An Experimental Investigation of Syntactic Priming and the Lexical Boost in Language Production

Lead Research Organisation: University of Dundee
Department Name: Psychology

Abstract

An essential part of producing a sentence involves the construction of its syntax. How this is done is a fundamental question in language production research. People often have choices how to express a message. For example, they can say either (1a) or (1b); both sentences express a similar meaning, but their syntactic structures are different.

1a. The driver showed the overall to the mechanic.
1b. The driver showed the mechanic the overall.

Research has shown that people tend to repeat recently spoken or heard structures: they tend to produce (1a) more often after hearing (2a) than (2b), and (1b) more often after (2b) than (2a), a finding termed "structural priming".

2a. The doctor gave the syringe to the nurse.
2b. The doctor gave the nurse the syringe.

Structural priming can inform us about the structural representations that people access during language production. One critical finding is that structural priming is stronger when the verb in the sentence to be produced is the same as in the previous sentence: the "lexical boost effect". For example, (1a) is more strongly primed by "The doctor showed the syringe to the nurse" than by (2a). There is currently much debate about the origin of the lexical boost and the precise circumstances under which it occurs.
Based on pilot studies, we suggest a hierarchical account, which assumes that hierarchical relations play an important role in the lexical boost: a structural representation (e.g., "the syringe to the nurse/the nurse the syringe") is associated not just with its own head (e.g., "gave"), but with any head of a structure in which the representation is embedded. This account makes novel predictions about the lexical boost. For example, despite the fact that "hesitated" in (3a) is not the head of "the syringe to the nurse/the nurse the syringe", priming should be larger when the prime (3a) and target to be produced (3b) contain the same verb ("hesitated") than when the verbs are different ("decided" and "hesitated"), because these verbs are the head of the larger structure "to show the overall to the mechanic/the mechanic the overall".

3a. The driver hesitated/decided to show the overall to the mechanic/the mechanic the overall. (prime)
3b. The doctor hesitated to give ... (target)

In contrast, repetition of "doctor" in (4) should not result in a lexical boost, because it is not the head of a structure that contains "the syringe to the nurse/the nurse the syringe".

4a. The driver/doctor showed the overall to the mechanic/the mechanic the overall. (prime)
4b. The doctor hesitated to give ... (target)

These predictions are different from existing accounts. The experiments will investigate which account makes the correct assumptions about how syntactic structures are mentally represented. We will ask approximately 40 participants to read aloud primes such as (3a, 4a) and then to provide a completion to targets such as (3b, 4b). Around 40 prime-target pairs will be tested.
A series of experiments will address several major questions. In addition to the question of whether the lexical boost occurs when the repeated word is the head of a structure in which the primed structure is embedded, we will ask whether the distance between the repeated word and the primed structure matters. Other questions will concern the duration of the boost and whether it is due to explicit memory of the repeated word from the prime. This is important for understanding language production and will contribute to language production models. Research so far has not provided clear answers to the question of what structural information is stored and with which words. Our study will lead to a better understanding of how people produce syntax and contribute to a more precise model of structural priming.

Planned Impact

Speakers often repeat the syntax of sentences, a phenomenon called "structural priming". When experimentally elicited, structural priming can inform us about the processes underlying sentence production. In the proposed project, we will focus on the "lexical boost effect", the finding that structural priming is stronger when specific words in the sentences are also repeated. This will inform us how syntactic structures are organised around specific words and allow us to test the main theories of structural priming. The ultimate aim of the project is to develop a more precise model of how speakers mentally represent syntax. The results will lead to a refinement of current models of language production. As explained in the "academic beneficiaries" section, this will benefit a wide variety of language research, including language production research, computational modelling, linguistic theory and memory research.

Non-academic impact
Although our project will primarily have academic beneficiaries, our understanding of how and when language users repeat sentence structures can be useful for the development of automatic language generation and comprehension systems such as in car navigation systems, automatized text summarising, automatic translation and augmentative alternative communication systems for people who cannot speak. First, if these systems repeat structure in a similar way as humans do, they will produce more naturalistic language. Second, repeating structures from a previous utterance could facilitate natural language generation, because the system does not need to generate structure from scratch. Natural language generation systems have the potential to increase both economic competitiveness and the effectiveness of public services by facilitating communication and information transfer. When used for augmentative alternative communication, they enhance people's well-being.

Professional skills
The project will have a direct impact on the skills of the staff involved. The postdoctoral researcher will develop several professional skills that are transferrable to other employment sectors. They include working in an international team (Dundee, Ghent, Umeå), coordinating the collaboration between the universities involved, giving presentations and writing articles, computer and statistical analysis skills, and publicising the work to both the scientific community and the wider public. The division of Psychology in Dundee has a research apprenticeship scheme that allows undergraduate students to assist in a research project. As part of this scheme, we will recruit students to help us on the project. This will give the postdoctoral researcher experience in supervision while the undergraduate students get experience in material construction, running experiments and data analysis.

