Introduction

The apprenticeship system (popular in the Middle ages) is possibly one of the earliest applications of implicit learning. In this system, the apprentice acquires skills of a craft from an expert or master teacher by working with him/her for a set period of time. At initial phases of "training" the apprentice, the master teacher would often demand the novice to perform tasks which seem to be irrelevant to the craft. Instead of immediately directing attention to the acquisition of the fundamental skills of the craft, the learner is given opportunities to gradually acquire intangibles through exposure. These intangibles, however may enhance the development of sensitivity towards the craft, or promote specific attitudes and ethical values crucial to the apprentice's subsequent career development. Through long-term exposure, the apprentice then develops a 'feel' or a 'sense' for the fundamentals of the craft. However, the apprentice may not be aware that subtleties of the craft have been acquired in the process.

A discussion of the apprenticeship system here serves a dual purpose. On one hand, it provides a historical context for the application of the concept of implicit learning to pedagogy; on the other, it raises the question whether the notion of implicit learning had been tacitly acquired and applied by masters of the apprenticeship system without any conscious attempts.

In this paper, I will explore issues of implicit learning as they relate to attention and memory in musical performance. This will involve a review of some of the evolving paradigms on implicit learning in cognitive psychology. Stemming from the results of Whittlesea & Dorken's (1993) research, I will then discuss the importance, applicability and implications of these issues as they relate to learning and teaching music (with particular focus on the areas of piano pedagogy and advanced piano performance). Finally, I will conclude by suggesting some possible directions of research in attention and memory as they relate to musical performance.

What is Implicit Learning?

Implicit learning refers to learning without focusing particular attention on the stimulus. Although the learner cannot consciously recall the stimulus, performance is influenced through exposure. As Reber's (1976) study suggested, when we are exposed to a series of stimuli which shares a common deep-structural relationship, we can become sensitive to the underlying relationship between these stimuli even though it may be extremely complex. We may not be conscious of our knowledge of this relationship, yet when we encounter a new stimulus, we are able to tell whether or not this new stimulus conforms to the relationship underlying the previously encountered stimuli (Reber, 1976). That is, even though we are unable to explicate this underlying structure (since we are not conscious of it), we are capable of making judgements based upon our understanding of this relationship and applying it to future situations. In other words, we have acquired knowledge through implicit learning.

Literature Review

Reber (1969, 1989), Lewicki (1985), Berry & Broadbent (1984) and a number of psychologists assume that implicit learning is a consequence of an "automatic abstraction" process. This process is considered "automatic" in that it is performed without the subject's awareness or consciousness. By "abstraction," the researchers are referring to a process that weighs the properties of a stimulus in terms of its relevance to the structure of the overall underlying relationship. Since this abstraction process is automatic, it is therefore independent of (i) how the subject processes the item, and (ii) the subject's intentions in learning. In other words, the automatic abstraction process assumes that the process or experience of learning is not important in implicit learning. This also implies that no attention and intention are required from the subject for implicit learning to occur.

To support their "automatic abstraction" postulate, Reber (1969) and Mathews et al. (1989) showed that subjects experience positive transfer of knowledge (transfer appropriate learning) when they are required to deal with a new set of items. That is, subjects can recognize that the new stimuli still conform to the old underlying relationship even though the surface elements of the stimuli have been altered.

Contrary to Reber (1969) and Mathews et al.'s (1989) views, Whittlesea & Dorken (1993) believe that the acquisition of implicit knowledge is contingent upon the circumstances in which learning occurs. Therefore, they propose an "episodic account of implicit learning" which claims that the encoding of structural information is dependent upon the particular processing experiences encountered. They also argue that depending upon the conditions of encoding [for instance, whether the encoding context matches with the testing context], implicit learning may occur with or without abstraction of the underlying structure. And in cases when structure is abstracted, there is the possibility of abstracting only the deep structure, or only the surface structure, or even both simultaneously (Brooks, 1978). They further contend that the depth of abstraction (whether deep or surface structure) depends upon whether (i) the subject has the intention to learn about the stimuli and whether (ii) he/she gives full attention to process them.

To illustrate their first argument, Whittlesea & Dorken conducted an experiment using two artificially constructed or finite-state grammars (please see Figure 1) Grammar A and Grammar B to generate two sets of exemplar items. Subjects are required to learn the structures of one set of items (generated by one grammar, say Grammar A) by spelling the items and the other set (generated by the other grammar, say Grammar B) by pronouncing the items. Subjects were then told that the items they had spelled previously were generated by one set of rules and those they pronounced earlier conformed to another set of rules. Subsequently, subjects were tested for their abilities to abstract elements underlying the general structure. In the testing phase, subjects were given items which they have to classify as whether they conform to the grammar of the previously spelled or pronounced items. (Before making a decision, they were required either to spell or pronounce the test item.) The results indicate that when the processing contexts of encoding and testing match (e.g. spell - spell), subjects can successfully classify the items (p = 0.66). According to the experimenters, this suggests that subjects may become sensitive to the deep structure without being aware of it. However, it was found that in the mismatched context, that is when the contexts of encoding and testing differ (e.g. spell - pronounce), subjects were still able to classify the items correctly above chance (p = 0.61), although there is a significant decrement in accuracy. This leads Whittlesea & Dorken to believe that sensitivity is due to specific encoding experiences encountered in the processing phase rather than an abstraction of the underlying structure.

