The Science of Beauty: What Makes Art Beautiful?

Show someone a painting they find beautiful while their head is inside an fMRI scanner, and a patch of the brain behind the forehead becomes more active in proportion to how strongly they say they feel it. The same patch responds to music they love. That finding, from a 2011 study of 21 volunteers, anchors the young science of beauty, a field called neuroaesthetics, which asks what happens in the brain when art moves us. The answers involve reward circuitry, pattern preferences, expectations, expertise, and personality, and they come with real limits.

Key takeaways

• A region called the medial orbitofrontal cortex tracks how beautiful people rate paintings and music, which suggests beauty is processed as a kind of value.
• Researchers describe aesthetic experience as three interacting systems: sensory-motor, emotion-valuation, and meaning-knowledge.
• Images with mid-range fractal complexity are often preferred, but the popular claim that Jackson Pollock "hit the optimal number" is more complicated than it sounds.
• Knowing where an image supposedly came from changes both ratings and brain activity, even when the image is identical.
• Brain scans correlate with beauty ratings; they do not explain why a particular work endures.

Why a science of beauty only began around 1999

For most of history, beauty belonged to philosophers. They argued about whether it lives in objects, in observers, or in the relationship between them, and they had no way to check. That changed when brain imaging became cheap and good enough to run on healthy volunteers who were simply looking at pictures.

The neuroscientist Semir Zeki of University College London coined the term "neuroesthetics" in 1999, the year of his book Inner Vision, and the field received a formal definition in 2002 as the scientific study of the neural bases for contemplating and creating art (Wikipedia: Neuroesthetics). In 1999 the neurologist Vilayanur Ramachandran and William Hirstein also proposed eight "laws" of artistic experience, such as peak shift (exaggerating the features that make something recognizable) and grouping. That theory is generally described as highly speculative. The field's more cautious branch, experimental neuroaesthetics, took a different route: show people art, ask them to rate it, and scan their brains.

That route has a built-in limitation worth stating early. Functional magnetic resonance imaging (fMRI) measures changes in blood oxygenation, an indirect stand-in for neural activity, with a time resolution of seconds and a spatial resolution of a few millimeters. It is a bit like judging a city's mood by satellite photographs of its electricity use. You can see which districts are busy, and you can notice when the pattern shifts. You cannot hear what anyone is saying.

With that caveat in mind, here is what the studies actually found.

What the scans show: a shared patch for beauty

Kawabata and Zeki ran an early version of the experiment in 2004. Volunteers rated paintings as beautiful, neutral, or ugly while being scanned, and activity in the medial orbitofrontal cortex (mOFC), a region on the underside of the frontal lobe, was higher for the paintings they rated beautiful. A second study, by Cela-Conde and colleagues in 2004, found that a different region, the dorsolateral prefrontal cortex, became active selectively for stimuli judged beautiful (Wikipedia: Neuroesthetics).

The study that gets cited most often is Ishizu and Zeki's 2011 paper in PLoS ONE. They scanned 21 healthy adults who had earlier rated paintings and musical excerpts. From those ratings, the team selected 30 paintings and 30 excerpts, each set divided into beautiful, indifferent, and ugly groups, and then recorded brain activity as the volunteers rated them again. The mOFC was the only cortical area active for both visual and musical beauty, and the two patterns overlapped almost completely in the same part of the region, which the authors call field A1. Activity there rose and fell with the declared intensity of the experience (Ishizu and Zeki 2011, PMC).

The authors proposed that this points to a faculty of beauty that does not depend on which sense delivers the stimulus. A fair analogy is a currency exchange counter. Paintings arrive in one currency (visual patterns) and songs in another (sound patterns), but both are converted at the same window into a common unit of "how good is this to me."

One detail deserves attention. Activity in the caudate nucleus, a structure involved in reward processing, appeared only for visual beauty, not musical beauty, and the authors said they had no current explanation. The authors also called their theory tentative, to be tested in other domains. The mOFC is involved in valuation more generally, so the safe reading of the result is that beauty engages the brain's valuation machinery, not that the mOFC is a dedicated "beauty center."

