For people with a language or cognitive impairment, the difference between "know" and "think" can be the difference between understanding and not understanding a sentence.

Zimmerer, V.C., Varley, R.A., Deamer, F., Hinzen, W. (2019). Factive and counterfactive interpretation of embedded clauses in aphasia and its relationship with lexical syntactic and general cognitive capacities. Journal of Neurolinguistics, 49, 29-44.

Çokal, D., Zimmerer, V., Varley, R., Watson, S., Hinzen W. (2019). Comprehension of embedded clauses in schizophrenia with and without formal thought disorder. The Journal of Nervous and Mental Disease, 207(5), 384-392.

You and your friend get ready for a day trip to the coast. You see your friend pack light clothing and sunscreen. “I know that it is sunny over there,” she says. You check the weather forecast. It is all clouds and rain. “No,” you say. “You just think that it is sunny.”

We were interested in processing of factive and non-factive verbs, which allow us to express the truth value of a proposition. Typically, factive verbs like “know” imply that the statement in the embedded sentence (e.g. “it is sunny”) is true. Non-factive verbs like “think” make no such implication. “I think it is sunny” can be true if it is raining, while under the same weather, “I know it is sunny” cannot. Modified, like “You just think it is sunny” in the example above, or using a counterfactive verb like in “John misremembered that it was sunny”, the implication is that the embedded sentence is not true. We are a cognitively flexible species that effortlessly navigate between the real, the uncertain and the imagined, so these linguistic constructions are important tools. We use them all the time: other examples for factive verbs are “see”, “reveal”, “show”, and “prove”; other examples for non-factives are “suggest”, “doubt”, “believe”, and “imply”.

Some people with cognitive disorders, such as schizophrenia, unreasonably consider the false to be true and the true to be false (delusions or paranoia). In people with a language impairment, for example caused by brain damage (aphasia), language use is restricted, and in some cases, the ability to produce or understand embedded sentences is affected. Our “Language and Mental Health” project was interested in language and thought in schizophrenia as well as aphasia, and within this context, factivity is an interesting phenomenon. I also believe that the project generated some ideas for clinical language testing (see final paragraph).

We know much about it from research on child language acquisition. The general consensus seems to be that children have difficulties understanding factivity until they are at least ca. 4 years old. Younger children have particular difficulties with non-factive verbs, e.g. “think”. Learning to interpret these sentences comes with a number of cognitive demands, and they seem to be larger when processing non-factives.

(1) The individual needs to understand what verbs like “know” and “think” even mean.

(2) The individual needs to understand the grammatical structure of sentence embedding. By itself, a sentence like “it is sunny” would likely be presupposed to be true, embedding it in a clause like “Peter thinks that [it is sunny].” changes the interpretation.

(3) The individual needs to simultaneously maintain contradicting thoughts or scenarios, for example that it is raining or not raining.

(4) Many of these verbs concern mental states, so the individual needs to understand the mental states of others, particularly that someone else’s perspective can differ from one’s own (theory of mind).

(5) Finally, there are more pragmatic factors. For example, someone may say “I think it’s raining” despite knowing that it’s raining, for example out of politeness.

Sentence picture matching task. Picture array for trial sentences  He knows that the bathroom is dirty  and  He thinks that the bathroom is clean .

Sentence picture matching task. Picture array for trial sentences He knows that the bathroom is dirty and He thinks that the bathroom is clean.

Project lead Wolfram Hinzen had worked on theories of factivity before. My role was to lead on the design of the experiment, collect data from people with aphasia and the control group, come up with a data analysis plan and write up the aphasia paper. We decided to design a sentence-picture matching test. The method requires the participant to listen to a sentence and then select from a number of pictures the one that matches it. Look at the three pictures. Which matches the sentence “The man knows that the bathroom is dirty”? Which matches “The man thinks that the bathroom is clean”? (Yes, I drew the pictures myself because we didn’t have the budget for an artist.)

We had two types of trials. In factive interpretation trials, selecting the correct picture required the participant to interpret the embedded sentence as true (e.g. “The man knows that the bathroom is dirty”, “It is clear to the woman that the stall is occupied.”). In counterfactive interpretation trials, the participant had to interpret the embedded sentence as false (e.g. “The man thinks that the bathroom is clean”, “It only seems to the man that the pool is safe”). We only presented non-factive verbs in trials where the correct picture showed the embedded proposition to be false.

Healthy controls had no problems with either condition. People with aphasia (mixed group) performed about as well as controls in factive interpretation trials. On counterfactive trials, they were significantly (and substantially) less accurate. In participants with schizophrenia, it mattered whether they had a thought disorder. Participants with schizophrenia without thought disorder performed about as well as controls. People with thought disorder were significantly less accurate in both conditions, worst of all in the counterfactive trials.

So for participants with aphasia or thought disorder, we saw that counterfactive interpretations were harder, which is what colleagues have reported in studies with young kids. To find out which aspect of the clinical groups’ cognitive profile drove performance, we correlated their accuracy in the experiment with other measures of language and cognition. People with aphasia who struggled with the sentences were those that performed poorly in other language tasks. There was also one interesting correlation between non-verbal intelligence and performance in counterfactive interpretation trials, suggesting that there is an overlap in cognitive resources required for both tasks. My assumption is that both require maintaining contradicting scenarios/solutions.

People with thought disorder underperform in a wider range of tests. Participants who had difficulties with our experiment had difficulties in multiple other tests of language, reasoning, executive function and non-verbal semantics. Results support the notion of formal thought disorder as a very general cognitive impairment.

I am happy with the two papers. They are the first to look at these important language phenomena in aphasia and schizophrenia. At the same time, I am quite aware that there is more about factivity we did not investigate than there is that we were able to. Future research can look at the rich language acquisition literature for motivation.

I think the work can be very relevant for language screening. Lately I have been thinking about the language constructions that are the “canaries in the coalmine”: Sentence types that are so demanding that they are particularly vulnerable to neurological change. They can therefore help reveal diseases such as dementia, or disorders such as depression or psychosis. I would count non- or counterfactives among these “canary constructions”, among center-embedded clauses and object clefts.

Click here for the testing materials.