Despite the recent developments that allowed neural networks to achieve impressive performance on a variety of applications, these models are intrinsically affected by the problem of overgeneralization, due to their partitioning of the full input space into the fixed set of target classes used during training. Thus it is possible for novel inputs belonging to categories unknown during training ...

0749-596X/$ see front matter 2010 Elsevier Inc doi:10.1016/j.jml.2010.12.005 ⇑ Corresponding author. Address: Department of P Princeton University, Princeton, NJ 08544, USA. E-mail address: [email protected] (S.-J. Lesli Generics are statements such as ‘‘tigers are striped’’ and ‘‘ducks lay eggs’’. They express general, though not universal or exceptionless, claims about kinds (Carlson & P...

The present study investigated how children learn that some verbs may appear in the figure-locative but not the ground-locative construction (e.g., Lisa poured water into the cup; *Lisa poured the cup with water), with some showing the opposite pattern (e.g., *Bart filled water into the cup; Bart filled the cup with water), and others appearing in both (Lisa sprayed water onto the flowers; Lisa...

In this paper we show that input frequency plays a role in the acquisition of word order, but only combined with other factors. Two constructions with similar input frequencies are investigated: embedded questions and (all) embedded clauses with negation. It is shown that the lack of input cues for the target word order in embedded clauses with negation leads children to overgeneralize main cla...