The table below collects the main imaging findings discussed in this article, with the limit that applies to each.

Study	Year	Design	Main finding	Limit
Kawabata and Zeki	2004	Paintings rated beautiful, neutral, or ugly during fMRI	mOFC more active for paintings rated beautiful	Ratings are subjective; images were chosen by researchers
Cela-Conde and colleagues	2004	Beauty judgments during imaging	Dorsolateral prefrontal cortex active selectively for beautiful stimuli	Region has many other roles
Kirk and colleagues	2009	Same paintings labeled "gallery" or "computer-generated"	Gallery label raised ratings and mOFC and prefrontal activity	Shows expectation effects, not a beauty "center"
Ishizu and Zeki	2011	21 adults, 30 paintings and 30 music excerpts	mOFC (field A1) active for both modalities and scaled with rated beauty	Authors call the theory tentative
Vessel, Starr, and Rubin	2012	16 observers, 109 paintings, rated 1 to 4 for "how strongly does this painting move you?"	Most moving works engaged default mode network regions	Small sample; arousal not measured

The aesthetic triad: three systems, not one

If one region handled beauty, a single scan would settle the matter. It does not, and Anjan Chatterjee of the University of Pennsylvania and Oshin Vartanian reviewed the evidence in a 2014 paper in Trends in Cognitive Sciences. They concluded that aesthetic experiences emerge from the interaction of three systems: sensory-motor, emotion-valuation, and meaning-knowledge (Wikipedia: Neuroesthetics). The proposal is often called the aesthetic triad.

The sensory-motor system handles the raw materials: color, form, line, composition, and the physical dynamism of an image. The emotion-valuation system produces the affective response. That is often pleasure or reward, but it also covers mixed feelings, which matters later in this article. The meaning-knowledge system brings in memory, context, and interpretation, including what you know about the artist, the period, or the subject.

A meal gives a workable analogy. The sensory-motor system is the ingredients on the plate. The emotion-valuation system is your appetite and the pleasure of the first bite. The meaning-knowledge system is knowing that this is your grandmother's recipe, or that the chef trained for 10 years in Kyoto. Change any one of the three and the whole experience changes, even if the other two stay put.

Consider a thought experiment with three viewers standing in front of the same still life. The first is a painter who notices the way a single highlight on a glass pulls the eye across the composition: her sensory-motor system is doing most of the work. The second has had a long day and finds the colors soothing in a way she could not explain: her emotion-valuation system leads. The third has just read that the painting was made during a siege, and sees the ripe fruit as an act of defiance: his meaning-knowledge system is in charge. All three might say the painting is beautiful, and for three different reasons. This is a hypothetical scenario, not a reported experiment, but it shows why a single score on a beauty scale hides so much.

The triad earns its keep because it predicts something a single-region theory cannot: dissociations. Two people can have nearly identical visual processing of the same canvas and still diverge sharply in how they feel about it, because the other two systems differ. It also explains why the same person can be unmoved by a painting on Monday and struck by it on Friday after reading about the artist's life.

When art moves you: the default mode network enters

Beauty and being moved are related but not identical. Edward Vessel, G. Gabrielle Starr, and Nava Rubin took on the second experience directly in a 2012 paper in Frontiers in Human Neuroscience. They recruited 16 observers at New York University and showed each of them 109 paintings for six seconds apiece in a 3-tesla scanner. After each image, observers rated on a scale of one to four how strongly the painting moved them (Vessel, Starr, and Rubin 2012).

For the works that earned the top rating, a set of frontal and subcortical regions showed a step-like increase in activity that was absent for the lower ratings. These included the left inferior frontal gyrus, the left lateral orbitofrontal cortex, the left mediodorsal thalamus, the left hippocampus, and the left substantia nigra. Parts of the default mode network (DMN) joined in as well. The DMN is a set of regions that tends to be more active when the mind turns inward, to memory, imagination, and thoughts about the self. The default mode network usually quiets down when attention is pulled toward an external task. In this study, the anterior medial prefrontal cortex dropped below baseline for most paintings, as expected, but stayed at baseline for the most moving ones.