Whittlesea & Dorken (1993) conducted another experiment to substantiate their argument that the depth of abstraction (that is, whether the deep and/or surface structure is abstracted) depends upon whether the subject pays full attention to the stimuli with the intent of learning about them or not. In this experiment, Whittlesea & Dorken purposely misled the subjects to believe that the study was intended to measure their memory for 3-digit numbers so as to prevent them from paying full attention to the training stimuli. Training stimuli were therefore presented as "distractors" to the subjects. Similar to the previous experiment, these "distractor" items were also generated by an artificial grammatical system (see figure 2). In between the number tasks, subjects were required to process the "distraction task" by repeating the letters out loud. Prior to testing, the subjects were told that all the "distractors" in fact were stimuli, and were generated by a complex set of rules. They were then shown the test items and were required to decide whether they comply to the rules or not. The results indicate that subjects were able to discriminate grammatical from non-grammatical items. This therefore suggests that subjects learned something from the experience through exposure, even though the underlying grammar and the training items were incidentally presented.

Episodic Processing and Musical Performance

In sum, Whittlesea & Dorken (1993) conclude from their study that implicit sensitivity to a deep underlying structure is due to specific encoding experiences encountered in the processing phase rather than an automatic abstraction. In other words, implicit sensitivity to the underlying structure is dependent upon the processing context, and is only an incidental by-product of episodic processing. Lastly, the depth of abstraction depends upon whether (i) the subject has the intention to learn about the stimuli and whether (ii) the subject gives full attention in processing them.

Since the processing context plays a significant role in memory, a consideration of the relationships between the experience(s) encountered in the processing and retrieval contexts may have significant implications for the disciplines of piano pedagogy and performance. Based upon this understanding, for instance, the performer may enhance his/her accuracy and reliability in retrieval by matching the practising (encoding) and performance (retrieval) contexts. Besides, an understanding of the relationship between the encoding and retrieval contexts may help reveal some potential problems that may surface in a live musical performance, and thus help us understand why some pianists 'fail' while others are more 'successful' in facing the realities of musical performance.

Differences between Item Recognition and Advanced Music Performance

What is involved in musical performance?

Before proceeding to a discussion of the implications involved, it is important to clarify the meaning of the term "memory" as it applies to the context of musical performance and pedagogy. "Memory" in the musical context refers not only to the ability to perform an entire piece of music without referring to the score, but it also encompasses the effective assimilation of a number of musical and extra-musical dimensions. Musical dimensions include timbre, pitch level, dynamics, rhythm, articulation, phrasing, harmonic relationships between musical parts, emotive tone, aesthetics and other aspects that are not easy to tease apart. Meanwhile extra-musical dimensions embody specific aural, tactile and kinaesthetic organisational relationships within the performer. In order to establish a communion with the audience through an artistic performance, these multiple dimensions must be simultaneously integrated into the technical and interpretive aspects of practice and performance.

Differences between Musical Performance and Item Recognition

Prior to transposing these psychological concepts to musical performance, it is essential to note the significant differences between the act of musical performance and that of recognizing items generated by an artificial grammar (as employed in Whittlesea & Dorken's study). First, the encoding experience provided by the experimental task is a unidimensional one, since the stimuli consist entirely of letter strings which are processed either by spelling or pronouncing. In contrast, as mentioned previously, a "live" musical performance involves multiple dimensions that are simultaneously integrated within its structure and executed in real-time. This process not only involves aesthetic and physical judgement, but also creativity of a high order. Secondly, recognition of the grammaticality of artificial words is by nature a passive task which differs from that of musical performance, which is an active "production task" based upon reflective and corrective feedback mechanisms in real-time. Lastly, the experience of musical performance entails a personalised reconstruction of the composer's intentions through the expression of emotion (ranging from placid calmness to unrestrained passion), whereas the exemplar items employed in the studies mentioned above are essentially "affectively neutral"- they do not involve or provoke any of the affective senses.

