Introduction

People learn a large variety of regularities in the environment during life. Many of these regularities are learned unintentionally, often without noticing that something has been learned. Motor reactions can be taken as examples: balancing, catching flying objects, etc. These actions are very complex from a computational point of view, but such «computations» are implemented outside of conscious awareness. Some higher-order skills which are learned unintentionally are language acquisition (Cleeremans, Destrebecqz, & Boyer, 1998; Perruchet, 2008; Winter & Reber, 1994) and expert knowledge (Berry & Dienes, 1993; Singley & Anderson, 1989). Such unintentional learning of complex regularities is called implicit learning.

The phenomenon of implicit learning has been studied since the mid-1960s, but a number of questions remain unanswered. The issue of the awareness of the tacit knowledge and its application are among these questions. It was accepted in the classical theories of implicit learning that the information about hidden regularities is acquired automatically, remains unconscious and manifests itself in behavior uncontrollably. However, more recent studies have shown that the knowledge acquired in standard experiments on implicit learning can actually become conscious: one of the classic books on the topic is even called How implicit is implicit learning? (Berry, 1997). It was also found that the manifestation of implicit knowledge in behavior can be mediated by other mental processes such as nonspecific subjective feelings.

Thus, since the 1970s, implicit learning theorists have discussed the nature of the phenomenon, taking virtually opposite positions. Decades of empirical studies have not led to a clear choice in favor of one of them. This fact suggests that radically different descriptions of implicit learning can refer to different phenomena, perhaps to different forms of learning. This paper will review the main theoretical positions, as well as supporting empirical data. The aim is to find out whether it is possible to carry the positions of the theorists not to one but to several types of learning, and thus treat them not as competing but as complementary descriptions.

In the description of the theoretical approaches to implicit learning, I will use some of the well-established terms in this field of study. Since the meaning of a term varies depending on the approach, some approximate definitions are provided which will be maintained throughout the paper. The representation of an object refers to someone’s mental state, allowing for a response to the object accordingly (differently from the responses to other objects for which there are different representations). Knowledge refers to the set of representations related to a specific situation. Knowledge can represent objects of the outer and inner worlds of the person and also to the relationships between objects. Learning relates to the acquisition of new knowledge. In behavioral terms, learning means that an individual starts to react in a new fashion to some class of objects. Thus, learning always results in acquired knowledge and is tested through the measurement of knowledge. This paper will primarily discuss knowledge, and the ways in which different learning processes lead to different kinds of knowledge. Awareness refers to a characteristic of knowledge: its subjective presentation to a person. Such knowledge may be verbalized and transferred to another person. Automaticity refers to the expression of some knowledge in behavior, without a persons conscious decision-making. That is, before the behavioral act, there is no conscious knowledge in a persons mind which can be formulated verbally as: «I possess some knowledge and now I will use it in my behavior». In contrast to automatic behavior, controlled behavior follows conscious decisions in the application of some knowledge.

The current paper consists of three parts. The first section will be an overview of theoretical approaches to implicit learning. In the second section, empirical evidence in support of these positions will be provided. The third section will show how different approaches and disparate experimental results can be considered in a consistent taxonomy of learning processes.

1. Theories of Implicit Learning

Four theoretical approaches to implicit learning will be considered in the paper. They are subdivided depending on how they answer two questions: 1) How is knowledge acquired during learning: consciously or unconsciously?; and 2) How is it applied in behavior: automatically or voluntarily (controlled)? These two questions differentiate existing scientific camps. Thus, a total of four approaches will be overviewed:

• automatic application of unconscious knowledge (completely unconscious learning);


• controlled application of conscious knowledge (completely conscious learning);


• controlled application of unconscious knowledge; and


• automatic application of conscious knowledge.

Automatic Application of Unconscious Knowledge

Reber can be called a pioneer of implicit learning research, as he published the first papers on artificial grammar learning in the 1960s (Reber, 1967). Rebers classical experiment consisted of two phases: learning and testing. In the first phase, Reber presented participants with about twenty letter strings constructed on the basis of certain rules (see Fig. 1). These rules determined the possible letter orders in the string. Participants were instructed to memorize the presented strings. After memorization, participants were informed that the strings they had seen were constructed according to some set of rules (artificial grammar), and that they would then be presented with new strings, which they should decide were consistent with the rules («grammatical» strings) or not («nongrammatical» strings). Participants usually responded correctly in 60-70% of trials (chance level was 50%). At the same time, they appeared to be unable to verbalize the rules of the grammar. On the basis of his first results, Reber concluded that people unconsciously learned the abstract structure of the artificial grammar, which allowed them to classify new strings successfully. Thus, acquired unconscious knowledge drives behavior automatically.

Rebers research was followed by the development of new experimental paradigms, which provided data similar to Rebers. Among them, dynamic systems control (Berry & Broadbent, 1984), sequence learning (Nissen & Bullemer, 1987) and perceptual learning (Lewicki, Czyzewska, & Hoffman, 1987) stand out.

The pioneers of implicit learning research suggested that it is supported by a powerful cognitive system separated from consciousness. Reber wrote about a phylogenetically older cognitive structure which is more robust, resistant to injuries and exhibits less individual variation than conscious information processing (Reber, 1993). In agreement with Reber, Lewicki also stressed that the results of unconscious processing are fundamentally unavailable to consciousness (Lewicki, Hill, & Czyzewska, 1997). Thus, these researchers believed that implicit knowledge is acquired unintentionally, is unconscious and drives behavior automatically.

A similar position is taken by some modern theorists. For example, Cleeremans believes that implicit learning is unavailable to consciousness and manifests itself in behavior automatically (Cleeremans, 2011). Cleeremans suggests that implicit learning takes place only in the initial stages of learning and its role in cognitive activity is much more modest than in the theories of Reber and Lewicki.

The theory of Ashby and his collaborators suggests the presence of two independent competitive cognitive structures; the theory is referred to as COVIS: Competition Between Verbal and Implicit Systems (Ashby, Alfonso- Reese, Türken, & Waldron, 1998). An explicit system is controlled by the person and effectively performs tasks associated with simple logical rules whenever objects can be classified by one clear basis. Meanwhile, an implicit system operates automatically and performs integrative analysis of complex material with several interconnected bases for object classification.

Implicit learning can therefore play different roles in the described approaches. It is a work of the old and powerful cognitive system in Rebers and Lewickis theories. In Ashby’s approach, it is a process competing with consciousness. According to Cleeremans, implicit learning is an initial stage of the learning process. However, all these approaches describe implicit learning as being completely unconscious with its manifestation in behavior considered to be fully automatic. In contrast to such a radical position, in the 1980s and 1990s a group of theorists presented an opposite perspective, according to which implicit learning is not different from any other type of learning, being completely conscious and controlled.

Controlled Application of Conscious Knowledge

A full-scale experimental, methodological and theoretical attack on the classical approach to implicit learning was launched in the 1990s. Both the notion of abstractness and the unawareness of knowledge acquired in experiments were challenged (Shanks & St. John, 1994). A group of theorists emerged who believed that all knowledge acquired by a person is available to consciousness (Dulany, 1997; Perruchet & Vinter, 2002). According to these researchers, representations arise only in consciousness, although they are supported by neural mechanisms that are themselves unconscious. Within this mentalistic framework, operations with representations (association links, use of representations in purposeful reasoning, etc.) can only take place with the participation of consciousness. Some rough adaptations to environmental regularities are possible in the nervous system, but they are not related to mind and, accordingly, to the problem of consciousness and learning (see also Dulany, Carlson & Dewey, 1984; Shanks, Wilkinson & Channon, 2003).

