Brain potentials reveal differential processing of masculine and feminine grammatical gender in native Spanish speakers

Abstract

Studies of Spanish grammatical gender have shown that native speakers exploit gender cues in determiners to facilitate speech processing and are sensitive to gender mismatches. However, past research has not considered attested distributional asymmetries between masculine and feminine gender, collapsing performance on trials with one or the other gender into a single analysis. We use event-related potentials to investigate whether masculine and feminine grammatical gender elicit qualitatively different brain responses. Forty monolingual Spanish speakers read sentences that were well-formed or contained determiner-noun gender violations. Half of the nouns were masculine and the other half were feminine. Consistent with previous research, brain responses varied along a continuum between LAN- and P600-dominant effects for both gender categories. However, results showed that individuals' ERP response dominance (LAN/P600) systematically differed across the two genders: participants who showed a LAN-dominant response to masculine-noun violations were more likely to show a P600 effect in response to feminine-noun violations. Correlations with individual difference measures further revealed that responses to masculine-noun violations were modulated by performance on the AX-CPT, a measure of cognitive control, whereas responses to feminine-noun violations were modulated by lexical knowledge, as indexed by verbal fluency. Together, the results demonstrate that even when processing features of language that belong to the same "natural class," native speakers can exhibit patterns of brain activity attuned to distributional patterns of language use. The inherent variability in native speaker processing is, therefore, an important factor when explaining purported deviations from the "native norm" reported in other types of populations.

Keywords: LAN; P600; Spanish; event-related potentials; language variation; morphosyntax.

FIGURE 1

Hypothetical illustration of feminine (yellow) and masculine (blue) category structure. Exemplars closer to the x- and y-intercept represent members whose phonemic makeup is more prototypical of the category, and those departing from it represent members sharing less phonological characteristics in common. While masculine and feminine gender categories are very similar numerically (masculine and feminine gender nouns are distributed approximately equally in Spanish; Bull, 1965); they nonetheless show varying degrees of schematicity: the masculine gender category is more schematic, allowing exemplars that are not closely similar to each other. Conversely, the feminine gender category is less schematic, showing higher degrees of similarity among its members

FIGURE 2

Trial sequence for the ERP sentence reading task. Each trial started with a fixation cross “+” presented at the center of the screen after which a sentence was presented word by word. Each word appeared in the middle of the screen for 300 ms with a 350 ms interstimulus interval. The target noun was always in sentence-medial position. Each sentence was followed by a picture, and participants were asked to respond whether the picture was semantically related or unrelated to the preceding sentence by, respectively, pressing “yes” or “no” on a button box. Pictures stayed on the screen until a response was recorded

FIGURE 3

AX-CPT trial sequence. A-cues occurred in 80% of sequence trials. Of these, only 10 trial sequences had non-target (i.e., non-X) probes

FIGURE 4

Grand-averaged ERPs time-locked to the onset of masculine (a) and feminine (b) target nouns at F3 and Pz electrodes for congruent (black) and incongruent (red) conditions. Negative is plotted up. Scalp topographic maps of the congruency effect in masculine (c) and feminine (d) target nouns. Differences were calculated by subtracting mean congruent noun amplitudes from mean incongruent noun amplitudes. Scale is from −2 (blue) to +2 (red) microvolts

FIGURE 5

Left: Grand-averaged ERPs time-locked to the onset of masculine (black) and feminine (red) determiners at F3 and Pz electrodes. Negative is plotted up. Right: Scalp topographic maps showing grand mean effects between 300–500 ms and 500–800 ms. Differences were calculated by subtracting mean feminine determiner amplitudes from mean masculine determiner amplitudes. Scale is from −2 (blue) to +2 (red) microvolts

FIGURE 6

Scatterplots showing the distribution of LAN (y axis) and P600 (x axis) effect amplitudes for masculine (left panel) and feminine (right panel) noun conditions. Each dot represents a data point from a single participant. Mean amplitudes were extracted from F3 and Pz electrode sites within the LAN and P600 windows, respectively. The solid lines indicate the best-fit regression line relating participants’ LAN and P600 effect magnitudes. The dashed lines represent equal LAN and P600 effect amplitudes. Participants above and left of the dashed line showed LAN-dominant responses to gender violations, while participants below and right of the dashed line showed primarily P600-dominant responses

FIGURE 7

Correlation between participants’ response magnitude indices for masculine and feminine nouns. Larger values indicate relatively greater ERP responses to the gender violations across both time windows, regardless of polarity

FIGURE 8

Violin plots visualizing the distribution of RDI values as a function of noun gender. More negative values reflect LAN-dominant responses to gender violations, whereas more positive values reflect P600-dominant responses. Values closer to zero reflect biphasic responses (i.e., relatively equal-sized LAN and P600 effects). The black dots represent the mean RDI value across participants and the black lines represent the 95% confidence intervals

FIGURE 9

Correlation between participants’ response dominance indices for masculine and feminine-noun conditions. More positive values indicate P600-dominant responses, whereas more negative values indicate LAN-dominant responses to the gender violations across both time windows, regardless of magnitude

FIGURE 10

Correlation between participants’ verbal fluency performance (total number of exemplars; x axis) and their RDI values (in microvolts) for feminine nouns (y axis). For verbal fluency, greater number of exemplars indicate better performance, while lower number of exemplars indicate poorer performance in the task. For RDI, more positive values indicate relatively P600-dominant responses, whereas more negative values indicate LAN-dominant responses to the gender violations across both time windows, regardless of magnitude

FIGURE 11

Correlation between participants’ average response time (ms) on AY trials (x axis) and their RDI values (in microvolts) for masculine nouns (y axis). On the x axis, larger values indicate slower RTs on Y-probes in AY trials. On the y axis, more positive values indicate relatively P600-dominant responses, whereas more negative values indicate LAN-dominant responses to the masculine-noun gender violations across both time windows, regardless of magnitude

FIGURE 12

A scheme of the purported systematic differences in determiner cue validity, category schematicity, and brain responses to grammatical gender violations for masculine and feminine nouns. Because feminine gender cues encoded in determiners categorically exclude masculine referents (but not vice versa), individuals’ predictions regarding the gender of an upcoming noun will be more reliable, and hence, stronger than those based on masculine cues. Violation of these stronger predictions thus results in the preponderance of LAN-dominant effects. Similarly, invalidly cued feminine nouns will elicit P600-dominant effects because feminine nouns have higher category coherence with relatively lower degrees of phonological schematicity relative to masculine nouns