“The idea of murder is represented a lot in the brain"
April 27, 2016 10:28 AM   Subscribe

I've been to lots of neuroimaging talks and Galant's is the only one that's ever truly impressed me.
posted by escabeche at 10:52 AM on April 27, 2016

Yeah, I saw Gallant talk in San Francisco a few years back, and his interests are not only limited to language. Here is his lab's web page with some of the more interesting work they get up to.

It's pretty impressive what can be done with a 7 Tesla magnet and big ideas.
posted by logicpunk at 11:32 AM on April 27, 2016

I'm very cautious about these results. You through enough words and data points and look for correlations and you will find some. The 'high' correlation performance between the projected results and the test results is around 0.5. I'm not an expert in these kinds of analyses, but if you are looking at thousands of 2x2x4 mm chunks of brain, you're bound to come up with something that looks 'interesting'.
posted by demiurge at 11:49 AM on April 27, 2016 [3 favorites]

Much better than most of the articles journalists jump on because it has the magic words "neuroscience" and "fMRI." The interactive link embedded in the text is pretty cool.

I had a pretty good idea that "families" of words hung out together in the brain because of language experiences I'd had. Here's the one that really makes me look stupid but also illustrates this general idea from the article. I knew about the musician Taj Mahal. I knew about the building Taj Mahal. But for twenty years, I never connected the two! Here's another one: I knew there was a sandwich shop named Subway, but it wasn't until I looked at the wallpaper (a subway map, if you've never been in one of their stores) that I realized the food chain's name was also the name used for an underground transportation system.

This kind of data sheds light on--and raises questions about--a lot of the epistemological and phenomenological questions that all of us, philosophers or not, wonder about.
posted by kozad at 11:53 AM on April 27, 2016

if you are looking at thousands of 2x2x4 mm chunks of brain, you're bound to come up with something that looks 'interesting'.

See the night sky. It has hunters and goats and rams and bulls and lions and virgins.
posted by bukvich at 11:53 AM on April 27, 2016 [1 favorite]

I'm also cautious, but similar words are found near each other in the brain, which suggests it's at least not completely random. The most surprising finding is probably that there is correlation in the word locations between people.

fMRI experiments have such a miserable history that papers should probably always include the same experimental setup repeated with dead salmon.
posted by ikalliom at 12:52 PM on April 27, 2016 [5 favorites]

but if you are looking at thousands of 2x2x4 mm chunks of brain, you're bound to come up with something that looks 'interesting'.
Scanning the paper a bit, their voxelwise regression models use a false discovery rate procedure to correct for multiple comparisons.
posted by adoarns at 2:50 PM on April 27, 2016

The problem with this kind of reporting is that it fails to accurately reflect the hypothetical nature of the entire conceptual framework that this experiment is based on.

to merely say something like
"For example, on the left-hand side of the brain, above the ear, is one of the tiny regions that represents the word “victim”.

This whole idea of "representation" of concepts by sections of the brain is itself extrememly controversial. the fMRI measures blood movement doesn't it? There is a whole slew of controversial suppositions that get you from:

a) blood is moving in this section when they say "cow" to this reported claim that:

b) this part of the brain "represents" the word "cow".

This whole idea of "representation" could be just the Phlogiston of neuro-psycho-science, we really just don't know.
posted by mary8nne at 3:04 PM on April 27, 2016 [6 favorites]

mary8nne: This whole idea of "representation" could be just the Phlogiston of neuro-psycho-science

Yes - the whole idea of the spatial extent of a meaning is problematic to begin with. Those are entirely different categories.
posted by rd45 at 4:30 AM on April 28, 2016 [1 favorite]

Oops, the post I intended to make on this very topic. Adding this respectfully into your thread.

The brain dictionary, a semantic atlas of the cerebral cortex

A marvelous, interactive brain map (slow to load but worth it) showing how words (in the form of voxels) are stored. | YouTube video explaining the map| Mapping semantic systems in the brain | Brain mapping.

This is your brain on The Moth. Scientists used Moth stories to make a sematic map. Alex Huth, Jack Gallant and colleagues set out to map the functional representations of semantic meaning in the human brain using voxel-based modelling of functional magnetic resonance imaging (fMRI) recordings made while subjects listened to natural narrative speech. They find that each semantic concept is represented in multiple semantic areas, and each semantic area represents multiple semantic concepts. The recovered semantic maps are largely consistent across subjects, however, providing the basis for a semantic atlas that can be used for future studies of language processing.

Gallant Lab on Twitter

Alexander Huth on Twitter
posted by nickyskye at 10:38 AM on April 28, 2016 [2 favorites]

Thank you so much, nickyskye. I shoulda checked with you first ;)
posted by Johnny Wallflower at 11:00 AM on April 28, 2016 [1 favorite]

« Older BioWare Goose Cam   |   Scots are mad for citrus! Newer »

This thread has been archived and is closed to new comments