The theorists who hold such a position do not deny the experimental facts of the acquisition of tacit knowledge, which is difficult to verbalize, and offer their own explanation of how such knowledge manifests itself in behavior. Nonetheless, a common feature in the approaches of these researchers is the idea that knowledge is acquired consciously, it is used under the control of consciousness, and there are no other alternative ways of information processing, such as unconscious (implicit) learning.

Controlled Application of Unconscious Knowledge

Many investigators’ points of view about implicit learning can be placed along the continuum that stretches from «completely unconscious» to «completely conscious». For example, some scholars paid attention to the fact that sometimes the application of implicit knowledge is mediated by conscious decisions. Mangan developed James’ idea of overtones, suggesting that such consciously experienced but hard-to-verbalise representations make it possible to obtain a generalized evaluation of the presence of some implicit knowledge relevant to a performed activity (Mangan, 2003). Price and Norman argue that the use of implicit knowledge is mediated by subjective experience, on which decisions are based (Price & Norman, 2008). According to Dienes and his colleagues, implicit learning is accompanied by the arising feeling of familiarity, which one learns to recognize and rely on when making a decision (Dienes, 2012; Scott & Dienes, 2008). Thus, representatives of this position do not deny the main point of the classical approach — the unconsciousness of learning process. However, they believe that one is not completely ignorant and still knows that she has learned something and can try to purposefully apply the hard-to-verbalize knowledge. Within such frameworks, it is assumed that the functional role of the access to implicit knowledge (albeit indirect) consists of the ability to control it (Price & Norman, 2008; Mangan, 2003; Koriat, 2007).

Automatic Application of Conscious Knowledge

One more «intermediate» approach was developed by Whittlesea and colleagues (Whittlesea & Dorken, 1997). They proposed the consideration of an example of well-developed skill application in unexpected situations. Imagine a person who is engaged in fencing. She slowly, explicitly learns the basic movements until she starts to execute them quickly and accurately. When she subsequently goes to a dance school, it may happen that she learns tango movements much faster than other students. She may not notice it, but a detailed analysis of her movements will show that she applies the skills consciously developed in the fencing school. Whittlesea believed that implicit learning works in a similar way. That is, in the first stage of artificial grammar learning one acquires some knowledge of a grammatical structure performing strings memorization. In the second stage, she is asked to perform a classification task, which, in her opinion, is not related to previous one, but the accuracy of string dassification is at the above chance level. A similar position was held in the earlier works of Dienes, whereby implicit knowledge was described in the context of «higher-order thoughts» theory (Dienes & Pemer, 2001). According to this approach, the representation A is conscious only when we have another representation B, whose content is the representation A. The perception of a red ball (representation A) is conscious only if we have the representation «I see a red ball» (representation B). According to that theory, implicit knowledge can be described as a representation which has no higher-order representation. It can be manifested in behavior, but we cannot report its content, as we simply do not know that it exists.

According to Bargh, behavioral priming (an effect of supraliminal, conscious stimuli on subsequent behavior of a person) works in the same way. In his opinion, it is a completely automatic process that does not involve any conscious decision making. Rather, the perception of meaning related to some social stereotype automatically launches a behavioral response (Bargh, 1994).

2. Empirical Data for Described Approaches

This section reviews the experimental evidence for each of the four listed approaches. An important moment in the history of implicit learning studies was the recognition of the fact that knowledge, acquired during the standard implicit learning procedure, is not one-dimensional and pure. People always demonstrate both implicit and explicit knowledge. Proponents of completely unconscious implicit learning, recognizing this fact, tried to prove that although people do have some amount of explicit knowledge, completely unconscious elements play a crucial role. Proponents of the notion of exclusively explicit learning tried to prove that tests for awareness detection are not sensitive enough. Some empirical results fit well with multiple theoretical approaches, so the data of several authors will be provided in different subsections.

Automatic Application of Unconscious Knowledge

Reber himself conducted a number of experiments that tested his initial hypotheses related to implicit learning. In his 1976 paper, it was shown that implicit knowledge manifests itself regardless of the individual’s purposes (Reber, 1976). Furthermore, if people are informed in the beginning of the training phase that the stimuli follow certain rules, they not only fail to improve their accuracy but they perform the test classification significantly worse. Hayes and Broadbent (1988) found a similar effect in a dynamic system control task. Participants learned to predict the behavior of an interactive system with complex interaction of some variables. For example, they had to hold the amount of production of a simulated factory at a given level, which depends on the input data (the number of employed workers) in a complex way. When participants were asked to explain their decisions, their performance decreased. Similar effects were obtained in the domain of problem solving (Ponomarev, 1976) and in social perception studies (Belova, 2004). Belova called this phenomenon «the verbalization effect».

Sequence learning (e.g., serial reaction time task) along with artificial grammar learning are among the standard laboratory methods for studying implicit learning. A typical experiment is as follows: there are several locations on a screen in which a target can appear (see Fig. 2). The task is to respond to its appearance as quickly as possible by pressing the key corresponding to the position in which the target appeared. Unbeknownst to participants, the sequence of target locations follows a certain regular pattern. Learning is manifested in the fact that participants respond faster and faster over time. If the regularity suddenly appears to be broken, a sharp slowdown in reaction times is observed. However, participants cannot report this regularity, and they often do not even notice its presence.

Destrebecqz and Cleeremans asked individuals who had participated in a serial reaction time task to generate the sequence of locations, first in a way which was consistent with the experimental regularity, and then in a way which violated it (Destrebecqz & Cleeremans, 2001). It was found that even under conditions when participants had to violate the sequence, they often unintentionally generated consistent sequences. That is, they were unable to control the expression of their implicit knowledge. Higham and colleagues presented participants at first with stimuli constructed on the basis of one grammar, and then with stimuli based on another grammar (Higham, Vokey, & Pritchard, 2000). In the test phase, participants had to mark the strings which were consistent with only one of these grammars (target grammar). They performed the task at above chance level, but in cases of mistakes they chose the strings of the non-target grammar significantly more often than completely nongrammatical strings. Thus, it was shown that people possess both controlled knowledge, which they can apply according to the experimenters instructions, and uncontrolled knowledge (of the second grammar), which is manifested against their will.

A number of researchers proposed that the inability to discriminate between correct and incorrect responses can indicate unawareness of applied knowledge (Dienes & Ferner, 2002). The ability to successfully monitor our own mental processes is usually called metacognitive sensitivity (Fleming & Lau, 2014), and this phenomenon has been important in the investigation of the nature of implicit learning. Chan asked participants to rate their confidence in their answers in every trial of a test phase in the artificial grammar learning experiment (Chan, 1992). Classification accuracy and confidence ratings did not correlate: on average, the participants were equally confident in their correct as well as their incorrect answers. Thus, the conclusion was made that people do not know when they answer correctly and when they do not, and therefore they apply their knowledge unconsciously. Today it is called «zero-correlation criterion». A lack of metacognitive sensitivity in artificial grammar learning was obtained in a number of subsequent works (Dienes, Altmann, Kwan, & Goode, 1995; Dienes & Altmann, 1997; Zizak & Reber, 2004). Dienes and colleagues (Dienes et al., 1995) asked participants to rate their confidence in every trial and then analyzed only those trials in which confidence was at the zero level; in these trials, participants believed that their likelihood of guessing correctly was not better than the flip of a coin. It turned out that responses in these trials occurred correctly at above chance level, which indicates the influence of implicit knowledge about which participants are unaware. This is usually referred as «guessing criterion».

