Bilingual Object Naming: A Connectionist Model

Shin-Yi Fang, Benjamin D. Zinszer, Barbara C. Malt, Ping LiFrontiers in Psychology 2016 · DOI

Introduction

Bilingual speakers often exhibit naming patterns that differ from those of monolingual speakers in either of their languages. Over time, the lexical boundaries in a bilingual's two languages tend to drift toward each other, a phenomenon known as semantic convergence. While empirical evidence for this convergence is well-documented, the underlying cognitive mechanisms have remained a subject of debate. This study employs a connectionist model to investigate how interactions between lexical representations drive this linguistic blending.

The Connectionist Architecture

To simulate the bilingual lexicon, researchers developed a multi-layer Self-Organizing Map (SOM). SOMs are artificial neural networks designed to map high-dimensional data onto a lower-dimensional grid, effectively grouping similar items together.

The multi-layer SOM architecture featuring Object, Orthographic, and Phonological maps connected via Hebbian learning.

The model consists of three primary components:

  1. Object Map (Semantic): Encodes physical and functional features (e.g., "has a lid," "made of glass").
  2. Phonological Map: Represents the auditory word forms in Dutch and French.
  3. Orthographic Map: Represents the visual spelling of words.

The critical innovation in this model is the implementation of lateral connections between word forms. Using Hebbian learning principles—where "cells that fire together, wire together"—the model associates words in different languages (e.g., Dutch fles and French bouteille) when they are simultaneously activated by the same object.

Training Data and Stimuli

The model was trained using empirical data from previous studies involving 73 household containers, such as bottles and jars. These objects were defined by 68 human-generated semantic features.

Example of a household container stimulus used to train the model's semantic and naming maps.

By comparing a "Standard Model" against versions where specific components (like orthography or lateral connections) were removed, the researchers could isolate the factors most responsible for convergence.

Key Findings

The simulation results closely mirrored human behavior, with the bilingual model achieving a 0.97 correlation between the naming patterns of its two languages.

  • Simultaneous Activation: Convergence is primarily driven by the co-activation of both languages. When a bilingual encounters an object, related terms in both languages are triggered, strengthening the cross-linguistic lexical bridges.
  • Simplified Categorization: The model suggests that bilinguals develop more streamlined mental maps, requiring fewer feature dimensions to distinguish between categories compared to monolinguals.
  • The Role of Lexical Links: Ablation tests revealed that removing lateral connections caused the most significant drop in performance. This confirms that direct links between the lexicons of different languages are the most critical factor in semantic drift.

Conclusion and Future Directions

This connectionist approach provides a robust framework for understanding how the bilingual brain manages two linguistic systems. While the current study focused on concrete household objects and simultaneous bilingualism, future research could expand this model to abstract concepts and sequential language learners. Understanding these dynamics is essential for a complete picture of how language immersion and code-switching shape the bilingual mind.

Summary updated

Abstract

Patterns of object naming often differ between languages, but bilingual speakers develop convergent naming patterns in their two languages that are distinct from those of monolingual speakers of each language. This convergence appears to reflect interactions between lexical representations for the two languages. In this study, we developed a self-organizing connectionist model to simulate semantic convergence in the bilingual lexicon and investigate the mechanisms underlying this semantic convergence. We examined the similarity of patterns in the simulated data to empirical data from past research, and we identified how semantic convergence was manifested in the simulated bilingual lexical knowledge. Furthermore, we created impaired models in which components of the network were removed so as to examine the importance of the relevant components on bilingual object naming. Our results demonstrate that connections between two languages' lexicons can be established through the simultaneous activations of related words in the two languages. These connections between languages allow the outputs of their lexicons to become more similar, that is, to converge. Our model provides a basis for future computational studies of how various input variables may affect bilingual naming patterns.
BibTeX
@inproceedings{fang2016bilingual,
  author = {Shin-Yi Fang and Benjamin D. Zinszer and Barbara C. Malt and Ping Li},
  title = {{Bilingual Object Naming: A Connectionist Model}},
  booktitle = {Frontiers in Psychology},
  year = {2016},
  doi = {10.3389/fpsyg.2016.00644},
}
Figure