Can sequence learning be implicit? New evidence with the process dissociation procedure

Can we learn without awareness? Although this issue has been extensively explored through studies of implicit learning, there is currently no agreement about the extent to which knowledge can be acquired and projected onto performance in an unconscious way. The controversy, like that surrounding implicit memory, seems to be at least in part attributable to unquestioned acceptance of the unrealistic assumption that tasks are process-pure--that is, that a given task exclusively involves either implicit or explicit knowledge. Methods such as the process dissociation procedure (PDP, Jacoby, 1991) have been developed to overcome the conceptual limitations of the process purity assumption but have seldom been used in the context of implicit learning research. In this paper, we show how the PDP can be applied to a free generation task so as to disentangle explicit and implicit sequence learning. Our results indicate that subjects who are denied preparation to the next stimulus nevertheless exhibit knowledge of the sequence through their reaction time performance despite remaining unable (1) to project this knowledge in a recognition task and (2) to refrain from expressing their knowledge when specifically instructed to do so. These findings provide strong evidence that sequence learning can be unconscious.     

Effects of a secondary task on "implicit" sequence learning: learning or performance?

Traditionally, implicit learning has been defined in terms of a lack of awareness of the process and products of learning. In the face of a number of conceptual and empirical difficulties with this definition, it has recently been suggested instead that the critical feature of implicit learning is that it proceeds without making any demands on attentional resources. As disconfirmatory evidence for this, we describe the results of two experiments which each used a sequential reaction time task. With a tone-counting secondary task, measures of sequence learning were significantly affected by whether training occurred under single- or dual-task conditions, regardless of whether testing took place under single- or dual-task conditions.     

Is implicit sequence learning impaired in schizophrenia? A meta-analysis

Cognition in schizophrenia seems to be characterized by impaired performance on most tests of explicit or declarative learning contrasting with relatively intact performance on most tests of implicit or procedural learning. At the same time there have been conflicting results for studies that have used the Serial Reaction Time (SRT) task to examine implicit learning in people with schizophrenia. In the present research, we used meta-analysis to clarify whether or not people with schizophrenia show impaired performance on the SRT task. A systematic review found nine studies published in peer review journals that had each compared the performance of a group of people with schizophrenia with healthy controls on the standard SRT task or a variant of it. The resulting meta-analysis represented the responses of 205 participants with schizophrenia and 159 healthy controls on the SRT task. The analysis found that participants with schizophrenia perform less well than controls reflected by a pooled effect size of 0.51. A secondary analysis of all nine studies found that they all reported a point estimate of the change in reaction time between sequence and random trials that was greater for the controls. We conclude that there is a moderate impairment in implicit sequence learning among people with schizophrenia and speculate on the implications of this for understanding this disorder. Suggestions for improving the methodological quality and statistical reporting of studies of this topic are made.     

Is statistical learning a mechanism?

Philosophers of science have offered several definitions of mechanism, most of which have biological or neuroscientific roots. In this paper, I consider whether these definitions apply equally well to cognitive science. I examine this question by looking at the case of statistical learning, which has been called a domain-general learning mechanism in the cognitive scientific literature. I argue that statistical learning does not constitute a mechanism in the philosophical sense of the term. This conclusion points to significant limitations in the scope of the mechanist philosophy when it comes to accounting for explanation in cognitive science.     

What types of learning are enhanced by a cued recall test?

In two experiments, we investigated what types of learning benefit from a cued recall test. After initial exposure to a word pair (A+B), subjects experienced either an intervening cued recall test (A→?) with feedback, or a restudy presentation (A→B). The final test could be cued recall in the same (A→?) or opposite (?→B) direction, or free recall of just the cues (Recall As) or just the targets (Recall Bs). All final tests revealed a benefit for testing as opposed to restudying. Tests produced a direct benefit for information that was retrieved on the intervening test (B). This benefit also “spilled over” to facilitate recall of information that was present on the test but not retrieved (A). Both theoretical and practical implications are discussed.     

Is there a general task switching ability?