The authors interpreted this as a possible gateway into the DMN, linking the sensory and emotional response to personal relevance. A rough analogy: ordinary viewing is like reading a map of somebody else's town, while being moved is finding your own street on it. The painting stops being only an object out there and becomes entangled with your own history.

One feature of this study makes it more interesting than a pure "reward" account: agreement between observers was low. The mean pairwise agreement on which paintings were moving was only 0.13, so the paintings that moved one person often left another unmoved. The authors also stated a limitation plainly: they did not explicitly measure physiological arousal.

Fractals, complexity, and the "just right" image

A separate line of research asks whether certain image statistics are inherently pleasing. The best-known example involves fractals, patterns that look similar at different scales, like a fern whose leaflets resemble the whole frond. Mathematicians describe such a pattern with a number called the fractal dimension, D. A smooth line has a D of 1, a pattern that fills a flat surface completely has a D of 2, and values in between describe how much of the plane the pattern's lines fill. You can think of D as a roughness dial: low values are sparse and calm, high values are dense and busy.

The physicist Richard Taylor measured the D values of Jackson Pollock's drip paintings. According to the review of the work in Frontiers in Human Neuroscience, an untitled 1945 painting had a D of 1.10, and the classic-period paintings reached about 1.7. A 1950 painting measured at 1.89 was one Pollock reportedly erased, after which he returned to canvases near 1.7 (Taylor and colleagues, Frontiers 2011).

Viewers, meanwhile, tend to prefer a lower, middle band. The same review reports that Spehar and colleagues (2003) tested 220 participants on mathematical, natural, and Pollock-style fractals and found preference peaked at a D between 1.3 and 1.5. Hagerhall and colleagues (2004) found that 119 participants preferred a D of 1.3 in skyline contours, and an earlier computer-generated study by Aks and Sprott (1996) also found a preferred value of 1.3. Taylor's own work with 120 participants found that 113 preferred fractal patterns to non-fractal ones.

Some studies add a physiological angle. In a 2006 study at NASA-Ames with 24 participants, mental tasks caused the smallest rise in stress when the viewed pattern had a D of 1.4. An EEG study by Hagerhall and colleagues (2008) with 32 participants found the largest changes, including maximal frontal alpha activity associated with relaxation, at a D of 1.3.

Source	Year	Sample	Finding
Aks and Sprott	1996	Computer-generated fractals	Preferred D about 1.3
Spehar and colleagues	2003	220 participants	Preference peaked at D 1.3 to 1.5
Hagerhall and colleagues	2004	119 participants, skyline contours	Most preferred D about 1.3
Taylor	2006	24 participants (NASA-Ames)	Smallest stress rise at D 1.4
Hagerhall and colleagues	2008	32 participants (EEG)	D 1.3 produced the largest changes

Notice what this does to the popular story. It is often said that Pollock "discovered" the mathematically ideal pattern, but his mature paintings sit near 1.7, above the 1.3 to 1.5 band where preference peaks in these studies. What the data support is narrower and still interesting: people consistently favor moderate complexity over both emptiness and clutter, and natural scenes happen to occupy that range. The number of stress-reduction percentages floating around the internet for this research should be treated with suspicion. The review reports a smallest rise in stress, not a percentage.

Other mathematical stories about beauty, such as the golden ratio in art and nature, raise the same question: what is the evidence that a ratio, rather than familiarity or convention, does the work? The fractal studies are more careful than most, and they still measure preference for simple stimuli, not for finished masterpieces.

