Quantitative and Qualitative Differences between Implicit and Explicit Sequence Learning

Introduction

Since the beginning of implicit learning research, the interpretation of available empirical evidence has flipped between a strong endorsement of an unconscious learning system and the denial of non-conscious acquisition

of new information (see, for instance, Reber, 1969; Wilkinson and Shanks, 2004). In this chapter, we would like to reflect on the origin of this contradiction and on the possible methodological and theoretical additions that can be made to the field in order to work this conflict out.

First of all, we would like to clarify the way in which we refer to the notion of implicit learning in this chapter. This notion can indeed be understood in different ways depending on whether one focuses on the processes or on the resulting knowledge involved during a learning episode. According to the first perspective, implicit learning refers to those situations in which some knowledge is acquired without intention to do so (unintentional learning) while the second meaning of the term relates to the acquisition of knowledge that is difficult to verbalize (unconscious learning). In this chapter, we will focus on this latter notion and discuss the conditions under which the knowledge acquired in sequence learning studies could or could not be described as unconscious.

Now that this clarification has been made, the next question arises: How can one demonstrate the acquisition of unconscious knowledge? Most studies have used a procedure consisting in (a) measuring learning in a first part of the experiment and (b) assessing conscious knowledge with another task in a second part of the experiment. In a sequence learning experiment, for instance, participants are first presented with a serial reaction time (SRT) task in which they have to press as fast as possible on a key corresponding to the spatial location of a visual target presented on a computer screen. Unknown to them, [Page 44]the sequence of locations follows a repeating pattern or some other sort of sequential regularities. Numerous studies have shown that participants can learn such a sequential pattern, as their reaction time (RT) improves more when the target follows the pattern rather than when it is random or when it follows a different pattern, in which case it increases dramatically (Cleeremans and McClelland, 1991; Reed and Johnson, 1994).

In a second phase, participants are informed of the existence of a sequential pattern and asked either to identify smaller sequences of the training material in a recognition task, or to reproduce this pattern in a generation task. These latter, direct tasks are used as measures of conscious knowledge. The rationale is that if participants are able to deploy their knowledge in these tasks, this knowledge must be described as conscious. This procedure is based on quantitative dissociation logic: implicit sequence learning would be demonstrated whenever a successful discrimination between regular and random trials is observed in the training phase while performance in generation or recognition remains at chance. Importantly, this logic depends on the exclusivity assumption (Reingold and Merikle, 1988), that is, the notion that the tasks used to measure awareness are only sensitive to conscious knowledge.

...