Korva is a scroll of the internet’s best ideas.
Language network shows up without language tasks
Mapping with tasks
Researchers have long mapped language with tasks. In one, people read sentences or matched lists of nonwords one word at a time. In another, they listen to intact speech or the same speech filtered so it sounds similar but carries no meaning. These very different tasks give basically the same set of regions: a distributed, mostly left-lateralized network across the temporal lobe and inferior frontal lobe.
An intrinsic network
That convergence suggests an intrinsic structure that works together in support of language. The new preprint takes the tasks out altogether and looks at how the regions covary over time during arbitrary tasks. The same network appears, so what was called a language network from task data is there in a deep sense, an intrinsic property of how the adult brain is organized in most people.
Is it selective?
Whether that network is selective for language is a controversial claim. One view holds that other general-purpose systems get recruited. A system that parses temporal structure could handle the steps of making a peanut butter sandwich, structure in music, and language alike. The other view holds that the network is selective for language, and this study is part of the evidence for it.
How can researchers see regions rise and fall together?
The method is called resting-state functional connectivity. It tracks slow fluctuations in the fMRI signal in different regions and measures how closely those fluctuations match. Bharat Biswal and colleagues reported in 1995 that left and right motor cortex showed correlated fluctuations in people lying still at rest. Regions that belong to the same system tend to rise and fall together.
Why compare sentences with lists of nonwords?
Nonword lists look and sound much like real language but carry no meaning or sentence structure. Subtracting the response to them from the response to sentences leaves the processing that depends on meaning and grammar. The same logic applies to filtered speech, which keeps the sound of speech while removing its content. This sentences-minus-nonwords contrast is widely used as a language localizer.
Why is the language network mostly left-sided?
The link between the left hemisphere and language goes back to Paul Broca, who reported in 1861 that a patient with damage to the left frontal lobe had lost the ability to speak. Carl Wernicke described a different left-hemisphere area tied to comprehension in 1874. In most right-handed people, language relies mainly on the left hemisphere. The right hemisphere has its own roles in language, such as tone of voice.
More from Wu Tsai Neurosciences Institute, Stanford