Fluency, familiarity, and why "easy" can be beautiful

A different explanation for pattern preferences does not require any special geometry. Rolf Reber, Norbert Schwarz, and Piotr Winkielman proposed in 2004 that aesthetic pleasure is a function of the perceiver's processing dynamics. In their processing fluency account, the easier an image is to process, the more positively people evaluate it. They reviewed variables such as figural goodness, figure-ground contrast, repetition, symmetry, and prototypicality, and argued that each influences judgments by changing how fluently the brain handles the stimulus. Priming, either visual or semantic, increased pleasure judgments as well (Reber, Schwarz, and Winkielman, Personality and Social Psychology Review, 2004).

Think of walking downhill. A fluent image lets perception coast. That is pleasant, and it explains why simple, balanced, well-organized pictures often score well. It also explains why repetition helps: the mere exposure effect, in which repeated contact with something makes us like it more, is one of the oldest findings in this area of psychology.

If fluency were the whole story, though, the most beloved art would be wallpaper. Psychologist Daniel Berlyne argued in the 1970s that pleasure peaks at a moderate level of "arousal potential," so images that are too simple bore and images that are too complex overwhelm. The resulting inverted U is the formal version of the Goldilocks intuition behind the fractal results. How cleanly the curve holds across different stimuli is debated. The practical lesson is that "moderate complexity" is a tendency in lab stimuli, not a recipe for a masterpiece.

Symmetry belongs in this section too. Reber and colleagues list it among the features that raise fluency, and people do tend to favor symmetrical forms. But the claim that mildly imperfect symmetry beats perfect symmetry, often tied to the Japanese idea of wabi-sabi (finding beauty in imperfection), is better described as a plausible idea than as a firm finding. It is a good description of how many people experience a hand-thrown bowl. It is a weaker claim about the average viewer in a scanner.

Labels, expectations, and the museum effect

Context is not decoration around the experience. It is an input. In a 2009 study in NeuroImage, Ulrich Kirk, Martin Skov, Oliver Hulme, Mark Christensen, and Zeki took a single database of artworks and randomly labeled each image as either from a gallery or computer-generated, while volunteers were scanned. Ratings were significantly higher for images carrying the gallery label, and activity in the medial orbitofrontal and prefrontal cortex tracked that boost. Activity in the temporal pole and entorhinal cortex tracked the context itself, independent of how much people liked the image (Kirk and colleagues, NeuroImage, 2009, as indexed in Europe PMC; the studies below are summarized on Wikipedia: Neuroesthetics).

Other studies point the same way. Lacey and colleagues (2011) found that the ventral striatum responded more to an image's "art status" than to its content. Huang and colleagues (2011) found that authentic Rembrandt portraits and copies evoked different neural responses when viewers were told which was which.

It is tempting to read these results as proof that art appreciation is snobbery. That conclusion goes too far. A gallery label is information: it tells you that curators, critics, and other viewers selected the work from a far larger pool, and that is a reasonable cue that the work rewards attention. A brain that raises its valuation on that cue is doing something sensible. What the experiments demonstrate is that "beauty" cannot be read off the image alone. It is computed from the image plus a prior expectation, and the same is true of wine, music, and food. This is why a work can feel different when you see it in a museum with a story attached and why prices in the art market are driven so strongly by provenance and reputation.

Expertise: the trained eye sees different things

Does knowing more about art change what you look at? One early experiment suggests so. Stine Vogt and Svein Magnussen compared nine artists with nine untrained participants in a 2007 study in Perception. Each person viewed 16 pictures, ranging from ordinary scenes to abstraction, with free scanning and memory instructions. Untrained viewers spent more time on human features and recognizable objects, while artists devoted more scanning time to structural and abstract features.

The sample was small, so the result is best read as a tendency, not a rule. Still, it fits a broader point. Experts and novices are often not looking at the same thing. A trained viewer may be tracking how a diagonal stabilizes a composition, how edges are handled, or how color temperature shifts across a surface, while a novice is asking what the picture is a picture of.

