Reasoning as we read: A psycholinguistic perspective on the comprehension of conditional statements.

Transcription

1 Reasoning as we read: A psycholinguistic perspective on the comprehension of conditional statements. Andrew Stewart Matthew Haigh

2 It s a matter of timing... Much work in the area of human language processing focuses on when different kinds of information influence comprehension, and how readers build the mental representations associated with text meaning (situation models, e.g., Zwaan et al., 1995).

3 Reasoning as we Read Comprehension of conditionals is non-trivial. Amongst other things, it involves determining whether the conditional is describing a possible future situation, is a counterfactual requiring undoing of the past. It involves determining the degree of belief of the conditional, deciding what speech act is being communicated (was that a promise?), and also the speaker s persuasiveness when engaged in the use of a slippery-slope argument. And lots more!

4 Focus of today s talk is on the past 6 years of research we ve carried out examining how quickly all these things happen (incl. nice new data on the processing of inducements and slippery slope arguments). Main interest on the moment-by-moment processing that accompanies the online comprehension of contextualised conditionals. Discussion as to what our results might mean for theories of how conditionals are produced and comprehended in everyday contexts.

5 Indicative vs. counterfactual conditionals in reading Stewart, Haigh and Kidd (QJEP, 2009) showed that the interpretation of counterfactual conditionals but not indicative conditionals is constrained by prior context. Reading times to (e.g.) If Darren had been athletic, he could probably have played on the rugby team longer when prior context mismatches the presupposition. Indicative conditionals (e.g.) If Darren is athletic, he probably plays on the rugby team always quick to read.

6 Interaction in critical region (last word of antecedent, first word of consequent).

7 How does believability of a conditional affect reading? Building on Oberauer & Wilhelm, (2003), and Over et al., (2007), Haigh, Stewart, & Connell (Mem & Cog, 2013) examined how different sources of probability affect reading of conditionals such as: If student tuition fees rise, then applications for university places will fall.

8 P(p) and P(q p) were the strongest predictors of reading times (last word of consequent).

9 All well and good but self-paced reading is a relatively coarse-grained measure of reading. Words appear one by one so normal reading processes are disrupted. Similar problem with examining event-related brain potentials during reading (e.g., Bonnefond & Van der Henst, 2013). So, if we want to measure how different sources of information influence the comprehension of conditionals during normal reading, how do we do it?

10 An eye-tracking in reading primer Eye-movements during reading consist of fixations (for about 250 msec. each) and saccades (where the eye jumps from one location to another). During reading 10-15% of all eye-movements are backwards (called regressions) and they allow the reader to (re)look at previously read text. When fixating at a point in a word, you can actually see about 4 characters to the left and about to the right of fixation. This is the perceptual span (McConkie & Rayner, 1975).

11

12 Region 1 Region 2 Region 3 When he was nearly thirteen, my brother Jem got his arm badly broken at the elbow Region 1 Region First pass = Second pass = 7 Total time = First pass = 6+8 Second pass = Regression path = Total time = } Expressed in ms. Regressions in = 7 Regressions in = 11, 12 Regressions out = 7 } Expressed as a % Early Processing: First pass, Regressions out Intermediate Processing: Regression path Late Processing: Second pass, Total time

13 Are readers rapidly sensitive to a conditional s utility? Haigh, Ferguson, & Stewart (2014) QJEP

14 Bonnefon (2009) Psych Review Bonnefon, Haigh, & Stewart (2013) JML Haigh, Ferguson, & Stewart (2014) QJEP Utility grid for a conditional promise: A father saying to his son: If you wash my car, then I ll pay you a fiver. {If h } {Then s + h } { actor utility target } Utility grid for a conditional threat: A traffic warden saying to motorist: If you park there, then I ll give you a ticket. {If h } {Then s - h } { actor utility target }

15 Alan had just presented his research paper to a meeting of leading physicists. During the coffee break he was called over by the Editor of the internationally renowned Journal of Physics. The Editor was very impressed by Alan s findings and said that they should be widely publicised / The Editor was very critical of Alan s findings and said that they were not valid. (a) As they parted, the Editor told Alan if you submit your paper to the Journal of Physics, then I will reject it outright. (b) As they parted, the Editor told Alan if you submit your paper to the Journal of Physics, then I will accept it outright. This comment made Alan consider his options carefully. Haigh, Ferguson, & Stewart (2014) QJEP

