2018 Friday Session C 0930
Friday, November 2, 2018 | Session C, Terrace Lounge | 9:30am
Facilitating quantifier acquisition: Training can eliminate children’s spreading errors
C. Roest, J. Spenader
Children (5;0-9;0) often reject universally quantified sentences like (1) with an extra object (Fig 1). These are “spreading” errors[1],[2][3], and they persist long after children have learned other semantic features of quantifiers, e.g. that (1) is false because of the extra girl in Fig. 2[4][5]. Puzzling, spreading experiments often show extreme response patterns: children either consistently spread or are adult-like[1],[4]. These patterns suggest acquisition involves a knowledge `leap’, perhaps triggered by exposure to a few highly informative examples. Previous training studies show simply correcting children’s errors (for second order theory-of-mind[6]), even in a single session (for prepositions[7].), can stimulate learning. Can training also eliminate quantifier spreading? We designed a study to test this.
Pre-test: 65 Dutch children between (Range 5;4-8;3, Mean:6;8, 24 girls) were tested individually for spreading. Children judged distributively quantified sentences in three conditions in a Picture Verification Task (PVT): Extra-Object Pictures (Fig. 1) and Extra-Subject Pictures (Fig 2) were presented with transitive sentences like (1) Incorporated sentences (2) were presented with other Extra Object Pictures (Fig 3), checking if spreading diminishes without an explicit object argument[5].
Pre-test Results: Mixed-effect model analysis showed Extra-Subject was significantly better than Extra-Obj Transitive (p<0.001) and Incorporated (p<0.1). Extra-Object Incorporated was significantly better than Extra-Obj Transitive (p<0.001) (See Table 1). 25 children spread in more than half the transitive Extra Object condition (Mean:6;6; 14 girls). See Table 2 for an overview.
Training: Within one week of the pretest, the 25 spreaders were trained in a 15 minute session. Children were asked to teach Tiger, a puppet. The experimenter distributed small toys to sets of paperdolls and asked Tiger to judge if sentences like (1) described the scene. Training included four situation types (One or multiple extra objects (Fig.4) , Extra subject (Fig. 5) and Correct cases). Tiger made numerous errors. The experimenter (i) corrects Tiger, emphasizing the need to check that each paperdoll has the mentioned object, (ii) emphasizing that any additional objects were “extra” and “irrelevant”. The child was encouraged to help correct Tiger. At conclusion, all children could explain correctly to Tiger why spreading responses are incorrect. Post-test Children did a PVT with Extra Object, Extra Subject and control conditions. Results: See Tables 3 and 4. Nine children made no spreading errors, and four children made only one. The remaining 12 children still spread.
Conclusion: Remarkably, spreading was eliminated in over half the children. For the remaining 12 children, an additional training may be sufficient. Training went extremely smoothly, strongly spreaders simply lacked relevant knowledge, or pragmatically misunderstood the pre-test[3]. However, because trained non-spreaders easily applied act-out knowledge to pictures in the post-test, we believe they actually learned not to spread. Our results also suggest how long a child spreads may be caused by chance: spreaders simply may not yet have encountered informative examples.
(499 words)