That difference helps explain the frustration many people feel with abstract art. When there is no recognizable subject to hold onto, a viewer who relies on subject matter has little to work with, while a viewer attuned to structure has plenty. Neither is wrong. They are using different channels, and the channel can be trained. The same logic applies to color theory, where a few hours of learning why a complementary accent works can change what a painting looks like afterward.

Why some beautiful art is sad, strange, or disturbing

Beauty and pleasure are not synonyms, and some of the most admired works are painful to look at. Aristotle noticed this problem about tragedy in the Poetics, where he argued that watching suffering on stage arouses pity and fear and produces a kind of emotional purging he called catharsis. Modern researchers have not settled why negative emotion within art can feel rewarding, and the answers on offer remain debated.

What the triad framework contributes is a way to see the puzzle clearly. The emotion-valuation system handles nuanced mixtures of feeling, not only pleasure. The meaning-knowledge system tells you that the scene is a painting and not an event, which can turn dread into awe. And the Vessel study shows that being moved is a distinct state with its own neural signature, not a rating of how pleasant something is. People rate paintings as moving that they would never call pretty.

This is also where the psychology of awe comes in. Awe is often described as the feeling of encountering something vast that challenges your existing mental frameworks, and it has a strong presence in descriptions of the most powerful aesthetic experiences. Music behaves the same way, and the overlap Ishizu and Zeki found between visual and musical beauty is one reason researchers compare the two. For more on the auditory side, see how music affects the brain.

The same emotional machinery explains why making art can be therapeutic as well as viewing it. That is a separate question with its own evidence base, covered in art therapy and the brain.

Why we disagree: personality, history, and the problem of one brain

If beauty has neural foundations, why do people argue about it? The Vessel study offers one part of the answer in its design: agreement between observers about which works were moving was very low. That is a result, not a flaw. It says individual variation is large enough that two people can look at 109 paintings and produce only weakly related lists.

Personality and background plausibly matter too, but the studies described above did not test them directly, so this article does not put numbers on their effect.

Personal history also matters, in ways that are difficult to study because they are unique to each person. A piece of music that was playing during an important event carries emotional weight that no analysis of its structure can capture. This is part of why the psychology of nostalgia overlaps with aesthetics: familiarity bundled with personal memory can outrank any formal quality of the object itself.

Even something as basic as color preference shows several layers at once. Some responses to hue may be widely shared, some are learned from culture, and some come from private associations. Sorting these out takes careful work, and the popular claims are often stronger than the evidence, as the discussion of color psychology shows.

What the science cannot tell you

A fair summary of the evidence needs a list of what it does not show.

• Correlation is not an explanation. An fMRI scan shows where activity rises when someone reports beauty. It does not show that this activity is what makes the experience beautiful, and the same regions respond to other kinds of value and reward.
• Samples are small. The studies described above enrolled 16, 21, 24, and 32 people. That is typical for the technique and also a reason for caution about sweeping claims.
• Stimuli are narrow. A reviewer of the field notes that stimuli are mostly Western, which can bias results through familiarity. A six-second glimpse of a painting on a screen is not an afternoon in a cathedral.
• Reports are filtered. Experiments depend on verbal ratings, and people may not report accurately what they feel or why.
• Grand theories outrun experiments. Critics point out that the field's broad claims are often wider than what its narrow experiments test.

Perhaps most important, a measurement of response is not a measurement of greatness. Science can tell you which images tend to be preferred in a lab, which brain regions track the preference, and how much expectations shift it. It cannot yet say why one painting from the 17th century keeps pulling people toward it in 2026 while thousands of technically similar ones do not. Part of that answer is surely cultural and historical, which is a different kind of explanation from the one a scanner provides.

The takeaway

Beauty in art is not a single property and not a single brain region. It is the outcome of three systems working together: one that reads the image, one that assigns it value, and one that places it in a web of knowledge, memory, and expectation. Some inputs to that process look shared across people, such as a taste for moderate complexity and a valuation circuit that responds to both a painting and a song. Others are learned, and the labeling and expertise studies show how readily they can move the result.