16 First Pass Regressions Out then I will accept it outright. Haigh, Ferguson, & Stewart (2014) QJEP

17 Are readers rapidly sensitive to whether the producer of a conditional has control over the consequent? Stewart, Haigh, & Ferguson (2013) JEP:LMC

18 Utility grid for a conditional promise: A father saying to his son: If you wash my car, then I ll pay you five pounds. {If h } {Then s + h } { actor utility target } Utility grid for a conditional tip: A father saying to his son: If you study hard, then you ll do well at school. {If h 0 A } {Then ω + h } { actor utility target } Stewart, Haigh, & Ferguson (2013) JEP:LMC

19 How does a reader s knowledge of speaker control influence processing of the conditional? If you submit your paper to the Journal of Physics, then I will publish it in the next issue. Felicitous if uttered by someone who has control over what gets published, but infelicitous if uttered by someone who does not. Stewart, Haigh, & Ferguson (2013) JEP:LMC

20 If you submit your paper to the Journal of Physics, then it stands a good chance of being published. Felicitous regardless of the control the speaker has (as the consequent does not require the speaker to have control). Does speaker control influence processing of the conditional as it is read? Does this influence occur early or late? Stewart, Haigh, & Ferguson (2013) JEP:LMC

21 Alan had just presented his research paper to a meeting of leading physicists. During the coffee break he was called over by the Editor of the internationally renowned Journal of Physics/by a junior colleague. The Editor/colleague was very impressed by Alan s findings and said that they should be widely publicised. (a) As they parted, the Editor/colleague told Alan if you submit your paper to the Journal of Physics, then I will publish it in the next issue. (b) As they parted, the Editor/colleague told Alan if you submit your paper to the Journal of Physics, then it stands a good chance of being published. This comment made Alan consider his options carefully. Stewart, Haigh, & Ferguson (2013) JEP:LMC

22 First Pass Regressions Out then I will publish it in the next issue. Stewart, Haigh, & Ferguson (2013) JEP:LMC

23 When does sensitivity to semantic flexibility of promise vs. threats arise? Haigh et al. (Acta Psychologica, 2011) showed that during reading, conditional threats can be referred back to as promises, but conditional promises can t be referred back to as threats. In 2 new experiments (Wood et al., under revision), people read the verb promise or threaten followed by a conditional promise or threat, or the phrase this promise or this threat following a conditional promise or threat.

24 Ian was at a builder's merchant to buy some paving slabs for a job. He approached the sales assistant intent on getting a good deal. She promised/threatened him "if you buy in bulk, then I'll give you our trade discount."/ "if you only buy a small amount, then I'll stop your trade discount." This comment helped Ian to make his decision. He thought about it for a while and then placed his order. 32 Ss, 32 Items.

25 She promised/threatened him "if you buy in bulk, then I'll give you our trade discount." "if you only buy a small amount, then I'll stop your trade discount."

26 Greater disruption to reading when a conditional promise follows the verb threaten compared to when a threat follows the verb promise suggesting that differences in the semantic scope of promises vs. threats influences conditional processing rapidly.

27 When the phrase this promise or this threat follows a conditional promise or threat, there is an immediate symmetrical penalty on the word promise or threat that turns into an asymmetrical penalty (with greater disruption following the word threat ) later in the sentence.

28 Results of this second experiment consistent with initial utility matching (takes about 300ms) followed by semantic flexibility in the comprehension of the anaphor this promise.

29 Slippery Slope Arguments (SSAs) If p, then q SSAs describe an initial proposal (P) and a predicted, undesirable consequence of this proposal (Q): If voluntary euthanasia is ever legalised, then it will ultimately lead to the legalisation of involuntary euthanasia. Haigh, Wood, & Stewart (submitted)

30 SSAs can be thought of as a negative consequentalist argument (following Corner et al., 2011, Bonnefon & Hilton, 2004). In a paraphrasing study we examined what SSAs are seen to reveal about the attitudes of the producer. 24 Ss presented with 24 SSAs and asked to write down what they think the producer believes. Haigh, Wood, & Stewart (submitted)

