# Pigeons learned in 15 days to sort breast histology images as benign or malignant, reaching nearly 85%

- Published: 2026-11-08 19:36 UTC
- Section: Life & Health › Medicine
- Tags: pigeons breast histology, Columba livia medical images, benign vs malignant tissue, mammogram microcalcifications, animal model of image perception, medical imaging evaluation, Richard Levenson UC Davis, majority vote accuracy
- URL: https://realtech.media/news/rt-b78b190973068dde862e434e_pigeons_learned_in/

**Birds trained with food reward separated benign from malignant breast tissue and handled new images, but on mammographic masses they only memorized the training set.**

![Pigeons learned in 15 days to sort breast histology images as benign or malignant, reaching nearly 85%](https://realtech.media/media/2026/11/319b242b4e10b1b6.jpg)
*U.S. Army photo by Monica K. Guthrie, Wikimedia Commons, Public domain*

In November 2015, researchers from the US, including Richard Levenson at the University of California, Davis, reported in PLoS ONE that pigeons (Columba livia) can learn to distinguish benign from malignant human breast histopathology [1]. The authors note that pigeons share many visual system properties with humans and that such a capability had not been documented before [1].

In the task, a bird sat in front of a touchscreen showing enlarged biopsy images and pecked a blue or a yellow button to choose between healthy tissue and breast cancer. A correct answer released a food reward. The images varied in size, tint and magnification.

## Histology: training and generalization

Within 15 days the pigeons went from chance level to nearly 85% correct answers each [1]. Shown images they had never seen, they kept classifying them correctly, which the authors read as learning of visual patterns rather than memorizing particular slides [1]. When the answers of all birds were pooled and the majority opinion was taken, accuracy rose to 99% [1].

As in humans, accuracy was modestly affected by the presence or absence of color and by the degree of image compression. According to the authors, further training reduced these effects [1].

## Mammography: calcifications and masses

On mammograms the pigeons detected microcalcifications, which are relevant to cancer [1]. On previously unseen images, however, the share of correct answers fell to around 72%, markedly below the histology results.

With suspicious mammographic densities (masses), a task that is difficult for humans as well, the birds only memorized the training images and could not generalize to novel examples [1].

The researchers do not propose entrusting diagnosis to pigeons. They see the birds as inexpensive observers for assessing medical imaging hardware, image processing and image analysis tools, and for studying how people perceive medical images [1].

## Sources

1. [PLoS ONE](https://doi.org/10.1371/journal.pone.0141357)
2. [science-et-vie.com](https://www.science-et-vie.com/nature-et-environnement/animaux/en-2015-des-chercheurs-ont-entraine-des-pigeons-pendant-15-jours-les-oiseaux-ont-appris-a-distinguer-des-tissus-cancereux-sur-des-images-medicales-265094.html)

---

Source: RealTech, realtech.media — https://realtech.media/news/rt-b78b190973068dde862e434e_pigeons_learned_in/