Discourse influences during parsing are delayed Keith Rayner, Simon Garrod,& Charles A. Perfetti Cognition, 45, 1992
Background… • How does context influence parsing strategy? • Initial parsing strategies are not influenced by context or pragmatics: Garden-Path Model • Initial parsing strategies are influenced by context/pragmatics: Interactive Model
Parsing in two different syntactically ambiguous texts • Garden-Path Model: we can resolve local ambiguity in the sentence syntax by relying solely on structural information • Minimal Attachment (MA) & Late Closure • (1) Prepositional Phrase (PP) • MA says #2 is more difficult to parse: • John put the book on the table and went to sleep. • John put the book on the table into his briefcase.
Parsing in two different syntactically ambiguous texts • (2) Reduced Relative Clause (RRC) • MA says #2 is more difficult to parse (and even more difficult than #2 in the aforementioned PP). • The woman rushed to the hospital and forgot her laundry. • The woman rushed to the hospital had a pretty baby girl.
Contextual Influences on Parsing? • John put the book on the table… • Perhaps context & postnominal referents help select among several possible candidates (e.g., multiple books in the room in various locations)… “Attention Selection” • Or, perhaps postnominal “re-orients” the reader to the appropriate context, thus enabling identification of the referent. “Attention Orientation” • #2 is the focus of this paper
Questions • Does establishing reference influence parsing decision? • What is the impact of attention-orienting postnominals that are either (1) within “discourse focus”, or (2) out of “discourse focus” • Does context influence parsing more on simpler text structures (i.e., PP vs. RRC)?
Participants • 18 Young Adults (read passages in context) • Another set of Ss read target sentences in isolation • No individual difference data reported (e.g., education, vocabulary, WM, etc.)
Sentence Materials • Two versions of target sentences: • PP & RRC • Three versions of each passage: • Minimal Attachment (MA) • Non-Minimal Attachment, Discourse Focused (NMA-F) • Non-Minimal Attachment, Discourse Non-Focused (NMA-NF) • 27 experimental & 15 filler passages each
PP - MA • Susan had real trouble trying to find the man who had promised to give her a lift into town. She looked for him in the office but without success. Eventually she discovered /the man/ in the car/ and they drove/ away into town/.
PP - NMA-NF • June always enjoyed theatrical parties and it was her ambition to be taken out by a real star. This evening she was particularly interested in a man who had arrived in a very expensive Italian sports car, but she couldn’t find him anywhere. She tried to circulate among the other guests in the hope of attracting someone’s attention. Eventually she discovered /the man/ in the car/ talking to one/ of her friends/.
PP - NMA-F • June always enjoyed theatrical parties and it was her ambition to be taken out by a real star. This evening she was particularly interested in a man who had arrived in a very expensive Italian sports car, but she couldn’t find him anywhere. Eventually she discovered /the man/ in the car/ talking to one/ of her friends/.
RR - MA • Jane loved to arrange fashion advertisements. At the back she put up a dais for her central figure, a model wearing one of the store’s most expensive garments. Around this she placed bunches of flowers. Then /the model/ perched on the dais/ and Jane took/ the picture/. Everything worked out very well.
RR - NMA-NF • Gillian had arrived late at her studio and faced a tight deadline for the fashion cover shot. In the center of the tableau she perched a model on a narrow dais. Around her she arranged some other people in various poses all observing a dangerous looking cut-out of a panther. Every time she thought she had the scene set someone would move out of position. Even the lights were acting up that day. Eventually it all seemed perfect. Then /the model/ perched on the dais/ lost her balance/. /She crashed/ into the rest of the tableau and everything had to be arranged over again.
RR - NMA-F • Gillian had arrived late at her studio and faced a tight deadline for the fashion cover shot. In the center of the tableau she perched a model on a narrow dais. Eventually it all seemed perfect. Then /the model/ perched on the dais/ lost her balance/. /She crashed/ into the rest of the tableau and everything had to be arranged over again.
Other details • Target regions always on the same line • “Attempt was made” to have the size (# of letters) of target regions 3 & 4 very similar… • Scored for (1) First-pass reading time and (2) total reading time, with (3) # of regressions counted as well • DV data reported in ms/character
Conclusions: PP • Context influences parsing only after the initial parse • Examining ONLY Total RT might suggest that discourse focus (NMA-F) might lead to context influencing immediate parsing, but first-pass RT suggests this isn’t the case • (MA rt < NMA rts)
Regressions • Similar as before: • More likely to regress from regions 3 & 4 to regions 2 & 3 in the two NMA conditions than the MA.
Conclusions: RR • Similar to PP: • MA rt < NMA rts on first-pass reading time • One noteable difference: • Total reading time suggested discourse focus might help improve RTs
Overall Conclusions • Pattern supports garden-path model: semantic information influences RT only after initial parsing decisions. • Evidence: PP & RR RTs for target region within NMA sentences > MA sentences (even when provided context to “help” them identify the NMA interpretation). • Context does influence ease of recovery from garden-pathing. • Evidence: Total reading times for NMA-Focus, esp. for RR
Questions… • Multiple referents in context, vs. just one? • Type of text (e.g., genre)? • Length of total passage?