What the Study Found
- For hours at a time, a 28-year-old man saw his cat’s eye region on human faces after surgery on his right temporal lobe.
- The swap ran both ways: a familiar woman’s eye shape and pupil size appeared within his cat’s face.
- Only parts of a face transferred, not whole faces; the rest of each viewed face looked normal, unlike a 1986 poodle case.
- He tied the borrowed features to particular people or his cat, which the authors read as pieces pulled from long-term visual memory.
After brain surgery, a 28-year-old man named Karl began seeing his cat’s eyes on other people’s faces. His brain, it seems, replays pieces of remembered faces, not only whole ones: since surgery on his right temporal lobe, the eye region of his cat has repeatedly shown up on the faces of people he meets, and the eyes of people he knows have shown up on his cat, often for hours at a time. The operation, two years ago, removed part of the back of the lobe, where a tangle of malformed blood vessels had formed. The journal Cortex gave his case four pages.
This is not the first time Cortex has published something like it. In 1986 the journal ran the case of a woman who, nine months after a stroke, looked at a poodle on another passenger’s lap on a bus and then saw the poodle’s face on everyone aboard, a swap that outlasted the ride and finally resolved 30 minutes later.
Eyes Borrowed from a Cat
Palinopsia is the persistence or return of something seen earlier into what a person sees later. When it lands on faces and distorts them, the condition has a longer name, prosopometamorphopsia (PMO). Most such distortions alter the shape or position of features, or the color and texture of skin (a pattern the report traces to earlier work), and whole-face swaps like the poodle one are rare. Roughly 275 people with distorted faces have contacted the authors’ lab over the past five years, and several of them have described face palinopsia too.
Karl’s version is finer grained than the poodle’s. Instead of a whole face, only particular parts of a facial region move across, sometimes along with their coloring and texture, while the rest of the viewed face looks normal.
The transfer runs both ways. In one direction, the eye region of a person Karl knows (eye shape and pupil size included) appears inside the face of his cat; in the other, human faces take on the cat’s eye region, with enlarged pupils, cat-like eye sockets and the shading and texture around them. Episodes often last several hours and occasionally all day. The faces still hold together, according to the report: the borrowed features blend in rather than sitting on top, and they shift naturally as Karl’s viewpoint or distance changes.
Drawing a Percept Nobody Else Can See
Words only go so far for a percept like this, so the authors fed Karl’s verbal descriptions and a sketch he drew into AI image tools and refined the results over several months, responding to his feedback, until he judged the pictures a match for what he sees. “If you know what the person is experiencing, you can much better research what they’re dealing with and try to figure out ways to control it,” says lead author Brad Duchaine, a professor of psychological and brain sciences at Dartmouth College.
So what, exactly, is crossing over? Not generic features, for a start. Karl identified the transferred parts as belonging to particular familiar people or to his cat, which the authors read as a sign that the borrowed pieces are retrieved from long-term visual memory. The swaps stand out most when one species’ features land on another species’ face, though Karl has also had human-to-human transfers often. Taken together, the observations suggest that percepts of this high-level kind can involve the persistent replay of facial components tied to one individual, and not only whole-face representations.
The design is an observational case report on one patient, with no datasets generated or analyzed, so the pictures record what Karl says he saw rather than anything measured. Karl alone judged the match, and the report describes no independent check on the pictures.
The press release and the paper are not quite in step. The release says Duchaine sketched the image that started the process and quotes him saying the case demonstrates that single features can be pulled from memory to shape what we perceive, while the paper credits a sketch Karl drew for the cat-eyed face and words its own conclusion as an indication, drawn from one patient.
Duchaine says the method could reach beyond this one case. Patients could help make such images, he says, and then use them to show clinicians, friends and family what faces look like to them. He adds that his lab has heard from people who were told they had a psychiatric problem, and in some cases were put on antipsychotics, when they were in fact having visual problems. The authors themselves expect photorealistic visualizations to become an increasingly useful way of documenting such experiences.
The team plans to keep working with Karl. Whether other people with distorted faces are replaying pieces of faces too is something a single report cannot say.
Reference
Duchaine, B., Mello, A., & Nosker, J. L. (2026). Forty years after poodle-face palinopsia: Human-to-cat and cat-to-human facial feature transfer. Cortex, 205, 496–499. https://doi.org/10.1016/j.cortex.2026.10.003
- Study type: Single-patient case report (Clinical Postcard) in the journal Cortex, a peer-reviewed journal per the press release; the article page itself does not state review status.
- Sample size: One patient, a 28-year-old man; no datasets were generated or analyzed.
- Method: Clinical observation and the patient’s own descriptions of recurrent percepts, plus AI-generated visualizations refined over several months until the patient judged them a match.
- Analytic approach: Descriptive. The authors infer from the patient naming the source of each borrowed feature that the features come from long-term visual memory.
- Duration: Distortions had persisted for two years; the visualizations were refined over several months.
- Funding / conflicts of interest: Hitchcock Foundation, which had no role in the report. The authors declare no competing interests. They used ChatGPT to generate the visualizations and improve the writing.
- Data availability: No datasets were generated or analyzed.
- Main limitation: Not author-stated: one patient, with the match between pictures and percepts judged by that patient alone; generalization to other people with face distortions is untested.
FAQ
Why were AI images used to show what Karl sees?
AI images were used to show what Karl sees because they can carry far more detail than the line drawing made for the poodle case. Karl guided the process by describing and sketching what he sees and by judging each revision until he thought it matched. The catch is that his judgment is the only check, so the pictures illustrate his recollection and do not measure it. Duchaine says patients could use such images to show clinicians, friends and family what faces look like to them.
Is Karl’s case the same as the 1986 poodle case?
Karl’s case is related to the 1986 poodle case but finer grained. In the poodle case, the dog’s face took the place of every passenger’s face on the bus. In Karl’s case, only particular parts of a facial region move across, sometimes with their coloring and texture, while the rest of the viewed face looks normal.
Do people with distorted faces usually see another face swapped in?
People with distorted faces usually do not see another face swapped in. According to the report, most distortions alter the shape or position of features, or the color and texture of skin, and whole-face swaps are rare. Karl’s eye-region swaps between his cat and people he knows are an unusual variation on that picture.
Can face distortions be mistaken for a psychiatric problem?
Face distortions have been mistaken for a psychiatric problem, according to Duchaine. He says his lab has heard from people who were told they had one, and in some cases were put on antipsychotics, when they were in fact having visual problems. That is his account in the press release, not a finding of the report.
Cite This Page