ABSTRACT
Amongst the numerous theoretical debates regarding implicit learning, one of the most discussed concerns the implicit nature of the information learned with the implicit learning paradigms. Some authors (e.g. Reber, 1993) argue that the information acquired in implicit learning situations is by nature different from the kind of information acquired in explicit learning tasks (in which, for example, learning is based on hypothesis-testing processes), while others reject the idea of two independent learning processes (Shanks & St John, 1994;). The problem of the implicit versus explicit nature of the information acquired in the implicit learning tasks has been classically investigated in normal subjects by assessing the acquired knowledge through an implicit and an explicit memory task; the arguments in favour of the idea that the knowledge is implicit (or unconscious) rest on the dissociations (or the lack of correlation) observed between performance in the implicit task and performance in the explicit task (e.g.; Reber & Lewis, 1977). However, the validity of the explicit memory tests used in these studies has been questioned by some authors. One of the criticisms concerns the fact that most of the explicit tests used do not fulfill the sensitivity and information criteria proposed by Shanks and St John (1994). The information criterion implies that the information accessed by the explicit test should be the same as the knowledge used by the subject in the implicit task; the sensitivity criteria means that the explicit test should be sensitive enough to the subject's conscious knowledge. Up to now, in the artificial grammar learning studies, there has been no agreement as to which memory test is better suited to assess the explicit knowledge reached by the subjects during the study phase. While some authors (e.g. Reber, 1989) consider that the subjects' inability to report verbally the rules (or some of the rules) of the grammar is sufficient to show that their knowledge is actually implicit, others argue that more sensitive measures have to be used. Actually, it is difficult to find any clear-cut evidence in favour of one or the other of the two opposite positions in the literature on implicit learning in normal subjects (even if some recent data seem to be promising; see, for example, the study of Dienes, Altmann, Kwan, & Goode, 1995, with the process dissociation procedure); this is notably due to the fact that no task can be considered as being process-pure and, therefore, the interpretation of the results is often ambiguous. For example, Perruchet and Pacteau (1990) have shown that a recognition task could be sensitive to the knowledge acquired by the subjects about the learning items. However, two kinds of mechanisms are thought to underlie recognition processes: (1) a conscious recollection mechanism; and (2) a mechanism based on familiarity, which could, at least partly, be underlain by implicit processes.
