Thought decoding in the abstract

Researchers have deciphered the abstract concepts people are thinking about – for example justice, truth and forgiveness – merely by analysing their brain scans.

Until now, this type of “thought decoding” has been largely confined to concrete concepts such as apple and hammer. The new findings, however, suggest slippery ideas that are not of the physical world also inhabit distinct parts of the brain.

The study is the work of psychologist Marcel Just and graduate student Robert Vargas from Carnegie Mellon University in the US.

It makes intuitive sense, they say, that physical or “concrete” objects, such as hammers and apples, will be represented in the brain similarly between people. Trade tools and fruit are, by nature, unambiguous.

It’s a contention born out in the science of neural decoding, where patterns of activity on brain scans are used to work out what someone is thinking.

Just, for example, has used brain scans to predict when a person is reading sentences that refer to concrete things, such as “the flood damaged the hospital”.

But given the fuzziness of abstract ideas like justice and ethics, intuitions cut the other way – could we really share common brain space for them too?

To find out, Vargas and Just put nine people in an MRI scanner and flashed an array of 28 abstract concepts at them, shown as words.

Those airy notions came from the worlds of maths (e.g. probability, multiplication), science (gravity, heat), society (gossip, intimidation), emotion (happiness, anger), law (ethics, crime), metaphysics (causality, necessity) and religion (deity, faith).

The pair then did a painstaking analysis of the scans that combined machine learning with something called multivoxel pattern analysis to look for similarities in the participants’ brain responses.

Abstract ideas, it turns out, are less mercurial than we thought.

The researchers were able to nail down the concepts people had in mind just by looking at their brain scans. And they did it with an average accuracy of .82 on a scale where 0.5 is mere chance.

“For me, the most exciting result of this study was that we were able to predict the neural activation patterns for individual abstract concepts across people,” says Vargas.

“It is wild to think that my concept of probability and spirituality is neurally similar to the next person’s, even if their experience of spirituality is different.”

The two then did a further analysis to map the brain’s representations of abstract concepts at a coarser “category” level.

They found some abstract ideas have social content, essential to forge understanding of things like gossip and intimidation. Other concepts lie on a continuum between the “internal” and “external” worlds, consciousness and gravity for example.

Perhaps most salient, though, is their finding that we tend to represent abstract ideas as words, in contrast to concrete objects which form “percepts” based on how they trigger the five senses.

“[A]bstract concepts both evoke less activation in regions associated with perceptual processing and evoke more activation in regions strongly associated with verbal processing,” the authors write in the journal Cerebral Cortex.

Which raises the prickly question of whether the science of “mind reading” has now advanced to include the abstract.

“It’s flashy to call this work mind reading,” says Just.

“For me, it is proof that we have identified some of the elements of the brain’s indexing system – verbal representation, externality/internality and the social dimension – that our brains use to code concepts that have no physical manifestation in the world.”