Poster #25 - Testing Young Children’s Preference for Unconfounded Evidence in a Hypothesis Testing Paradigm

• In Event: 1-061 - Poster Session 02
  In Poster Session: PS 02 Section - Cognitive Processes

Integrative Statement

The ability to generate unconfounded and informative evidence to test a given hypothesis is a critical skill for scientific reasoning. A recent study by Köksal-Tuncer and Sodian (in press) demonstrated that children selectively generate more disconfirming and unconfounded evidence than confirming and confounded evidence in response to false causal claims in a counterargumentation paradigm. The study showed that young children use unconfounded over confounded evidence as a means to falsify a claim when they already formed a correct causal belief based on their earlier interaction with informative evidence. It is an open question whether children would show such preference for unconfounded evidence to test causal hypotheses when they do not know the true cause. In the present study, we investigated preschoolers’ preference for unconfounded evidence in a hypothesis testing task. Participants were 32 preschoolers (18 females, Mage = 5;5, range = 3;11–6;8) recruited from preschools in a large German city. In a “blicket detector” paradigm (Gopnik & Sobel, 2000), children were presented with a music box which was activated when certain objects were put on it. Objects had two salient features, namely bigness (big vs. small) and having a black dot on them or not. First, children were presented with confounded evidence: the objects that activated the music were big and had a black dot on them, and the objects that did not activate the music were small and did not have a black dot. Two puppets were used to present two different hypotheses. One of the puppets suggested that the causal feature was bigness and the other puppet suggested the causal feature was having a black dot. In two evidence generation trials, children were asked to choose objects to find out which puppet was right. In each trial, children were given two novel object options, one confounded and one unconfounded. For the confounded objects, the two features were either present or absent together (i.e., big with a black dot and small without a black dot), and for the unconfounded objects, the two features appear in isolation (i.e., small with a black dot and big without a black dot). Results revealed that children did not show any preference for the unconfounded objects (Binomial test, chance level .50) both in the first (p = .377) and in the second trial (p = .110). In the first trial, 19 children (59%) chose the unconfounded object, and in the second trial, 11 children (34%) did so. Considering the similarity of the design of the present study with the counterargumentation task (Köksal-Tuncer & Sodian, in press), these findings suggest that generating unconfounded evidence to test a hypothesis is harder than doing so to falsify a hypothesis that one already knows to be false. Thus, young children’s scientific reasoning abilities may not be limited by an inability to distinguish between confounded and unconfounded evidence, but by the demands of tasks that require them to consider hypotheses of unknown truth value.

Authors

• Özgün Köksal-Tuncer, LMU Munich
  Presenting Author

• Andrea Saffran, LMU Munich
  Non-Presenting Author

• Beate Sodian, LMU Munich
  Non-Presenting Author