Donate | Print Page | Contact Us | Report Abuse | Sign In | Register
News & Information: Featured Content

When the cat barks and the guitar has a bow

Friday, July 17, 2015   (0 Comments)
Share |
Stephan Lewandowsky

When the cat barks and the guitar has a bow: Neurocognitive signatures of processing perplexing text


“A mouse was looking for something to eat while a bigger animal was waiting to hunt it.” What’s your best guess about which animal was lurking over the unfortunate mouse’s shoulder? I suspect you would be surprised if the next sentence in the story were “the bigger animal barked loudly.” Most of us would expect that bigger animal to meow rather than bark.

Whenever we process text, we constantly “seek to construct an integrated, coherent, and accurate mental representation of the state of affairs described by the text”. To do so, we need to go beyond the literal text in front of us, and we need to resort to general world knowledge to generate inferences about what might be happening next. So, when an unspecified larger animal is watching a mouse, real world knowledge tells us that it most likely is a cat.

If the cat suddenly barks, we are faced with a dilemma because the situation we imagined turns out to have been false. We must revise and update our understanding of the situation—in our heads, the imagined cat has to turn into a dog. This type of updating of our unfolding understanding of a narrative is known as “inferential revision”: It is inferential because we inferred the presence of a cat based on prior knowledge without being told there is a cat, and it is a revision because our initial inference was wrong.

recent article in the Psychonomic Society’s journal Memory and Cognitionexamined the role of working memory in inferential revision. Researcher Ana Pérez and colleagues were interested in the extent to which a person’s ability to revise inferences during narrative comprehension was associated with their working memory capacity. Some prior research suggested that people with lower working memory capacity had greater difficulty with inhibiting of the outdated content (“cat”) that is required during revision to the new information (“dog”) than people with higher working memory capacity.

Pérez and colleagues examined event-related brain potential (ERP) markers in addition to conventional behavioral measures. ERPs are a long-standing tool in experimental cognitive psychology and they offer the distinct advantage of providing a measure of cognitive activity without requiring a behavioral response. (If you are interested in ERPs, in an earlier post I examined how ERPs can tell us something about mindfulness.)

ERP measures were particularly useful in the present case because the brain’s signal in response to a stimulus can be broken into several distinct components, some of which are reliable fingerprints of specific cognitive activity. In particular, a component known as N400 (“N” for negative, and “400” for the delay in milliseconds after stimulus presentation) appears to be an index of the ease with which the meaning of a word can be integrated into the current situation model.

Participants in the study by Pérez and colleagues read short texts whose first few sentences set the stage to permit some fairly strong inferences. For example, consider the following 2 sentences:

(1) “Dan was a gypsy who had played flamenco since childhood.”

(2) “Today, he is giving a recital of his favorite works.”

Now suppose you start reading the third sentence:

(3) “His instrument is made of maple wood …”

What is the instrument? And what do you expect to read in place of the “…” in the remainder of the sentence?

Chances are that the cues in sentence (1), namely “gypsy” and “flamenco”, generated an expectation that the instrument would be a guitar. So, if the full version of sentence (3) were:

(3a) “His instrument is made of maple wood, with a beautiful curved body,”

then this would not require a revision of your inference. By contrast, if sentence (3) were:

(3b) “His instrument is made of maple wood, with a matching bow”,

Then this would require a revision of your inference because in the same way that cats do not bark, guitars don’t come with a bow.

In addition to ERPs, Pérez and colleagues measured reading times to the critical sentence, which was presented in three different variants: A no-revise version (3a above), a revise version (3b) and a neutral version that did not rely or involve any of the inferences triggered by the first sentence (not shown above).

The results were pretty straightforward. At a behavioral level, it was found that the revise version took nearly 300 milliseconds longer to read than the neutral or no-revise versions. This confirms that the violation of expectations generated by the “bow” tied up people’s cognitive processes as they had to update their inferences and revise their mental model of the narrative. This slowing was unrelated to working memory capacity: the time cost due to the required revision was identical for people with low and high working memory capacities. (Working memory capacity had been established by a pretest, and participants were selected for their particularly high or low capacity.)

The ERP analysis focused on a further sentence (4) that was identical for all conditions and that contained a disambiguating word:

(4) “The public was delighted to hear Dan playing the violin.”

By the time readers encounter the word “violin”, processing should vary with the extent of inferential revision readers were able to achieve during that extra 300 milliseconds of processing during the presentation of the revision sentence (3b). Did the “bow” knock out the “guitar” and replace it with “violin”? If so, then this should be reflected in the N400 component of the ERP: Recall that this component indicates the ease with which people can integrate a word into their unfolding model of a narrative.

The results are shown in the figure below:


The figure shows that the N400 component was much reduced when the word “violin” was preceded by a sentence that pointed towards that instrument (revise condition). This indicates that the word “violin” no longer presented a challenge to readers who had already revised their mental model with the revision sentence.

The size of this effect differed with working memory capacity: The high memory group showed larger negativity in the neutral and no revise conditions compared to the revise condition, suggesting that they experienced difficulties with integrating the word “violin” into their situation model when their current model (based on the previous sentences) was still focused on the idea of “guitar”. Conversely, the high memory group found the word “violin” very easy to process after the revise condition.

Those differences between conditions were attenuated for the low memory group: They did not struggle as much when the word “violin” popped out of nowhere, and they did not find the “violin” terribly easy when the situation had already been facilitated by the revise sentence. This pattern suggests that low working memory readers seem to have somewhat greater difficulty integrating new information into the situation model.

Intriguingly, those differences between the two working memory groups did not show up in the behavioural data: Recall that reading times did not differ between the memory groups. It follows that ERPs can offer a window into cognitive processing that purely behavioral measures sometimes fail to provide.

2424 American Lane • Madison, WI 53704-3102 USA
Phone: +1 608-441-1070 • Fax: +1 608-443-2474 • Email: info@psychonomic.org

Use of Articles
Legal Notice