How Incidental Sequence Learning Creates Reportable Knowledge: the Role of Unexpected Events
Total Page:16
File Type:pdf, Size:1020Kb
Journal of Experimental Psychology: Copyright 2008 by the American Psychological Association Learning, Memory, and Cognition 0278-7393/08/$12.00 DOI: 10.1037/a0012942 2008, Vol. 34, No. 5, 1011–1026 How Incidental Sequence Learning Creates Reportable Knowledge: The Role of Unexpected Events Dennis Ru¨nger Peter A. Frensch Berlin–Brandenburg Academy of Sciences and Humanities Humboldt-Universita¨t zu Berlin Research on incidental sequence learning typically is concerned with the characteristics of implicit or nonconscious learning. In this article, the authors aim to elucidate the cognitive mechanisms that contribute to the generation of explicit, reportable sequence knowledge. According to the unexpected- event hypothesis (P. A. Frensch, H. Haider, D. Ru¨nger, U. Neugebauer, S. Voigt, & J. Werg, 2003), individuals acquire reportable knowledge when they search for the cause of an experienced deviation from the expected task performance. The authors experimentally induced unexpected events by disrupt- ing the sequence learning process with a modified serial reaction time task and found that, unlike random transfer sequences, a systematic transfer sequence increased the availability of reportable sequence knowledge. The lack of a facilitative effect of random sequences is explained by the detrimental effect of random events on the presumed search process that generates reportable knowledge. This view is corroborated in a final experiment in which the facilitative effect of systematic transfer blocks is offset by a concurrent secondary task that was introduced to interfere with the search process during transfer. Keywords: sequence learning, serial reaction time task, reportable knowledge, explicit knowledge, unexpected events Individuals can learn about the sequential structure of environ- able to provide a verbal description of the sequential regularity mental events incidentally, that is, without prior intention to ap- after the training phase. This and related findings beget the ques- prehend a sequential regularity. For the most part, empirical re- tion of which cognitive mechanisms might be responsible for the search on sequence learning has been concerned with the generation of explicit knowledge during incidental sequence learn- characteristics of implicit learning. Sequence learning is said to be ing. However, researchers have only recently begun to approach implicit when it occurs in the absence of conscious or explicit this particular issue theoretically; pertinent empirical research is knowledge about the sequential regularity (cf. Erdelyi, 2004). still largely absent from the literature. Little theoretical value has been attached to the ubiquitous finding The purpose of this report is to test a central prediction of a that incidental sequence learning also creates explicit knowl- theoretical framework advanced by Frensch et al. (2003) to explain edge—at least in some participants. Moreover, a review of the the acquisition of explicit, reportable knowledge in incidental literature suggests that the generation of explicit sequence knowl- learning situations. The framework was dubbed the unexpected- edge varies systematically across experimental conditions (Fren- event hypothesis in reference to the central theoretical notion that sch et al., 2003). For example, Frensch, Lin, and Buchner (1998; unexpected events can trigger the generation of reportable knowl- Experiments 2a and 2b) manipulated the amount of training, the edge. Before we delineate the hypothesis in greater detail, we type of training (i.e., single or dual task training), and the type of review some alternative theoretical accounts that have been of- sequence to be learned with the serial reaction time (SRT) task fered in the literature. (Nissen & Bullemer, 1987). On each trial with the task, a partic- It is possible to identify two broad classes of theories on the ipant responds to a target that appears in one of several screen generation of explicit sequence knowledge—single-system and positions by pressing a spatially compatible response key. Re- multiple-system accounts. According to the single-system view, sponse locations on consecutive trials conform to a fixed sequence implicit and explicit knowledge are rooted in the same set of that is continuously recycled throughout the training phase. All learning mechanisms (e.g., Cleeremans, 2006; Cleeremans & three factors were found to influence whether participants were Jime´nez, 2002; Kinder & Shanks, 2003; Perruchet & Vinter, 2002; Shanks, Wilkinson, & Channon, 2003). In its most stringent form, the single-system view rejects the notion of separable knowledge bases altogether, that is, the distinction between implicit