That structure resolves the old argument about whether beauty is in the object or the observer. It is in the interaction, and the interaction can change. A viewer who learns to notice more has not talked themselves into liking something; they have given their valuation system better input. The sensible stance toward the science is interest without overreach: brain imaging has opened real questions about why art matters to us, and it has not closed any of the hard ones.

FAQ

Q: Is beauty in the eye of the beholder, or is it universal? A: The evidence supports both, in different layers. Some patterns of preference, such as moderate fractal complexity, recur across studies, and the same orbitofrontal region tracked beauty for paintings and music in one study. Individual agreement about which paintings are moving was very low in another, and labels and expertise measurably shift ratings. A shared biological foundation with large individual variation on top is the most defensible summary.

Q: What is neuroaesthetics? A: Neuroaesthetics is the scientific study of how the brain responds to art and beauty. Semir Zeki coined the term "neuroesthetics" in 1999, and it was formally defined in 2002 as the study of the neural bases for contemplating and creating art. It uses fMRI, EEG, and behavioral experiments, and critics note that its experiments are narrower than its biggest theories.

Q: Which part of the brain responds to beautiful art? A: The medial orbitofrontal cortex is the region most consistently reported. In Ishizu and Zeki's 2011 study of 21 adults, it was the only cortical area active for both visual and musical beauty, and its activity scaled with rated intensity. It is also involved in valuation more broadly, so it is better described as part of the brain's value system than as a dedicated beauty center.

Q: Did Jackson Pollock paint the mathematically ideal pattern? A: Not exactly. His classic paintings measure at a fractal dimension near 1.7, while viewers in several studies preferred a lower, mid-range value of about 1.3 to 1.5. The more accurate claim is that Pollock's work is fractal, like natural forms, and that people generally prefer moderate complexity. The stress-reduction percentages often quoted online are not supported in the review cited here.

Q: Does knowing about art make it more beautiful? A: Often, yes. In a 2009 experiment, identical images received higher ratings and stronger orbitofrontal activity when labeled as coming from a gallery. In a separate study, artists and untrained viewers spent their viewing time on different features. Both results suggest that knowledge changes the input to the brain's valuation system, though the studies are small and the effect sizes vary.

Q: Why can sad or disturbing art feel beautiful? A: Researchers describe the emotion-valuation system as handling mixed feelings, not only pleasure, and the meaning-knowledge system as framing the scene as art. Being moved is also a distinct state: in the 2012 Vessel study, the most moving paintings engaged brain regions that lower-rated ones did not. Why negative emotion can be rewarding in art is still debated.

Sources

• Ishizu, T., and Zeki, S. (2011). Toward a Brain-Based Theory of Beauty. PLoS ONE 6(7): e21852.
• Vessel, E. A., Starr, G. G., and Rubin, N. (2012). The brain on art: intense aesthetic experience activates the default mode network. Frontiers in Human Neuroscience.
• Taylor, R. P., and colleagues (2011). Perceptual and physiological responses to Jackson Pollock's fractals. Frontiers in Human Neuroscience.
• Neuroesthetics overview, used for the history of the field and the summaries of the Kawabata, Cela-Conde, Kirk, Lacey, and Huang studies.
• Chatterjee, A., and Vartanian, O. (2014). Neuroaesthetics. Trends in Cognitive Sciences.
• Reber, R., Schwarz, N., and Winkielman, P. (2004). Processing fluency and aesthetic pleasure. Personality and Social Psychology Review.
• Vogt, S., and Magnussen, S. (2007). Expertise in pictorial perception: eye-movement patterns and visual memory in artists and laymen. Perception 36(1): 91-100. doi:10.1068/p5262.
• Kirk, U., Skov, M., Hulme, O., Christensen, M. S., and Zeki, S. (2009). Modulation of aesthetic value by semantic context: an fMRI study. NeuroImage.

---

Read next:

• How Perspective Tricks the Eye
• The Psychology of Creativity: How the Brain Generates New Ideas
• How Pigments Create Color in Paintings