This content is archived, it could be outdated.
Large overlap in grammatical processing for speaking and comprehending
August 31, 2011In ancient times, the Morning Star and the Evening Star were believed to be two different celestial bodies. It was Parmenides of Elea (around 500 B.C.) who discovered that they in fact were the same planet. A similar type of discovery was recently made in the area of natural language processing. According to mainstream psycholinguistics, the cognitive processes that enable speakers to produce grammatically well-formed sentences are largely different from those underlying the grammatical analysis of perceived sentences during language comprehension. But this view has now been challenged.
Large overlap in processing mechanisms
MPI researcher Gerard Kempen and his colleagues Nomi Olsthoorn and Simone Sprenger have discovered that the mechanism responsible for constructing the grammatical form of sentences during speaking and writing, called grammatical encoding, overlaps to a large extent with the mechanism that analyses the structure of perceived sentences as part of the comprehension process, grammatical decoding. Using a new method called 'grammatical multitasking', language users had to carry out a decoding task quasi-simultaneously with an encoding task. While reading a printed input sentence, they had to produce a spoken output sentence that had the same meaning as the input sentence, but a different grammatical form (i.e., a syntactic paraphrase).
The results indicate that in the decoding process people spontaneously treated the syntactic structure of the paraphrase as if it was the structure of the input sentence. Blatant syntax errors in the sentence being read were systematically overlooked if the erroneous words fitted with the structure encoded for the paraphrase. Kempen explains: "Apparently, the grammatical decoding and encoding processes share a common workspace for syntactic structure formation and can hardly avoid operating on one another's products. This means that the cognitive mechanisms for grammatical encoding and grammatical decoding are overlapping to a large extent: a finding that has widespread consequences for psycholinguistic theories of grammatical encoding and decoding."
Bronze by Juan Munoz, Spain
No two things at the same time
"All these phenomena seem to suggest that language users are capable of language production and language comprehension at physically the same point in time (true simultaneity), supporting our intuitive belief that grammatical encoding and grammatical decoding are subserved by separate cognitive mechanisms: One mechanism cannot do two things at the same time. However, if our cognitive system hosts only one 'grammatical coder', we must assume that the apparent simultaneity is superficial and that, in reality, the coder is performing encoding and decoding tasks alternately." In their paper, the authors argue that there is no hard evidence for true simultaneity of grammatical encoding and decoding.
Minimal overlap in time
Kempen also refers to a publication of MPI researchers Tanya Stivers, Nick Enfield, Stephen Levinson et al. (PNAS 2009), in which they show that the temporal overlap between dialogue turns is minimal and is restricted to intervals where the speaker who interrupts could have anticipated and predicted the grammatical structure of the other speaker’s contribution and thus only needs to encode his or her own contribution (see also our news archive).
Link to the publication.