Abstract
In recent decades, many eye-tracking studies have demonstrated that both languages of bilingual speakers are activated while processing phonological input in only one. To date, there have been no eye-tracking co-activation studies assessing word recognition among trilinguals. The present research investigates co-activation in all three languages of 48 Russian (Heritage Language)/Hebrew (Societal Language)/English (Third Language) speakers using a trilingual visual world paradigm experiment. The results paint a picture of a highly interactive multilingual lexicon, in line with findings from prior studies on bilingualism. Although accuracy was not affected by competition conditions, reaction times and eye-fixation proportions showed slow-down and distraction in the presence of cross-linguistic competitors, albeit to different extents across the three experiments, evidencing effects of language dominance and acquisition order. This study makes considerable contributions to our understanding of the dynamics of trilingual language processing and discusses findings in the context of existing bilingual processing models.
Introduction
Psycholinguistic research has long established that bilinguals do not simply "switch off" one language when using another. Instead, they experience phonological co-activation: hearing a word in one language automatically triggers similar-sounding candidates in the other. For instance, a Russian-English bilingual hearing "marker" might briefly activate the Russian word marka (stamp). However, as the world becomes increasingly multilingual, a critical question arises: how does this "mental juggling" scale when a third language (L3) is involved?
This study by Clara Fridman and Natalia Meir explores the dynamics of competition in trilingual speakers of Russian (Heritage Language, HL), Hebrew (Societal Language, SL), and English (Third Language, L3). By mapping how these three systems interact, the researchers aim to determine if language dominance and acquisition order act as gatekeepers for cross-linguistic interference.
Methodology and Participants
The researchers recruited 48 trilingual adults (average age 27.3) living in Israel. To establish a clear profile of their linguistic abilities, participants provided self-ratings and completed the Multilingual Naming Task (MINT) to objectively measure vocabulary size. As shown in the figure below, Hebrew was consistently the strongest language, followed by English and then Russian.
The core of the study utilized the Visual World Paradigm (VWP) paired with high-resolution eye-tracking (1,000 Hz). In each trial, participants viewed four images while hearing a target word in one of their three languages. Crucially, experimental trials included "cross-linguistic competitors"—images whose names in a different language sounded similar to the target word. For example, while hearing the English target "bear" [ber], the display might include a picture of a "knee" (berex [beʁeχ] in Hebrew). Eye-fixations on these competitors served as a direct measure of non-target language activation.
Key Results
The study revealed a highly interactive, yet asymmetrical, mental lexicon. While accuracy remained at ceiling levels (near 100%) across all conditions—indicating that co-activation does not typically cause overt errors—processing effort was significantly impacted.
The Dominance Shield
When performing the task in Hebrew—the dominant societal language for most participants—there was virtually no evidence of co-activation from Russian or English. High proficiency appears to create a "shield," allowing the brain to efficiently suppress weaker languages during dominant-language processing.
The Weakest Link and L3 Vulnerability
In contrast, when processing Russian (the heritage language), participants experienced significant interference from both Hebrew and English. Interestingly, English (the L3) showed a unique vulnerability: despite participants often having higher objective proficiency in English than Russian, they were significantly distracted by Hebrew competitors during the English task. Reaction times in the English task were delayed by approximately 327ms when a Hebrew competitor was present.
Discussion and Conclusion
These findings align with the BLINCS model, suggesting that the multilingual lexicon is a massive, interconnected network where activation spreads automatically based on phonological similarity. The study highlights that the "juggling" of three languages is governed primarily by dominance: stronger languages interfere with weaker ones, but the reverse is less common.
While the study provides robust evidence for trilingual co-activation, the authors note that the fixed task order and the use of repeated stimuli are limitations that future research should address. Nonetheless, this work significantly advances our understanding of how the human brain manages the complex interplay of multiple linguistic systems, confirming that the "Third Language" is particularly susceptible to the gravitational pull of a dominant societal language.
Summary updated