Poster #26 - Mapping Language to Real-World Event Probabilities Across Middle Childhood

• In Event: Poster Session 16

Abstract

As listeners, epistemic modal expressions are vital for calibrating our expectations and actions: told that it 'might' rain---where 'might' is an epistemic modal verb---you will be less likely to grab your raincoat than if you’re told that it 'should' or 'will.' Using others’ modal expressions to inform our behavior requires understanding (1) how modal verbs relate to one another (i.e., 'should' is stronger than 'might'), and (2) how modal verbs relate to real-world probabilities. Previous studies suggest that older preschoolers understand that modal verbs can be ordered on a continuum, but leave open how they map that scale onto event probabilities (Hirst & Weil, 1982; Noveck, et al., 1996; Ozturk & Papafragou, 2015). We use a child-friendly task to present children and adults with probabilistic events at varying levels of uncertainty, and elicit their preferences across a range of epistemic modal verbs (could, can, should, will, won’t).

Experiment 1. Children (N= 44, 4-7 years, M_age=5.5) and adults (N=21, M_age=20) were familiarized with images of a gumball machine containing green and purple marbles. Each critical trial displayed a machine with 20%, 80% or 100% green marbles (Fig.1). Two aliens described the chance of getting a green marble using contrasting modal statements (e.g., a green marble “should” versus “could come out”), and children pointed to who “got it right.”

Results. A mixed effects logistic regression on children’s stronger modal selection (will>should>could) revealed a significant effect of probability (X2(1)=28.77, p<.001), age (X2(1)=6.25, p=.01) and their interaction (X2(1)=21.14, p<.001): older children exhibited a greater stronger-modal preference to describe high-probability events. Children notably diverged from adults in their understanding of 'will,' which adults reserved exclusively for 100% probabilities, while children often applied it to 80% outcomes.

Thus, while children may know the relative strength of different modals, they do not seem to make adults’ categorical distinction between probabilistic and deterministic modals, despite being able to recognize whether an outcome is deterministic. Alternatively, rather than a problem with their linguistic understanding, if children interpreted getting a green marble as more desirable, their tendency to use 'will' for 80% probabilities might reflect an optimism bias.

Experiment 2 sought to dissociate these possibilities by testing an additional deterministic outcome (0%), and modal ('won’t').

Experiment 2. Forty-eight children (M_age=8.1; 16 6—7.5yrs, 17 7.5—9yrs, 15 9—10.5yrs) and adults (N=19,M_age=20) participated.

Results. While almost all children preferred 'won’t' for 0%, and 'could' for 20%, they were again more variable in the modal they preferred to describe an 80% outcome, with 6-9-year-olds at chance between 'could,' 'should,' and 'will.' That children did not reliably choose the higher-probability modal suggests that an optimism bias is unlikely to explain our results.

Future work will investigate the asymmetry across the epistemic modal and probability rankings, where children were more systematically adult-like in describing lower-probability outcomes (0%, 20%), relative to higher probability ones (80%, 100%). For example, it may be that children are doing something sensible: while adults reserve 'will' in our task for deterministic outcomes, they may more often use 'will' to refer to probabilistic outcomes when speaking to children than they do 'won’t.'

Authors

• Ruthe Foushee, University of Chicago
  Presenting Author

• Ariel Starr, University of Washington

• Jon Wehry, University of California - Berkeley

• Jeff Anderson

• Grace Horton

• Shaun O'Grady, University of California, Berkeley

• Mahesh Srinivasan, University of California - Berkeley