The basis of transfer in artificial grammar learning

In two experiments, we examined the extent to which knowledge of sequential dependencies and/or patterns of repeating elements is used during transfer in artificial grammar learning. According to one view of transfer, learners abstract the grammar's sequential dependencies and then learn a mapping to new vocabulary at test (Dienes, Altmann, & Gao, 1999). Elements that are repeated have no special status on this view, and so a logical prediction is that learners should transfer as well after exposure to a grammar without repetitions as after exposure to a grammar with them. On another view, repetition structure is the very basis of transfer (Brooks & Vokey, 1991; Mathews & Roussel, 1997). Learners were trained on grammars with or without repeating elements to test these competing views. Learners demonstrated considerable knowledge of sequential dependencies in their training vocabulary but did not use such knowledge to transfer to a new vocabulary. Transfer only occurred in the presence of repetition structure, demonstrating this to be the basis of transfer.     

Two modes of transfer in artificial grammar learning

Participants can transfer grammatical knowledge acquired implicitly in 1 vocabulary to new sequences instantiated in both the same and a novel vocabulary. Two principal theories have been advanced to account for these effects. One suggests that sequential dependencies form the basis for cross-domain transfer (e.g., Z. Dienes, G. T. M. Altmann, & S. J. Gao, 1999). Another argues that a form of episodic memory known as abstract analogy is sufficient (e.g., L. R. Brooks & J. R. Vokey, 1991). Three experiments reveal the contributions of the 2. In Experiment 1 sequential dependencies form the only basis for transfer. Experiment 2 demonstrates that this process is impaired by a change in the distributional properties of the language. Experiment 3 demonstrates that abstract analogy of repetition structure is relatively immune to such a change. These findings inform theories of artificial grammar learning and the transfer of grammatical knowledge.     

Abstract analogies and positive transfer in artificial grammar learning.

Following Brooks and Vokey (1991), we show that positive transfer to new items generated from an artificial grammar in which the vocabulary has been changed from training to test can be based on "abstract analogy" to specific training items (specific similarity) rather than abstraction of a grammar and symbol remapping rules, even with remapping unique to each test item. The results confirm that transcendence over symbols provides no support for the implicit learning of abstract structure. Ironically, they also show that the effect of specific similarity does not depend on surface characteristics of the items, but the residual effect of grammaticality does.     

Syntactic structure and artificial grammar learning: the learnability of embedded hierarchical structures

Embedded hierarchical structures, such as "the rat the cat ate was brown", constitute a core generative property of a natural language theory. Several recent studies have reported learning of hierarchical embeddings in artificial grammar learning (AGL) tasks, and described the functional specificity of Broca's area for processing such structures. In two experiments, we investigated whether alternative strategies can explain the learning success in these studies. We trained participants on hierarchical sequences, and found no evidence for the learning of hierarchical embeddings in test situations identical to those from other studies in the literature. Instead, participants appeared to solve the task by exploiting surface distinctions between legal and illegal sequences, and applying strategies such as counting or repetition detection. We suggest alternative interpretations for the observed activation of Broca's area, in terms of the application of calculation rules or of a differential role of working memory. We claim that the learnability of hierarchical embeddings in AGL tasks remains to be demonstrated.     

Effects of grammar complexity on artificial grammar learning

The present study identified two aspects of complexity that have been manipulated in the implicit learning literature and investigated how they affect implicit and explicit learning of artificial grammars. Ten finite state grammars were used to vary complexity. The results indicated that dependency length is more relevant to the complexity of a structure than is the number of associations that have to be learned. Although implicit learning led to better performance on a grammaticality judgment test than did explicit learning, it was negatively affected by increasing complexity: Performance decreased as there was an increase in the number of previous letters that had to be taken into account to determine whether or not the next letter was a grammatical continuation. In particular, the results suggested that implicit learning of higher order dependencies is hampered by the presence of longer dependencies. Knowledge of first-order dependencies was acquired regardless of complexity and learning mode.     

