Studies of AGL have frequently used training and test stimuli that might provide multiple cues for learning, raising the question what subjects have actually learned. Using a selected subset of studies on humans and non‐human animals, Trotter et al. demonstrate how a meta‐analysis can be used to identify relevant experimental variables, providing a first step in asssessing the relative contribution of design features of grammars as well as of species‐specific effects on AGL.

The suitability of the artificial grammar learning paradigm to capture relevant aspects of the acquisition of linguistic structures has been empirically tested in a number of EEG studies. Some have shown a syntax-related P600 component, but it has not been ruled out that the AGL P600 effect is a response to surface features rather than the underlying syntax structure. Therefore, in this study, we controlled for the surface characteristics of the test sequences and recorded the EEG before (...) and after exposure to a grammar. After exposure, a typical, centroparietal P600 effect was elicited by grammatical violations and not by unfamiliar subsequences, suggesting that the AGL P600 effect signals a response to structural irregularities. Moreover, preference and grammaticality classification showed a qualitatively similar ERP profile, strengthening the idea that the implicit structural mere-exposure paradigm in combination with preference classification is a suitable alternative to the traditional grammaticality classification test. (shrink)

In response to concerns with existing procedures for measuring strategic control over implicit knowledge in artificial grammar learning , 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 , or had the target grammar varying randomly from trial to trial 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. (shrink)

It has been suggested that external and/or internal limitations paradoxically may lead to superior learning, that is, the concepts of starting small and less is more (Elman, ; Newport, ). In this paper, we explore the type of incremental ordering during training that might help learning, and what mechanism explains this facilitation. We report four artificial grammar learning experiments with human participants. In Experiments 1a and 1b we found a beneficial effect of starting small using (...) two types of simple recursive grammars: right‐branching and center‐embedding, with recursive embedded clauses in fixed positions and fixed length. This effect was replicated in Experiment 2 (N = 100). In Experiment 3 and 4, we used a more complex center‐embedded grammar with recursive loops in variable positions, producing strings of variable length. When participants were presented an incremental ordering of training stimuli, as in natural language, they were better able to generalize their knowledge of simple units to more complex units when the training input “grew” according to structural complexity, compared to when it “grew” according to string length. Overall, the results suggest that starting small confers an advantage for learning complex center‐embedded structures when the input is organized according to structural complexity. (shrink)

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 . 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 . 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. (shrink)

Levelt, reviewing the AGL field from a psycholinguistic perspective, identifies various gaps and makes a number of concrete suggestions for improving several currently used experimental designs. He raises the question whether artificial (and natural) grammar learning is about detecting ‘rules’, as is commonly assumed, or rather the detection of a set of ‘constraints’. He cautions the community to not ignore ‘semantics’, and recommends to consider less artificial tasks, that may be needed for learning more complex (...) rules by human or nonhuman animals. (shrink)

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 expressed their subjective awareness of rules using one of five different scales: confidence ratings , post-decision wagering , feeling of warmth , rule awareness , and continuous scale . 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 a finer assessment of subjective states of awareness. (shrink)

The question of whether studies of human learning provide evidence for distinct conscious and unconscious influences remains as controversial today as ever. Much of this controversy arises from the use of the logic of dissociation. The controversy has prompted the use of an alternative approach that places conscious and unconscious influences on memory retrieval in opposition. Here we ask whether evidence acquired via the logic of opposition requires a dual-process account or whether it can be accommodated within a single (...) similarity-based account. We report simulations using a simple neural network model of two artificial grammar learning experiments reported by Higham, Vokey, and Pritchard that dissociated conscious and unconscious influences on classification. The simulations demonstrate that opposition logic is insufficient to distinguish between single- and multiple-system models. (shrink)

Zuidema et al. illustrate how empirical AGL studies can benefit from computational models and techniques. Computational models can help clarifying theories, and thus in delineating research questions, but also in facilitating experimental design, stimulus generation, and data analysis. The authors show, with a series of examples, how computational modeling can be integrated with empirical AGL approaches, and how model selection techniques can indicate the most likely model to explain experimental outcomes.

