Auditory Implicit Learning, and Its Transfer to and from Visual Implicit Learning
Christopher D. Green
Philip R. Groff
© 1996 by Christopher D. Green & Philip R. Groff
Reber and others have shown that the passive learning of synthetic grammars (“implicit learning”) is a robust phenomenon when visual stimulus materials are employed. It was the main aim of this study to discover if the same effects occur in the auditory modality, and then to determine if such learning can be transferred from the visual to the auditory mode, and vice versa. In the present study, first, the standard effect was replicated with visual material (Experiment I). Second the effect was also shown to occur when the same material was presented to the auditory modality (Experiment II). It was then shown that implicitly learned material can be transferred from the visual to the auditory modality (Experiment III) and from the auditory to the visual modality (Experiment IV). The implications of the results are discussed with respect to the debate about the “abstractness” or “concreteness” of the mental representation of the material learned.
Auditory Implicit Learning, and Its Transfer to and from Visual Implicit Learning
In the standard “implicit learning” paradigm (e.g., Reber, 1967, 1969, 1976, 1993) subjects are asked to memorize visually-presented sets of approximately 20 strings of three to eight written letters. Although these letter strings appear to be random, they are in fact the product of a Finite State Machine (FSM) such as in Figure 1.
Figure 1. Finite State Machine used by Reber (1967) and in the present study.
To produce a string with an FSM one begins at the “start” arrow at the left, and traces around the diagram in the direction of the arrows until one reaches the “end” arrow on the right. Each transition from one state (circle) to another generates a letter which is added to the end of the string generated thus far. For instance, some of the letter strings possible with this particular FSM include:
TTS TPTS TTXVS
VVS VXVS VVPS
Reber (1967) found that his subjects could learn the letter strings generated by the FSMs (“grammatical” strings) at a faster rate than they could learn truly random letter strings (see Figure 2). Moreover, he showed that although subjects were unable to articulate the rules underlying the grammatical strings when told that such rules existed, they were able to distinguish previously-unseen strings that had been generated by the same FSM from truly random strings at a rate of nearly 80%.
Figure 2. Implicit learning curves from Reber (1967). All subjects improved with experience, but the group learning the grammatical strings was faster overall. The interaction was also significant.
Reber (1969) also showed that subjects who have learned the strings of a given FSM are able to learn more quickly strings made of new symbols if they are generated by an FSM of the same structure (e.g., replace all the Ts in Figure 1 with Qs, all the Vs with Bs, etc.) than strings made of the same symbols, but generated by a different FSM. In addition, Reber (1976) showed that subjects who are told at the outset of the procedure that the strings are governed by a complex set of rules and are instructed to explicitly figure out what the rules governing the strings are do worse at both the learning task and the identification of new strings than subject who are not do not try to figure out the rules.
These effects have been replicated and elaborated upon many times in the intervening decades and, although no one seriously doubts the effect, per se, some have challenged Reber’s interpretation; viz., that subjects unconsciously learn the abstract structure (“grammar”) of the strings simply by learning the individual strings themselves during the study phase of the procedure. These critics include Brooks (1978); Dulany, Carlson, and Dewey (1984); Perruchet and Pacteau (1990); Shanks and St. John (1993); as well as Redington and Chater (1996). Reber (e.g., 1993), however, continues to defend his belief in unconscious abstract learning.
It is not the main aim of the present paper to enter into the debate over the correct theoretical interpretation of the phenomenon but rather to extend it to a new domain. Although implicit learning of visuo-verbal material has been demonstrated many times, few have ever investigated whether it occurs in modalities other than the visuo-verbal. When investigators have gone beyond these bounds, they have typically extended the phenomenon into other visual areas (see, e.g., Brooks, 1978; Lewicki, Hill, & Bizot, 1988). It was the main aim of the present study, however, to reach into the auditory modality. Specificlly, we sought (1) to discover whether implicit learning is present in the auditory modality; and (2) to see if such learning can be transferred from one sensory modality to another, i.e., if grammatical letter strings are memorized in the visual mode, can subjects then distinguish grammatical from ungrammatical strings in the auditory mode, and vice versa?