Practical Examples and Applications to the Areas of Piano Pedagogy and Performance

Since the efficiency of memory in coping with unanticipated situations is highly dependent upon the specific coding experiences encountered, it is useful for musical performers (particularly solo pianists) who must deal with innumerable contingencies on the concert stage to examine their approaches to musical practice. By revealing the conditions under which the music is being processed, we may better understand why certain approaches to practice are more effective in enhancing subsequent memory than others. For example, a recent study of the relationship between the roles of depth of processing and elaboration in recall of musical passages learned through sight-reading (Segalowitz, Cohen, Prieur & Chan, 1995) suggests that increased elaboration favours recall of deeply processed musical material but hinders recall of shallow processed musical material. This may explain why practising a musical passage by meaningless repetition is less efficient than conferring a particular meaning or meanings to the passage, or by providing various related contexts for processing.

For similar reasons, a number of world renowned solo pianists recommend manipulating the processing contexts in practise to enrich the encoding experiences to enhance subsequent recall. For example, Walter Gieseking and Jorge Bolet (Mach, 1988. pp. 21) mentioned the importance of visualisation of the music score, and practising away from the instrument. As Bolet describes, "Yet when it comes to memorising the music, I do perhaps ninety-five to ninety-eight percent of it away from the piano. I look at the score, study it, go through it in my mind, and piece it all together" (Mach, 1988. pp.29). This process of visualisation apparently serves as an additional cue for retrieval by providing an encoding experience beyond auditory and sensory cues established by physically playing on the keyboard. Certainly, the approach mentioned above is only one of many ways to provide a specific and elaborated context to the coding experience. Other pedagogues and pianists, such as Alicia de Larrocha, and Leon Fleisher (Mach, 1988. pp. 34) recommend other aids to memorisation, including knowing the details of each musical phrase (such as cadences and form), and making use of kinaesthetic (such as physical choreography at the keyboard) and aural cues (chord progressions and contrapuntal melodic relationships) to facilitate memory.

Since recall of a stimulus can be significantly affected when the context in which it was presented differs from that of retrieval, it may be helpful to establish conditions of practice that resemble that of actual performance. By way of illustration, it may be useful to mention some problems encountered by my piano students.

One of my piano students has been working on Chopin's Valse Brilliante in E-flat Major for the past few months. He dedicated a considerable amount of time practising the piece slowly and carefully. Despite the painstaking preparation, he ran into a number of technical difficulties as he attempted to play it during our lesson. Part of the reason is that he became extremely nervous as he was required to "perform" for an audience. My presence evidently presented a difference in context(s) between the encoding phase and the retrieval phase adequate to disrupt his performance [encoding specificity]. As a consequence, his tempo increased and accelerated to such a point that his playing went out of control. Since he has neither practised nor played through the piece at such a fast tempo, this sudden increase in speed (involuntarily induced by nervousness) created a context very different from that when the piece was practised. For similar reasons, the change in performance setting and emotional state of the performer probably also led to a disruption in memory [again due to encoding specificity], and naturally the performance suffered. Based upon this understanding, I think it would be more productive for him to practise the piece at various tempi (ranging from the slowest to the fastest) to ensure that all possible episodic representations [various levels of processing] have been attended to in the encoding phase. In so doing, it would be less likely for contingencies encountered in the "retrieval phase" to disrupt the performance. By the same token, it is advisable for students to try to perform the piece in a series of rehearsals and informal performances as a preparation for handling the excitement and anticipation prior to performing publicly. In so doing, one might be better able to cope with the unexpected demands of performing for an audience through (i) establishing a practice context that is similar to the actual performance context, as well as (ii) becoming acquainted with the pressures of public performance through implicit learning.

I also notice that developing pianists often isolate technically difficult passages from musical pieces and practice them repeatedly devoid of emotion. This approach can become counter-productive. First, when the pianist's attention is directed only towards the technical aspects of the piece, the depth of abstraction from the music would be less comparable to processing the technical, musical and emotive aspects in conjunction (as one may have more reference cues for retrieval). As a consequence, memory reliability may suffer under challenging performance situations. Secondly, practising without full integration of emotion would reinforce an encoding context which is significantly different in nature from that of the final performance which entails emotion and aesthetic subtleties.

It may be useful here to list some of the common situations in which the conditions or contexts of practice differ significantly from that of performance, and as a consequence this discrepancy between processing and retrieval contexts may lead to problems in memory. (Figure 3)

It is also important to mention that the notation of the musical text can trigger specific associations with the context in which the specific musical symbols were first acquired [constant mapping] (Prieur, 1996). For example, a large interval can be perceived automatically by the student as a technical problem involving an awkward stretch of the fingers and hands rather than as a harmonic blend of tones. As a result of constant mapping, this non-reflective conditioning may present unnecessary difficulties which can limit the alternative possibilities available to the performer in expressing his/her aesthetic and interpretative intentions and in perceiving the music altogether.

Figure 3.

