Listeners Predict What’s Coming Next


by Faith Hickman Brynie

April 24, 2009

“After the tragic plane crash, they had to decide where to bury the survivors.”

What was your response to that statement? Did the incorrect word at the end of the sentence bother you? Probably not. Many people quickly understand that survivors should have been victims, and they move on without missing a beat. They know what the sentence means despite the error.

According to neurolinguists, this seamless correction is just one of many examples of the brain’s “messy” manner of understanding language, of how we actively anticipate and assume meaning despite the literal arrangement of words and text.

Words ≠ Meaning

“Language comprehension looks pretty easy,” says Jos J.A. van Berkum, who researches how the brain processes language at the Max Planck Institute for Psycholinguistics and the Donders Centre for Cognitive Neuroimaging in the Netherlands. “Although what you’re taking in is a bunch of letters and sounds, what you really perceive—if all goes well—is meaning.”

Meaning intrigues van Berkum, who published a review of his research (pdf) in the December 2008 issue of Current Directions in Psychological Science. His studies show that the brain draws on more than words, grammar and syntax when interpreting written or spoken language. For instance, research shows that readers and listeners anticipate meaning, making the earliest possible interpretation of a sentence. Thus, we perceive the intent of the plane-crash sentence milliseconds before we even encounter the wrong word.

“When interpreting language, people don’t just slavishly follow the syntax. The brain takes shortcuts, using its statistical knowledge of what we tend to talk about [and how] to make a quick-and-dirty analysis,” van Berkum says.

Unexpected words

Van Berkum suspects that people guess how an unfolding sentence will continue. To test that hypothesis, he has used event-related potentials (ERPs) to measure the brain’s response to a variety of sentences. ERPs are averages calculated from electroencephalogram (EEG) data. They reveal changes in the voltage of brain waves elicited by a stimulus such as a tone or a word. “Because of their speed,” van Berkum says, “ERPs allow us to trace the very rapid processes involved in language comprehension.”

Suppose our example read, “After the tragic plane crash, they planned a celebration.” Listeners and readers need only a fraction of a second to discover that a word does not fit the preceding text. In the brain, the contextually inappropriate word “celebration” elicits an increase in the size of the N400 component of the ERP signal. That N400 effect, the classic response to incoherent language, occurs very quickly, developing 150–250 milliseconds after we hear a confusing word and peaking around 400 milliseconds afterward (thus the 400 in N400). With spoken language, this response is so fast that it occurs even before the end of the word is uttered. In this example, the first syllable “cel” alone can induce the N400 effect in the brain of the listener.

The N400 effect does not appear, however, in examples like our first one, when the meaning is clear despite an inappropriate word. In such cases, the brain responds with a later shift in ERP patterns, occurring 500–600 milliseconds after the word is spoken. The shift suggests that readers and listeners ultimately detect the anomaly but adjust their comprehension to override it.

Inferring the referent

Did you notice something else grammatically wrong with both sentences about the plane crash? Probably not. Most people read past the they without a pause, although nothing tells us whether those doing the burying or celebrating were government officials or family members or the actual crash survivors. You perceive the meaning without knowing the pronoun’s referent. To the language-interpreting brain, it doesn’t matter that they are faceless, nameless people.

But van Berkum’s studies reveal that referents can be important in other cases. When pronouns are ambiguous, the brain responds with another kind of ERP shift, the Nref effect. This is a sustained negative shift in the ERP signal from the brain’s frontal regions.

Consider these sentences:

“David shot at Linda as he jumped over the fence.”

“David shot at John as he jumped over the fence.”

In the first sentence, the meaning is clear, because Linda is a female name, and a male was jumping. Therefore, David must have been shooting and jumping simultaneously. In the second sentence, however, the pronoun he might refer to either David or John. The brain responds to the ambiguity with the Nref effect.

While the N400 effect in response to incoherent language is largest in networks in the back of the brain, the Nref effect is greatest at the front. “We can infer [from this difference] that different aspects of interpretation are handled by at least partly different networks in the brain,” van Berkum says. Recent studies using functional magnetic resonance imaging (fMRI) support that distinction, he notes.

Yet another ERP shift appears when the meaning of a word leads a listener or reader to expect a certain referent. Consider this example: ‘‘David praised Linda because . . . .’’ Unconsciously, people expect the next word to be “she.” The brain anticipates Linda, not David, as the referent. If the next word is “he,” the brain responds with a P600 effect, an ERP shift typically seen in response to a grammatical error. There is no grammatical error, but the expectation for a female pronoun is so strong that the brain responds as if an error had been made.

“People don’t just stick to what is given. They anticipate what might be said, about whom,” van Berkum says. “We are unconsciously yet continuously predicting what might be talked about next.”

The opportunistic brain

Psycholinguists might like language comprehension to occur in neatly ordered steps, van Berkum says, but his research reveals that the linguistic brain is “messy and opportunistic.” The brain processes any partial cue to meaning in a matter of milliseconds.

Because other processes in the brain may prove just as messy, van Berkum’s work is drawing attention from researchers in diverse fields outside linguistics. “[His work] has far-reaching implications for understanding the neural basis of numerous neuropsychiatric and developmental disorders,” says Gina Kuperberg, an associate professor of psychology at Tufts University and a psychiatrist at Massachusetts General Hospital.

 “Van Berkum’s research program is tremendously significant,” she says. “His team is using cutting-edge techniques of cognitive neuroscience—event-related potentials, functional magnetic resonance imaging and magnetoencephalography—to explore the most fundamental questions about the brain basis of human language, attitude and thought.”