A brief history of our language for the brain: Vocabulary, dictionary, and poetry

To model how the brain works, we first need a list of its parts and some rules for putting them together—in other words, we need a language for the brain. The way that neuroscientists are developing this language is not unlike how a tiny human learns to speak. At the earliest stage, both a child and the field of neuroscience must gather words into a vocabulary (Panel A below). Like an infant who starts picking out words as the distinctive sounds it hears in a stream of speech, neuroscientists at the turn of the 20th century drew boundaries around brain regions according to the cellular structures they observed under the microscope. Toddlers go on to grasp the meaning of words, entering them into a mental dictionary; similarly, neuroscientists have mapped out the roles that brain regions play in language, vision, and other processes by having people perform tasks while their brains were scanned for indicators of neural activity (Panel B). As children string words into sentences, grammar becomes necessary, and neuroscientists have likewise turned to network models to explain how brain regions with similar activity patterns link together (Panel C).

The parallel between language development and the field of neuroscience begins to break down when we examine the brain’s network structure more closely. After brain regions are linked based on how well their activation levels track together over time, we see that the patterns between links are less straightforward than the grammatical parsing of a sentence. Recent research has shown that we can divide brain networks into distinct but tightly interconnected subnetworks. Are these subnetworks like the phrases that make up a sentence, or something else? In a typical sentence, meaning is built up in a predictable sequence from a phrase for a noun (“A red block…”) to a phrase for a verb (“…fell from a tower.”). In the brain, by contrast, mental processes unfold simultaneously as each subnetwork remains at least a little bit engaged at all times. Recent research led by Maxwell Bertolero has proposed that the brain can make sense of this cacophony by relying on a few central “connector hub” regions that coordinate interactions between subnetworks. For example, a connector hub may signal between a visual subnetwork, which perceives a red cube, and a language subnetwork, which maps thoughts to words, so that a person can call the toy in their hand a “block.” Thanks to connector hubs, what the brain “speaks” as it processes information need not be a linear sentence—it’s more like poetry.

Before we unpack this metaphor, let’s first examine the approach that Bertolero et al. took to parse the grammar of brain function. The researchers scanned the brains of nearly 500 adults with functional magnetic resonance imaging to view signals of neural activity. They mapped networks for every participant by looking for which brain regions showed similar activity patterns: we assume that if activity in two regions fluctuates up and down in concert over time, the regions are signaling with one another. To understand this idea, imagine you are sitting through a lecture and see two classmates sneak out their cell phones at about the same time over and over again. Even without reading their text messages, you might guess they are “signaling” to each other because their cell phone “activity” levels appear to be related. To demonstrate that brain networks support a wide range of behaviors, Bertolero et al. constructed computational models that used network features to predict how well an individual performed on tests of memory, language, math, problem solving, and social reasoning (Figure 2a in Bertolero et al.).

While most brain signaling takes place within subnetworks, Bertolero et al. hypothesized that connector hubs would coordinate the subnetworks with one another by signaling among them. Connector hubs were defined as brain regions with a relatively even distribution of links across multiple subnetworks (Figure 1e in Bertolero et al.), an arrangement that would position them to integrate neural signals originating throughout the brain. The connector hubs of the classroom would be the social butterflies who seem to be friends with everyone, no matter their social clique. The researchers found that having connector hubs with highly “diverse” links across many different subnetworks was related to how accurately an individual responded during a task. In addition, a “modularity” metric was calculated: the more similar the activity patterns within subnetworks, and the more different the patterns among subnetworks, the higher the modularity. Bertolero et al. showed that when modularity increased in relation to the diversity of a brain region’s links, so did task performance (Figure 3e in Bertolero et al.). In other words, to the extent that connector hubs are able to bridge between subnetworks, a modular network may be an ideal setup for messaging in brain.

We still do not know how the brain would work in the absence of connector hubs, so it remains to be proven whether they are essential. Nonetheless, the evolutionary implications of this research are an enticing area for speculation. It may seem curious that the brain would break itself down into subnetworks and thereby create the need for special regions to piece its processes together. One theory is that modular organization shortens the distance that most signals have to travel, increasing the speed of processing and reducing energy consumption. This same principle would explain why people congregate into cities—it means a shorter distance to travel to centrally located schools and hangouts, burning less fuel than if the population were spread out evenly across the land. Even though connector hubs have far-reaching links that burn through calories, they are small in number and carefully positioned to convey necessary bits of information between the subnetworks. Connector hubs are a bit like the postal workers that deliver specialty items you can’t buy in your local shopping mall. All in all, the human brain may achieve its optimal balance of efficiency and intelligence thanks to subnetworks that facilitate modularity and connector hubs that integrate across them.

So, what does this network model for brain function have to do with poetry? Let’s consider this question in the context of a few lines by Emily Dickinson:

Tell all the truth but tell it slant—

Success in Circuit lies

Too bright for our infirm Delight

The Truth’s superb surprise

As Lightning to the Children eased

With explanation kind

The Truth must dazzle gradually

Or every man be blind—

The usual word order is flipped in phrases like “Success in Circuit lies,” with the result that meaning unfolds circuitously. As we discover in the penultimate line, this grammatical construction reflects the key tenet of the poem—that “Truth” can be observed only when it is uncovered with gradual ease. To a greater degree than in everyday speech, the structure of the poem and its function directly inform one another. We should also note that the structure of the human brain evolved in tandem with its function. In the maps generated by Bertolero et al., we can see that functionally related neurons are huddled together into the regions that come together into subnetworks, with connector hubs situated (not by pure chance) in central regions between them (Figure 4 in Bertolero et al.). Our rich mental lives emerge from pulses through this network that, with the resolution afforded by neuroimaging, are caught in flashes as self-aware as Dickinson’s poem by scientists who seek the brain’s own Truth.

Edited by Manasi Iyer

References

Bertolero, M.A., Yeo, B.T.T., Bassett, D.S. & D’Esposito, M.D. A mechanistic model of connector hubs, modularity, and cognition. Nat. Hum. Behav. 2, 765-777 (2018).