Cortical recycling in high-level visual cortex during childhood development

Abstract

Human ventral temporal cortex contains category-selective regions that respond preferentially to ecologically relevant categories such as faces, bodies, places and words and that are causally involved in the perception of these categories. How do these regions develop during childhood? We used functional magnetic resonance imaging to measure longitudinal development of category selectivity in school-age children over 1 to 5 years. We discovered that, from young childhood to the teens, face- and word-selective regions in ventral temporal cortex expand and become more category selective, but limb-selective regions shrink and lose their preference for limbs. Critically, as a child develops, increases in face and word selectivity are directly linked to decreases in limb selectivity, revealing that during childhood, limb selectivity in ventral temporal cortex is repurposed into word and face selectivity. These data provide evidence for cortical recycling during childhood development. This has important implications for understanding typical as well as atypical brain development and necessitates a rethinking of how cortical function develops during childhood.

Extended Data Figure 1.. Control analyses examining developmental increases and decreases in category-selective activation in lateral VTC.

LMM slopes indicating change in category-selective activation volume per month (n=128 sessions, 29 children). Error bars: 95% CI. If the CI does not cross the y=0 line, this indicates that the slope is significantly different than 0 (before FDR-correction). (A) Slopes for the age predictor for models in which adult faces, child faces, limbs, and words are excluded as control categories from the contrast. No effects survive FDR-correction. (B) Slopes for the age predictor for models including both age and time series signal-to-noise ratio (tSNR) as predictors. Significant development after FDR-correction (p<0.05) is indicated by asterisks. (C) Slopes for the main analysis (filled bars), the contrast control (open bars with gray outline) and the tSNR control (open bars with red outline) are overlaid to illustrate the changes in effect size across the different analyses. Full statistics in Supplementary Tables 3–4,6–7. Related to Fig. 1.

Extended Data Figure 2.. Control analyses examining functional changes underlying the development of category-selective ROIs.

Left panel: Colored bars: Slopes of LMMs indicating changes in selectivity by age for all 10 categories in emerging and waning ROIs. Open bars with gray outline: LMM slopes for contrasts in which adult faces, child faces, limbs, and words are excluded as control categories in contrasts. Error bars: 95% CI. If the CI does not cross the y=0 line, this indicates that the slope is significantly different than 0 (before FDR-correction). Asterisks: significant development after FDR-correction (p<0.05) for colored bars, circles: significant after FDR-correction for open bars. Right panel: Response amplitudes for 5-9-year-olds and 13-17-year-olds. Lighter colors indicate younger ages. One functional session per child is included per boxplot. Boxplots show the 75% and 25% percentiles (colored areas) and median (horizontal lines). Whiskers extend to the most extreme data points not considered outliers (minimum, maximum). Crosses: outliers (values more than 1.5 times the interquartile range away from the bottom or top of the box). Black diamonds: LMM prediction for the response at the mean age of each age group. (A) Left emerging pOTS-words. Left panel: n=24 (112 sessions). (B) Left waning OTS-limbs. n=26, 122 sessions. (C) Left emerging pFus-faces. n=22, 105 sessions. (D) Right waning OTS-limbs. n=21, 100 sessions. (E) Right emerging pFus-faces. n=21, 96 sessions. Full statistics in Supplementary Tables 9–10. As we observed a significant decrease in limb-selectivity in emerging parts of word- and face-selective regions and word-, face- and limb-selective regions neighbor, we tested if emerging parts of word- and face-selective regions overlap with waning parts of the limb-selective regions. However, the overlap between the developing parts of the ROIs (difference between initial and end ROIs) assessed by the dice coefficient (DC) was small (overlap between developing parts of OTS-limbs and pFus-faces, left: DC=0.025±0.01 (mean± SD), n=22; right: DC=0.026±0.01, n=18; overlap between developing parts of left OTS-limbs and pOTS-words: DC=0.006±SD=0.004, n=21). Related to Fig 3.

Extended Data Figure 3.. Developmental changes in word-, face-, and limb-selectivity are also linked in the right hemisphere.

(A) Limb-selectivity vs face- and word-selectivity in the waning right OTS-limbs. (B) Face-selectivity vs. limb- and word-selectivity in the emerging right pFus faces. Left: Model prediction for 5-9-year-olds and 13-17-year-olds for the selectivity that defines the ROI as a function of the selectivity to the other two variables. Middle: Individual participant data visualized in 3D. In each panel the variable on the z-axis is related to the x- and y-variables. LMM βs, 95%-CIs, t-values, df, and p-values are shown on top. Full statistics are reported in Table S11. Orange arrows: Individual child data. Blue arrows: LMM, same as left panel. Right: Rotated version of the plots in the middle column to increase visibility of changes along the horizontal axes. Related to Fig. 4.

Extended Data Figure 4.. Pairwise preferences of the ROI-defining category in the right hemisphere.

In each plot we show the pairwise preference of the ROI-defining category vs. each of the other two developing categories as a function of age. Thin lines: individual participant data showing the pairwise preference from the initial to end session. Gray line: LMM prediction of pairwise preference based on data from all sessions. Shaded gray: 95%-CI. LMM results (intercept: β_intercept and slope: β_age (rate of change in preference, t/month) and their significance are reported under each panel. (A) Waning right OTS-limbs (n=21 subjects, n=100 sessions). Left: Limbs vs faces. Right: Limbs vs Words. (B) Emerging right pFus-faces (n=21 subjects, n=96 sessions). Left: Faces vs limbs. Right: Faces vs Words. Related to Fig. 5. Statistics in Supplementary Table 15.