Theories of artificial grammar learning

Artificial grammar learning (AGL) is one of the most commonly used paradigms for the study of implicit learning and the contrast between rules, similarity, and associative learning. Despite five decades of extensive research, however, a satisfactory theoretical consensus has not been forthcoming. Theoretical accounts of AGL are reviewed, together with relevant human experimental and neuroscience data. The author concludes that satisfactory understanding of AGL requires (a) an understanding of implicit knowledge as knowledge that is not consciously activated at the time of a cognitive operation; this could be because the corresponding representations are impoverished or they cannot be concurrently supported in working memory with other representations or operations, and (b) adopting a frequency-independent view of rule knowledge and contrasting rule knowledge with specific similarity and associative learning (co-occurrence) knowledge.     

Impaired artificial grammar learning in agrammatism

It is often assumed that language is supported by domain-specific neural mechanisms, in part based on neuropsychological data from aphasia. If, however, language relies on domain-general mechanisms, it would be expected that deficits in non-linguistic cognitive processing should co-occur with aphasia. In this paper, we report a study of sequential learning by agrammatic aphasic patients and control participants matched for age, socio-economic status and non-verbal intelligence. Participants were first exposed to strings derived from an artificial grammar after which they were asked to classify a set of new strings, some of which were generated by the same grammar whereas others were not. Although both groups of participants performed well in the training phase of the experiment, only the control participants were able to classify novel test items better than chance. The results show that breakdown of language in agrammatic aphasia is associated with an impairment in artificial grammar learning, indicating damage to domain-general neural mechanisms subserving both language and sequential learning.     

Measuring strategic control in artificial grammar learning

In response to concerns with existing procedures for measuring strategic control over implicit knowledge in artificial grammar learning (AGL), we introduce a more stringent measurement procedure. After two separate training blocks which each consisted of letter strings derived from a different grammar, participants either judged the grammaticality of novel letter strings with respect to only one of these two grammars (pure-block condition), or had the target grammar varying randomly from trial to trial (novel mixed-block condition) which required a higher degree of conscious flexible control. Random variation in the colour and font of letters was introduced to disguise the nature of the rule and reduce explicit learning. Strategic control was observed both in the pure-block and mixed-block conditions, and even among participants who did not realise the rule was based on letter identity. This indicated detailed strategic control in the absence of explicit learning.     

Visual feature learning in artificial grammar classification

The Artificial Grammar Learning task has been used extensively to assess individuals' implicit learning capabilities. Previous work suggests that participants implicitly acquire rule-based knowledge as well as exemplar-specific knowledge in this task. This study investigated whether exemplar-specific knowledge acquired in this task is based on the visual features of the exemplars. When a change in the font and case occurred between study and test, there was no effect on sensitivity to grammatical rules in classification judgments. However, such a change did virtually eliminate sensitivity to training frequencies of letter bigrams and trigrams (chunk strength) in classification judgments. Performance of a secondary task during study eliminated this font sensitivity and generally reduced the contribution of chunk strength knowledge. The results are consistent with the idea that perceptual fluency makes a contribution to artificial grammar judgments.     

Recoding and representation in artificial grammar learning

We apply an exemplar model of memory to explain performance in the artificial grammar task. The model blends the convolution-based method for representation developed in Jones and Mewhort’s BEAGLE model of semantic memory (Psychological Review 114:1–37, 2007) with the storage and retrieval assumptions in Hintzman’s MINERVA 2 model of episodic memory (Behavior Research Methods, Instruments, and Computers, 16:96–101, 1984). The model captures differences in encoding to fit data from two experiments that document the influence of encoding on implicit learning. We provide code so that researchers can adapt the model and techniques to their own experiments.     

