Imagine a vibrant sunset — the light slowly passing the horizon in a symphony of color before revealing a stunning, starry sky. As it turns out, beautiful imagery such as this, according to a study reported in the scientific journal PNAS Nexus in December, may in fact be easier on the brain through a lessened consumption of energy compared to other varieties of images and scenes.
While it is of only modest mass compared to the rest of the body, the brain is the most energy-hungry organ in it — using about 20% of the body’s total energy and 44% of that energy being used to power its visual system. When our eyes are actively viewing a particular scene or image, a series of electrical and chemical activities requiring both glucose and oxygen to function are triggered. While the simple act of viewing doesn’t typically feel like a daunting task, this exchange of energy for information has been suggested by evolution to be an area adapted to avoid the wasting of energy based on certain images, such as the use of heuristics — mental shortcuts that focus on the use of feeling to guide decision-making.
In order to test their hypothesis developed from this evolutionary basis that a liking to certain aesthetics is inversely related to the metabolic cost of neutral representations (the energy spent on viewing said image), researchers Tikai Tang, William A. Cunningham, and Dirk B. Walther from the University of Toronto launched a study utilizing both artificial neural networks and human subjects. To analyze the energy consumed in the visual process, the researchers turned to an MRI dataset where four individuals viewed 5,000 images while being actively monitored — the measurements of oxygen consumption throughout the brain acting as indicators of metabolic activity. In addition to this data, images were run through the artificial neural network known as VGG-19—a convolutional neural network model that aims to mimic the layers of the visual cortex of the brain—that was measured for metabolic cost based on the number of neurons activated when presented the series of images.
Results from these exercises were put into context through aesthetic ratings of the images that were collected from 1118 participants, recruited using Amazon Mechanical Turk, who responded in an online survey and were asked to score each picture on a scale of five. The metabolic cost estimates were compared to these survey results to reveal that, in both the case of human and neural network data, the effort required to process images was inversely proportional to their aesthetic ratings — supporting the team’s initial hypothesis. In fact, not only images that were deemed to be more pleasing but also those that featured less complicated aspects were subject to greater energy efficiency — backing the “processing fluency” theory in psychology that describes how information processed more quickly generates a greater positive response. The results were found to be most significant in higher-level visual regions of the brain, such as the fusiform face area and its equivalent on the neural network (responsible for recognizing faces), suggesting energy savings occur mostly at these more advanced stages of visual process as opposed to catering toward lower-level functions such as edge and contrast detection.
What this study helps begin to illuminate is how our brains value certain images: While a pure white wall may be efficient to look at due to its simple nature, that lack of complexity also makes it difficult to generate interest. According to these results, an “interesting” image or scene is one that both has relative simplicity and a degree of information to give way to certain complexity that may derive the feeling of awe in appreciating it as something beautiful. Beyond this, a previous study demonstrated that people also tend to like faces and even cars that are closer to the average rather than those that were more unique — a phenomenon explained by the idea that outliers require the brain to spend more energy to update internal models of what both cars and faces look like.
While the inherent qualities of what makes particular images or scenes “beautiful” are not yet quite defined by this study, further research may seek to better understand both this, as well as the dynamic between cognitive rewards that are derived from taking greater energy to understand certain, more complex images that have the ability to outweigh their simple visual costs.
