Question

Explain the concept of “statistical learning” and why it is important. How does it relate to “rule learning”, as discussed by Marcus. At what age do babies show evidence of statistical learning and how has this been demonstrated? What about rule learning? Why is it more than just memorizing the exact sound sequences or syllables that they have heard?

Answer 1

Statistical learning refers to the ability to group elements in the environment by using probabilities or likelihood of occurrences. Statistical learning in the context of language development is the primary method by which infants identify patterns in their native language. This is achieved by identifying units of the input, discovering what features of the input predict other features and making categorisations of features that are likely to occur together. Infants, as young as 8 months old are able to display evidence of statistical learning.

According to the rule based learning proposed by Marcus and his colleagues, language acquisition among infants involve the representation, extraction and generalisation of abstract algebraic rules. Because rule based learning posits that when infants hear speech sounds, they can learn rules that govern their combination, it is comparable to saristical learning. Evidence for rule based learning has been observed in 7 month old infants. The design of the artificial language task used in the experiments by Marcus and his colleagues ensured that this discrimination could not be performed by counting, by a system that is sensitive only to transitional probabilities, or by a popular class of simple neural network models.