Participants were tested on two analogous task switching paradigms involving Shape/Size tasks and Vertical/Horizontal tasks, respectively, and three measures of psychometric intelligence, tapping fluid, crystallized and perceptual speed abilities. The paradigms produced similar patterns of group mean reaction times (RTs) and the vast majority of the participants showed switching cost (switch RT minus repeat RT), mixing cost (repeat RT minus single-task RT) and congruency effects. The shared intra-individual variance across paradigms and with psychometric intelligence served as criteria for general ability. Structural equations modeling indicated that switching cost with ample preparation ("residual cost") and mixing cost met these criteria. However, switching cost with little preparation and congruency effects were predominantly paradigm specific.     

What matters in implicit task sequence learning: perceptual stimulus features, task sets, or correlated streams of information?

Implicit task sequence learning may be attributed to learning the order of perceptual stimulus features associated with the task sequence, learning a series of automatic task set activations, or learning an integrated sequence that derives from 2 correlated streams of information. In the present study, our purpose was to distinguish among these 3 possibilities. In 4 separate experiments, we replicated and extended a previous study by Heuer, Schmidtke, and Kleinsorge (2001). The presence or absence of a sequence of tasks, as well as that of a sequence of different task-to-response mappings, was manipulated independently within experiments. Evidence of implicit sequence learning was found only when correlated sequences of tasks and mappings were present. No sequence learning effects were found when only a single task sequence or a single mapping sequence was present, even when the structure of the single sequence was identical to the structure of the integrated sequence of task-mapping combinations. These results suggest that implicit task sequence learning is not dependent on either perceptual learning of stimulus features or automatic task-set activation per se. Rather, it appears to be driven by correlated streams of information.     

Taking patterns for chunks: is there any evidence of chunk learning in continuous serial reaction-time tasks?

When exposed to a regular sequence, people learn to exploit its predictable structure. There have been two major ways of thinking about learning under these conditions: either as the acquisition of general statistical information about the transition probabilities displayed by the sequence or as a process of memorizing and using separate chunks that can later become progressively composed with extended practice. Even though chunk learning has been adopted by some theories of skill acquisition as their main building block, the evidence for chunk formation is scarce in some areas, and is especially so in the continuous serial reaction-time (SRT) task, which has become a major research tool in the study of implicit learning. This article presents a reappraisal, replication and extension of an experiment that stands so far as one of the few alleged demonstrations of chunk learning in the SRT task (Koch and Hoffmann, Psychological Res., 63:22–35, 2000). It shows that the effects which were taken as evidence for chunk learning can indeed be obtained before any systematic training and thus surely reflect a preexistent tendency rather than a learned outcome. Further analyses of the effects after extended practice confirm that this tendency remains essentially unchanged over continuous training unlike what could be expected from a chunk-based account of sequence learning.     

Does learning a complex task have to be complex? A study in learning decomposition

Many theories of skill acquisition have had considerable success in addressing the fine details of learning in relatively simple tasks, but can they scale up to complex tasks that are more typical of human learning in the real world? Some theories argue for scalability by making the implicit assumption that complex tasks consist of many smaller parts, which are learned according to basic learning principles. Surprisingly, there has been rather sparse empirical testing of this crucial assumption. In this article, we examine this assumption directly by decomposing the learning in the Kanfer-Ackerman Air-Traffic Controller Task (Ackerman, 1988) from the learning at the global level all the way down to the learning at the keystroke level. First, we reanalyze the data from Ackerman (1988) and show that the learning in this complex task does indeed reflect the learning of smaller parts at the keystroke level. Second, in a follow-up eye-tracking experiment, we show that a large portion of the learning at the keystroke level reflects the learning even at a lower, i.e., attentional level.     

