Overuse of familiar phrases by individuals with Williams syndrome masks differences in language processing.
This work has been fun on so many levels. First, it connects my current research with my very early experiences with language sciences. Williams syndrome was often used as an example to illustrate how language functions as an independent cognitive “module” - or, even better, an “organ” - which doesn’t always behave like other functions of the brain/mind. I was academically raised with that belief, and now it was time to scrutinize it.
Second, the people with whom I worked were fantastic. This paper is based on a secondary data analysis conducted by two then-final-year undergraduate students, Ioana Sederias and Ariane Krakovitch, who I supervised. These were the years of the Covid lockdowns - dark and stressful at least for me - and working with these two smart and fun students was one of the few bright spots during that time.
Finally, the project itself went remarkably smoothly. Our predictions came naturally from previous observations and the data supported them strongly. Some confused head-scratching is fairly typical when one first looks at results, but not here. The data tell a very clear story. And this is the story:
Williams syndrome (WS) is a rare genetic disorder which, just like Down syndrome, comes with a rather specific phenotype, which affects facial features as well as body and cognitive functions. People with WS typically have learning disabilities (they often score very low in IQ tests compared to their age group). Some earlier research, claimed that language is preserved in WS, with individuals producing lots of well-formed sentences fluently. WS was seen as a demonstration of language being a self-contained system, likely evolved independently of other capacities, playing by its own rules, and merely translating rather than assisting thoughts. I guess that if you studied linguistics from 1990 to 2005 and didn’t engage with this subject since, this may still be how you think language works.
Even at that time, there was good evidence that language in WS is far from preserved. Speakers with WS make more grammatical errors, can struggle with understanding complex sentences and don’t always use words in the right context. The appearance of intact language is strong, however. So, what’s really going on?
Previously, I helped with data analysis of a study led by Vesna Stojanovik, from the University of Reading. We found that when people with WS process short, structured sequences of pseudo-words (artificial grammar learning), they were more likely to pick up on repeating chunks of stimuli rather than the rules underlying the sequences. We saw the same behaviour in younger typically developing children, but by the time children reach the same age as those with WS, they generally begin to grasp the rules. This suggests that processing abilities change as part of typical development, but that this change is much delayed or much weaker in WS.
I wondered whether these processing differences would manifest in language production. In language science, we can distinguish between two types of processing: In “analytic” processing, individual words are retrieved and combined, enabling us to understand the contribution of each word to the utterance and to produce sentences and phrases we have never heard before. In “holistic” processing, word combinations, including larger phrases and sentences, are treated as one unit. These formulaic chunks are expressions we have heard very often or are idiomatic in meaning. Many are conversational phrases. With enough of them available, we may get by in familiar situations. But if we rely on formulaic language too much, we can lose our ability to use language precisely and creatively.
If it seems like each blog post on my work touches on this idea, it’s because these concepts characterize most of my research. We know that formulaic language is easier as people with aphasia or dementia use it more often. Now, we are seeing a similar pattern in WS.
Vesna provided us with “Frog Story” picture book descriptions from participants with WS, typically developing controls of the same age, and younger typically developing controls which performed similarly to the WS group on language tests. We compared these groups on a number of variables, focusing on measures of language and of familiarity of expressions, the latter serving as an indicator of formulaic language (or holistic processing). We could identify developmental delays (differences between WS and chronological-age-matched controls) and aspects of language which were more fundamentally different (differences between WS and both control groups).
On many linguistic features, we identified a delay in WS participants compared to their age-matched peers. However, in terms of using familiar expressions, the WS group stood out significantly from both control groups. Their speech showed stronger collocations—meaning they relied on words that are often produced together, i.e. familiar, formulaic chunks of language.
Some researchers previously assumed that language in WS looks preserved because speakers depend heavily familiar chunks. I believe our paper is the first systematic analysis of this phenomenon in WS and produces evidence of a new quality. We explain not only how language processing in WS is different, but also how overreliance on formulaic language can leave the impression of typical language capacity.