Subjects exposed to strings of letters generated by a finite state grammar can later classify grammatical and nongrammatical test strings, even though they cannot adequately say what the rules of the grammar are (e.g., Reber, 1989). The MINERVA 2 (Hintzman, 1986) and Medin and Schaffer (1978) memory‐array models and a number of connectionist outoassociator models are tested against experimental data by deriving mainly parameter‐free predictions from the models of the rank order of classification difficulty of test strings. The (...) importance of different assumptions regarding the coding of features (How should the absence of a feature be coded? Should single letters or digrams be coded?), the learning rule used (Hebb rule vs. delta rule), and the connectivity (Should features be predicted only by previous features in the string, or by all features simultaneously?) is investigated by determining the performance of the models with and without each assumption. Only one class of connectionist model (the simultaneous delta rule) passes all the tests. It is shown that this class of model can be regarded by abstracting a set of representative but incomplete rules of the grammar. (shrink)

Gervain et al. discuss both behavioral and neurophysiological AGL studies that investigate rule and structure learning processes in infants. The paper provides an overview of all the major AGL paradigms used to date to investigate infant learning abilities at the level of morpho‐phonology and syntax from a very early age onwards. Gervain et al. also discuss the implications of the results for a general theory of natural language acquisition.

Gervain et al. discuss both behavioral and neurophysiological AGL studies that investigate rule and structure learning processes in infants. The paper provides an overview of all the major AGL paradigms used to date to investigate infant learning abilities at the level of morpho‐phonology and syntax from a very early age onwards. Gervain et al. also discuss the implications of the results for a general theory of natural language acquisition.

Three experiments investigated whether a motor-linked measure (string typing speed) and an judgment-linked measure (grammatical judgment of strings) accessed the same implicit learning mechanisms in the artificial grammar learning task. Participants first studied grammatical strings through observation or through responding to each letter by typing it and then performed typing and grammatical judgment tests. Grammatical judgment test performance was better after observation than after respond learning, whereas typing test performance on higher order relations was worse (...) after observation than after respond learning (Experiment 1). Participants transferred grammatical knowledge across letter sets on the grammatical judgment test, but not on the typing test (Experiment 2). Typing speed did not differ for hits (grammatical strings classified by participants as grammatical) and misses (grammatical strings classified as nongrammatical, Experiment 3). These results are consistent with typing and grammatical judgment tests tapping independent mechanisms and indicate that implicit learning may consist of many different forms of learning rather than being a unitary learning mechanism. (shrink)

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. (shrink)

Following neural network simulations of the two experiments of Higham, Vokey, and Pritchard , Tunney and Shanks argued that the opposition logic advocated by Higham et al. was incapable of distinguishing between single and multiple influences on performance of artificial grammar learning and more generally. We show that their simulations do not support their conclusions. We also provide different neural network simulations that do simulate the essential results of Higham et al.

Much of what we know and love about music is based on implicitly acquired mental representations of musical pitches and the relationships between them. While previous studies have shown that these mental representations of music can be acquired rapidly and can influence preference, it is still unclear which aspects of music influence learning and preference formation. This article reports two experiments that use an artificial musical system to examine two questions: (1) which aspects of music matter most for (...) learning, and (2) which aspects of music matter most for preference formation. Two aspects of music are tested: melody and harmony. In Experiment 1 we tested the learning and liking of a new musical system that is manipulated melodically so that only some of the possible conditional probabilities between successive notes are presented. In Experiment 2 we administered the same tests for learning and liking, but we used a musical system that is manipulated harmonically to eliminate the property of harmonic whole-integer ratios between pitches. Results show that disrupting melody (Experiment 1) disabled the learning of music without disrupting preference formation, whereas disrupting harmony (Experiment 2) does not affect learning and memory but disrupts preference formation. Results point to a possible dissociation between learning and preference in musical knowledge. (shrink)

Page argues that localist models can be applied to a number of problems that are difficult for distributed models. However, it is easy to find examples where the opposite is true. This commentary illustrates the superiority of distributed models in the domain of artificial grammar learning, a paradigm widely used to investigate implicit learning.

Norman et al. reported that participants exposed in succession to two artificial grammars could be able to learn implicitly about them, and could apply their knowledge strategically to select which string corresponds to one of these two grammars. In this commentary, I identify an artifact that could account for the learning obtained not only in this study, but also in some previous studies using the same procedures. I claim that more methodological control is needed before jumping to conclusions (...) on the kind of strategic control that could be achieved unconsciously. (shrink)

Participants were unknowingly exposed to complex regularities in a working memory task. The existence of implicit knowledge was subsequently inferred from a preference for stimuli with similar grammatical regularities. Several affective traits have been shown to influence AGL performance positively, many of which are related to a tendency for automatic responding. We therefore tested whether the mindfulness trait predicted a reduction of grammatically congruent preferences, and used emotional primes to explore the influence of affect. Mindfulness was shown to correlate negatively (...) with grammatically congruent responses. Negative primes were shown to result in faster and more negative evaluations. We conclude that grammatically congruent preference ratings rely on habitual responses, and that our findings provide empirical evidence for the non-reactive disposition of the mindfulness trait. (shrink)

Are rules processes or similarity processes the default for acquisition of grammatical knowledge during natural second language acquisition? Whereas Pothos argues similarity processes are the default in the many areas he reviews, including artificial grammar learning and first language development, I suggest, citing evidence, that in second language acquisition of grammatical morphology “rules processes” may be the default.

