The rise and fall of rapid occipito-temporal sensitivity to letters: Transient specialization through elementary school

Abstract

Letters, foundational units of alphabetic writing systems, are quintessential to human culture. The ability to read, indispensable to perform in today's society, necessitates a reorganization of visual cortex for fast letter recognition, but the developmental course of this process has not yet been characterized. Here, we show the emergence of visual sensitivity to letters across five electroencephalography measurements from kindergarten and throughout elementary school and relate this development to emerging reading skills. We examined the visual N1, the electrophysiological correlate of ventral occipito-temporal cortex activation in 65 children at varying familial risk for dyslexia. N1 letter sensitivity emerged in first grade, when letter sound knowledge gains were most pronounced and decayed shortly after when letter knowledge is consolidated, showing an inverted U-shaped development. This trajectory can be interpreted within an interactive framework that underscores the influence of top-down predictions. While the N1 amplitudes to letters correlated with letter sound knowledge at the beginning of learning, no association between the early N1 letter response and later reading skills was found. In summary, the current findings provide an important reference point for our neuroscientific understanding of learning trajectories and the process of visual specialization during skill learning.

Keywords: Development; ERP; Familial risk for dyslexia; Letter processing; Visual N1.

Fig. 1

Cognitive performance across measurements. Y-axis in the letter knowledge (green) panel indicates the number of correctly pronounced letter sounds (upper and lower case items summed); in reading, the number of correctly read pseudowords (orange) and words (red) within a minute; in RAN letters (blue) the score is the number of letters named per second. Error bars indicate mean and 95 % CIs. Asterisks in x-axis indicate significant differences between pairs of measurements (p < 0.05). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

EEG results per time point (T1 to T5). Left panel: ERPs (μV) for letters (black lines) and false fonts (red lines) with ribbons indicating 95 % CIs (within-subject). The N1 interval is highlighted in yellow and the horizontal orange bars show p <  0.05 in t-tests per data point (FDR corrected). Middle panel: scalp topographical maps for letters and false fonts (μV). Right panel. GFP for both conditions (black lines), letters (dashed black lines) and false fonts (dashed red lines) in μV. Box near the x-axis shows TANOVA results for the N1 interval (yellow-orange gradient shows ps < 0.05 in permutation tests). Scalp maps in the right show the t values as t-test maps for the difference between letters and false fonts. ERPs = event-related potentials; CIs = confidence intervals; FDR = false discovery rate; GFP = global field power; TANOVA = topographic analysis of variance; LET = letters; FF = false fonts; LOT = left occipito-temporal; ROT = right occipito-temporal. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

N1 mean amplitudes (μV) for letters (black) and false fonts (red) for left and right clusters (top and bottom row, respectively) per time point. Those time points with significant condition effects are highlighted in dark green. The shape of the scatter points indicates whether a participant was classified as typical (circle) or poor reader (triangle); square shapes indicate that the participant was not classified as no reading scores were available. Error bars within the density plots show the mean and 95 % CI. LOT = left occipito-temporal; ROT = right occipito-temporal; LET = letters; FF = false fonts. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

Differences in N1 mean amplitudes between letters and false fonts per time point. The shape of the scatter points indicates whether a participant was classified as typical (circle) or poor reader (triangle); square shapes indicate that the participant was not classified as no reading scores were available. Error bars within the density plot show the mean and 95 % CIs. More negative values indicate stronger (negative) amplitudes for letters vs false fonts. LOT = left occipito-temporal; ROT = right occipito-temporal.

Fig. 5

Top panels. Linear regression with letter sound knowledge at T2 as dependent variable and N1 amplitudes (μV) for letters at T2 (left and right clusters) as predictor. Bottom panel. Linear regression with familial risk score as dependent variable and T2 letter-false font N1 amplitude differences (μV) in the left cluster as predictor. The shape of the scatter points indicates whether a participant was classified as typical (circle) or poor reader (triangle); square shapes indicate that the participant was not classified as no reading scores were available. Box plots are displayed in the margins. LOT = left occipito-temporal; ROT = right occipito-temporal.

Fig. 6

Schematic of a model for visual specialization, reflected in vOTC activity to letters (orange) and words (blue), with advancing reading expertise. Vertical lines represent learning milestones. Specialization to single letters is proposed to peak when letters are learned and decay afterwards, and specialization to words is proposed to peak later and then decline slower, persisting over time at more subtle levels. Arrow heads indicate the changing contributions from phonological and/or semantic areas. Phonological input would be stronger with initial learning and when grapheme-phoneme decoding is the main reading strategy. Semantic influence would be stronger with more advanced sight word reading strategies. Illustrations on the right margin show the location of the vOTC (red highlight) and the N1 response as the main electrophysiological correlate of its activity. vOTC; ventral occipital-temporal cortex. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)