Implicit learning and statistical learning: one phenomenon, two approaches Pierre Perruchet and Sebastien Pacton Universite ´ de Bourgogne, LEAD/CNRS, Po ˆ le AAFE, Esplanade Erasme, 21000 Dijon, France The domain-general learning mechanisms elicited in incidental learning situations are of potential interest in many research fields, including language acquisition, object knowledge formation and motor learning. They have been the focus of studies on implicit learning for nearly 40 years. Stemming from a different research tradition, studies on statistical learning carried out in the past 10 years after the seminal studies by Saffran and collaborators, appear to be closely related, and the similarity between the two approaches is strengthened further by their recent evolution. However, implicit learning and statistical learning research favor different interpretations, focusing on the formation of chunks and statistical computations, respectively. We examine these differing approaches and suggest that this divergence opens up a major theoretical challenge for future studies. Introduction There is no doubt that many of our most fundamental abilities, whether they concern language, perception, motor skill, or social behavior, reflect some kind of adaptation to the regularities of the world that evolves without intention to learn, and without a clear awareness of what we know. This ubiquitous phenomenon was called ‘implicit learning’ (IL) by Reber [1,2] 40 years ago. Since then, several studies have explored this form of learning with several experimental paradigms (mainly finite-state grammars and serial reaction time tasks; for reviews, see [3,4]). Originating from a different research tradition, the term ‘statistical learning’ (SL) was proposed 10 years ago by Saffran and collaborators [4] to designate the ability of infants to discover the words embedded in a continuous artificial language, and this field of research is now growing exponentially. There are obvious similarities between SL and IL. As in IL, participants in SL experiments are faced with structured material without being instructed to learn. They learn merely from exposure to positive instances, without engaging in analytical processes or hypothesis-testing strategies. Researchers have pointed out that SL proceeds automatically [5–8], incidentally [9], spontaneously [6], or by simple observation [9], and that participants in SL settings were unaware of the statistical structure of the material [7]. This article first describes how recent evolution in IL and SL research fields has made them closer to one another, leading to a growing number of cross-references and to the occasional use of the two expressions as synonymous. Conway and Christiansen [10] even now propose the term ‘implicit statistical learning’ to cover the two domains. However, we then go on to show that beyond the similarity of paradigms and results, the two domains emphasize different interpretations of the data. We suggest that this divergence, which has not been high- lighted as yet, opens up a deep challenge for future studies. The recent evolution of IL and SL studies Ten years ago, it seemed possible to contrast IL and SL on their main issues of interest, namely syntax acquisition and lexicon formation, respectively. Indeed, the to-be- learned material used in artificial grammar learning research is typically governed by rules, that is by organizing principles which are independent of the specific material used in a given instance. If participants learned the rules, then this form of learning would be out of the scope of SL studies, in which the notion of rules is a priori irrelevant. However, research from the past few years has made it increasingly clear that participants in artificial grammar learning experiments do not need to extract the rules to perform well, even in situations involving transfer across surface forms (Box 1). In addition, the artificial grammar learning paradigms tend to be now supplanted by other paradigms, such as the serial reaction-time tasks, in which a description of the materials in terms of rules appears less appropriate. Another initial difference between the two domains was that IL research used a large variety of situations involving different sensory modalities and response systems, whereas SL originally focused on the early stage of language acquisition. However, more recently research on SL has progressively broadened its scope of investigation. The syllables used in the first studies have been replaced by tones with the same results [11,12].A parallel literature has evolved with visual shapes [6–8], or even tactile stimuli [13]. Perhaps even more importantly, Corresponding author: Perruchet, P. (pierre.perruchet@u-bourgogne.fr). Review TRENDS in Cognitive Sciences Vol.10 No.5 May 2006 www.sciencedirect.com 1364-6613/$ - see front matter Q 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.tics.2006.03.006