<<

Processing relative : Discourse or frequency? Douglas Roland, Carolyn O’Meara, Hongoak Yun and Gail Mauner University at Buffalo droland, ckomeara, hyun3, [email protected]

Introduction Object relative clauses in discourse Results

Object relative clauses are generally more difficult Fox and Thompson (1990) argue for a discourse explanation of the Object relativeNeutral clauses discourse slower topic in neutral context to process than relative clauses prevalence of pronominal object relative clauses in spoken 650 English: Object Relative The lady that visited the banker enjoyed the dinner very Subject ☺ 600 much. • Referents need to be relevant/grounded to the listener at the point

are introduced 550 The lady that the banker visited enjoyed the dinner very • Inanimate subject of sentence less likely to be grounded (since humans much. 500 normally discuss themselves) RT (ms) Difficulty is reduced/eliminated when the • Relative clauses are used to make the modified phrase relevant by 450 embedded NP is pronominal (e.g., Gordon et al., relating back to a given discourse referent 400 2001; Reali & Christiansen, in press; Warren & • Nearly always done via Gibson, 2002) 350 • Results in pronominal relative clauses The lady that visited the banker enjoyed … the banker visited 1st person pronoun reverses SR bias • Usually referring back to a human (since humans normally discuss Object relative slower than subject relative clause themselves) Difference between RT means (neutral condition) The lady that visited me enjoyed the meal. at relative clause and main in neutral context • Combination of inanimate modified NP and animate embedded 40 Relative Clause Combined Region ☺ The lady that visited enjoyed the meal. NP/pronoun results in object relative clauses Main Verb Region 20 nd Examples from Switchboard 2 person pronoun reverses SR bias 0 … the types of music that I don't like are opera and, uh, screaming heavy The consultant that called emphasized the need for -20 metal. additional funding. -40 … some of the Nautilus equipment that I started seeing at the that ☺ The consultant that you called emphasized the need for used to go to was really interesting. -60 additional funding. … and at the same time, the budget sent to Congress has tax and fee Magnitudeof Difference (ms) -80 3rd person pronoun reverses SR bias increases -100

According to the Taylors, the landlord that telephoned them Appropriate discourse context eliminates object Hypothesis Good discourse topic offered a nice apartment. relative difficulty during relative clause If 650 ☺ According to the Taylors, the landlord that they telephoned Object Relative Clause offered a nice apartment. Subject Relative Clause Object relative clauses are typically used for grounding the 600

modified NP in the discourse context 550 Frequency? And 500

Reali and Christiansen (in press) argue for a RT (ms) Subject relative clauses are typically used for other purposes 450 frequency-based account: such as supplying additional information about the modified NP 400 • In general, subject relative clauses are much more Then frequent than object relative clauses 350 • But, object relative clauses are more frequent when Object relative clauses will incur more of a processing penalty The lady that visited the banker enjoyed … the embedded NP is pronominal when the embedded NP is not grounded than subject relative the banker visited clauses – such as in processing relative clauses with full NPs Object relatives still more difficult at main verb • Pronominal object relative advantage is due to the Difference between RT means (topic condition) higher frequency of that + pronoun + verb chunks and no preceding context region 40 than that + verb + pronoun chunks Relative Clause Combined Region Object relative clauses will be easier to process when the Main Verb Region 20 embedded NP refers back to the contextual topic – such as in 0 Distribution of relative clauses by type of the Reali and Christiansen experiments where the referent of the embedded NP embedded pronoun is either explicitly mentioned (e.g., Taylors) -20 50,000 Subject Relative Clause or is easily invoked (e.g., you, I) -40 Object Relative Clause 40,000 -60

Magnitude of Difference(ms) -80 30,000 Full NP Experimental Design subject relative has advantage -100 20,000 Pronominal NP • Participant-paced moving window paradigm object relative has advantage Reading times for embedded NP 10,000 • 2x2 design 540

Count in AmericanCount in NationalCorpus Neutral discourse topic - • Subject vs. object relative clause NP is discourse topic (datafromChristiansen, Reali & press) in Other embedded NP Pronominal embedded NP 520

• Appropriate discourse context vs. neutral discourse context 500 * Pronominal object relative clauses are most common in spoken English, especially when the • Sample items 480 460 relative clause modifies the subject of the • Appropriate discourse context RT (ms) sentence 440 The banker was friendly. The lady that visited the banker enjoyed the DistributionDistribution subjectof subject modifying modifying relative relative clauses clauses meal. 420 inin spoken spoken English English 400 The banker was friendly. The lady that the banker visited enjoyed the Subject Relative Clause Object Relative Clause 250 meal. Subject Relative Clause Object Relative Clause 200 • Neutral context Conclusions

150 There was a dinner party Saturday night. The lady that visited the Nearly all subject modifying • Providing an appropriate discourse context can relative clauses are banker enjoyed the meal. pronominal object relative eliminate the processing difficulty found in the 100 clauses There was a dinner party Saturday night. The lady that the banker relative clause region of sentences containing Count in SwitchboardCount in 50 visited enjoyed the meal. object relative clauses (the same region where

0 Reali & Christiansen found the advantage for Other embedded NP Pronominal embedded NP pronominal object relative clauses) Distribution of subject modifying relative References Acknowledgements • The pronominal relative clause effect may be in clauses in spoken English part due to the better fit between the typical 300 Fox, B. A. and Thompson, S. A. (1990). A We would like to thank: discourse use of object relatives and Subject Relative (Intransitive) discourse explanation of the grammar of • The University at Buffalo Research Foundation 250 Subject Relative (Full NP) relative clauses in English conversation. experimental contexts in which they appeared Subject Relative (Pronominal) Language, 66(2), 297–316. • Jean-Pierre Koenig, Breton Bienvenue, and the 200 Object Relative (Full NP) Gordon, P. C., Hendrick, R., and Johnson, M. other members of the Psycholinguistics Lab at the • The typical difficulties found in processing Object Relative (Pronominal) (2001). Memory interference during language University at Buffalo 150 processing. Journal of Experimental object relative clauses may be in part due to the • The Computational Psycholinguistics Lab at the Psychology: Learning, Memory, and Cognition, University of Buffalo comparative un-naturalness of the experimental 100 27(6), 1411-1423. Count in Switchboard Reali, F. and Christiansen, M. H. (in press). contexts in which they appear clauses Switchboard in 50 Processing of relative clauses is made easier

Count of subject modifying relative by frequency of occurrence. Journal of Memory • Subject relative clauses, which typically serve 0 and Language. Animate Subject Inanimate Subject Warren, T. and Gibson, E. (2002). The influence of different discourse functions than object relative referential processing on sentence complexity. clauses, may not be affected in the same way Cognition, 85, 79-112.