EXPERIMENT I: Written implicit learning
The first experiment was simply an attempt to replicate the standard implicit learning effect using written material.
The subjects were 10 undergraduate students (6 female, 4 male) who had been recruited from psychology classes. They were paid a small amount for their participation.
The subjects were told that they were in a learning experiment and were instructed to memorize 18 strings of six to eight letters each. They were seated at normal reading distance (approx. 30 cm) from a 14 in. computer monitor. The strings were presented one at a time at the center of the screen. After viewing three strings for five seconds each, the subjects were asked to try to copy them out exactly by hand. If they made any errors, they were shown the same set of three strings again for five seconds each and asked to try to copy them again. This procedure was repeated until they were able to reproduce the three strings exactly. When the three strings were correctly reproduced, the number of repetitions required was recorded, and work on a new set of three strings was begun. This process continued until all eighteen strings had been learned perfectly.
The five subjects in the experimental group memorized grammatical letter strings, governed by the FSM in Figure 1. The five subjects in the control group memorized 18 random letter strings. To make the two sets of strings maximally indistinguishable from each other, the random strings began and ended with the same letters as the grammatical ones. They also only contained the letters found in the grammatical strings, though in random rather than grammatical order.
After the learning phase of the experiment was completed, the subjects in both groups were told that the strings they had just memorized had been generated by a complex set of rules. They were asked if they could articulate the rules governing the strings. Their answers were recorded on paper. Finally, they were shown 32 new strings on the computer monitor, one at a time, and asked to try to decide which of these had been generated by the same set of rules as had the strings they had previously learned. The number of correct identifications (grammatical or not) was recorded. The subjects were then debriefed, paid, and thanked for their participation.
Results and Discussion
A two-way ANOVA with one repeated factor (2 independent grammaticality conditions ´ 6 triplets of strings to be learned by each subject) was used to analyze the data. In the learning phase of the experiment, both groups became increasingly good at the task of memorizing the strings across sets (F(5, 40)=4.22, p=.004). The subjects in the grammatical group, however, learned the strings faster overall than those in the ungrammatical group (F(1, 8)=22.12, p=.002). There was no interaction between learning sets and group (i.e., grammatical or random) (F(5, 40)<1.0, NS). These results are plotted in Figure 3.
Figure 3. Implicit learning curves from Experiment I. Both groups improved with experience, but the grammatical group learned significantly more quickly than the random group. There was no significant interaction.
In response to the question about what rules governed the strings, there were no apparent differences in the answers given by subjects in the two groups. Typically they noted that there were no vowels in the strings, or they were able to name some of the starting or ending letters (which were the same for both groups). Little specific information about the specific structure of the strings was forthcoming. These results are typical of those found in all the experiments of this study and, as such, they are not discussed further in this paper.
In the test phase of the experiment, the subjects who had been in the grammatical group during the learning phase were able to correctly identify 76% of the strings (this included both identifying grammatical strings as such and identifying ungrammatical groups as such). Those who had been in the ungrammatical group were, as expected, only able to identify 50% (see Figure 4). This difference was highly significant (t(8)=3.55, p=.008), despite the small sample size.
Figure 4. Results of identification task in Experiment I. The grammatical group correctly identified significantly more new strings (as being grammatical or not) than the random group.
The traditional phenomena of implicit learning were clearly replicated in the present result.
EXPERIMENT II: Auditory implicit learning
In the second experiment, we attempted to show that the effect obtained with written material could also be obtained with auditory material.
Subjects and Procedure
Ten undergraduate students (7 female, 3 male), recruited from psychology classes and paid a small fee were presented with exactly the same materials as in Experiment I, but in the auditory rather than the visual mode. The stimuli were made using commercially-available sound-recording software, and were stored in .wav format on the hard disk of a desktop microcomputer. They were played back to the subjects through two small speakers, one to either side of the subject.
