ABSTRACT
Inspired by Syrett (2013), three experiments explored children’s ability to distinguish attributives (e.g., “three-pound strawberries,” where MPs as adjectives signal reference to attributes) versus pseudopartitives (e.g., “three pounds of strawberries,” where MPs combine with of to signal part-whole relations). Given the systematic nature of the syntax-semantics mapping, we asked whether children are able to use syntax to interpret how entities are quantified. In Experiment 1, four- and five-year-olds were asked to choose between two characters for the one who was selling appropriate items matching an attributive or pseudopartitive expression. In Experiment 2, children of the same age heard items described with a phrase using either an attributive, a pseudopartitive, “each” (“each weighs three pounds”), or “all together” (“all together they weigh three pounds”). At test, with some items removed, children were asked whether the same phrase applied to the remaining items (e.g., “Does Dora still have three-pound strawberries?”). Children did not distinguish between attributives and pseudopartitives but did so for “each” and “all together.” Experiment 3 extends the age range with a third experimental design. Children heard “each” or “all together” descriptions (e.g., “each strawberry weighs three pounds”) and judged, at test, which of two characters “said it better” (i.e., “Mickey says ‘these are two pounds of strawberries,’ but Donald says ‘these are two-pound strawberries.’”). Children under 6 were at chance. Together, the three experiments suggest that despite its systematicity, children do not automatically appreciate the mapping between syntax and semantics.
Notes
1 The children in the mass nouns followed by count nouns group correctly rejected the puppet’s utterance 63.9% of the time. Syrett did not report whether this percentage is statistically above 50% chance. Instead, the percentage she reported as being statistically above chance was the percentage after removing children who had an (incorrect) response bias to accept the puppet’s utterances. However, it is unclear whether removing such children is justified, as doing so would favor a false positive.
2 With a small N and with each children being tested on only four questions per condition, we conducted distribution-free nonparametric tests. However, we included the more common parametric t-tests, as Syrett (Citation2013) had done.
3 The effect of block order (MPs first or quantifiers first) was not included in the model because initial analysis showed that order did not affect tendency to accept or reject the linguistic expression.