Wittenberg, E. 1 , Barner, D. 1 & Levy, R. 2
1 University of California, San Diego
2 Massachusetts Institute of Technology
Many accounts of how people interpret scalar adjectives like ?tall? focus on how statistical distributions determine thresholds. But are people also sensitive to subtle syntactic and pragmatic cues when interpreting scalar terms? If so, do these effects on thresholds persist over time?
To test this, we developed a two-step task: First, an alien creature (Fiba) asked participants to identify the ?red tall building? in a ten-building display (16 critical trials: tall/wide/big/long, 48 fillers). However, red buildings were of only small or medium heights, such that they should not be judged as tall unless their tallness is presupposed. In the prompt, we varied adjective order (syntax) and whether the question presupposed a {red,tall} building (pragmatics).
Participants then counted buildings Fiba would count as tall in a subsequent display to determine whether syntax and pragmatics in the first step affected their thresholds.
People counted more tall buildings when the first task presupposed a {red,tall} building. Also, this effect was stronger in the red-tall order than in the tall-red order, suggesting that tall-red was partly interpreted as ?tall-for-a-red?. Data from a Verbal-Forced-Choice task support this.
This suggests that adjective order and presupposition influence scalar threshold computation, and the threshold effect persists over time.