Is research painting a biased picture of implicit learning? The dangers of methodological purity in scientific debate

The properties of implicit learning in natural settings are contrasted with those found in research. It is suggested that the search for pure cases and the necessity of finding features that clearly discriminate the two types of learning lead to bias in our estimation of the power of implicit processes. In more natural settings, such as face recognition, object perception, and natural language processing, implicit processes operate with flexibility and adapt to changes in environmental conditions. It is suggested that the search for pure cases of implicit processes has led to focusing on relevant but atypical examples of these processes. Additional research that emphasizes high levels of skill in control of complex systems may reveal greater adaptive power of implicit processes. However, such research may require less methodological purity and more emphasis on synthesis of theoretical ideas rather than analysis into pure cases.     

Does temporal contiguity moderate contingency learning in a speeded performance task?

In four experiments, we varied the time between the onset of distracting nonwords and target colour words in a word-word version of the colour-word contingency learning paradigm. Contingencies were created by pairing a distractor nonword more often with one target colour word than with other colour words. A contingency effect corresponds to faster responses to the target colour word on high-contingency trials (i.e., distractor nonword followed by the target colour word with which it appears most often) than on low-contingency trials (i.e., distractor nonword followed by a target colour word with which it appears only occasionally). Roughly equivalent-sized contingency effects were found at stimulus-onset asynchronies (SOAs) of 50, 250, and 450?ms in Experiment 1, and 50, 500, and 1,000?ms in Experiment 2. In Experiment 3, a contingency effect was observed at SOAs of -50, -200, and -350?ms. In Experiment 4, interstimulus interval (ISI) was varied along with SOA, and learning was equivalent for 200-, 700-, and 1,200-ms SOAs. Together, these experiments suggest that the distracting stimulus does not need to be presented in close temporal contiguity with the response to induce learning. Relations to past research on causal judgement and implications for further contingency learning research are discussed.     

Does temporal contiguity moderate contingency learning in a speeded performance task?

In four experiments, we varied the time between the onset of distracting nonwords and target colour words in a word–word version of the colour–word contingency learning paradigm. Contingencies were created by pairing a distractor nonword more often with one target colour word than with other colour words. A contingency effect corresponds to faster responses to the target colour word on high-contingency trials (i.e., distractor nonword followed by the target colour word with which it appears most often) than on low-contingency trials (i.e., distractor nonword followed by a target colour word with which it appears only occasionally). Roughly equivalent-sized contingency effects were found at stimulus-onset asynchronies (SOAs) of 50, 250, and 450 ms in Experiment 1, and 50, 500, and 1,000 ms in Experiment 2. In Experiment 3, a contingency effect was observed at SOAs of –50, –200, and –350 ms. In Experiment 4, interstimulus interval (ISI) was varied along with SOA, and learning was equivalent for 200-, 700-, and 1,200-ms SOAs. Together, these experiments suggest that the distracting stimulus does not need to be presented in close temporal contiguity with the response to induce learning. Relations to past research on causal judgement and implications for further contingency learning research are discussed.     

Why is learning to read a hard task for some children?

In this brief review of some of my research on reading disability, I argue that a child’s development into literacy has two major ontogenetic roots, one involving early informal literacy socialization and one related to phonological awareness. Although failure in reading acquisition might be a question of cultural deprivation, the dynamic interaction between genetic dispositions and environment must be acknowledged, especially in a society providing rich sources of print exposure. The crucial transition in literacy development involves a step from implicit to explicit control of the phonemic units of language. When the cipher of the alphabetic system is understood the child is equipped with a powerful instrument for self teaching. A primary source of reading disability is then related to problems of segmenting the speech stream into phonemic units. The poor functioning of the phonological module seems to be a circumscribed deficit not related to general cognitive ability. The causal and reciprocal relationship between phonological awareness and learning to read has been demonstrated in my research over 20 years. Some of these studies are briefly reviewed in this article. The preventive and remedial implications of the findings are rather obvious.     

Can expectancies produce placebo effects for implicit learning?