31 Carly utters: If voluntary euthanasia is ever legalised, then it will ultimately lead to the legalisation of involuntary euthanasia. ~ 77% of responses indicate that participants inferred the speaker had a negative attitude towards the antecedent information. Participant 2: Carly disagrees with voluntary euthanasia Participant 3: Carly does not think voluntary euthanasia should be legalised, as it could lead to murder. Participant 16: Carly thinks the risks associated with the escalation of the laws is not worth legalising voluntary euthanasia. [sic] Participant 19: Carly opposes voluntary euthanasia. Haigh, Wood, & Stewart (submitted)

32 Eye-tracking experiment 24 participants read 24 SSAs in one of three conditions: Speaker was known to be against the antecedent proposal (Consistent) vs Speaker was known to support the antecedent proposal (Inconsistent) vs Speaker s position towards the antecedent proposal was unclear (Neutral). Haigh, Wood, & Stewart (submitted)

33 Key analysis regions were the emboldened text in the Antecedent and Consequent. Haigh, Wood, & Stewart (submitted)

34 Regression path reading times Haigh, Wood, & Stewart (submitted)

35 The SSA reading time data are compatible with a view that readers have difficulty understanding a SSA when it goes against what it known about the producer s attitudes (cf. our 2009 QJEP paper). Readers are rapidly sensitive to the rhetorical function of SSAs and what they reveal about the producer s attitudes with respect to the antecedent proposition. Haigh, Wood, & Stewart (submitted)

36 What does it all mean? Readers are rapidly sensitive to the grammatical mood of a conditional (subjunctive vs indicative), the conditional probability of a conditional, a conditional s utility, the extent to which the speaker has control over the conditional consequent action, the speech act communicated by a conditional, the semantic flexibility associated with the category of promises, and the rhetorical function of SSAs (and what they reveal about the attitudes of the speaker).

37 Are conditionals special? The conditional form seems to act as a handy package which allows complex information to be conveyed in a manageable (and processable) way. Bonnefon s (2009) utility grid framework provides a nice starting point for capturing in a helpful notation the features that might form the building blocks of the search for meaning that occurs when conditionals are read.

38 Strong evidence that many semantically and pragmatically rich sources of information influence the comprehension of conditionals at an early opportunity. Evans (2005) highlighted the need to consider the communicative function of conditionals in order to best understand them. To do this, we need to also understand how conditionals are processed (and what sources of information are used to inform this). Fits with the spirit of the New Paradigm view of conditionals (and with Hadley and Trippas, in press, view of Default Interventionist accounts - parallel influences of a range of factors). Contextualised conditionals examined using temporally sensitive measures could be a key theme of future research.

39 Thanks to: University of Manchester Jeff Wood Elizabeth le-luan Helen Wray University of Kent Heather Ferguson Lancaster University Louise Connell Toulouse School of Economics Jean-François Bonnefon

40 Our recent papers on conditionals and reading. Bonnefon, J.-F., Haigh, M., & Stewart, A.J. (2013). Utility templates for the interpretation of conditional statements. Journal of Memory and Language, 68, Haigh, M., Ferguson, H.J., & Stewart, A.J. (2014). An eye-tracking investigation into readers sensitivity to expected versus actual utility in the comprehension of indicative conditionals. Quarterly Journal of Experimental Psychology, 67, Haigh, M., & Stewart, A.J. (2011). The influence of clause order, congruency and probability on the processing of conditionals. Thinking and Reasoning, 17, Haigh, M., Stewart, A.J., & Connell, L. (2013). Reasoning as we read: Establishing the probability of causal conditionals. Memory & Cognition, 41, Haigh, M., Stewart, A.J., Wood, J., & Connell, L. (2011). Conditional advice and inducements: are readers sensitive to implicit speech acts during comprehension? Acta Psychologica, 136, Stewart, A.J., Haigh, M., & Ferguson, H.J. (2013). Sensitivity to speaker control in the online comprehension of conditional tips and promises: an eye-tracking study. Journal of Experimental Psychology: Learning, Memory, and Cognition, 39, Stewart, A.J., Haigh, M., & Kidd, E. (2009). An investigation into the online processing of counterfactual and indicative conditionals. Quarterly Journal of Experimental Psychology, 62,

41 QJEP extra slides

42 Regression Path Time then I will accept it outright.

43 JEP extra slides

44 Regression Path Time then I will publish it in the next issue.

45 Total Reading Time then I will publish it in the next issue.