Different neurophysiological mechanisms underlying word and rule extraction from speech

Abstract

The initial process of identifying words from spoken language and the detection of more subtle regularities underlying their structure are mandatory processes for language acquisition. Little is known about the cognitive mechanisms that allow us to extract these two types of information and their specific time-course of acquisition following initial contact with a new language. We report time-related electrophysiological changes that occurred while participants learned an artificial language. These changes strongly correlated with the discovery of the structural rules embedded in the words. These changes were clearly different from those related to word learning and occurred during the first minutes of exposition. There is a functional distinction in the nature of the electrophysiological signals during acquisition: an increase in negativity (N400) in the central electrodes is related to word-learning and development of a frontal positivity (P2) is related to rule-learning. In addition, the results of an online implicit and a post-learning test indicate that, once the rules of the language have been acquired, new words following the rule are processed as words of the language. By contrast, new words violating the rule induce syntax-related electrophysiological responses when inserted online in the stream (an early frontal negativity followed by a late posterior positivity) and clear lexical effects when presented in isolation (N400 modulation). The present study provides direct evidence suggesting that the mechanisms to extract words and structural dependencies from continuous speech are functionally segregated. When these mechanisms are engaged, the electrophysiological marker associated with rule-learning appears very quickly, during the earliest phases of exposition to a new language.

Figure 1. Illustration of the experimental procedure.

A. Illustration of the experimental sequence for each language, highlighting the underlying structure of the artificial language. The “_” represents the 25 ms pause between words. After a learning phase lasting four minutes, an online test (violation phase) was administered in which new-words, either following the rule or violating it, appeared at random positions in the stream. B. Illustration of the recognition phase. In order to determine whether the participants had learned the words and rules of the language, an offline behavioural test (recognition phase) was administered after the violation phase. Half of the streams were tested for word acquisition; rule-learning was evaluated in the other half using a two alternative forced-choice test. Event-related responses were recorded throughout the whole sequence (learning, violation and recognition phases). Each participant was presented with a total of eight languages, thus eight sequences as the one presented here.

Figure 2. Grand average ERPs at frontal (Fz) and central (Cz) electrode locations for language streams.

A. ERP averages comparing the first and second minute of exposition. The ERP signature of the average of words from their onset for the first and second-minute blocks pooled across the four languages is shown. Words in the language streams developed an N400 component during the second minute relative to the first minute. The topography of the difference waveform (subtracting the second from the first minute) showed a central scalp distribution at 50 ms, around the peak of the component (370–420 ms). B. ERP averages comparing the third and first minute. An increase in the amplitude of the P2 component was observed from the third minute. The corresponding difference waveform (third minute minus the first minute) showed a right frontal distribution at 50 ms around the peak of the component (140–190 ms). C. Mean voltage at 50 ms around the peak of the components for the N400 and P2 effects (370–420 and 140–190 ms, respectively) as a function of time at Fz (where both modulations were significant).

Figure 3. Modulation of the ERP components as a function of rule learning performance.

A. Correlation between the mean amplitude of the P2 component at Fz in the third minute of learning (at the 120–220 ms time-window) and the performance on the rule-learning test (N = 20). B. Percentage (± s.e.m.) of correct recognition in the word-learning and rule-learning tests for the groups of good and poor learners (n = 8, in each group). C. ERP averages of the language conditions for each group at a frontal location (Fz), showing the evolution of the differences between groups over the time of exposition (first, second and third minute). While a noticeable increase in the P2 component is shown across time for the good-learners, no modulation is observed for the poor-learners.

Figure 4. Illustration of the ERP results for the violation phase.

A. Left panel: ERP averages from the onset of the presentation of words and new words that violated the previously acquired rule (non-words). An early Mismatch Negativity (MMN) appears, which indicates automatic detection of the rule violation. This negativity is followed by a late positive component (LPC) that could be assimilated into a P600 syntactic component. Right panel: ERP averages from the onset of the presentation of words and new words that violated the previously acquired rule (non-words). B. The difference waveform (subtracting non-words from words) has an MMN effect peaking around 190 ms after the onset of the non-word presentation and a fronto-central distribution. The LPC shows a more left lateralised parieto-occipital distribution that peaks around 720 ms after onset.

Figure 5. Illustration of the ERP results for the recognition phase A.

ERPs averaged from the onset of the presentation of each word in the offline recognition test. While a clear long lasting N400 effect is observed when comparing words and non-words, rule-words did not differ from words. B. Scalp distribution of the N400 effect for non-words compared to words and compared to rule-words. The same topographical distribution of the N400 effect is observed between 350 and 550 ms peaking at fronto-central locations.