Participants with memory disorders classify strings with the same accuracy as intact participants, but they are unable to indicate the letter combinations from the learning phase in a special recognition test (Knowlton, Squire, Paulsen, Swerdlow, & Swenson, 1996). Intact participants performed this task successfully, showing the presence of explicit knowledge. A similar result was obtained in a sequence learning paradigm (Reber & Squire, 1998). Preserved implicit learning was repeatedly demonstrated in patients with Alzheimer’s disease (Smith, Siegelt, & McDowall, 2001; Peigneux, Meulemans, Van Der Linden, Salmon, & Petit, 1999; Reber & Squire, 1999).

The ability to acquire unconscious knowledge and to apply it in behavior is evident from a large amount of experimental data. Most often, it demonstrates independence of implicit learning from the conscious processing system or even contradictions between them. It was shown in the different tasks that intentional conscious search of implicit regularities can lead to decreased performance (Reber, 1976; Hayes & Broadbent, 1988; Ponomarev, 1976; Belova, 2004). People often cannot control the application of implicit knowledge: it is expressed even if it is prohibited by the task instruction (Destrebecqz & Cleeremans, 2001; Higham et al., 2000). Researchers explain this based on the assumption that there is no access to implicit knowledge, and therefore no control. People cannot distinguish between their correct and incorrect responses (Chan, 1992; Dienes & Altmann, 1997; Zizak & Reber, 2004; Dienes et al., 1995). It is also consistent with this interpretation: the lack of access to the implicit processing system makes it impossible to evaluate its performance. Finally, clinical data show that people with severe memory disorders are able to learn implicitly (Knowlton et al., 1996; Reber & Squire, 1998; Reber & Squire, 1999; Smith, Siegert, & McDowall, 2001; Peigneux, Meulemans, Van Der Linden, Salmon, & Petit, 1999). This suggests that implicit learning is not connected with conscious processing. In the next subsection, we will look at the opponents of the classical approach to implicit learning and the empirical data they provide.

Controlled Application of Conscious Knowledge

Dulany argued that people do not learn any complex abstract structures during learning: they form some explicit rules that partially match the rules of the artificial grammar. In Dulany and colleagues’ experiments, participants pointed to the fragments of test stimuli that made them grammatical or nongrammatical (Dulany, Carlson, & Dewey, 1984). It was found that participants form their own rules on the presence or absence of some elements, they follow these rules, and it leads to above-chance performance. Perruchet and Pacteau, in a different experiment, showed that it is enough to simply memorize some letter combinations from the learning phase to classify test strings at above the chance level (Perruchet & Pacteau, 1990).

Shanks and his colleagues showed in a number of studies that confidence in answers does correlate with accuracy (for example, if a binary scale is used instead of a continuous one), claiming that it proves awareness during the learning process (Tunney & Shanks, 2003; Tunney, 2005).

The results of Destrebecqz and Cleeremans on the automaticity of implicit knowledge application (2001) were not replicated in other studies: participants were able not to follow the learned regularity when asked to do so (Norman, Price, & Duff, 2006; Wilkinson & Shanks, 2004).

A separate line of studies has been dedicated to demonstrating that dissociations between learning and other measures, presumably associated with consciousness (recognition, for instance), are not always manifested and can be explained by the work of a single cognitive system. For example, Shanks and his associates successfully modeled these dissociations in a sequence learning paradigm with a single-system computational model (Shanks, Wilkinson, & Channon, 2003).

The critics of completely unconscious and automatic learning often followed their opponents, suggesting alternative explanations, pointing to experimental designs flaws or demonstrating the non-replicability of results. It was shown that participants can explicitly learn not the entire grammar but its fragments (Dulany et al., 1984) or short letter combinations (Perruchet & Pacteau, 1990), which provides the level of accuracy observed in the experiments. In some conditions, the correlation between accuracy and confidence ratings still occurs (Tunney & Shanks, 2003; Tunney, 2005). The results demonstrating the inability to ignore existent implicit knowledge have not been replicated in later studies (Norman et al., 2006; Wilkinson & Shanks, 2004). The dissociations between learning and, for example, recall can be modeled by a single-system computational models (Shanks et al., 2003), which indicates the redundancy of describing the additional implicit block in the cognitive system. In the following subsection, the data supporting the intermediate approaches will be provided.

Controlled Application of Unconscious Knowledge

The representatives of the controlled application of unconscious knowledge framework have tried to prove that people are not aware of the content of their implicit knowledge, but they have indirect access to it. The aim of these research
ers was to detect the conscious markers of such access and to show that they allow for controlling implicit knowledge and applying it in accordance with a persons conscious objectives. As mentioned above, a number of studies have shown that, during artificial grammar learning, participants often know when they are right or wrong (Tunney & Shanks, 2003; Tunney, 2005). Such a result was obtained and repeatedly replicated by the pioneers of the method (Dienes & Berry, 1997; Scott & Dienes, 2008).

The correlation between accuracy and confidence can indicate that relevant implicit knowledge is somehow reflected in a persons consciousness. There are a number of opinions on what kind of experience this refers to: fluency, a feeling of familiarity, pleasantness, etc. These are explored individually below.

The experience of processing fluency occurs when someone once again perceives certain stimuli, or when stimuli are perceived which are similar to those that were perceived earlier (Jacoby & Dallas, 1981). In implicit learning studies, Buchner (1994) showed that stimuli consistent with learned implicit regularity are read from the screen faster. Kinder and colleagues demonstrated that the faster a noisy stimulus is detected, the more likely it will be called grammatical (Kinder, Shanks, Cock, & Tunney, 2003).

The mere exposure effect refers to the fact that objects that were presented to a participant several times are evaluated as more pleasant than the objects that were presented for the first time (Bomstein, 1989). Gordon and Holyoak (1983) obtained the structural mere exposure effect: new stimuli with a structure similar to those presented earlier were liked more than new stimuli that were inconsistent with the original structure. This result was replicated in a number of subsequent works (Newell & Bright, 2001; Zizak & Reber, 2004).

In an experiment of Scott and Dienes (2008) participants in a test phase rated the familiarity of the stimuli (in relation to the learning phase), and classified them as grammatical or not. The correlation of familiarity ratings and grammaticality judgments was .68. Furthermore, it turned out that in the initial stages of the experiment, the feeling of familiarity correlated with classification decisions (the higher the feeling of familiarity, the more likely the string is classified as grammatical); however, at the level of verbal reports, participants were rarely aware of the fact that they rely on the feeling of familiarity. They realized it more and more as they passed more experimental trials.

In Norman and colleagues’ experiment, participants performed a simple sequence learning task (Norman, Price, Duff, & Mentzoni, 2007). At the end of the experiment, participants were unable to verbally report the regularity that determined stimuli locations, but they could accurately predict the position of the next stimulus in a special test. The authors suggested that such behavior is possible due to the presence of access to existent implicit knowledge through conscious subjective experiences which participants were unable to verbalize; the authors call it «fringe feelings» after James and Mangan.