Results and Discussion
As in Experiment I, both groups showed improvement on the memorization task as they gained experience with it (F(1, 8)=11.96, p=.009). Also, the subjects in the grammatical group learned faster, on average, than those in the random group (F(5, 40)=18.60, p<.0005). An interaction trend can be clearly seen in the graph (Figure 5), but it did not reach significance, perhaps because of the small sample size (F(5, 40)=1.58, NS).
Figure 5. Implicit learning curves from Experiment II. Both groups improved with experience, but the grammatical group learned significantly more quickly than the random group. There was no significant interaction.
On hearing the 32 new strings, the subjects who had been in the grammatical group in the learning phase were able to correctly identify, as being grammatical or not, 79% of the strings. Those who had been in the random group were only able to identify 53% (see Figure 6). This difference was highly significant (t(8)=7.85, p<.0005).
Figure 6. Results of identification task in Experiment II. The grammatical group correctly identified significantly more new strings than the random group.
Clearly, the effect shown in Experiment I in the visual modality was obtained with approximately equal strength in the auditory modality. One minor difference was that it typically took the subjects one or two trials longer to learn the sets of auditory strings than it had taken them to learn the sets of written strings.
EXPERIMENT III: Written-to-auditory transfer
Finally, we came to the question of cross-modal transfer; can subjects who learn strings visually then use the information they have picked up to make correct decisions about auditory stimuli?
Subjects and Procedure
The subjects were, again, 10 undergraduate students (6 female, 4 male) recruited from psychology classes and paid a small fee for their participation. The instructions and procedure were identical to those in Experiments I and II, but for the following details. Five of the subjects were asked to learn six sets of three grammatical strings of written letters, and five others were asked to learn six sets of three random strings of written letters. Then both groups were told that the strings followed certain complex rules and asked to try to identify which of 32 new auditory strings followed the same set of rules.
Results and Discussion
As in the other experiments both groups showed significant improvement in memorization over the course of the procedure (F(5, 40)=5.24, p=.001). Similarly, the grammatical group was significant better at learning the strings than the random group (F(1, 8)=7.8, p=.023). There was no interaction (F(5, 40)<1.0, NS) (see Figure 7).
Figure 7. Implicit learning curves from Experiment III. Both groups improved with experience, but the grammatical group learned significantly more quickly than the random group. There was no significant interaction.
On the decision task, subjects in the grammatical group correctly identified the grammaticality of 68% of the new letter strings. Those in the ungrammatical group correctly identified only 51% (see Figure 8). This difference was significant (t(8)=2.44, p=.04).
Figure 8. Results of identification task in Experiment III. The grammatical group correctly identified significantly more new strings than the random group.
The results show there to have been cross-modal implicit learning from the visual to the auditory mode. The next question was whether the information learned could be transferred from the auditory to the visual modality.
EXPERIMENT IV: Auditory-to-written transfer
The aim of the fourth experiment was to test whether implicit learning can be transferred from the auditory to the written modality as well.
Subjects and Procedure
Thirty undergraduate subjects (19 female, 11 male) were recruited from psychology classes and paid a small fee for their participation. The subjects were asked to memorize 18 auditory letter strings (in 6 sets of 3). Half them heard grammatical strings, and half heard random strings. After learning the auditory strings the subjects were asked to identify which of a set of 32 new written strings governed by the same FSM as those they had learned in the auditory study phase of the experiment.
Results and Discussion
The subjects showed a significant improvement in learning over the course of the procedure (F(5, 140)=6.397, p<.0005), and there was a significant difference between the two groups (F(1, 28)=5.718, p=.02). The interaction was also significant (F(5, 140)=6.305, p<.0005). This seems to reflect the fact that this time only the group learning grammatical strings improved over the course of the learning phase (in addition to the increase power provided by using 30 subjects).
Figure 9. Implicit learning curves from Experiment IV. Both groups improved with experience, but the grammatical group learned significantly more quickly than the random group. There was also a significant interaction.
