Have you ever wondered how your brain manages to think about words? How you can see or hear a word and just know what it means? Well, these are questions that have been plaguing scientists for years. How does word meaning present itself in the brain? Previous studies have tried and figured out models of how the brain activates for individual words, but they didn’t account for the neurobiology that plays a role in how these words and their meanings present themselves in the brain. The scientists were trying to fit text to brain instead of brain to text – meaning doesn’t come from the words themselves, it’s your brain that assigns meaning based on experiences, senses, motion, social cognition, and many other factors. The complexity of word meaning makes it extremely difficult to try and encompass all of that in a tangible model.

Nonetheless, science persists. Scientists have found a large network in the brain that includes multiple different neural control systems and stores/receives conceptual information and have since been trying to identify what the network contains, and the knowledge organized within it. These studies have led to ideas about creating a computational model that is able to predict neuron activation relative to semantic content. As of 2017, some scientists seem to have accomplished just that.

In combining ideas from previous studies, scientists used a 65-dimensional system to assess each word so that each word could have features that encompass all of the neural processes/attributes they considered as contributing to semantic processing. These attributes consisted of sensory, motor, affective and cognitive information. Attributes were weighted differently per word, according to the average of how people rated words as normally associated with each attribute. Weighting the attributes according human perception of their word association allows us to account for more of the complexity involved in word meaning than was previously accounted for. It also gives meaning weight to the words that allows the computational model to work appropriately.

Here’s the thing: do words just appear in your mind by themselves? When you’re reading this article, are you reading it one word at a time or allowing the sentences or phrases to help you fully understand what’s being said? Assuming it’s the latter, processing and predicting the meanings of individual words becomes less natural to do than trying to assume the meaning of sentences and phrases. Taking this into account, scientists decided that in predicting the patterns of neural activation for meaning, it would be more natural to do it from whole sentences instead of isolated words.

In the experiment they conducted, scientists had 14 participants read a sentence word for word and think about the overall meaning of each sentence while in a functional MRI (fMRI) scanner. They took the fMRI representations of the sentences and grouped together sentences that repeated the words of interest, using the rest of the words in the sentence as contextual information which they eventually found could be filtered out. In summing together these functions and associating them with the attributes previously discussed, they were able to predict neural activations for specific words. The predicted activation for the verb “play” is shown in Figure 1.

Figure 1: synthesized and predicted neural activation for the word “play” across neural modalities

This is the first study of its kind to apply a 65-dimensional model of word meaning that captures the range of the human experience, including abstract concepts, as a basis for predicting neural activation patterns of single words. Many of the studies that preceded it had ambiguous neurobiological interpretations, which left specific brain processes that could potentially be involved in semantic processing undefined inn their models. This experiment and model encompass and prioritizes neurobiology, allowing it to be accurate in defining/combining more aspects and complexities of the brain when processing semantic content. Though the neural semantic code/representation may not include all the attributes used in the model, this model does provide a way to predict the parts of the brain that contribute to semantic representation.

So, yes, you just might be able to look at an fMRI scan of a person thinking of a word and its meaning and be able to read their mind!

References:

Andrew James Anderson, Jeffrey R. Binder, Leonardo Fernandino, Colin J. Humphries, Lisa L. Conant, Mario Aguilar, Xixi Wang, Donias Doko, Rajeev D. S. Raizada, Predicting Neural Activity Patterns Associated with Sentences Using a Neurobiologically Motivated Model of Semantic Representation, Cerebral Cortex, Volume 27, Issue 9, September 2017, Pages 4379–4395, https://doi.org/10.1093/cercor/bhw240

Binder JR, Conant LL, Humphries CJ, Fernandino L, Simons S, Aguilar M, Desai R. 2016. Toward a brain-based componential semantic representation. Cogn Neuropsychol. http://dx. doi.org/10.1080/02643294.2016.1147426.

Mitchell TM, Shinkareva SV, Carlson A, Chang K-M, Malave VL, Mason RA, Just MA. 2008. Predicting human brain activity associated with the meaning of nouns. Science . 320:1191–1195.