Tuning of the visual word processing system: distinct developmental ERP and fMRI effects

Abstract

Visual tuning for words vs. symbol strings yields complementary increases of fast occipito-temporal activity (N1 or N170) in the event-related potential (ERP), and posterior-anterior gradients of increasing word-specific activity with functional magnetic resonance imaging (fMRI) in the visual word form system (VWFS). However, correlation of these coarse ERP and fMRI tuning responses seems limited to the most anterior part of the VWFS in adult and adolescent readers (Brem et al. [ 2006]: Neuroimage 29:822-837). We thus focused on fMRI tuning gradients of young readers with their more pronounced ERP print tuning, and compared developmental aspects of ERP and fMRI response tuning in the VWFS. Children (10.3 y, n = 19), adolescents (16.2 y, n = 13) and adults (25.2 y, n = 18) were tested with the same implicit reading paradigm using counterbalanced ERP and fMRI imaging. The word-specific occipito-temporal N1 specialization, its corresponding source activity, as well as the integrated source activity (0-700 ms) were most prominent in children and showed a marked decrease with age. The posterior-anterior fMRI gradient of word-specific activity instead which was fully established in children did not develop further, but exhibited a dependence on reading skills independent of age. To conclude, prominent developmental dissociation of the ERP and fMRI tuning patterns emerged despite convergent VWFS localization. The ERP response may selectively reflect fast visual aspects of print specialization, which become less important with age, while the fMRI response seems dominated by integrated task- and reading-related activations in the same regions.

Figure 1

Waveforms of GFP (top), left (O1′: middle) and right (O2′: bottom) occipito‐temporal electrodes to words (solid lines) and symbol strings (dotted lines) in children (red), adolescents (blue) and adults (black). Clearly visible are major decreases in amplitude (and latency) as well as the pronounced decrease in the difference between words and symbols with age.

Figure 2

Maps (left) of the group grand averages for the potential fields (μV) in the N1 interval for words and symbol strings. t‐Maps show the t‐values of the statistical difference between the potential field maps to words and symbol strings. Maps and t‐maps are shown for children (top), adolescents (middle), and adults (bottom). The distinct difference in N1 amplitude and topography between words and symbol strings is highly consistent as evident in the t‐maps. The statistical comparison of the sLORETA source activity (non‐normalized) to words and to symbols revealed regions with stronger sources for each group in the occipito‐temporal cortex to words (word‐specific activity) in the N1 interval. Regions coloured in red exhibit t‐values corresponding to a P < 0.01 corrected for multiple comparisons. Adolescents, due to the smaller group size exhibited less occipito‐temporal activation, however when lowering the threshold to P < 0.05 (corrected) a similar activation pattern as in children and adults was found (not shown here).

Figure 3

A: fMRI analysis showing word activation (words vs. baseline) for each group at the corrected threshold of P(FWE) < 0.05, k = 20. B: Word‐specific activation (words–symbol strings) is illustrated for each group at P(FDR) < 0.01, k = 20. C: The statistical comparison of sLORETA source estimations to words and symbol strings (non‐normalized) revealed regions (red) with stronger sources to words (P < 0.01 corrected) in the time‐range from 0 to 700 ms for each group. The axial slices in A and B are illustrated at the corresponding activation maxima.

Figure 4

A: Plots illustrating the normalized sLORETA mean activity to words (red) and symbol strings (blue) in the five analyzed left hemispheric ROIs within the VWFS, in the time range from 0 to 700 ms. ROI1 (R1) represents the most posterior, R5 the most anterior ROI. The activity is more pronounced to words in all anterior ROIs and age related decreases of word activity characterize the most posterior ROI (R1). B: fMRI ROI analyses corresponding to the ROIs chosen for sLORETA analyses. Clearly recognizable is the gradient of the relative increase in word‐specific activity from posterior to anterior ROIs. In contrast to ERP analyses, the overall activity to symbol strings (blue) is more pronounced than to words (red), especially in posterior ROIs. No age related changes in activity were found. At the bottom two pictures illustrate the location of the five ROIs used for sLORETA (bottom view) and fMRI (sagittal slice) analyses.

Figure 5

Partial regression plots showing (A) the decrease in word‐specific activity (% signal change fMRI) with increasing reading performance (speed) in ROI2, independent of age; (B) the decrease in word‐specific N1 amplitudes with increasing age, while controlling for reading (speed).