Development of the macaque face-patch system

Margaret S. Livingstone, Justin L. Vincent, Michael J. Arcaro, Krishna Srihasam, Peter F. Schade, Tristram SavageNature Communications 2017 · DOI

TL;DR

This study uses longitudinal fMRI in infant macaques to reveal that the specialized brain regions for face recognition (face patches) are not fully formed at birth but emerge and stabilize within the first year of life. This development is driven more by the refinement of existing neural structures through experience, specifically by reducing responses to non-face stimuli, rather than solely by the emergence of new face-specific circuits.

What Problem Does This Paper Address?

The ability of humans and other primates to recognize faces is remarkably proficient, even emerging in infancy. However, the developmental trajectory of the neural mechanisms underlying this crucial ability, particularly in the inferotemporal cortex (IT), has remained largely unknown. While some theories suggest innate, hardwired mechanisms for face processing, others emphasize the role of early experience in shaping these specialized brain areas. This paper aims to bridge this gap by longitudinally tracking the development of face-selective regions in macaques from infancy to early adulthood, investigating when and how these specialized "face patches" emerge and mature.

Methods

The researchers employed blood-volume functional MRI (fMRI) to monitor brain activity in infant macaques over an extended period, from 1 month to 2 years of age. This longitudinal approach allowed them to observe developmental changes in real-time. They used a contrast agent (MION) to enhance signal-to-noise and measure blood volume directly, enabling reliable detection of neural responses. The study compared responses to various visual stimuli, including static images of monkey faces, scrambled monkey faces (which share low-level features with faces but lack semantic meaning), and rectilinear objects. For younger infants, where attention to static images was limited, they also used movie clips of monkey interactions, pixelated faces, and natural scenes. Statistical analyses involved calculating beta coefficients for different stimulus contrasts and using D-prime sensitivity indices to quantify the ability of brain regions to distinguish between categories.

Datasets

The study utilized a dataset of four infant macaque monkeys scanned longitudinally. The visual stimuli consisted of:

  • Blocks of static images: monkey faces, scrambled monkey faces, rectilinear objects.
  • Movie clips: monkeys grooming with prominent faces, same movies with faces pixelated, inanimate dynamic scenes (traffic, nature time-lapse).

No new benchmarks were introduced; the study focused on characterizing the developmental emergence of existing functional domains.

Key Insights and Ideas

  • Face patches aren't born, they're made (over time): Contrary to a purely innate model, the study demonstrates that specialized face-processing regions (face patches) in the macaque IT are not fully formed at birth but emerge and stabilize over the first year of development.
  • Pruning, not just growth: The refinement of face selectivity appears to be driven significantly by a decrease in responsiveness to non-face stimuli, suggesting a "pruning" mechanism where less relevant connections are weakened.
  • Early proto-organization: Even at 1 month of age, there's evidence of a rudimentary organization in IT that is broadly tuned to face-like stimuli, suggesting an innate "proto-map" that is then sculpted by experience.
  • Low-level features matter early on: The early, broad selectivity observed in infant macaques is not purely semantic but is also driven by low-level visual features shared between faces and scrambled faces, indicating that the initial organization is sensitive to basic visual properties.
  • Subcortical structures don't lead the way: The study found no evidence of early face selectivity in subcortical structures like the LGN, pulvinar, or amygdala, challenging hypotheses that these areas drive early cortical face processing.

Key Results

  • Emergence of face selectivity: Face-selective patches in the IT emerged around 200 days of age and remained remarkably stable thereafter.
  • Early proto-organization: At 1 month of age, movie stimuli revealed some face-selective activity in the STS, in locations that would later become face patches. This early selectivity was weaker and less specific than that observed later.
  • Refinement through experience: The selectivity for faces over objects increased over time, primarily due to a decrease in responsiveness to objects and scrambled faces.
  • Stability of organization: Once face patches emerged, their spatial organization was highly stable across subsequent scanning sessions.
  • No early subcortical selectivity: Visual responsiveness was detected in the LGN as early as 10 days, but no category-specific selectivity was found in the LGN, pulvinar, or amygdala throughout development.

Limitations and Future Work

  • Movie stimuli confounds: The authors acknowledge that movie stimuli, while necessary for capturing early attention, introduce confounds like motion and biological motion that static images avoid. This limits the interpretation of early movie-based selectivity as purely face-specific.
  • fMRI sensitivity: While MION contrast agent improved sensitivity, there's a discussion about potential discrepancies with electrophysiological studies that report earlier face-selective neurons. The authors suggest fMRI might reflect population-level properties that are slower or less reliable in infants.
  • Distinguishing innate models: The study cannot definitively distinguish between two innate models: a broad, innate semantic template for faces or an innate low-level proto-map biased towards certain features (like curvature). Future work could explore this by manipulating specific low-level features more systematically.
  • Role of eye gaze: The refinement of the proto-organization is suggested to be influenced by eye gaze patterns, an area that could be further investigated.
  • Unnatural categories: The study's previous work on forming domains for unnatural categories suggests that experience plays a significant role. Further research could explore how the interaction between innate predispositions and specific types of experience shapes these developing visual circuits.

Summary updated

Abstract

Face recognition is highly proficient in humans and other social primates; it emerges in infancy, but the development of the neural mechanisms supporting this behaviour is largely unknown. We use blood-volume functional MRI to monitor longitudinally the responsiveness to faces, scrambled faces, and objects in macaque inferotemporal cortex (IT) from 1 month to 2 years of age. During this time selective responsiveness to monkey faces emerges. Some functional organization is present at 1 month; face-selective patches emerge over the first year of development, and are remarkably stable once they emerge. Face selectivity is refined by a decreasing responsiveness to non-face stimuli.
BibTeX
@inproceedings{livingstone2017development,
  author = {Margaret S. Livingstone and Justin L. Vincent and Michael J. Arcaro and Krishna Srihasam and Peter F. Schade and Tristram Savage},
  title = {{Development of the macaque face-patch system}},
  booktitle = {Nature Communications},
  year = {2017},
  doi = {10.1038/ncomms14897},
}
Figure