Outreach
In the past, the PI has had press coverage in newspapers and the local radio. We will liaise with the Press Office at the University of Dundee to attract coverage in the media. In order to publicise our work to the wider public, we also aim to present at "Café Science", which is a series of scientific talks to the public in Dundee, and contribute to exhibitions at the Science Festival organised by the Dundee Science Centre. A website interfacing with Twitter and Facebook will be used to publicise our research to the wider public. The aim of these outreach activities is to inform the public interested in understanding how language is shaped by the mind and how current developments in language science inform us about this.

Publications

10 25 50
publication icon
Carminati M (2019) An investigation into the lexical boost with nonhead nouns in Journal of Memory and Language

publication icon
Van Gompel RPG. (2017) Structural priming in bilinguals in Bilingualism: Language and Cognition

 
Description The aim of the project is to investigate the lexical boost in structural priming, which informs us about the structural representations that people use during language production. The lexical boost refers to the finding that structural priming (the tendency to repeat the same structure across utterances) is stronger when particular words in the sentences are also repeated. It informs us how structural representations are associated with words in the sentence.

At the approximate half-way point of the project, we have two main findings. First, structures do not appear to be associated with nouns in the sentence that are not the head of those structures, because structural priming is no larger when such nouns are repeated between sentences than when they are not. For example, following prime sentence (1a), participants tend to produce more target completions of (2) with double object structures (e.g., the man the book), whereas following (1b) they tend to produce more prepositional object prime completions (e.g., the book to the man), but critically, this structural priming effect is no stronger when "boy" is repeated between prime and target than when it is not. Thus, there is no lexical boost from the repetition of the subject of the sentence.

1a. The boy will hand the celebrity the present. (double object structure)
1b. The boy will hand the present to the celebrity. (prepositional object structure)

2. The boy/lawyer will show

We have conducted a series of experiments that corroborate this finding using different priming methods and testing the repetition of different nouns in the sentence (e.g., boy, celebrity, present). This absence of a lexical boost with the nouns in the sentence contrasts with the boost that is observed with the repetition of the syntactic head verb (e.g., hand/show), suggesting that the double object/prepositional object structure is stored with its head verb, but not with the nouns in the sentence.

Our second main finding is that the lexical boost also does not occur with other verbs in the sentence that are not the syntactic head of the primed structure. In a series of experiments, we investigated whether the repetition of a matrix verb (hesitated/decided) between prime (3) and target (4) boosted structural priming.

3a. The boy hesitated to hand the celebrity the present. (double object structure)
3b. The boy hesitated to hand the present to the celebrity. (prepositional object structure)

4. The lawyer hesitated/decided to show

We found no clear evidence that matrix verb repetition boosted priming, suggesting that the primed structures are not associated with verbs in the sentence that are not their syntactic licensing head.

Together, the experiments lead to the conclusion that syntactic structures are associated with their syntactic heads, but not with any other words in the sentence. This supports the residual activation model by Pickering and Branigan (1998), which makes exactly this assumption. In contrast, our results provide evidence against other structural priming models such as by Chang, Dell and Bock (2006) and Reitter, Keller and Moore (2011), which assume that a lexical boost should occur with the repetition of any content word in the sentence.
Exploitation Route Given that we find different results with different priming methods, other researchers may investigate how the choice of method affects other structural priming findings that have been reported in the literature.
Sectors Education,Electronics

URL https://sites.dundee.ac.uk/langprolab/
 
Description Collaboration between Roger van Gompel (University of Dundee) and Leila Kantola (University of Umea) 
Organisation Umea University
Country Sweden 
Sector Academic/University 
PI Contribution My contribution to the long-standing research collaboration with Leila Kantola from the University of Umea has been in terms of experimental ideas, design and analysis of a series of psycholinguistic studies on reference and sentence production.
Collaborator Contribution Leila Kantola's contribution has been in terms of intellectual input and carrying out experimental work.
Impact Wakeford, L.J., Kantola, L., & Van Gompel, R. P. G. (2019). Lexical boost from the subject noun: The influence of task. Poster presented at the 32nd CUNY conference on human sentence processing, Boulder, CO, March 2019. Wakeford, L.J., Kantola, L., & Van Gompel, R. P. G. (2018). Lexical boost from the matrix verb. Poster presented at the 24th AMLaP conference, Berlin, September 2018. Kantola, L., & Van Gompel, R. P. G. (2016). Is anaphoric reference cooperative? Quarterly Journal of Experimental Psychology, 69, 1109-1128. Kantola, L. & Van Gompel, R.P.G. (2011). Between- and within-language priming is the same: Evidence for shared bilingual representations. Memory & Cognition, 39, 276-290.
Start Year 2008
 
Description Collaboration between Roger van Gompel (University of Dundee) and Rob Hartsuiker (University of Ghent) 
Organisation University of Ghent
Country Belgium 
Sector Academic/University 
PI Contribution My contribution to the collaboration with Rob Hartsuiker (University of Ghent) has been in terms of experiment ideas, design and analysis of studies on structural priming.
Collaborator Contribution Rob Hartsuiker 's (University of Ghent) contribution to the collaboration has been in terms of making a laboratory available for conducting experiments.
Impact No outputs yet
Start Year 2019