Convergent Grammar

What structure do images share?

Portrait study 1. Source image and extracted relationships between figure, ground, focal point, and negative space.
A source portrait with geometric relationships traced over it.
Layered geometric traces from the portrait study.
Geometric traces from the portrait study.

A portrait.
A spatial proposition.

Figures, focal points, and negative space can be described through relationships. The study extracts those relationships and compares them across images, asking what remains when style and subject change.

Portrait study 2. Source image and extracted relationships between figure, ground, focal point, and negative space.
A second portrait, read through the same geometric reference.
Portrait study 3. Source image and extracted relationships between figure, ground, focal point, and negative space.
Different subjects, a shared method of comparison.

What survives
a change of image?

Portraiture is the reference for a wider inquiry into landscapes, interiors, garments, photographs, and generated images. The current study finds that grammar follows the spatial problem a genre addresses.

A collection of source portraits from the working corpus.
Portraits in the working corpus.

Project details

Period
2026–ongoing
Location
New York
Form
Computational study
Status
In development
About the project

A computational study of latent visual grammar across portraiture and adjacent image regimes.

Across centuries, cultures, and media, do painted portraits share an underlying spatial grammar? The study began with portraiture and now extends to landscapes, interiors, garments, photographs, and machine-generated images.

It looks for compositional structure that survives a change of artist, period, style, or mode of image making.

Method

Extraction

  1. Open collection images
  2. Genre classification
  3. Geometric relationships

Comparison

  1. Vector primitives
  2. Comparison across the corpus
  3. Convergence and divergence

The pipeline extracts relationships between figures, horizons, focal points, and negative space. Each documentation plate pairs a source portrait with its extracted spatial grammar.

The study’s finding so far is that grammar follows the spatial problem a genre addresses. Portrait and landscape grammars diverge; portrait and figure-study grammars converge.

Research scope

Portraiture is the most densely sampled genre and the reference grammar for every other scan. The work widens into other image types and sources, including photographic archives, contemporary image libraries, and machine-generated imagery. Each source tests whether the grammar survives a change of regime.

Research record
Reference
Portraiture
Corpus
Open collections and adjacent image sources
Output
Diagrams and a planned longitudinal paper
Stage
Active research
Research & connections
All work →