The controlled application of implicit knowledge was demonstrated in studies where participants learned several grammars and then had to purposefully use their knowledge of only one of them. In a two-grammar design (as mentioned in the study of Higham), Dienes and colleagues showed that participants could intentionally choose strings which were consistent with only one of the learned grammars (Dienes et al., 1995).

In the same paper, Dienes and colleagues demonstrated that if participants do not know about the relationship between the learning and test phases, classification accuracy decreases to a chance level (Dienes et al., 1995, Experiment 5).

A number of studies (especially of the sequence learning paradigm) explored the role of attention in implicit learning. Researchers asked participants to perform some concurrent task that distracts attention from the main task. It could be random number generation, counting musical tones, reversed counting, etc. Some studies showed that attention is needed not for the acquisition of implicit knowledge, but for its application (Frensch, Lin, & Buchner, 1998; Jiang & Leung, 2005). That is, when attention is distracted in the test phase, the application of implicit knowledge is impaired. Distracting attention in the learning phase did not have the same effect.

In accordance with theorists arguing for completely conscious learning, proponents of the idea of controlled application of unconscious knowledge point to the fact that in most cases there is still a correlation between classification accuracy and confidence ratings in artificial grammar learning (Dienes & Berry, 1997; Scott & Dienes, 2008). However, the views on the basis of this correlation differ between the two approaches. Proponents of completely conscious learning say that this basis is the conscious knowledge of a learned regularity. Proponents of controlled application of unconscious knowledge say that these are some indirect signals from unconscious knowledge. These signals can have various forms. It was shown that participants can rely on the experienced fluency (Buchner, 1994; Kinder et al., 2003), on the pleasantness of stimuli (Gordon & Holyoak, 1983; Newell & Bright, 2001; Zizak & Reber, 2004), and on the feeling of familiarity (Scott & Dienes, 2008). At the same time, reliance on such fringe feelings can grow within the course of the experiment (Scott & Dienes, 2008). It was demonstrated that the ability to make predictions on the basis of learned regularity is not related to the awareness of it (Norman et al., 2007).

Another similarity of the representatives of these two approaches is that they emphasize the ability to arbitrarily control the application of the learned regularities (Dienes et al., 1995); however, Dienes and colleagues emphasize the unconsciousness of the grammar knowledge. The reliance on attentional resources for successful test performance is also consistent with the idea of controlled application of implicit knowledge (Frensch et al., 1998; Jiang & Leung, 2005). Also, the absence of any effects of attentional deficits in the learning phase indicates an unconscious learning process (but the issue is still disputable; see Jimenez & Vazquez, 2005).

The verbalization effect described in the previous section is consistent with the idea of two processing systems: implicit and explicit. It is likely that it can be associated with approaches which assume that people rely on fringe feelings; the need to verbalize their decision may interfere with other bases for decision-making. However, the proponents of these approaches have not addressed this issue.

Automatic Application of Conscious Knowledge

Ponomarev described an effect very similar to implicit learning (Ponomarev, 1960, 1976). He studied how the introduction of a hint during the different stages of a creative task solution affects participants’ behavior. In one of his experiments, participants were given the «polytypic panel» task. They had to put a set of bars on the panel according to certain rules. Participants easily solved this task and then they were given another task — a maze. The optimal path in the maze repeated the shape of the final locations of the bars in the «polytypic panel» task. Ponomarev found that in normal conditions, participants made on average 70-80 false turns passing the maze, but after solving the «polytypic panel» they made less than 10 errors. At the same time, the verbalization effect came to play: if subjects were required to explain their decisions in a maze, the number of errors increased dramatically (Ponomarev, 1976). The locations of bars in the first task were completely explicit for participants, but they did not realize that this experience affected performance in the next task.

Suprathreshold priming data can also give support to this approach. In social psychology, several studies were conducted in which consciously-perceived stimuli affected peoples behavior without their intention. For example, if participants were presented with some form of words semantically related to old age, they started to walk slower (Bargh, Chen, & Burrows, 1996). Words associated with the concept of «library» made people behave more quietly (Aarts & Dijksterhuis, 2003), and primes associated with «hostility» made people more aggressive (Carver, Ganellen, Froming & Chambers, 1983). Bargh argues that such behavior is not mediated by any deliberate decision making, since in most cases measurable behavior is observed when people think the experiment is over. Such behavior can be controlled only if a person knows about the experimental influences, and attentional resources are needed to overcome the learned stereotype. Attempts to overcome the learned response in attentional load can lead to more frequent execution of this response than with the instruction not to execute it (Wegner, 1994).

Thus, it was shown that although the knowledge that determines behavior in a given situation can be conscious, the very fact of its influence may not be realized by the person and may be unrelated to his or her intentions. Most of the effects of behavioral priming were obtained in the situation where participants did not know that they took part in the experiment or did not know what behavior was measured (Bargh et al., 1996; Aarts & Dijksterhuis, 2003; Carver et al., 1983). The intervention of consciousness in this process can decrease performance (Ponomarev, 1976). Such learning is hardly controlled, especially in a condition of attentional load (Wegner, 1994).

Summary of Empirical Results

The existent data does not allow us to make a final conclusion in favor of one of the alternative approaches. However, we cannot say that the efforts of researchers have been fruitless. At the moment, most authors agree that human knowledge and learning has many manifestations, and some of the factors that determine the occurrence of various processes of learning are known. Reber and other early investigators of implicit learning showed that individuals cannot report the rules of an artificial grammar (Reber, 1967,1989). Dulany and colleagues demonstrated that tests not requiring verbalization indicate conscious knowledge of some fragments of the grammar (Dulany et al., 1984). Perruchet and Pacteau showed that the forms of acquired implicit knowledge can be fundamentally different: the classification of whole strings in the test phase of an experiment can be accounted for by conscious learning of stimuli fragments in the learning phase (Perruchet & Pacteau, 1990). However, this effect is not always manifested and it has been suggested that this result is due to the specific selection of stimulus material (Gomez & Schvaneveldt, 1994). It also does not explain the ability of participants to classify strings correctly in the transfer experiment (Dienes & Altmann, 1997). Transfer experiments involve a standard learning procedure, with letter strings that follow a set of grammatical rules. However, during the test phase, stimuli are composed from a new set of letters which follow the same grammar rules.

The purposeful application of implicit knowledge still remainsatopicofdiscussion.Inalmostidentical experiments, Higham (Higham et al., 2000) and Dienes (Dienes et al., 1995) obtained different results. In Higham’s experiment, participants chose strings of a «forbidden» grammar more often than completely nongrammatical strings (which indicates the automatic application of implicit knowledge), but this effect was not found in Dienes and colleagues’ experiment. However, there were some differences in the procedures and stimuli between these two experiments (see discussion in Higham’s paper).

Participants classification accuracy correlates with response confidence, indicating the conscious application of implicit knowledge (Dienes et al., 1995; Tunney & Shanks, 2003; Scott & Dienes, 2008), although it does not always appear (Chan, 1992). In transfer experiments, for example, such correlation usually disappears (Dienes & Altmann, 1997).

Confidence ratings are presumably related to other subjective experiences which arise in learning situations: three examples are an experience of fluency (Buchner, 1994; Kinder et al., 2003), pleasantness (Gordon & Holyoak, 1983; Zizak & Reber, 2004), or familiarity of the stimuli (Scott & Dienes, 2008). These phenomena appear to be equally sensitive to the same experimental conditions. For example, if some stimuli unfamiliar to participants were used, neither a structural mere exposure effect nor a confidence-accuracy correlation were observed (Zizak & Reber, 2004). Neither effects were obtained in transfer experiments (Newell & Bright, 2001). Scott and Dienes found a correlation between familiarity of stimuli and confidence in classification of them (Scott & Dienes, 2008). This suggests that a feeling of familiarity, a structural mere exposure effect and metacognitive sensitivity may have a common source.

