How Probable?

In this week’s readings, Rickford reiterates a number of unfortunate yet unsurprising facts regarding the disadvantages faced by those in the African American community in the U.S. With historically poorer performance than their ‘White’ counterparts on standardised tests as illustrated through statistics, he demonstrates that more could potentially be done in order to increase test scores of African American youths. To this end he cites the now-defunct Bridge programme as a positive case: easing speakers of AAVE into the habit of using SE, or elevating the status of AAVE in general society tends to effect a visible upward trend in the aforementioned scores.

He naturally takes a language-based stance on this issue and examines data with respect to linguistics departments in various institutions. In my opinion, Rickford’s most salient observation was that the problem of a paucity of African American linguists employed in tertiary institute linguistics departments begins at the roots: few undergraduates enrolled in linguistics programmes are Black, causing a vicious cycle to take hold.

On the other hand, Lupyan adopts a cognitive-science-based standpoint in order to demonstrate that the cognitivist point-of-view stands on shaky ground. Instead, his experiments are a confirmation that although natural language contains words that are intended to denote a general formal category of similar ideas, the human brain tends to impose a typicality bias on such thoughts. That is, we would nevertheless be more likely to associate the general term with a small number of particular instances of things within that category. For instance, a word stimulus of ‘triangle’ or ‘dog’ might trigger notions of equilateral triangles or golden retrievers, respectively. Each category has an idealised perceptual state, or prototype.

In any case, this gives natural language special characteristics. Without language, it would be impossible to access each category as a class of items in itself due to a lack of stimulus. Even more interestingly, language itself makes the existence of such general categories possible to begin with, for if not, our experiences would take on a far more specific and individual nature. Thus, language is causal on multiple levels.

The idea of each category having a prototype fascinates me. I would like to apply this further to Amy Perfors’ expansion upon Bayesian models of cognition. While Bayes writes about prior and posterior hypotheses (a binary) to which probability calculations for mental processes should be applied, she suggests an expansion of that model to one which has multiple hypothesis spaces which call upon one another. I believe that this may be likened to the idea of general categories in the case of Lupyan: how could we apply a probability function to the selection of a word from a hypothesis space (category) in order to classify it as a prototype? It would be valuable to then tie a discussion linking Bayes’ and Lupyan’s research to one which discusses learning through perception, and how experiential learning of concepts and frequency of exposure determine probability functions for each atom in a hypothesis space.