Dienes & Perner argue that volitional control in artificial grammar learning is best understood in terms of the distinction between implicit and explicit knowledge representations. We maintain that direct, explicit access to knowledge organised in a hierarchy of implicitness/explicitness is neither necessary nor sufficient to explain volitional control. People can invoke volitional control when their knowledge is implicit, as in the case of artificial grammar learning, and they can invoke volitional control when any part (...) of their knowledge representation is implicit, as can be seen by examining “feeling of knowing” phenomena. (shrink)

We explore three methods for measuring the conscious status of knowledge using the artificial grammar learning paradigm. We show wagering is no more sensitive to conscious knowledge than simple verbal confidence reports but is affected by risk aversion. When people wager rather than give verbal confidence they are less ready to indicate high confidence. We introduce a “no-loss gambling” method which is insensitive to risk aversion. We show that when people are just as ready to bet on (...) a genuine random process as their own classification decisions, their classifications are still above baseline, indicating knowledge participants are not aware of having. Our results have methodological implications for any study investigating whether people are aware of knowing. (shrink)

In this article, we explore the extraction of recursive nested structure in the processing of binary sequences. Our aim was to determine whether humans learn the higher-order regularities of a highly simplified input where only sequential-order information marks the hierarchical structure. To this end, we implemented a sequence generated by the Fibonacci grammar in a serial reaction time task. This deterministic grammar generates aperiodic but self-similar sequences. The combination of these two properties allowed us to evaluate hierarchical (...) class='Hi'>learning while controlling for the use of low-level strategies like detecting recurring patterns. The deterministic aspect of the grammar allowed us to predict precisely which points in the sequence should be subject to anticipation. Results showed that participants’ pattern of anticipation could not be accounted for by “flat” statistical learning processes and was consistent with them anticipating upcoming points based on hierarchical assumptions. We also found that participants were sensitive to the structure constituency, suggesting that they organized the signal into embedded constituents. We hypothesized that the participants built this structure by merging recursively deterministic transitions. (shrink)

According to classical arguments, language learning is both facilitated and constrained by cognitive biases. These biases are reflected in linguistic typology—the distribution of linguistic patterns across the world's languages—and can be probed with artificial grammar experiments on child and adult learners. Beginning with a widely successful approach to typology (Optimality Theory), and adapting techniques from computational approaches to statistical learning, we develop a Bayesian model of cognitive biases and show that it accounts for the detailed pattern (...) of results of artificial grammar experiments on noun-phrase word order (Culbertson, Smolensky, & Legendre, 2012). Our proposal has several novel properties that distinguish it from prior work in the domains of linguistic theory, computational cognitive science, and machine learning. This study illustrates how ideas from these domains can be synthesized into a model of language learning in which biases range in strength from hard (absolute) to soft (statistical), and in which language-specific and domain-general biases combine to account for data from the macro-level scale of typological distribution to the micro-level scale of learning by individuals. (shrink)

In this article, we develop a hierarchical Bayesian model of learning in a general type of artificial language‐learning experiment in which learners are exposed to a mixture of grammars representing the variation present in real learners’ input, particularly at times of language change. The modeling goal is to formalize and quantify hypothesized learning biases. The test case is an experiment (Culbertson, Smolensky, & Legendre, 2012) targeting the learning of word‐order patterns in the nominal domain. The (...) model identifies internal biases of the experimental participants, providing evidence that learners impose (possibly arbitrary) properties on the grammars they learn, potentially resulting in the cross‐linguistic regularities known as typological universals. Learners exposed to mixtures of artificial grammars tended to shift those mixtures in certain ways rather than others; the model reveals how learners’ inferences are systematically affected by specific prior biases. These biases are in line with a typological generalization—Greenberg's Universal 18—which bans a particular word‐order pattern relating nouns, adjectives, and numerals. (shrink)

Cognitive Science, Volume 46, Issue 2, February 2022.

A significant part of everyday learning occurs incidentally — a process typically described as implicit learning. A central issue in this and germane domains such as language acquisition is the extent to which performance depends on the acquisition and deployment of abstract rules. In an attempt to address this question, we show that the apparent use of such rules in a simple categorisation task of artificial grammar strings, as reported by Shanks, Johnstone, and Staggs (1997), can (...) be simulated by means of a simple recurrent network, and may thus turn out not be incompatible with the acquisition of statistical regularities rooted in the processing of exemplars of the presented material. (shrink)