Tuesday, May 16, 2017

Implicit learning and statistical learning: One phenomenon, two approaches

Perruchet, P. & Pacton, S. (2006) Implicit learning and statistical learning: One phenomenon, two approaches. Trends in Cognitive Sciences, 10(5).

People have a distinct ability to detect patterns and this ability has been explained using two theories: implicit learning and statistical learning. Implicit learning theory suggests that after sufficient exposure to a pattern one automatically deduces a rule regarding the formation of that pattern which can then be accessed consciously. In contrast, statistical learning suggests no explicit rule learning is necessary and rather, that our brains perform statistical computations to predict the likelihood that a certain stimulus will occur given the occurrence of another stimulus. These theories have huge implications in the world of language acquisition, where words and grammatical structures must be intuited from continuous sequences of syllables.

In the present paper, Perruchet and Pacton (2006) make it clear that while the origins of implicit and statistical learning theories are vastly different, they have converged to a single goal: to explain general learning through domain-general processing. Further, Perruchet and Pacton (2006) suggest two explanations which attempt to integrate the two theories.

The first explanation suggests all levels of learning are done with chunks based on the frequency with which they occur. This model claims that statistical patterns, which statistical learning theory predicts are used to acquire information about a pattern, are only a byproduct of the frequency with which different chunks occur. The plausibility of this theory is supported by competitive chunking models like PARSER (Perruchet & Vinter, 1998, as cited in Perruchet and Pacton) which posits that chunks which maintain cognitive representation are those which are most frequent.

The second theory suggests that statistical computations are carried out to form the chunks and then these chunks are used in a competitive model. As Perruchet and Pacton explain, this explanation has the benefit of explaining why the stimulus preceding and following a string can impact the way it is remembered, a fact not well explained by competitive chunking models of learning like PARSER. The paper concludes that there is insufficient evidence to support the superiority of either the pure chunking or the statistical chunking model of learning. It is suggested the distinction between statistical and implicit learning is important in determining the involvement of consciousness in learning, as implicit learning theory suggests chunks are manipulated consciously whereas statistical learning theory would indicate that chunk knowledge would be held in the unconscious.

Blogger; Braxton Murphy is an undergraduate research student working under the supervision of Dr. Lisa Archibald


  1. This comment has been removed by a blog administrator.

  2. Thanks you for sharing the article. The data that you provided in the blog is infromative and effectve. Through you blog I gained so much knowledge. Also check my collection at machine learning online course Blog