The 24 items used in experiments 1 and 2 were used, modified as described above. The position of the four pictures on the screen was pseudo-randomised. Protein Tyrosine Kinase inhibitor The items were presented to participants in either one of two pseudo-randomised orders.

The task took between 15 and 20 min to administer and was part of an experimental session that lasted around 40 min for adult participants and 30 min for children. The session also involved the two verbal and non-verbal IQ selection measures for children. The experiments took place in a relatively quiet room in the children’s school, or at the university for adults. The participants were 15 5-year-old English-speaking children (mean age Bortezomib 5;7; range 5;1–6;1), recruited from primary schools in Cambridge, UK, as well as 10 adults, students of various subjects at the University of Cambridge (mean age 23;9; range 19;9–26;3).

One child was removed and replaced in the sample on the grounds of low performance in the selection measures. Adults performed at ceiling with only one error in a non-scalar condition. The children’s performance was as presented in Table 2. Between-group comparisons (Mann–Whitney U) revealed that children did not perform significantly differently than adults in any condition (all U < 2.5, p > .05). Focusing on the children, a Friedman’s ANOVA reveals no significant pairwise differences between conditions (χ2(3) = .84, p > .1). This suggests that any difficulty children had was general to all conditions of the task, rather than specific to the conditions contrasting on informativeness. We investigated this further by analysing the children’s erroneous responses for the critical conditions (‘some’ and single RAS p21 protein activator 1 noun phrase). The 17% of erroneous responses for ‘some’

were distributed over all the other three pictures on display (7% for the true but underinformative picture, 7% for the picture with the correct quantity but the incorrect object, and 3% for that with the incorrect quantifier and object). A similar pattern arose for the non-scalars (9% errors distributed as 4%, 4%, and 1% for the true but underinformative, false single object, and false two objects respectively). These findings further document that 5- to 6-year-old children are sensitive to informativeness. Crucially, there is no significant difference between the children’s performance when the selection is based exclusively on logical meaning (for ‘all’ and conjoined noun phrases) and when it is also reliant on informativeness (‘some’ and single noun phrases)3.