Goal relevance and artificial grammar learning

This investigation used a newly developed artificial grammar learning (AGL) paradigm in which participants were exposed to sequences of stimuli that varied in two dimensions (colours and letters) that were superimposed on each other. Variation within each dimension was determined by a different grammar. The results of two studies strongly suggest that implicit learning in AGL depends on the goal relevance of the to-be-learned dimension. Specifically, when only one of the two stimulus dimensions was relevant for their task (Experiment 1) participants learned the structure underlying the relevant, but not that of the irrelevant dimension. However, when both dimensions were relevant, both structures were learned (Experiment 2). These findings suggest that implicit learning occurs only in dimensions to which we are attuned. Based on the present results and on those of Eitam, Hassin, and Schul (2008) we suggest that focusing on goal relevance may provide new insights into the mechanisms underlying implicit learning.     

The role of similarity in artificial grammar learning

The authors examine the role of similarity in artificial grammar learning (AGL; A. S. Reber, 1989). A standard finite-state language was used to create stimuli that were arrangements of embedded geometric shapes (Experiment 1), connected lines (Experiment 2), and sequences of shapes (Experiment 3). Main effects for well-known predictors from the literature (grammaticality, associative global and anchor chunk strength, novel global and anchor chunk strength, length of items, and edit distance) were observed, thus replicating previous work. However, the authors extend previous research by using a widely known similarity-based exemplar model of categorization (the generalized context model; R. M. Nosofsky, 1989) to fit grammaticality judgments, by nested regression analyses. The results suggest that any explanation of AGL that is based on the existing theories is incomplete without a similarity process as well. Also, the results provide a foundation for further interpreting AGL in the wider context of categorization research.     

Conscious and unconscious thought in artificial grammar learning

Unconscious Thought Theory posits that a period of distraction after information acquisition leads to unconscious processing which enhances decision making relative to conscious deliberation or immediate choice (Dijksterhuis, 2004). Support thus far has been mixed. In the present study, artificial grammar learning was used in order to produce measurable amounts of conscious and unconscious knowledge. Intermediate phases were introduced between training and testing. Participants engaged in conscious deliberation of grammar rules, were distracted for the same period of time, or progressed immediately from training to testing. No differences in accuracy were found between intermediate phase groups acting on decisions made with meta-cognitive awareness (either feeling-based intuitive responding or conscious rule- or recollection-based responding). However, the accuracy of guess responses was significantly higher after distraction relative to immediate progression or conscious deliberation. The results suggest any beneficial effects of 'unconscious thought' may not always transfer to conscious awareness.     

Individual behavior in learning of an artificial grammar

Artificial grammar learning (AGL) is a widely used experimental paradigm that investigates how syntactic structures are processed. After a familiarization phase, participants have to distinguish strings consistent with a set of grammatical rules from strings that violate these rules. Many experiments report performance solely at a group level and as the total number of correct judgments. This report describes a systematic approach for investigating individual performance and a range of different behaviors. Participants were exposed to strings of the nonfinite grammar A ⁿ B ⁿ . To distinguish grammatical from ungrammatical strings, participants had to pay attention to local dependencies while comparing the number of stimuli from each class. Individual participants showed substantially different behavioral patterns despite exposure to the same stimuli. The results were replicated across auditory and visual sensory modalities. It is suggested that an analysis that looks at individual differences grants new insights into the processes involved in AGL. It also provides a solid basis from which to investigate sequence-processing abilities in special populations, such as patients with neurological lesions.     

Rule-based and exemplar-based classification in artificial grammar learning

In this study, we examined the induction of syntactic rules, given the presentation of letter strings generated from a finite-state grammar. Our primary interest was whether application of abstracted syntax or analogy to remembered exemplars could serve as a basis for judgments of grammaticality of novel stimuli. The grammatical status of test items and their objective similarity to studied exemplars were manipulated independently to investigate whether rule-based or instance-based information was a more important determinant of classification performance. When group data were examined, the results indicated that both factors were equally important in influencing grammaticality judgments about novel letter strings. There were, however, large individual differences in the magnitude of grammatical status effects, with a subgroup of subjects clearly using a classification strategy other than analogy to remembered exemplars. The results offer qualified support for the hypothesis (Reber & Allen, 1978) that rule-based information can be implicitly abstracted given limited experience with richly structured stimulus domains, and these results are inconsistent with a strong version of the instance-based model of categorization.     