The placebo effect is an important phenomenon whereby real changes occur in response to an otherwise inert intervention. Despite increasing research attention, it remains unclear exactly which processes are amenable to placebo effects. The current study tested whether an instructional manipulation could produce placebo effects on a nonconscious cognitive task, namely implicit learning. Four hundred and sixty-four university students completed a visual search task while smelling an odor or no odor, in alternating blocks. Unknown to them, the task contained a contingency whereby on half the trials the target's location was cued by the pattern of distractors, which was achieved by repeating some configurations of targets and distractors. Prior to the task, participants received positive, negative, or no information about the odor's possible effects on performance. Those given positive information demonstrated faster reaction times on cued trials than other participants. Those given negative information showed slower reaction times on cued trials compared with participants given no information. Further, the cuing effect appeared to be nonconscious, with participants' ability to recognize the repeated configurations equivalent to chance and no evidence that performance on a recognition test was related to the magnitude of the cuing effect. This suggests that instructional manipulations can produce placebo effects on some nonconscious processes.     

Is language necessary for human spatial reorientation? Reconsidering evidence from dual task paradigms

Being able to reorient to the spatial environment after disorientation is a basic adaptive challenge. There is clear evidence that reorientation uses geometric information about the shape of the surrounding space. However, there has been controversy concerning whether use of geometry is a modular function, and whether use of features is dependent on human language. A key argument for the role of language comes from shadowing findings where adults engaged in a linguistic task during reorientation ignored a colored wall feature and only used geometric information to reorient [Hermer-Vazquez, L., Spelke, E., & Katsnelson, A. (1999). Sources of flexibility in human cognition: Dual task studies of space and language. Cognitive Psychology, 39, 3-36]. We report three studies showing: (a) that the results of Hermer-Vazques et al. [Hermer-Vazquez, L., Spelke, E., & Katsnelson, A. (1999). Sources of flexibility in human cognition: Dual task studies of space and language. Cognitive Psychology, 39, 3-36] are obtained in incidental learning but not with explicit instructions, (b) that a spatial task impedes use of features at least as much as a verbal shadowing task, and (c) that neither secondary task impedes use of features in a room larger than that used by Hermer-Vazquez et al. These results suggest that language is not necessary for successful use of features in reorientation. In fact, whether or not there is an encapsulated geometric module is currently unsettled. The current findings support an alternative to modularity; the adaptive combination view hypothesizes that geometric and featural information are utilized in varying degrees, dependent upon the certainty and variance with which the two kinds of information are encoded, along with their salience and perceived usefulness.     

How well do elderly people cope with uncertainty in a learning task?

The relation between age, task complexity and learning performance in a Multiple Cue Probability Learning task was studied by systematically varying the level of uncertainty present in the task, keeping constant the direction of relationships. Four age groups were constituted: young adults (mean age = 21), middle-aged adults (45), elderly people (69) and very elderly people (81). Five uncertainty levels were considered: predictability = 0.96, 0.80, 0.64, 0.48, and 0.32. All relationships involved were direct ones. A strong effect of uncertainty on 'control', a measure of the subject's consistency with respect to a linear model, was found. This effect was essentially a linear one. To each decrement in predictability of the task corresponded an equal decrement in participants' level of control. This level of decrement was the same, regardless of the age of the participant. It can be concluded that elderly people cope with uncertainty in probability learning tasks as well as young adults.     

An old problem: how can we distinguish between conscious and unconscious knowledge acquired in an implicit learning task?

A long lasting debate in the field of implicit learning is whether participants can learn without acquiring conscious knowledge. One crucial problem is that no clear criterion exists allowing to identify participants who possess explicit knowledge. Here, we propose a method to diagnose during a serial reaction time task those participants who acquire conscious knowledge. We first validated this method by using Stroop-like material during training. Then we assessed participants’ knowledge with the Inclusion/Exclusion task (Experiment 1) and the wagering task (Experiment 2). Both experiments confirmed that for participants diagnosed as having acquired conscious knowledge about the underlying sequence the Stroop congruency effect disappeared, whereas for participants not diagnosed as possessing conscious knowledge it only slightly decreased. In addition, both experiments revealed that only participants diagnosed as conscious were able to strategically use their acquired knowledge. Thus, our method allows to reliably distinguish between participants with and without conscious knowledge.