Whether it’s news headlines or WhatsApp messages, modern people are bombarded with short bits of text. Now researchers say they’ve picked out at a glance how we get their nucleus.
Prof Liina Pylkkanen, co-author of the study from New York University, said most theories of language processing assume words are understood one by one, in sequence, before being combined to give the meaning of the whole sentence .
“From this perspective, language processing really shouldn’t work at a glance, since there just isn’t enough time for all the sequential processing of words and their combination into a larger representation,” she said.
However, the research offers fresh insights, revealing that we can detect certain sentence structures in as little as 125 milliseconds (ms) – a time frame similar to the blink of an eye.
Pylkkanen said: “We still don’t know exactly how this ultra-fast structure detection is possible, but the general hypothesis is that when something you perceive matches really well with what you know about – in this case we’re talking about knowledge of the grammar – this knowledge of top down can help you identify the stimulus very quickly.
“So just as your own car is quickly identifiable in a parking lot, certain language structures are quickly identifiable and can then give rise to a rapid effect of syntax in the brain.”
The team says the findings point to parallels with the way we perceive visual scenes, with Pylkanen noting that the results could have practical uses for the designers of digital media, as well as advertisers and designers of road signs.
Write in the journal Science AdvancesPylkkanen and colleagues report how they used a non-invasive scanning device to measure the brain activity of 36 participants.
Each participant received an initial three-word sentence flashed for 300 ms, followed by a second sentence that was either identical or differed by one word. Participants were asked to indicate whether the sentences matched, with the experiment repeated using different starting sentences.
The results show that participants made faster and more accurate judgments about whether the sentences matched when they contained a subject, verb and object – such as “nurses clean wounds” – than when they contained a list of nouns such as “hearts lung liver”.
What’s more, participants’ brain activity increased rapidly in response to a beginning sentence with a subject, verb and object, with activity detected in the left-middle temporal cortex within 130 ms – about 50 ms faster than for a list proper nouns.
A similar rapid response was observed when starting sentences with a subject, verb and object were changed so that they contained an agreement error – for example “nurses clean wounds” – or became improbable, for example “nurses clean wounds”.
But the effect disappeared when the sentences got less typical structures, for example “nurses clean wounds” or “nurses clean wounds”. Pylkkanen said this suggests that these sequences do not activate the same sentence recognition system.
While the authors note that they focused on English, and add that rapid at-a-glance comprehension may rely on other features in different languages, they say the study offers new insights.
“The earliest stage of a glance comprehension appears to be more structure than meaning driven,” they write.