On the decision task, the subjects in the grammatical group correctly identified the grammaticality of 75% of the new letter strings. Those in the random group correctly identified only 49%. This difference was highly significant (t(28)=6.22, p<.0005).
Figure 10. Results of identification task in Experiment IV. The grammatical group correctly identified significantly more new strings than the random group.
These results show there to be cross-modal implicit learning from the auditory to the visual mode.
Past research has repeatedly shown that the phenomenon dubbed “implicit learning” by Reber is robust in the visual modality when using letter strings (hereafter “visuo-verbal mode”). In the present study, we set out to discover (1) whether such learning is possible in the auditory modality, and (2) whether such learning, if obtainable, can be transferred from the written to the auditory modality and vice versa. The answer to both questions was found to be strongly in the affirmative. Auditory implicit learning seems to be as robust as written implicit learning, and strong transfer of such learning occurs in both auditory-to-written and written-to-auditory directions. Indeed cross-modal implicit learning seems to be generally just as strong as the conventional intra-modal sort.
One interesting avenue for future research would be to move beyond the verbal realm and explore implicit learning other sense modalities (e.g., visuo-spatial, auditory-musical), and the transfer among these various modalities. Although it was not the main point of this paper to assess Reber’s position that the subjects unconsciously learn abstract information duringthe implicit learning procedure, it is interesting to note that consistent evidence of transfer among widely diverse modalities would provide confirmation of his claim that the information learned is relatively abstract in nature. If subjects were only learning concrete bigrams or trigrams contained within the strings, or if they were relying on visual “analogies,” as has been argued by some, one would expect them to do much more poorly on the test phase of the procedure is the concrete features of the stimuli had been radically changed. Because in this study we have only changed from visual to auditory representations of the same stimulus objects (viz., letters), no strong confirmation of this hypothesis has been provided here. Nevertheless, it would seem to be a small step on the way to demonstrating the abstractness of the information mentally represented when one is engaged in the implicit learning procedure.
Brooks, L. (1978). Nonanalytic
concept formation and memory for instances. In
Dulany, D. E., Carlson, R. A., & Dewey, G. I. (1984). A case of syntactical learning an judgement: How conscious and how abstract? Journal of Experimental Psychology: General, 113, 541-555.
Lewicki, P., Hill, T., & Bizot, E. (1988). Acquisition of procedural knowledge about a pattern of stimuli that cannot be articulated. Cognitive Psychology, 20, 24-37.
Perruchet, P. & Pacteau, C. (1990). Synthetic grammar learning: Implicit rule abstraction or explicit fragmentary knowledge? Journal of Experimental Psychology: General, 119, 264-275.
Reber, A. S. (1967). Implicit learning of artificial grammars. Journal of Verbal Learning and Verbal Behavior, 6, 855-863.
Reber, A. S. (1969). Transfer of syntactic structure in synthetic languages. Journal of Experimental Psychology, 81, 115-119.
Reber, A. S. (1976). Implicit learning of synthetic languages: The role of instructional set. Journal of Experimental Psychology: Human Learning and Memory, 2, 88-94
Reber, A. S.
learning and tacit knowledge: An essay on the cognitive unconscious.
Redington, M. & Chater, N. (1996). Transfer in Artificial Grammar Learning: A Reevaluation. Journal of Experimental Psychology: General, 125, 123–138.
Shanks, D. R. & St. John, M. (1993). Characteristics of dissociable human learning systems. Behavioral and Brain Sciences, 17, 367-447.
This research was conducted with the assistance of an Individual Research Grant from the Natural Sciences and Engineering Research Council of Canada. Correspondence concerning this research, including the computer materials used therein, may be addressed to Christopher D. Green, Department of Psychology, York University, Toronto, Ontario, M3J 1P3, CANADA; e-mail: firstname.lastname@example.org; World Wide Web homepage: http://www.yorku.ca/faculty/academic/christo/.