The memory processes underpinning complex communication: the association of non-adjacent words
Lead Research Organisation:
University of York
Department Name: Language and Linguistic Science
Abstract
Comprehension requires us to associate words: consider utterance (1), in which the comprehender must link 'film' with 'watched'. Such grammatical association-between words separated from each other by other words-is central to the combinatory and expressive power of language, and underpins humans' capacity for complex communication.
1. What film do you reckon Jo watched last night?
How do we associate the noun 'film' and the verb 'watched'? One theory holds that we maintain information about 'film' actively in working memory to guide the search for a suitable partnering verb. Another suggests we retrieve 'film' from declarative memory as a suitable object after processing 'watched'. Evidence is found for both word 'maintenance' and 'retrieval', with neither ultimately conclusive.
Wagers and Phillips (2014) advanced the maintenance/retrieval debate by suggesting that some word 'features'-information constituting words, such as word class (e.g. noun, verb), gender (masculine/feminine), animacy ((in)animate), or number (singular/plural)-are maintained while others are retrieved.
Under this 'hybrid' model, word class information about 'film' is maintained in working memory and used by the parser to locate its canonical comprehension position after 'watched' (c.f. Jo watched a film). Semantic information, however, decays after initial processing and must be retrieved from declarative memory when the parser encounters 'watched'. The features retrieved are reintegrated with the features maintained, allowing comprehension. This account combines the low comprehension error rate of maintenance models with the reduced processing demands on working memory of retrieval models.
A series of self-paced reading studies by Ness and Meltzer-Asscher (2017;2019) found that a noun's grammatical animacy (masculine vs feminine) is a feature held in working memory during the parsing of incoming linguistic material. In my MA thesis (York, 2020), I adapted and extended Ness and Meltzer-Asscher's experimental paradigm, finding no evidence that a noun's number (singular vs plural) is a feature maintained in working memory, and, by extension, is a feature that undergoes retrieval.
This PhD will continue efforts to establish which features of words are maintained in working memory to guide the establishment of relationships between words. When word information is 'dropped' by the parser, this work will chart the time-course of this feature decay. I will achieve this by replicating the kinds of reading studies conducted by Ness and Meltzer-Asscher and in my MA thesis with manipulations of word information (gender, case, word class etc.) and variations in sentence structure and length. This work will build a comprehensive and invaluable understanding of the mental parser's hierarchy of application regarding the use of word features to guide relationship establishment between words.
With differences in grammar, it is likely that speakers of different languages rely on maintenance of different word features to guide associations. I will conduct research on Russian speakers, in addition to English, to build a cross-linguistic picture of word association.
The benefits of such work are substantial. It will add clarity to the longstanding debate around information maintenance and retrieval in sentence processing. Equally importantly, a better understanding of the applications of word features will bring new dimensions to an array of related sentence processing fields. It will also benefit linguistic inquiry writ large: by investigating the parser's ability to associate distant words, this research will elucidate the memory processes ultimately underpinning all complex linguistic communication.
This research requires collection of psycholinguistic experimental data. I will use two experimental techniques: self-paced reading (conducted over the Internet) and eye-tracking while reading using the eye-tracker in York's Linguistics lab.
1. What film do you reckon Jo watched last night?
How do we associate the noun 'film' and the verb 'watched'? One theory holds that we maintain information about 'film' actively in working memory to guide the search for a suitable partnering verb. Another suggests we retrieve 'film' from declarative memory as a suitable object after processing 'watched'. Evidence is found for both word 'maintenance' and 'retrieval', with neither ultimately conclusive.
Wagers and Phillips (2014) advanced the maintenance/retrieval debate by suggesting that some word 'features'-information constituting words, such as word class (e.g. noun, verb), gender (masculine/feminine), animacy ((in)animate), or number (singular/plural)-are maintained while others are retrieved.
Under this 'hybrid' model, word class information about 'film' is maintained in working memory and used by the parser to locate its canonical comprehension position after 'watched' (c.f. Jo watched a film). Semantic information, however, decays after initial processing and must be retrieved from declarative memory when the parser encounters 'watched'. The features retrieved are reintegrated with the features maintained, allowing comprehension. This account combines the low comprehension error rate of maintenance models with the reduced processing demands on working memory of retrieval models.
A series of self-paced reading studies by Ness and Meltzer-Asscher (2017;2019) found that a noun's grammatical animacy (masculine vs feminine) is a feature held in working memory during the parsing of incoming linguistic material. In my MA thesis (York, 2020), I adapted and extended Ness and Meltzer-Asscher's experimental paradigm, finding no evidence that a noun's number (singular vs plural) is a feature maintained in working memory, and, by extension, is a feature that undergoes retrieval.
This PhD will continue efforts to establish which features of words are maintained in working memory to guide the establishment of relationships between words. When word information is 'dropped' by the parser, this work will chart the time-course of this feature decay. I will achieve this by replicating the kinds of reading studies conducted by Ness and Meltzer-Asscher and in my MA thesis with manipulations of word information (gender, case, word class etc.) and variations in sentence structure and length. This work will build a comprehensive and invaluable understanding of the mental parser's hierarchy of application regarding the use of word features to guide relationship establishment between words.
With differences in grammar, it is likely that speakers of different languages rely on maintenance of different word features to guide associations. I will conduct research on Russian speakers, in addition to English, to build a cross-linguistic picture of word association.
The benefits of such work are substantial. It will add clarity to the longstanding debate around information maintenance and retrieval in sentence processing. Equally importantly, a better understanding of the applications of word features will bring new dimensions to an array of related sentence processing fields. It will also benefit linguistic inquiry writ large: by investigating the parser's ability to associate distant words, this research will elucidate the memory processes ultimately underpinning all complex linguistic communication.
This research requires collection of psycholinguistic experimental data. I will use two experimental techniques: self-paced reading (conducted over the Internet) and eye-tracking while reading using the eye-tracker in York's Linguistics lab.
Organisations
People |
ORCID iD |
| James Abbott (Student) |