Some experiments are not replicable. For example, the data on the expression of implicit knowledge contrary to the task instruction (Destrebecqz & Cleeremans, 2001) was not replicated in several studies (Norman et al., 2006; Wilkinson & Shanks, 2004). Bargh’s experiment is also criticized for poor replicability (see for example Doyen, Klein, Pichon, & Cleeremans, 2012) as well as other experiments on behavior priming (see Newell & Shanks, 2014 for a review).

A large amount of data has been accumulated that require some generalization. As the four described approaches to implicit learning are supported by a solid amount of empirical data, a new framework had to be elaborated which describes learning mechanisms and the conditions in which they can result in the behaviors that are captured in experiments. Dienes and colleagues laid a foundation for such a framework, which will be discussed in the following section.

3. Integrative Approach by Dienes and Scott

Dienes and ferner (2002), and then Dienes and Scott (2005) divided knowledge into two types according to its content. After interaction with an environment, people can learn some things about the relationships between objects in that environment: a) the structure of relations between objects; and b) whether a new situation is consistent with this structure. The first type of knowledge can be verbalized and allows an individual to share acquired experience with another person. The second type of knowledge allows a person to evaluate a new situation correctly. Possessing the second type of knowledge, one can say WHAT is the case (to classify correctly the situation according to whether objects in this situation correspond to the learned structure). Possessing the first type of knowledge, one can explain WHY she classified the situation one way or another. Dienes and Scott call the first type of knowledge (the knowledge of the structure of learned material) structural knowledge, and the second type (the knowledge of whether a new situation is consistent with this structure) judgment knowledge.

It seems obvious that judgment knowledge cannot exist without structural knowledge: if one does not know the basis of classification, how can she classify it accurately? However, there is one more factor which renders the relationship between the two types of knowledge to be not so trivial: awareness of knowledge. Someone possesses conscious judgment knowledge when she claims that she certainly knows to which class the given situation belongs, and in fact her judgment is correct (e.g., a person is sure that the presented string is grammatical and that is the case). Conscious structural knowledge takes place when one can describe the structure of objects which has been learned (e.g., to report the rules of the grammar). Thus there are two variables in the proposed scheme which can characterize human knowledge: structural/judgment and conscious/unconscious. Four types of knowledge logically follow from these two variables:

• unconscious structural and unconscious judgment knowledge;


• unconscious structural and conscious judgment knowledge;


• conscious structural and unconscious judgment knowledge; and


• conscious structural and conscious judgment knowledge.

It should be noted that Dienes did not consider all four logically possible situations. The third case, conscious structural and unconscious judgment knowledge, was not discussed in his papers. The following review of the studies of Dienes and his associates will include a suggestion as to why one of the possible situations is ignored by them.

In their experiments on artificial grammar learning, Dienes and Scott (2005) applied the decision strategy attribution test that they had developed. In every trial, a participant had to declare on what basis she made her decision about string classification:

1. Guessing (the same as flipping a coin);


2. Intuition (not guessing; believing that an answer is correct without being able to explain why);


3. Conscious knowledge of the grammar (can explain, if needed);


4. Recollection of learning strings or their fragments.

Then authors analyzed classification accuracy on each of these attributions. For example, answers attributed to intuition were analyzed together. In every attribution, participants could classify strings at the level of chance or better. If a participant performed above the chance level, reporting that she was guessing (attribution A), Dienes and Scott conclude that there is structural and judgment knowledge, but both remain unconscious, as subjectively the participant does not use her knowledge, considering that the relevant knowledge is missing. If a participant performs at above the chance level when relying on intuition (attribution B), the conclusion is that there are conscious judgment knowledge (the person is aware that she answers accurately) and unconscious structural knowledge (as she cannot explain why exactly she classifies strings one way or another). If a participant performs at an above-chance level when relying on conscious knowledge of the rules or recollection (attributions C and D), the conclusion is that there are conscious judgment and structural knowledge. In their experiments, Dienes and Scott obtained abovechance accuracy for all of the attributions, demonstrating the presence of several knowledge types in artificial grammar learning.

The researchers also showed that selected types of knowledge have different properties, suggesting a qualitative difference between them. This in turn confirms the theoretical views of the authors of this taxonomy, as shown in the results below.

Attributions A and В reflect the application of unconscious structural knowledge, while attributions C and D demonstrate the application of conscious structural knowledge. Certain differences were detected in experiments between trials with conscious and unconscious structural knowledge.

1. When participants apply conscious structural knowledge (when attributions C and D were given), accuracy is higher than when unconscious structural knowledge (attributions A and B) is applied (Dienes & Scott, 2005, Experiment 1).


2. In the application of conscious structural knowledge (attributions C and D) participants more often repeatedly misclassify stimuli (Dienes & Scott, 2005, Experiment 1).
   Such a consistency of erroneous responses is often considered as a sign of conscious processing (Reber, 1989; see also Allakhverdov, 2009 and Andriyanova, 2014).


3. An additional task distracting attention reduces the amount of conscious structural knowledge, measured by the proportion of trials with attributions C and D, while the instruction to search for the rules of the grammar increases it (Dienes & Scott, 2005, Experiment 2). In addition, the rule-search instruction and attentional load together (i.e., their interaction) only reduces the accuracy of conscious structural knowledge application (Dienes & Scott, 2005, Experiment 2).


4. Response times using unconscious structural knowledge are longer than those of conscious structural knowledge (Mealor & Dienes, 2012). This difference is negatively correlated with confidence (the smaller the confidence in the answer, the longer it takes), but if confidence is controlled, the effect still remains.


5. When assessing the contribution of different factors in the way the participant classifies the string, it appears that the objective similarity between test strings and learning strings almost completely determines the participant’s responses. But in trials in which participants used a conscious structural knowledge, the additional contribution of grammaticality itself is manifested, which cannot be reduced to any objective similarity of strings (Scott & Dienes, 2008).

Thus, Dienes, Scott and their associates tried to demonstrate that their classification reflects real distinctions between types of knowledge that people acquire and use. Dienes argues that the proposed decision strategy attribution test allows researchers to find the real difference in knowledge applied by people (Dienes, 2012). At the same time, according to Dienes, this awareness measure is useful for psychologists as it assesses the awareness of structural knowledge, while other subjective measures of awareness (for example, confidence ratings) measure awareness of judgment knowledge. That is, they give a positive result in a situation where a person knows how to classify stimuli correctly, but does not know why, and therefore does not have a conscious structural knowledge. Dienes expressed the same criticism in relation to tests in which the ability to control the knowledge manifestation is interpreted as an awareness measure (Destrebecqz & Cleeremans, 2001). According to Dienes, it is also possible with conscious judgment knowledge alone.

