Sunday, November 13, 2016

Scalar Implicature, Inductive, and Deductive Concept Learning

Both readings are studies on scalar implicature in children. These studies expanded on the fact that young children have difficulty understanding scalar implicature such as "some" and "all." Barner's study discovered that part of the difficulty is the lack of alternatives to the scalar implicatures such as "some."

This reminds me of the difference between inductive and deductive learning. Most concept learning by children is deductive- meaning that it starts with a hypothesis and based on evidence reaches a conclusion. Children in most scenarios do not learn by induction - starting with a broad generalization based on some specific instances. For example, learning via induction what a swan is is through seeing a few white swans and immediately coming to the conclusion that all swans are white. This can be inaccurate in many cases if there is not enough evidence. For example, when children are learning the concept of a dog, through induction they might assume all dogs look like a beagle. If a child were to learn through deduction what the concept of a dog is, they start by using a general hypothesis of what a dog is: for example, it has four legs and barks, has a tail, etc. Further evidence of different breeds of dogs refines this hypothesis. While there are issues with induction as a way to learn, it is more accurate than deduction through the use of counter examples. If we were start with knowing a beagle is a dog, and given more examples of dogs and counter examples of what exactly is not a dog, this, eventually, will give us an accurate hypothesis of what a dog is.

This is similar to learning scalar implicature. Children use deductive learning in learning what scalar implicature is - "some" vs "all" and etc. However, it is the lack of induction type learning that involves negative examples that is hard for children to grasp. The lack of counter examples makes their concept learning of scalar implicatures much less accurate. While it takes much more time to learn scalar implicatures through induction type learning, eventually adults achieve it and can use them accurately.

2 comments:

  1. Hi Nicole! I find the cognitive neuroscience connection from SymSys that you draw regarding children using deduction rather than induction in scalar implicature quite interesting. It makes sense that without negative alternatives gained through the implementation of repetitive induction-type learning, children would really struggle with instances of ad-hoc scalar implicature. However, according to Barner's research, scalar implication using numerical-based quantifiers seems to be quite simple for adults and children alike. Why do you think this exception exists? Do children understand number-based implicatures through induction, as adults do, or through a different method?

    ReplyDelete
  2. Hi Nicole, thanks very much for your great comment. I enjoyed your observation on how scalar implicatures grow in adults but I was wondering what your thoughts were on consistencies with other languages? Do you also feel that children speaking different languages struggle with scalar implicatures in a similar way and if so do you think the time frame for their eventual mastery of scalars is similar?

    ReplyDelete