and Dennis Ru¨nger, Berlin–Brandenburg Academy of Sciences and Human- explicit knowledge. It is assumed that all markers of learning, be ities, Berlin, Germany; Peter A. Frensch, Department of Psychology, it faster responses to sequentially structured stimuli or verbal Humboldt-Universita¨t zu Berlin, Berlin, Germany. descriptions of the sequential regularity, provide different expres- This research was supported by Federal Ministry of Education and Research Grant 01GWS061. sions of the same underlying memory representations. Learning Correspondence concerning this article should be addressed to Dennis increases the quality of representations, which, in turn, leads to Ru¨nger, Berlin–Brandenburg Academy of Sciences and Humanities, Inter- improved performance in all available measures of learning (e.g., disciplinary Research Group “Functions of Consciousness,” Ja¨gerstr. 22/ Perruchet & Amorim, 1992; Perruchet, Bigand, & Benoit-Gonin, 23, D-10117 Berlin, Germany. E-mail: [email protected] 1997). 1011 1012 RU¨ NGER AND FRENSCH Weaker versions of the single-system view allow for a partial then might, or might not, lead to discovery and, subsequently, to decoupling of alternative measures of sequence learning. For ex- verbal report of the regularity (cf. Dienes & Perner, 1999). ample, according to Shanks and collaborators (Shanks, 2005; According to the unexpected-event hypothesis, the search for Shanks & Perruchet, 2002; Shanks et al., 2003), response time the cause of an observed unexpected event will likely lead to (RT) priming and recognition, two indicators of sequence learning, discovery of the regularity when the observed unexpected event is reflect different transformation processes of the same underlying related to the incidentally experienced regularity and when no knowledge representations. The transformation of memory repre- other immediately attributable causes for the unexpected event sentations into manual responses might be affected less by random exist. Consider, for example, a typical experiment with the SRT noise, for instance, than the transformation of the same represen- task. At some point during the training phase, the regular response tations into recognition judgments (e.g., Shanks et al., 2003). sequence is briefly replaced by random sequences, so that se- Therefore, it is conceivable that experimental manipulations that quence learning can be assessed by contrasting average RTs to affect the quality of memory representations influence different regular and random response locations. When viewed in light of measures of learning to different extents. In this way, empirical the unexpected-event hypothesis, this standard indirect test of dissociations may emerge that are typically interpreted as evidence sequence learning takes on a new significance: Over the course for the existence of separable implicit and explicit knowledge of training, implicit sequence learning improves task performance bases. by reducing both average RT and error rate. When the participant According to Cleeremans (2006; Cleeremans & Jime´nez, 2002) is transferred to random response locations, responses are likely to sequence learning is a mandatory consequence of performing a become slower and more error-prone due to the unpredictability of sequentially structured task such as the SRT task. Learning pro- target locations. We surmise that this sudden deterioration in task duces, over time, increasingly strong, stable, and distinct repre- performance, if registered by the participant, might constitute an sentations of the underlying sequential regularity. However, ensu- unexpected event that can trigger the generation of explicit se- ing high quality of a representation is not a sufficient condition for quence knowledge. conscious awareness of its content. For knowledge to be explicit, If interpolated random sequences can indeed function as a it has to be re-represented in a metarepresentation (Cleeremans, trigger for the generation of explicit knowledge, one should expect to find more explicit knowledge in this situation compared with an 2006; cf. Dienes & Perner, 1999). Cleeremans’ framework thus experimental condition in which sequence learning proceeds with- demarcates explicit and implicit knowledge according to the pres- out disruption. Preliminary support for this hypothesis comes from ence or absence of relevant metaknowledge. Nevertheless, the a study by Buchner, Steffens, Erdfelder, and Rothkegel (1997). framework qualifies as a single-system account of sequence learn- Buchner and colleagues used an auditory version of the SRT task ing because metarepresentations are produced by the same learn- in which participants discriminated between four different tones ing mechanisms in the same representational systems as their instead of target locations. Sequences of tones (and their