to monolingual Dutch speakers.
Methods
The researchers employed a comparative corpus analysis. They collected two sets of Dutch texts:
- Group A (FWD): Compositions written by 400 francophone 18-year-olds with approximately 6 years of Dutch training.
- Group B (SWD): Extracts from contemporary Dutch literature by native Dutch and Flemish authors.
They then performed a computational analysis of these texts, focusing on:
- Token and Type Counts: Analyzing the total number of words (tokens) and unique words (types) in each corpus.
- Logarithmic Measures: Using logarithmic measures of vocabulary size (LogTypes) and the proportion of foreign words (LogF.T.) to normalize comparisons.
- Word-Length Analysis: Examining the distribution of word lengths for both types and specific word categories (like composita).
- Statistical Tests: Employing chi-square tests to determine the significance of observed differences.
- Interference Model: Proposing a conceptual model where a "lexical morpheme store" contains conceptual symbols, and words are "fitted" into sentence structures based on semantic requirements. This model suggests a "tape" with tracks for Dutch and French, where bilinguals might "fill in blanks" with French words.
Datasets
- Francophone Written Dutch (FWD) Corpus: Approximately 100,000 words compiled from Dutch compositions written by 400 francophone high school students. This corpus was divided into two parts of roughly 50,000 words each.
- Standard Written Dutch (SWD) Corpus: Approximately 10,000 words compiled from extracts of literary works by 10 contemporary Dutch authors.
Key Insights and Ideas
- "Keeping Pace" with Diversity: Francophone bilinguals use significantly more foreign (French) words than monolingual Dutch speakers. The authors hypothesize this is not due to a lack of vocabulary but rather an effort to maintain a "normal" level of overall vocabulary diversity, essentially "keeping pace" with the expected diversity in native speech.
- Conceptual Lexical Store: The paper proposes a model where lexical items are stored as "conceptual symbols." This suggests that the choice of words is driven by semantic needs before the actual linguistic form is selected, and that these conceptual symbols are ordered, possibly by frequency or semantic relationships.
- Ordered Lexical Access: The analysis of word frequency distributions (Zipf-law) strongly supports the idea that words in the lexical store are ordered. This ordering, likely based on frequency, allows for more efficient retrieval of words.
- French Loanwords as a "Basic Vocabulary": The study suggests that French loanwords in FWD texts form a kind of "basic vocabulary" that appears consistently as text length increases, particularly above the first 3,000 words. This implies a systematic integration rather than random borrowing.
- Interplay of Lexical and Syntactic Levels: The paper hints that lexical interference is closely linked to higher syntactic levels, suggesting that phenomena like the formation of Dutch composita from French genitive constructions are not purely lexical but involve broader language production mechanisms.
Key Results
- Francophone bilinguals use more than twice as many words as monolingual Dutch speakers.
- The percentage of foreign words in FWD texts is significantly higher (11.85%) compared to SWD texts (5.38%).
- While the overall word length of tokens is similar between FWD and SWD, the average word length of types (unique words) is longer in FWD (7.85) than in SWD (7.03).
- The vocabulary growth rate in FWD shows a stable ratio of new foreign words to the total vocabulary (around 1/10) until a vocabulary of 3,000 items is reached, after which it increases considerably.
- The rank-frequency distribution of words closely fits the theoretical Zipf-Mandelbrot distribution, with a correlation coefficient of 0.993, supporting the hypothesis of an ordered lexical store.
Limitations and Future Work
- Heuristic Simplification: The authors acknowledge that their model, particularly the assumption that word-formational rules are purely syntactic and that grammatical morphemes are borrowed, is a heuristic over-simplification.
- Orthographic Idiosyncrasies: The analysis of "Part 2" of the FWD corpus, which was not manually pre-edited for orthographic mistakes, led to a larger "basic vocabulary" size (3,000-3,500 words) compared to what might be expected, suggesting that such idiosyncrasies were counted as new types by the computer.
- Future Work: The paper suggests that further investigation into the ordering principles of the lexical store, beyond frequency, and the interaction between lexical and syntactic levels in bilingual speech production, would be valuable. The "pigeon-holing" of semantic information is presented as a promising avenue for understanding this ordering.
Summary updated