Role of selective attention in artificial grammar learning

To investigate the role of selective attention in artificial grammar (AG) learning, participants were presented with “GLOCAL” strings—that is, chains of compound global and local letters. The global and local levels instantiated different grammars. The results of this experiment revealed that participants learned only the grammar for the level to which they attended. The participants were not even able to choose presented but unattended strings themselves. These results show that selective attention plays a critical role in AG learning.     

Stimulus set size and statistical coverage of the grammar in artificial grammar learning

Adults and children acquire knowledge of the structure of their environment on the basis of repeated exposure to samples of structured stimuli. In the study of inductive learning, a straightforward issue is how much sample information is needed to learn the structure. The present study distinguishes between two measures for the amount of information in the sample: set size and the extent to which the set of exemplars statistically covers the underlying structure. In an artificial grammar learning experiment, learning was affected by the sample’s statistical coverage of the grammar, but not by its mere size. Our result suggests an alternative explanation of the set size effects on learning found in previous studies (McAndrews & Moscovitch, 1985; Meulemans & Van der Linden, 1997), because, as we argue, set size was confounded with statistical coverage in these studies. nt]mis|This research was supported by a grant from the Netherlands Organization for Scientific Research. We thank Jarry Porsius for his help with the data analyses.     

Intentional control based on familiarity in artificial grammar learning

It is commonly held that implicit learning is based largely on familiarity. It is also commonly held that familiarity is not affected by intentions. It follows that people should not be able to use familiarity to distinguish strings from two different implicitly learned grammars. In two experiments, subjects were trained on two grammars and then asked to endorse strings from only one of the grammars. Subjects also rated how familiar each string felt and reported whether or not they used familiarity to make their grammaticality judgment. We found subjects could endorse the strings of just one grammar and ignore the strings from the other. Importantly, when subjects said they were using familiarity, the rated familiarity for test strings consistent with their chosen grammar was greater than that for strings from the other grammar. Familiarity, subjectively defined, is sensitive to intentions and can play a key role in strategic control.     

Subjective measures of consciousness in artificial grammar learning task

Consciousness can be measured in various ways, but different measures often yield different conclusions about the extent to which awareness relates to performance. Here, we compare five different subjective measures of awareness in the context of an artificial grammar learning task. Participants (N=217) expressed their subjective awareness of rules using one of five different scales: confidence ratings (CRs), post-decision wagering (PDW), feeling of warmth (FOW), rule awareness (RAS), and continuous scale (SDS). All scales were equally sensitive to conscious knowledge. PDW, however, was affected by risk aversion, and both RAS and SDS applied different minimal criteria for rule awareness. CR seems to capture the largest range of consciousness, but failed to indicate unconscious knowledge with the guessing criterion. We close by discussing the theoretical implications of scale sensitivity and propose that CR's unique features enable (in conjunction with RAS and FOW) a finer assessment of subjective states of awareness.     

The subjective experience of remembering in artificial grammar learning

The question of what processes are involved in artificial grammar learning has been the subject of a great deal of debate for nearly four decades. Neuropsychological and some behavioural data have found evidence for episodic memory processes such as recollection and familiarity in artificial grammar learning. To date all of this evidence has been found using objective techniques that do not rely on subjective reports. However, recollection and familiarity are associated with distinct phenomenal states that can be measured using the well-known “remember”–“know” procedure. This paper reports three experiments in which evidence was found for recollection and familiarity-based influences in artificial grammar learning using subjective reports of the experience of remembering. Subjective reports were not related to the similarity of the test items to the study items, suggesting that they do not merely reflect levels of confidence. The data are discussed with reference to multiple process models and a modified signal-detection model.