Early neurophysiological indices of second language morphosyntax learning

Abstract

Humans show variable degrees of success in acquiring a second language (L2). In many cases, morphological and syntactic knowledge remain deficient, although some learners succeed in reaching nativelike levels, even if they begin acquiring their L2 relatively late. In this study, we use psycholinguistic, online language proficiency tests and a neurophysiological index of syntactic processing, the syntactic mismatch negativity (sMMN) to local agreement violations, to compare behavioural and neurophysiological markers of grammar processing between native speakers (NS) of English and non-native speakers (NNS). Variable grammar proficiency was measured by psycholinguistic tests. When NS heard ungrammatical word sequences lacking agreement between subject and verb (e.g. *we kicks), the MMN was enhanced compared with syntactically legal sentences (e.g. he kicks). More proficient NNS also showed this difference, but less proficient NNS did not. The main cortical sources of the MMN responses were localised in bilateral superior temporal areas, where, crucially, source strength of grammar-related neuronal activity correlated significantly with grammatical proficiency of individual L2 speakers as revealed by the psycholinguistic tests. As our results show similar, early MMN indices to morpho-syntactic agreement violations among both native speakers and non-native speakers with high grammar proficiency, they appear consistent with the use of similar brain mechanisms for at least certain aspects of L1 and L2 grammars.

Keywords: ERP/F; L2 acquisition; MEG; MMN; Morphosyntax.

Fig. 1

Examples of stimuli. Panels on the left show a schematic illustration of the phoneme combinations used in the experiments, with the constituent phonemes of the three displayed example items highlighted in yellow. Panels on the right show spectrograms of the respective audio information. The onset of the critical inflectional suffix (and zero point of epoch) is marked with a vertical, yellow line for each inflectional condition. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

Means and standard error bars for behavioural tasks, A: shadowing experiment, B: GJT latencies, C: GJT accuracy (converted to d-prime).

Fig. 3

Time and topography of MMNm responses averaged across all conditions. A: MMNm topographies as shown by planar gradiometer (local RMS values calculated from gradiometer pairs are displayed in fT/m) and magnetometer recordings (in fT) in the first 365 ms after acoustic divergence. B: Average topography of the MMNm in the time window 80–335 ms, as recorded through magnetometers and gradiometers. Arrows on the latter reflect direction of field gradients. Parallelograms index gradiometers delivering large signals in the present task, which were therefore selected for further signal space analysis. Each parallelogram indicates a group of sensors whose average formed one region of interest in the sensor-space ANOVA. C: Time course of the MMNm response recorded through all large signal gradiometers (RMS values from ROIs indexed in b). Borders of time window used for signal space analysis, which capture both early MMN peaks, are indicated by vertical dashed lines.

Fig. 4

MMNm waveforms for the grammatical/ungrammatical he/*we –s conditions (first column in blue/*red) and the he/we –ed control conditions (second column in blue/green) as recorded in native speakers (top panels), high proficient non-natives (middle panels), and low proficiency non-natives (bottom panels). y-axes give magnetic field gradient in fT/m and x-axes give time in milliseconds. Bar charts in the right column show mean magnetic field gradient amplitudes taken within the 80–335 ms time window (again indicated by dashed lines on the waveforms) for each of the four conditions (with the ungrammatical one shown in red). Error bars represent standard error. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 5

sMMNm sources and their relationship to psycholinguistic performance (values in arbitrary units, a.u.) A: Main sources (yellow circles) of the MMNm calculated across all conditions and across the 80–335 ms time window. The lateral sections of cortex have been removed to clearly show activation maxima. B: Strength of the left (top) and right (bottom) superior temporal main sources of the sMMNm (calculated as the difference MMNm for *we –s minus MMNm for he –s) plotted against psycholinguistic performance as captured by PC1. Nonnative speakers showed significant correlations between left- and right-hemispheric sMMN source strength and their individual performance on psycholinguistic tests. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)