The main aim of this paper is to demonstrate that Dienes’ classification can include previously described approaches to implicit learning. Dienes himself believes that only situations of unconscious structural knowledge should be investigated in relation to implicit learning research. The current paper suggests that Dienes’ framework has more potential applications. We can account for some rare but important cases of conscious structural and unconscious judgment knowledge, and unite the fields of cognitive and social psychology using common terms and descriptions. The framework even opens the opportunity for dialog between implicit learning researchers and those who think that there are no «implicit» learning processes. Two properties of knowledge in Dienes’ framework and two criteria on which the approaches were identified at the beginning of this paper (acquired knowledge: conscious/unconscious; and application of knowledge: controlled/automatic), may refer to the same phenomena in reality (see Table 1). That is, researchers with different approaches to implicit learning might study different manifestations of it. As noted above, the third type of knowledge was not discussed by Dienes and colleagues. The reason for this, apparently, is that the proposed measurement procedure (decision strategy attribution test) cannot identify such a situation: one cannot report that she does not know which answer is correct if she relies on conscious structural knowledge.

The properties of the four selected types of knowledge are listed below. For convenience, I will use the terms of Dienes and Scott (2005). The properties are shown in Table 2, and some disputable details are hereby discussed.

1. Unconscious structural and judgment knowledge (completely unconscious learning). Such learning occurs when someone does not know that something has been learned or existent knowledge cannot be applied due to attentional load or distraction. As a result, there is complete unawareness of the learned regularity and corresponding knowledge. There is no metacognitive sensitivity, but we have above-chance performance when we think that we merely guess. No attentional resources are needed since there is no deliberate decision-making. This also causes an inability to apply knowledge purposefully.


2. Unconscious structural and conscious judgment knowledge (controlled application of unconscious knowledge). Such learning occurs when one knows that she has learned something and possesses attentional resources for the situation evaluation. Two unique properties of this type of knowledge, caused by the absence of conscious structural knowledge, are that 1) conscious judgment knowledge can be attributed to some emotional feelings (structural mere exposure effect: Newel & Bright, 2001; Zizak & Reber, 2004), and 2) response accuracy is increased if participants are forced to rely on their subjective feelings, to «trust their intuition» (Kinder et al., 2003). We certainly do not know which types of knowledge are affected by the verbalization effect. However, we can expect that it can be observed in the situation of unconscious structural and conscious judgment knowledge, as the need to verbalize one’s own decisions can interfere with the reliance on experienced judgment knowledge.


3. Conscious structural and unconscious judgment knowledge (automatic application of conscious knowledge). Such a situation is possible when someone applies conscious
   knowledge, but does not know that it is relevant to the actual task and that she did, in fact, apply it. The properties of this type of knowledge are very similar to the properties of completely unconscious knowledge, but the structure of applied knowledge was perceived at the conscious level: if presented to the participant, she recognizes it. The verbalization effect in this case should be explained by other causes than in the previous section. In this type of knowledge, any kind of drawing attention to the performed activities has negative consequences (Wegner, 1994). We cannot tell anything about metacognitive sensitivity in this case (Item 4 in Table 2), as the relevant studies have not been conducted. Assuming that metacognitive sensitivity is based on conscious judgment knowledge, one can hypothesize its absence.


4. Conscious structural and judgment knowledge (completely conscious learning). This happens when one is informed about true patterns in the environment, or she finds them by herself, and the new task is perceived as relevant to this knowledge. This type of knowledge has quite expected properties. In the implicit learning research domain, this type of knowledge can be observed in artificial grammar learning experiments, when participants are aware of some important elements of the grammar rules (Dulany et al., 1984). It often happens with biconditional grammars (a small set of rules of «if, then» type: Shanks, Johnstone, & Staggs, 1997). The cases with clear classification criteria are also related to these combination of knowledge components (Waldron & Ashby, 2001).

Conclusions

Theoretical disputes have continued since the discovery of the implicit learning phenomenon. Empirical evidence in support of diametrically opposed theories is constantly emerging. In this situation, the main goal of researchers should be the rethinking of their positions, and the analysis of the reasons why the development of the scientific field extends in different directions. The current paper aimed to serve this purpose. Dienes and Scotts framework is suggested to be used as a reference point for implicit learning researchers, because it provides useful terms and taxonomy. This framework can be logically extended to include cases not considered by aforementioned authors, such as the incidental application of explicit knowledge. Popular approaches to implicit learning and supporting empirical data can be related to different classes in this extended taxonomy. These different types of learning have different properties, some of which I have provided above, summarizing the accessible empirical data. It is worthwhile to continue the list later on.

The ultimate purpose of the work in this direction should be the description of the exact mechanisms underlying the proposed taxonomy. From my point of view, the mechanisms of the emergence of judgment knowledge can be proposed by considering cognition as the work of two independent processing systems; for example, implicit/explicit, logical/intuitive, and so on (Ashby et al., 1998; Dienes, 2012; Allakhverdov & Gershkovich, 2010). The consistency between their outcomes evokes the feelings discussed above (fluency, pleasantness, familiarity, etc.) (Allakhverdov, 2009; Chetverikov, 2014). This signal can be the basis for judgment knowledge. The investigation of such mechanisms should be the target of further studies. In this paper, I have limited the discussion to the observed phenomena and the proposed classification, which, hopefully, will help researchers to more accurately categorize the learning processes found in their studies, and consequently, to better understand each other.

References

Aarts, H., & Dijksterhuis, A. (2003). The silence of the library: environment, situational norm, and social behavior. Journal of Personality and Social Psychology, 84 (1), 18-28. doi: 10.1037/0022-3514.84,1.18

Allakhverdov, V. (2009). The role of consciousness in human cognitive activity. In V. P. Z. &. V. Petrenko (Eds.), Psychology in Russia: State of the art. Scientific yearbook (pp. 124-140). Moscow: Russian Psychological Society.

Allakhverdov, V.M., & Gershkovich, V. A. (2010). Does Consciousness Exist? — In What Sense? Integrative Psychological and Behavioral Science, 44 (4), 340-347. doi: 10.1007/ S12124-010-9133-8

Andriyanova, N.V. (2014). [Regular errors in the learning process: features and possibility to predict], Vestnik of Saint- Petersburg University. Series 16: Psychology. Education,. (Russian).

Ashby, F. G., Alfonso-Reese, L.A., Türken, A. U., & Waldron, E.M. (1998). A neuropsychological theory of multiple systems in category learning. Psychological Review, 105 (3), 442-481. doi: 10.1037/0033-295X. 105.3.442

8argh, J. (1994). The four horsemen of automaticity: Intention, awareness, efficiency, and control as separate issues. In R. S. Wyer Jr., & T. K. Srull (Eds.), Handbook of social cognition. Vol. 1: Basic processes (pp. 1-40). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

8 argh, J. A., Chen, M.,&Burrows, L. (1996). Automaticity of social behavior: Direct effects of trait construct and stereotype activation on action. Journal of Personality and Social Psychology, 71 (2), 230-244. doi: 10.1037/0022-3514.71.2.230

Belova, S. S. (2004). [Subjective evaluation of the Other’s intelligence: verbalization effect]. In D. V. Lyusin, & D. V. Ushakov (Eds.), Social intelligence: theory, assessment, investigation (pp. 39-62). Moscow: IP RAS. (Russian).

