Figure 3.

An adult chooses a linguistic expression w₁ associated with concept r₁ that they want to communicate. At t₁ the child doesn’t know whether the novel word w₁ refers to r₁ or r₂ and for now the best she can do is remember the associations between the scene and the words. At t₂ she hears w₁ with one object r₁ familiar from t₁ and one new object r₃. Cross-situational learning works by repeatedly recording the associations between language and the context in which it is used. Over time, the signal (an intentional word-referent pair) is more strongly represented than the noise (an unintentional word-referent pair). Items that appear in the hashed box are the raw data on which the child makes the cross-situational associations.