1. On the PROCESSING of “might”
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36
On the PROCESSING of “might”
PLC 36
Dimka Atanassov, Florian Schwarz & John Trueswell
University
of Pennsylvania

2. Literal vs. Pragmatically enriched meaning
Did Mary eat the cookies?
(1) Mary ate some of the cookies
a. Mary ate some, but not all of the cookies
b. In fact, Mary ate all of them
(Grice, 1975): speakers are expected to be as informative as necessary, but no more than that (quantity maxim).
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

3. Implicatures and scales
Mary ate some of the cookies
Some and possibly all (literal)
Some but not all (pragmatically enriched)
Would you happen to know where Mary is?
(2) Mary might be in her office
Quantity
none
some
all
might
must
Certainty
Gadzar (1979) and Horn (1972
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

4. Visual world paradigm
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

5. Previous psycholinguistics work: evidence for delayed implicatures
Huang & Snedeker (2011) : Evidence for delayed implicature processing (600ms)
Some
Two
target
target
distractor
distractor
all
three
target
target
distractor
distractor
Point to the girl that has [some/two/all/three] of the ice cream sandwiches.
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

6. Previous psycholinguistics work: evidence for immediate implicatures
Grodner et al. (2010): some is processed immediately.
Degen & Tanenhaus 2012: speed of computing some impl. depends on naturalness of some and its lexical alternatives
Early summa condition
Late summa condition
summa
target
summa target
no way to identify target prior to phonological disambiguation.
Click on the girl who has summa the balls/ alla the balloons/nunna the items.
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

7. More work on implicatures:
Camp “Rapid”
Camp “Delayed”
Sedivy et al., 1999
Degen & Tanenhaus 2012
Bott & Noveck, 2004
Noveck & Posada, 2003
Huang & Snedeker, 2009a
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

8. Present study: Compares might to must instead of some to all
Instead of using prerecorded stimuli incorporates target utterances within a natural conversation with a confederate
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

9. A game of guessing with a confederate
Subject and confederate take turns guessing shapes on a screen (Building on a design by Brown-Schmidt et al., 2008; Brown-Schmidt, 2009)
Guesser: each time sees only part of the display and has to guess whatever is hidden
Verifier: sees entire display, including whatever is hidden for the verifier, and has to mark guesses as “correct” or “incorrect”
Part I: subject guesses; Part 2: confederate guesses
Target sentences incorporated as part of the confederate’s “guesses”.
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

10. The game-rules: 4 types of shapes 4 possible colors
Row: all different shapes
Col: all different colors
Col: all same colors
4 types of shapes
4 possible colors
In each row either all shapes are of the same type or all are different
In each column either all items have same color or all have different color
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

11. Example of a target trial
Two conditions for every scene:
confederate utters (1) followed by (2a) or (2b)
1. Hmm…in the upper left there is a red square.
2a.There must be a red square located in the upper right.
2b. There might be a red square located in the bottom left.
Target
Competitor
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

12. Example of an “incorrect guess”
½ target items had guesses for “might” that happened to be incorrect (but consistent with rules)
There must be a green square located in the upper right.
There might be a green square located in the bottom left. (but in reality it’s a green heart)
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

13. Procedure
14 subjects, all native English speakers and undergraduates at UPenn participated for course credit (2 were taken out due to high track loss)
Stimuli presented in Experiment Builder
Eye tracking data collected in Eyelink 1k eye tracker
Sampling rate: 1 kHz, re-sampled offline to 100Hz
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

14. Results:
Target Advantage for entire sentence (500ms prior to modal onset until 2 sec after onset of right/left)
Target Advantage:
looks to target- Looks to competitor.
Target= shape that can be guessed with certainty (must)
Competitor= shape that cannot be guessed with certainty (might)
Avg. disambiguation
time
Modal onset
1s after mo
There must be a red square located in the upper right
might bottom left
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

15. Interpretation: Implicature processing is delayed by 800ms
Results (continued):
Correct Guesses, Ambiguity period
Ambiguity period: from modal onset (marked as 0) until 200ms after disambiguation (onset of upper right/ bottom left) A B
Significant interaction between condition and time window :
by item p <.01;
by subject p <.05
Interpretation: Implicature processing is delayed by 800ms
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

16. Conclusion
Evidence for delayed processing times of scalar implicatures (800ms)
Comparable to previous results on all vs some (Huang & Snedeker)
Evidence based on
different implicature triggers (must &might)
using natural discourse context
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36

17. Thank You!
And many thanks to our confederates, Aviad Eilam and David Faber!
Dimka Atanassov, Florian Schwarz and John Trueswell, UPenn PLC 36