Berry, D. C. (Ed.). (1997). How implicit is implicit learning? Debates in psychology. New York: Oxford University Press. 2 doi: 10.1093/acprof:oso/9780198523512.001.0001

Berry, D., & Dienes, Z. (1993). Implicit learning: Theoretical and empirical issues. Hove, Hillsdale: Lawrence Erlbaum Associates. 2

Berry, D. C., & Broadbent, D. E. (1984). On the relationship between taskperformance and associatedverbalizable knowledge. The Quarterly Journal of Experimental Psychology, 36 (2), 209-231. doi: 10.1080/14640748408402156

Bornstein, R.F. (1989). Exposure and affect: Overview and metaanalysis of research, 1968-1987. Psychological Bulletin, 106 (2), 265-289. doi: 10.1037/0033-2909.106.2.265

Brooks, L. R. (1978). Nonanalytic concept formation and memory for instances. In Cognition and concepts (pp. 169-211). Hillsdale, NJ: Erlbaum.

Buchner, A. (1994). Indirect effects of synthetic grammar learning in an identification task. Journal of Experimental Psychology: Learning Memory, and Cognition, 20 (3), 550-566. doi: 10.1037/0278-7393.20.3.550

Carver, C. S., Ganellen, R. J., Framing, W. J., & Chambers, W. (1983). Modeling: An analysis in terms of category accessibility. Journal of Experimental Social Psychology, 19 (5), 403-421. doi: 10.1016/0022-1031(83190019-7

Chan, C. (1991). Implicit cognitive processes: theoretical issues and applications in computer systems design. Unpublished doctoral dissertation, University of Oxford.

Chetverikov, A. (2014). Warmth of familiarity and chill of error: Affective consequences of recognition decisions. Cognition & Emotion, 28 (3), 385-415. doi: 10.1080/02699 931.2013.833085

Cleeremans, A. (2011). The radical plasticity thesis: how the brain learns to be conscious. Frontiers in Psychology, 2 (86), 1-12. doi: 10.3389/fpsyg.2011,00086

Cleeremans, A., Destrebecqz, A., & Boyer, M. (1998). Implicit learning: News from the front. Trends in Cognitive Sciences, 2 (10), 406-416. doi: 10.1016/81364-6613(98101232-7

Destrebecqz, A., 8r Cleeremans, A. (2001). Can sequence learning be implicit? New evidence with the process dissociation procedure. Psychonomic Bulletin & Review, 8 (2), 343-350. doi: 10.3758/BF03196171

Dienes, Z. (2012). Conscious versus unconscious learning of structure. In P. Rebuschat, & J. Williams (Eds.), Statistical learning and language acquisition: Vol. 1 (pp. 337-364). Bos- ton/Berlin: Mouton de Gruyter Publishers.

Dienes, Z., & Altmann, G. (1997). Transfer of implicit knowledge across domains: How implicit and how abstract. In D. Berry (Ed.), How implicit is implicit learning? Debates in psychology: Vol. 5 (pp. 107-123). Oxford: Oxford University Press, doi: 10.1093/acprof:oso/9780198523512.003.0005

Dienes, Z., Altmann, G., Kwan, L., & Goode, A. (1995). Unconscious knowledge of artificial grammars is applied strategically. Journal of Experimental Psychology: Learning Memory, and Cognition, 21 (5), 1322-1338. doi: 10.1037/0278-7393.21.5.1322

Dienes, Z., & Berry, D. (1997). Implicit learning: Below the subjective threshold. Psychonomic Bulletin & Review, 4 (1), 3-23. doi: 10.3758/BF03210769

Dienes, Z., & Perner, J. (2002). A theory of the implicit nature of implicit learning. In A. Cleeremans, & R. French (Eds.), Implicit learning and consciousness: An empirical, philosophical, and computational consensus in the making (pp. 68-92). Psychology Press.

Dienes, Z., & Perner, J. (2001). When knowledge is unconscious because of conscious knowledge and vice versa. In Proceedings of the twenty-third annual conference of the Cognitive Science Society (pp. 1-4).

Dienes, Z., & Scott, R. (2005). Measuring unconscious knowledge: Distinguishing structural knowledge and judgment knowledge. Psychological Research, 69 (5-6), 338-351. doi: 10.1007/S00426-004-0208-3

Doyen, S., Klein, O., Pichon, С.-L., & Cleeremans, A. (2012). Behavioral priming: it’s all in the mind, but whose mind?. PloS one, 7 (1), e29081. doi: 10.1371/journal.pone.0029081

Dulany, D.E. (1997). Consciousness in the explicit (deliberative) and implicit (evocative). In J. D. Cohen, & J. W. Schooler (Eds.), Scientific approaches to the study of consciousness (pp. 179-212). New Jersey: Erlbaum.

Dulany, D.E., Carlson, R.A., & Dewey, G.I. (1984). A case of syntactical learning and judgment: How conscious and how abstract?. Journal of Experimental Psychology: General, 113 (4), 541-555. doi: 10.1037/0096-3445.113.4.541

Fleming, S.M., & Lau, H.C. (2014). How to measure metacognition. Name: Frontiers in Human Neuroscience, 8 (443). doi: 10.3389/fhhum.2014.00443

Frensch, P.A., Lin, J., & Buchner, A. (1998). Learning versus behavioral expression of the learned: The effects of a secondary tone-counting task on implicit learning in the serial reaction task. Psychological Research, 61 (2), 83-98. doi: 10.1007/ S004260050015

Gomez, R.L., & Schvaneveldt, R.W. (1994). What is learned from artificial grammars? Transfer tests of simple association. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20 (2), 396-410. doi: 10.1037//0278-7393.20.2.396

Gordon, P. C., & Holyoak, K. J. (1983). Implicit learning and generalization of the «mere exposure» effect. Journal of Personality and Social Psychology, 45 (3), 492-500. doi: 10.1037/0022-3514.45.3.492

Hayes, N. A., & Broadbent, D. E. (1988). Two modes of learning for interactive tasks. Cognition, 28 (3), 249-276. doi: 10.1016/0010-0277(88)90015-7

Higham, P.A., Vokey, J.R., & Pritchard, J. L. (2000). Beyond dissociation logic: Evidence for controlled and automatic influences in artificial grammar learning. Journal of Experimental Psychology: General, 129 (4), 457-470. doi: 10.1037/0096-3445.129.4.457

Jacoby, L. L., & Dallas, M. (1981). On the relationship between autobiographical memory and perceptual learning. Journal of Experimental Psychology: General, 110 (3), 306-340. doi: 10.1037/0096-3445.110.3.306

Jiang, V., 8r Leung, A. W. (2005). Implicit learning of ignored visual context. Psychonomic Bulletin & Review, 12 (1), 100-106. doi: 10.3758/BF03196353

Jimenez, L., & Vazquez, G. A. (2005). Sequence learning under dual-task conditions: Alternatives to a resource-based account. Psychological Research, 69 (5-6), 352-368. doi: 10.1007/S00426-004-0210-9

Kinder, A., Shanks, D.R., Cock, J., & Tunney, R. J. (2003). Recollection, fluency, and the explicit/implicit distinction in artificial grammar learning. Journal of Experimental Psychology: General, 132 (4), 551-565. doi: 10.1037/0096-3445.132.4.551

Knowlton, B. J., Squire, L. R., Paulsen, J. S., Swerdlow, N. R., & Swenson, M. (1996). Dissociations within nondedarative memory in Huntingtons disease. Neuropsychology, 10 (4), 538-548. doi: 10.1037//0894-4105.10.4,538

Lewicki, P., Czyzewska, M., & Hoffman, H. (1987). Unconscious acquisition of complex procedural knowledge. Journal ofExperimentalPsychology:LearningMemory,andCognition, 13 (4), 523-530. doi: 10.1037//0278-7393.13.4.523

Lewicki, P., Hill, T, & Czyzewska, M. (1997). Hidden covariation detection: A fundamental and ubiquitous phenomenon. Journal of Experimental Psychology. Learning Memory, and Cognition, 23 (1), 221-228. doi: 10.1037/0278-7393.23.1.221

Mangan, B. (2003). The conscious «fringe»: Bringing William James up to date. In J. N. B. Baars W. P. Banks (Ed.), Essential sources in the scientific study of consciousness (pp. 741-759). Cambridge, MA: MIT Press.

