Michael Jackson's classic transformation from a panther to a human during the music video "Black and White" may cause confusion upon first viewing for good reason. Now neuroscientists have used a similarly disruptive trick to show how the brain recognizes objects under changing conditions.
Their new experiments succeeded in retraining monkey brains to blur the distinction between two objects – a Dalmatian dog image and a rhinoceros image. Such results provide the strongest proof yet of how the brain relies upon time-dependent image sequences to learn how to recognize objects.
"Don't think of the visual system as explicitly learning about object differences," said James DiCarlo, a neuroscientist at MIT. "Instead, think of the visual system as explicitly learning to group images that likely arise from the same source."
Latest Videos From
The mind can usually get away with the assumption that images appearing rapidly one after the other – like a flipbook – belong to the same source. Researchers suspect that the learning process, called "temporal contiguity," trains the brain to recognize objects, such as humans, from different distances, different angles and in both well-lit and shadowy situations.
Confusing monkeys
DiCarlo and Nuo Li, a brain researcher at MIT and lead author on the study, decided to test the theory by intentionally confusing the learning process. Their work is detailed in the Sept. 23 issue of the journal Neuron.
The study monitored spikes in activity among brain cells in the inferior temporal cortex of the brain – an area suspected of having activity related to object recognition. Two Rhesus monkeys served as the test subjects, so that the researchers could replicate their experimental results.
Both monkeys watched a Dalmatian dog's image become larger or smaller on a computer screen, as though approaching or moving away from view. The twist came when the dog suddenly morphed into a rhinoceros as it became larger on the screen.
The temporal contiguity response proved so powerful that it reshaped brain activity patterns after a few hundred viewings. Brain cells that usually fired up in response to the large Dalmatian image eventually began to see activity spikes in response to the large rhino image as well. A few switched over entirely and responded even more strongly to the large rhino image.
Pattern recognition
Of course, Dalmatians don't typically change their spots on a whim and become rhinos. The unusual experiment simply exploited the temporal contiguity process to reveal how it works to help monkeys (and likely humans) identify objects.
The learning process may apply beyond specific objects, such as Dalmatians, and could simultaneously train the brain to recognize broader categories, such as all dog breeds. The evidence so far supports that theory, even if it has yet to undergo full testing.
"One probably does not need to see every type of dog in every position or size to have the learning we describe build a useful representation of dogs," said Li and DiCarlo in an e-mail to LiveScience.
When the rules meet the real world
Temporal contiguity works well as a quick-and-dirty rule of thumb for the brain, as long as objects in this world don't spontaneously change identity. But what happens to a monkey or human who experiences a disruption of their object recognition learning outside the lab?
Such "natural experiments" in the real world seem hard to find, even without accounting for the technology needed to monitor such cases.
"This is unknown at the level of detail that we think is interesting, because we do not (yet) each wear cameras to record all the images that we see (and eye tracking would be required as well)," Li and DiCarlo explained.
The neuroscientists did come up with one possible scenario. A person who spent a significant part of his or her waking life watching "rapid-fire" TV – where each frame barely contains any action or movement – could theoretically have more object recognition problems.
For now, Li and DiCarlo plan to see just how the Dalmatian-rhino confusion affects monkeys' perceptual judgments, by training the animals to make behavioral signals in response to the Dalmatian or rhino images.
Live Science Plus
