Graham Publishes, Presents New Work – Hobart and William Smith Colleges \
The HWS Update

Graham Publishes, Presents New Work

Assistant Professor of Psychology Daniel Graham recently published work on aesthetic preferences in the leading journal Vision Research and presented research on this and related topics at Imperial College London.

In the paper “Preference for luminance histogram regularities in natural scenes,” Graham and his colleagues report findings showing the correlation between aesthetic preference and a particular pattern of light and dark in an image. In particular, they show in the paper that greyscale images that are well-balanced between light and dark are preferred over images that have more dark than light, or more light than dark — “one of the only ‘aesthetic universals’ (properties that are preferred by most or all observers) that seems to exist in humans,” Graham says.

At the Workshop on Cultural and Social Analytics at Imperial College London’s Data Science Institute in June, Graham presented past work as well as the newly published work in a talk “Visual Structure in Artwork,” focusing on the typical patterns of light and dark that artists include in their work.

“It might sound strange to say that art has a typical structure but it does. For example, on average it is a good bet that neighboring points in an art image will look similar. Not every point follows this rule but on average it is a good bet,” says Graham, who showed this in his Ph.D. work in a study of a large database of digitized paintings from the Herbert F. Johnson Museum of Art at Cornell University.

This common visual structure is also common in the natural visual world — not coincidentally, Graham says, since “it is something the human visual system has adapted to in the way it processes visual information. In fact, images that don’t have this structure are basically imperceptible, precisely because the visual system is not set up to handle them.”

As applied to art, “this finding suggests that art that violates the rule of neighboring points looking similar is very uncommon (though there are a tiny number of exceptions), and that artists are unlikely to make random or blurry art in the future,” Graham says. “So artists — even abstract artists, as we’ve found — reproduce the basic structure typical of the natural world.”

In his new work, however, Graham has shown that “some aspects of the natural world cannot — and perhaps should not — be reproduced. In particular, the range of light intensities in the natural world cannot be fully captured on canvas.”

In looking for the connections between aesthetics and the distribution of light intensities, Graham and his colleagues manipulated those distributions in artistic photographs and other natural scene images “so that only the proportion of light and dark was affected (not the average brightness or contrast).”

“We found — quite to our surprise — that versions of the images most like what we see in art (symmetrical, bell curve-like distributions) were preferred over versions of the same images that were more like they would appear in nature, which are asymmetrical,” Graham says. “This was true across every stimulus condition and participant group. It suggests that artists have figured out a particularly good way of using the light intensities available to them. In a sense the artist could be doing some of the early visual processing the brain would otherwise do, since neurons in the visual system also has a very limited range of firing. Perhaps, then, we like images that are easier for the brain to process.”

Organized by the evolutionary biologist Armand Leroi and computer scientist Yike Guo, the Workshop on Cultural and Social Analytics brought together a number of social scientists, natural scientists, computer scientists, artists and others to explore the acquisition, analysis and explanation of large amounts of cultural and social data.

“What’s kind of funny is how much this group takes for granted the notion of interdisciplinary work spanning the sciences and humanities, since there is still resistance to these endeavors in both the sciences and humanities,” Graham says, adding that “on the other hand, there is certainly a recognition in this group that data on its own — in the form of Twitter messages, high-definition art images, pop music databases, etc. — are only a starting point. Everyone at the conference agreed that the sorts of questions we ask as scientists ultimately have to be guided by an understanding of cultural output in context, even if we tend to use rather simplistic measures of culture in a first attempt.”

Graham joined the Hobart and William Smith faculty in the fall of 2012. He earned his B.A. in physics from Middlebury College. Graham went on to study at Cornell University, where he received his M.S. in physics and his Ph.D. in psychology.

This summer, Graham, Catherine Forman ’16, and Margaret Coldiron ’77 — Deputy Head of World Performance at the University of Essex’s East 15 Acting School — are collaborating on a new research project involving the study of face representations in masks.