Scientific American just posted an article about a study correlating learning language to mathematical statistics.

Cognitive scientist Bob McMurray of the University of Iowa set up a relatively simple mathematical model of word learning on a commonly available spreadsheet, assessing the potential to learn each of some 200 words. He set a numerical threshold at which a given word would be considered learned, operating under the assumption that it would take kids time to learn each word, they could pick up multiple words at the same time, and that some words were more difficult to process.

Color me doubtful. It looks to me like they are applying a simple mathematical to an overly complex subject and calling correlation causation. Why I disagree after the jump.

Their study showed that, after a certain amount of time, there was an explosion of “learned” words. This supposedly accounts for the explosion of language skill seen by most kids (the transition from incoherent babbling to chatting away like there is no tommorow). They compared language to jelly beans filling jars. Children collect words until their mental “jelly bean jar” is filled, at which point the word is learned and usable in daily vocabulary.

The finding casts doubt on the necessity of any particular special function acquired during childhood that enables kids to master language. “These specialized mechanisms are working on the margins,” McMurray says. But “to explain the big picture, it’s much, much simpler. … Anytime you have more difficult than easy words [the learning curve] will have this property.”

The key point about their research is that the bell curve is created because of the discrepancy between “difficulty” in words.

Which is just plain silly. Imagine you are a newborn infant. For the past 12 months, the only sensory information your auditory regions have experienced is muffled sounds from the outside world. One day BAM! you are outside and experiencing a cacophony of sounds. To the neurons in your brain, every sound is some combination of inputs from the neurons in your ear. What is to say that the particular combination of sounds from the word “neuroscience” is any more “difficult” than the sounds from the word “baby”? Taking it a step further, there is no practical discrepancy between a car sound and a spoken word in terms of difficulty.

What I believe the study is lacking is the fact that language isn’t about memorizing words and that words do not have any inherent difficulty. I can spend all day memorizing French words but not learn their context, use or meaning.

I will concede that some words may be more difficult to pronounce because they require a more complex maneuver of your vocal chords and mouth. But the words themselves, and therefore the comprehension of the child, are not inherently any more difficult than others. Granted, the article did not link the real study (I wish they would) but it seems like the whole thing was an overly simplistic view of language skills.

This is also one of the problems I have been trying to wrap my head around while working on DN. To an infant, all inputs are fundamentally identical. How does an infant begin to discriminate between various inputs? How does an infant decide that “mama” means the woman holding it? Some people would argue that positive reinforcement (gentle tone, happy words, encouragement) teach the infant right from wrong and how to speak. But how do you positively encourage an infant when positive reinforcement (at a molecular level) is the same as negative encouragement as both are merely sets of input? It boggles my mind and should prove quite interesting if/when DN gets to that point.