The Goldilocks Corollary to the Input Hypothesis

Share on FacebookShare on Google+Tweet about this on TwitterShare on LinkedInEmail this to someone

Zoe M. Flack, Jessica S. Horst. Two sides to every story: Children learn words better from one storybook page at a time. Infant and Child Development, 2017; e2047 DOI: 10.1002/icd.2047 (paywall)

The Input Hypothesis (more generally referred to now as the Comprehension Hypothesis) states that we acquire language by understanding messages (Krashen, 1981, 1982). How exactly do we “understand messages”?

Making sense of language can be done by using (a) what we already know about the world (background knowledge), (b) what we already know about the language we are acquiring (linguistic knowledge), and/or (c) clues given us by the environment in which the language is found (context).

When it comes to the clues or context we use in understanding messages, Krashen (1999) offers these four additional classifications:

  1. “Partly determining”: This is a context that helps us understand the meaning of something new in our input. Partly determining context gives us a good clue as to the meaning of the input (e.g. a new word), but does not (crucially) tell us everything.
  2. “Underdetermining”: This is a neutral or poor context that doesn’t help us understand the meaning of novel input.
  3. “Overdetermining”: This is context that is so rich we can in effect ignore the novel input, since it doesn’t add anything to the message we want to understand.
  4. Deceptive: This is a context that leads us astray, that makes us think the input means something it does not.

The application of the Goldilocks analogy is, of course, that we want our context to be neither too cold (underdetermining) nor too hot (overdetermining), but just right (partly determining). (“Deceptive” context doesn’t fit our fairy tale analogy, but is also quite rare outside of researchers’ experiments, which are themselves often flights into fantasy.)

Flack and Horst (2017) is a good illustration of the problem of Krashen’s second category, underdetermining context. The researchers read a set of storybooks to a group of three-year-olds. Some of the storybooks had illustrations on both sides of the page, some on just one side. When children listened to stories read from the books with illustrations on both sides of the page, they didn’t acquire as many new words as when there were illustrations on just one side of the page.

The likely reason is that putting illustrations on both sides of the page makes the context for the novel word underdetermined – kids don’t know which illustration the novel word goes with, and hence the illustrations don’t help them acquire the word’s meaning.

The solution, the researchers found, was either to put illustrations on just one side of the page, or to point out to the children which page was being referred to as the story was being read. This was done by making a simple “sweeping hand gesture” (p. 6).

Both strategies make the input partly determined, providing the right amount of context to facilitate acquisition. (The researchers actually rely on a “Cognitive Overload Theory” for their explanation of the results, but their data are consistent with our interpretation as well.)

Flack and Horst cite previous research that has found that too few illustrations can also underdetermine the context for word learning. Young children struggle with new word meanings when only one object is present, apparently preferring to know what the novel word is not as well as what it is. There is an upside-down U-shaped distribution for optimal context, as the researchers noted that “optimal word learning tasks must be not too hard, but not too easy” (p. 2)

I would add only that “hard” and “easy” here should be interpreted as referring not to effort, but to the amount of determination of the context itself.

Concurrent translation – when everything is presented in two languages –  is another example of (in most cases) inappropriate context for language acquisition. When you know everything you are about to read or hear is provided in your own language, you may simply ignore the second language input. The context is overdetermining. This is what we typically do when we watch subtitled movies and TV shows.

Note: Don’t confuse translated subtitles with captions, where the words are given to you in the same language as the audio. Captioning is helpful in many cases and often facilitates acquisition.



Share on FacebookShare on Google+Tweet about this on TwitterShare on LinkedInEmail this to someone