Meaning first: a case for language-independent access to word meaning in the bilingual brain

Abstract

This study aimed to determine how deeply a word is processed in the bilingual brain before the word's language membership plays a role in lexical selection. In two ERP experiments, balanced Spanish-English bilinguals read lists of words and pseudowords in Spanish and English, and performed in each language (1) a language-specific lexical decision task, e.g., respond to real words in Spanish, and (2) a language-specific category decision tasks, e.g., respond to Spanish words that refer to a person. In Experiment 1, infrequent words elicited larger negativity between 350 and 650 ms post-stimulus onset for both target and non-target languages. This indicates that language membership did not block lexical access of non-target words, contrary to previous findings. In Experiment 2, we measured the onset of the target-category P300 as a way of determining if words from the non-target language were temporarily treated as targets. When Spanish was the target language, the ERP waveforms diverged early based on semantic category (people versus non-people), indicating that non-target 'English people' words were briefly treated as potential targets. This finding indicates that meaning was accessed prior to using language membership for lexical selection. However, when English was the target language, the waveforms diverged first based on language (Spanish versus English) then semantic category. We argue that the order in which meaning or language membership are accessed may be based on the frequency of use of a bilingual's languages: the more frequently a language is used (English was more frequently used herein), the faster the words are identified as members of the language, and the greater interference it causes when it is not the target language. In brief, these findings make the case for a moment in processing when language membership matters less than meaning.

Figure 1

The BIA+ model (van Heuven & Dijkstra, 2010).

Figure 2

Diagram representing the 26-electrode placement in geodesic array.

Figure 3

Grand average ERPs for 4 medial electrodes to high-frequency and low-frequency words in four sets of stimuli: Spanish targets, English targets, Spanish non-targets, and English non-targets. Negative is plotted upward. A significant difference (highlighted in green) in amplitude for all conditions was found between 350–650 msec, with low-frequency words more negative than high-frequency words.

Figure 4

Topographic scalp plots of difference wave activity (low-frequency words minus high-frequency words) for the four sets of stimuli in Experiment 1. The latency was where a main effect of frequency was observed for the Spanish target words, English target words and Spanish non-target words, and where an interaction of frequency and electrode was identified for the English non-target words.

Figure 5

Four categories of real words, and pseudo-words derived from the two languages were used in Experiment 2. In the English session, only the ‘English people’ words (e.g., sister) were the targets. In the Spanish session, only the ‘Spanish people’ words (e.g., rey) were the targets.

Figure 6

Grand average ERPs of the four types of words in English and Spanish session.

Figure 7

Topographic scalp plots of difference wave activity to show the Language and Category effect in English and Spanish session from 100 to 600 msec in Experiment 2.