Mealor, A., & Dienes, Z. (2012). No-loss gambling shows the speed of the unconscious. Consciousness and Cognition, 21 (1), 228-237. doi: 10.1016/j.concog.20U.12.001

Newell, B. R., 8k Bright, J. E. (2001). The relationship between the structural mere exposure effect and the implicit learning process. The Quarterly Journal of Experimental Psychology: Section A, 54 (4), 1087-1104. doi: 10.1080/02724980042000525

Newell, B. R., 8k 8hanks, D. R. (2014). Prime numbers: Anchoring and its implications for theories of behavior priming. Social Cognition, 32 (Supplement), 88-108. doi: 10.1521/ soco.2014.32.supp.88

Nissen, M. J., 8k Bullemer, P. (1987). Attentional requirements of learning: Evidence from performance measures. Cognitive Psychology, 19 (1), 1-32. doi: 10.1016/0010-0285(87)90002-8

Norman, E., Price, M.C., & Duff, S.C. (2006). Fringe consciousness in sequence learning: The influence of individual differences. Consciousness and Cognition, 15 (4), 723-760. doi: 10.1016/j.concog.2005.06.003

Norman, E., Price, M. C., Duff, 8. C., & Mentzoni, R.A. (2007). Gradations of awareness in a modified sequence learning task. Consciousness and Cognition, 16 (4), 809-837. doi: 10.1016/j.concog.2007.02.004

Peigneux, P., Meulemans, T, Van der Linden, M., Salmon, E., & Petit, H. (1999). Exploration of implicit artificial grammar learning in Parkinsons disease. Acta Neurologica Belgica, 99, 107-117.

Perruchet, P. (2008). Implicit learning. In Cognitive psychology of memory. Vol. 2 of Learning and memory: A comprehensive reference (pp. 597-621). Oxford: Elsevier.

Perruchet, P., & Pacteau, C. (1990). Synthetic grammar learning: Implicit rule abstraction or explicit fragmentary knowledge? Journal of Experimental Psychology: General, 119 (3), 264-275. doi: 10.1037/0096-3445.119.3.264

Perruchet, P., & Vinter, A. (2002). The Self-Organizing Consciousness. Behavioral and Brain Sciences, 25 (3), 297-388.

Ponomarev, Y. A. (1976). [Psychology of creativity]. Moscow: Nauka. (Russian).

Ponomarev, Y. A. (1960). [Psychology of creative thinking]. Moscow: APN of the RSFSR. (Russian).

Price, M. G., 8k Norman, E. (2008). Intuitive decisions on the fringes of consciousness: Are they conscious and does it matter. Judgment and Decision Making, 3 (1), 28-41.

Reber, A. (1993). Implicit learning and tacit knowledge: An essay on the cognitive unconscious. New York: Oxford University Press.

Reber, A. S. (1976). Implicit learning of synthetic languages: The role of instructional set. Journal of Experimental Psychology: Human Learning & Memory, 2 (1), 88-94. doi: 10.1037/0278-7393.2.1.88

Reber, A. S. (1967). Implicit learning of artificial grammars. Journal of Verbal Learning and Verbal Behavior, 6 (6), 855-863. doi: 10.1016/S0022-5371(67)80149-X

Reber, P. J., & Squire, L. R. (1999). Intact learning of artificial grammars and intact category learning by patients with Parkinsons disease. Behavioral Neuroscience, 113 (2), 235-242. doi: 10.1037/0735-7044.113.2.235

Reber, P. J.,        & Squire, L. R. (1998). Encapsulation

of implicit and explicit memory in sequence learning. Journal of Cognitive Neuroscience, 10 (2), 248-263. doi: 10.1162/089892998562681

Scott, R. B., 8k Dienes, Z. (2010). Knowledge applied to new domains: The unconscious succeeds where the conscious fails. Consciousness and Cognition, 19 (1), 391-398. doi: 10.1016/j.concog.2009.11.009

Scott, R.B., & Dienes, Z. (2008). The conscious, the unconscious, and familiarity. Journal ofExperimental Psychology: Learning Memory, and Cognition, 34 (5), 1264-1288. doi: 10.1037/ aOO12943

Shanks, D.R., Johnstone, T, & Staggs, L. (1997). Abstraction processes in artificial grammar learning. The Quarterly Journal ofExperimental Psychology. Section A: Human Experimental Psychology, 50 (1), 216-252. doi: 10.1080/713755680

Shanks, D. R., & St John, M. F. (1994). Characteristics of dissociable human learning systems. Behavioral and Brain Sciences, 17 (03), 367-395. doi: 10.1017/80140525X00035032

Shanks, D. R., Wilkinson, L., & Channon, S. (2003). Relationship between priming and recognition in deterministic and probabilistic sequence learning. Journal ofExperimental Psychology: Learning, Memory, and Cognition, 29 (2), 248-261. doi: 10.1037/0278-7393.29.2.248

Singley, M.K., & Anderson, J.R. (1989). The transfer of cognitive skill. Cambridge, MA: Harvard University Press. 2

Smith, J., Siegert, R.J., McDowall, J., & Abernethy, D. (2001). Preserved implicit learning on both the serial reaction time task and artificial grammar in patients with Parkinsons disease. Brain and Cognition, 45 (3), 378-391. doi: 10.1006/ brcg.2001.1286

Tunney, R.J. (2005). Sources of confidence judgments in implicit cognition. Psychonomic Bulletin & Review, 12 (2), 367-373. doi: 10.3758/BF03196386

Tunney, R. J., & Shanks, D. R. (2003). Subjective measures of awareness and implicit cognition. Memory & Cognition, 31 (7), 1060-1071. doi: 10.3758/BF03196127

Waldron, E.M., & Ashby, EG. (2001). The effects of concurrent task interference on category learning: Evidence for multiple category learning systems. Psychonomic Bulletin & Review, 8 (1), 168-176. doi: 10.3758/BF03196154

Wegner, D. M. (1994). Ironic processes of mental control. Psychological Review, 101 (1), 34-52. doi: 10.1037/0033-295X.101.1.34

Whittlesea, B.W, &Dorken, M.D. (1997). Implicit learning: Indirect, not unconscious. Psychonomic Bulletin & Review, 4 (1), 63-67. doi: 10.3758/BF03210775

Wilkinson, L., & Shanks, D. R. (2004). Intentional control and implicit sequence learning. Journal of Experimental Psychology: Learning Memory, and Cognition, 30 (2), 354-369. doi: 10.1037/0278-7393.30.2.354

Winter, B., & Reber, A.S. (1994). Implicit learning and the acquisition of natural languages. In N. C. Ellis (Ed.), Implicit and explicit learning of languages (pp. 115-145). London: Academic Press.

ZizakjD.M., & Reber, A.S. (2004). Implicit preferences: The role(s) of familiarity in the structural mere exposure effect. Consciousness and Cognition, 13 (2), 336-362. doi: 10.1016/j.concog.2003.12,003