Practice Conditions (Encoding Processes)	Performance Conditions (Retrieval Processes)	Possible Solutions
Relying on only one mode of perception (e.g. tactile memory)	Retrieval by another mode of perception (e.g. visual memory) or perhaps by synergistically combining various modes of perception	Experimenting by varying the combination of sensory modes
Slow tempo	Fast tempo	Practise with variations of tempi
Devoid of emotion	Full range of emotions	Practise with a full range of emotions
Absence of audience	Presence of audience	(Open rehearsals) Mock recitals with audience
Practice from memory	Performing with the score may create uncertainties (e.g. when accompanying)	Performing by memory

Based upon Whittlesea & Dorken's study which suggests that the experience of how a specific task is being processed has a significant effect on subsequent retrieval (implicit learning) , I recommend these alternative approaches to piano practice to maximize productivity:

(1) Instead of aimlessly repeating a musical passage until it "gets into one's fingers", one may change one's processing strategy by directing one's attention towards the deeper meaning(s) of the passage. This, for instance, can be achieved by establishing a musically oriented and internally derived cue, which aesthetically organises the multiple dimensions involved in musical performance by providing a relationship and linkage between various musical parts. This cue in turn provides a simplified processing context, within which the performer can effectively articulate his/her musical intentions. This cue can also help to focus the performer's attention on a specific task that is meaningfully relevant to the musical context. Meanwhile, memory retention can also be facilitated through implicit learning.

(2) The processes mentioned above may also be extended to establish a synergistic relationship between the aesthetic experience of the music and the perception of multiple sensory modes (tactile, kinaesthetic, visual and auditory senses) to facilitate recall.

Future Directions for Research

From the discussion of memory and attention and their relationship to advanced musical performance, the issues that follow seem to be particularly amenable to interdisciplinary research:

The inter-relationship between kinaesthetic, auditory, tactile and related factors that coalesce into a single performing entity (perhaps a phenomenon roughly analogous to synaesthesia) and how these factors relate to memory and attention. For instance:

- Whether retrieval can be facilitated by combining multiple sensory modes? And if so, what are the most efficient combinations?

- How does this multi-modal combination interact with the encoding and retrieval contexts?

- How can attentional resources be efficiently allocated to the various modes of encoding involved (kinaesthetic, tactile, auditory, etc.) in order to maximize musical performance?

- How does the virtuoso performer efficiently shift his/her attention from one mode of processing to another to optimize performance? And what is (are) the relationship(s) between the efficiency and/or flexibility of attentional shifting and the musician's level of expertise?

Berry, D.C., & Broadbent, D.E. (1984). On the relationship between task performance and associated verbalizable knowledge. Quarterly Journal of Experimental Psychology, 35, 209-231.

Brooks, L.R. (1978). Non-analytic concept formation and memory for instances. In E. Rosch & B.B. Lloyd (Eds.), Cognition and categorization (pp. 169-211). Hillsdale, NJ: Erlbaum.

Craik, F.I.M., & Lockhart, R. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behaviour, 11, 671- 684.

Craik , F.I.M., & Tulving, E. (1975). Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104, 268- 294.

Hastroudi, S. (1983). Type of semantic elaboration and recall. Memory and Cognition, 11(5), 476-484.

Lewicki, P. & Hill, Y. (1985). Nonconscious biasing effects of single instances on subsequent judgements. Journal of Personality and Social Psychology, 48, 563- 574.

Mach, E. (1980). Great contemporary pianists speak for themselves. Dover Publications, Inc. New York.

Mathews, R. & Roussel, L. (1993) Automatic abstraction of stimulus structure from episodes: Comment on Whittlesea & Dorken. Journal of Experimental Psychology: General, 122(3), 397-400.

Petrovsky, A.V. and Yaroshevsky M.G. (1987). A concise psychological dictionary. New York: International Publishers.

Prieur, Thierry. (1996). Personal communication and thesis in progress.

Reimer, B. & Wright, J. (1992) On the Nature of Musical Experience. University Press of Colorado.

Reber, A.S. (1989). Implicit learning and tacit knowledge. Journal of Experimental Psychology: General, 118, 219-235.

Segalowitz,N., Cohen, P., Prieur, T., & Chan, A. P.Y., (1995). Keyboard Recall Memory: the effects of elaboration versus depth of processing. (Submitted for Conference on Musical Perception and Cognition, Montreal, August, 1996.)

Tulving, E. & Thompson, D. (1971). Retrieval processes in recognition memory: effects of associate context. Journal of Experimental Psychology, 87, 116-124.

Whittlesea, B.& Dorken, M. (1993). Incidentally, things in general are particularly determined: an episodic-processing account of implicit learning. Journal of Experimental Psychology: General, 122(2), 227-248.

Whittlesea, B. & Dorken, M. (1993). Potential power of coding particular experiences: Reply to Mathews and Roussel. Journal of Experimental Psychology: General, 122(3), 401-404.