Fig. 1.. Developmental increases and decreases in category-selective activation in lateral VTC.

(A) Volume of word-, limb- and child-face-selective activation by age. Each dot is a session and colored by participant. Red line: Linear mixed model (LMM) prediction of category-selective activation by age. Shaded gray: 95% confidence interval (CI). All scatterplots in Supplementary Fig. 2. (B) Lateral and medial VTC on the inflated cortical surface of a 5-year-old. (C) LMM slopes indicating change in category-selective activation volume per month (n=128 sessions, 29 children). Error bars: 95% CI. If the CI does not cross the y=0 line, the slope is significantly different than 0 before FDR-correction. Asterisks: Significant development after FDR-correction (p<0.05). (D) Category-selective activation by age group. The boxplots illustrate the average category-selective volume for two age groups: 5-9- and 13-17-year-olds. One session per child is included per boxplot (see legend for number of sessions). The diamonds represent the LMM estimate for category-selective volume for the mean age of each age group. LMMs are based on data from all sessions, same as panels A,C. Boxplots show the 75% and 25% percentiles (colored areas) and the median (horizontal lines). Whiskers extend to the most extreme data points not considered outliers (minimum, maximum). Crosses: outliers (values more than 1.5 times the interquartile range away from the bottom or top of the box). All categories in Supplementary Fig. 3. Acronyms: N: numbers; W: words; L: limbs; B: bodies; AF: adult faces; CF: child faces, C: Cars, SI: String instruments, H: Houses, Cor: Corridors.

Fig. 2.. Development of category-selective ROIs.

Initial ROIs (colored) and end ROIs (outline) in 3 example children: (A) left pOTS-words at age 10 and 15, (B) left OTS-limbs at age 11 and 13, (C) left pFus-faces at age 9 and 14. MFS: mid fusiform sulcus; OTS: occipito-temporal sulcus. (D) LMM slopes: change in volume of category-selective regions per month. Error bars: 95% CI. If the CI does not cross the y=0 line, this indicates that the slope is significantly different than 0 (before FDR-correction). Asterisks: significant development, p<0.05, FDR-corrected. The number of sessions per ROI is as follows: left mOTS-words: n=103, right mOTS-words: n=35, left pOTS-words: n=119, right pOTS-words: n=73, left OTS-limbs: n=126, right OTS-limbs: n=126, left mFus-faces: n=107, right mFus-faces: n=102, left pFus-faces: n=120, right pFus-faces: n=98.

Fig. 3.. Age-related increases in word- and face-selectivity parallel decreases in limb-selectivity in the developing regions.

(A-C) Left: LMM slopes indicating changes in selectivity by age. Error bars: 95% CI. If the CI does not cross the y=0 line, the slope is significantly different than 0 (before FDR-correction). Asterisks: Significant development after FDR-correction (p<0.05). All ROIs and categories in Extended Data Fig. 2. Full statistics in Supplementary Tables 9–10. Right: Response amplitudes for the 10 categories. The boxplots illustrate response amplitudes by age group for 5-9-year-olds and 13-17-year-olds. One session per child is included per boxplot; Legend indicates number of sessions. LMMs on response amplitudes using data from all available sessions for each ROI were used to estimate development of response amplitudes. Diamonds represent LMM estimates for response amplitudes for the mean age of each age group. Full statistics in Supplementary Table 11. Boxplots show the 75% and 25% percentiles (colored areas), and median (horizontal lines). Whiskers extend to the most extreme data points not considered outliers (minimum, maximum). Crosses: outliers (values more than 1.5 times the interquartile range away from the bottom or top of the box).

Fig. 4.. Developmental changes in word-, face-, and limb-selectivity are linked.

Left: LMM prediction (circle) of category selectivity to words (A), limbs (B), and faces (C) vs. selectivity to the other two categories in 5-9-year-olds and 13-17-year-olds. Middle: Individual participant data visualized in 3D. In each panel the variable on the z-axis is related to the x- and y-variables, this relationship is quantified in the model; βs, 95%-CI, t-values, degrees of freedom, and p-values are shown at the top. Full statistics in Supplementary Table 13; Orange arrows: Individual child data. Blue arrows: LMM prediction (same as left panel). Right: Rotated version of the plots in the middle column to enhance visibility of positive and negative values along the other horizontal axis. Right hemisphere in Extended Data Fig. 3.

Fig. 5.. In emerging category-selective ROIs, selectivity flips from preference to limbs at age 5 to preference to either words or faces, respectively, later in childhood.

In each plot we show the pairwise preference of the ROI-defining category vs. each of the other two developing categories as a function of age. Thin lines: individual participant data showing the pairwise preference from the initial to end session. Gray line: LMM prediction of pairwise preference based on data from all sessions. Shaded gray: 95%-CI. LMM results (intercept: β_intercept and slope: β_age (rate of change in preference, t/month) and their significance are reported under each panel. (A) Emerging left pOTS-words (n=24 subjects, n=112 sessions). Left: Words vs. faces (red). Right: Words vs. limbs (yellow). (B) Waning left OTS-limbs (n=26 subjects, n=122 sessions). Left: Limbs vs. faces (red). Right: Limbs vs. Words (blue). (C) Emerging pFus-faces (n=22 subjects, n=105 sessions). Left: Faces vs. limbs (yellow). Right: Faces vs. Words (blue). Right hemisphere data in Extended Data Fig. 4. Statistics in Supplementary Table 15.