<p>This thesis explores the feasibility of Artificial Grammar (AG) recognition using spiking neural networks. A biologically inspired minicolumn model is designed as the base computational unit. Two network topographies are defined with different ideologies. Both networks consists of minicolumn models, referred to as nodes, connected with excitatory and inhibitory connections. The first network contains nodes for every bigram and trigram producible by the grammar’s finite state machine (FSM). The second network has only nodes required to identify unique internal states of the FSM. The networks produce predictable activity for tested input strings. Future work to improve the performance of the networks is discussed. The modeling framework developed can be used by neurophysiological research to implement network layouts and compare simulated performance characteristics to actual subject performance.</p>

abstract: Incidental learning of sequential information occurs in visual, auditory and tactile domains. It occurs throughout our lifetime and even in nonhuman species. It is likely to be one of the most important foundations for the development of normal learning. To date, there is no agreement as to how incidental learning occurs. The goal of the present set of experiments is to determine if visual sequential information is learned in terms of abstract rules or stimulus-specific details. Two experiments test the extent to which interaction with the stimuli can influence the information that is encoded by the learner. The results of both experiments support the claim that stimulus and domain specific details directly shape what is learned, through a process of tuning the neuromuscular systems involved in the interaction between the learner and the materials. / Dissertation/Thesis / M.A. Psychology 2013

This thesis explores the feasibility of Artificial Grammar (AG) recognition using spiking neural networks. A biologically inspired minicolumn model is designed as the base computational unit. Two network topographies are defined with different ideologies. Both networks consists of minicolumn models, referred to as nodes, connected with excitatory and inhibitory connections. The first network contains nodes for every bigram and trigram producible by the grammar’s finite state machine (FSM). The second network has only nodes required to identify unique internal states of the FSM. The networks produce predictable activity for tested input strings. Future work to improve the performance of the networks is discussed. The modeling framework developed can be used by neurophysiological research to implement network layouts and compare simulated performance characteristics to actual subject performance.

L'apprentissage implicite est une acquisition d'information complexe sans intention d'apprendre. Le but de cette thèse est de déterminer comment des régularités temporelles peuvent influencer l'apprentissage implicite d'une grammaire artificielle basée sur des structures de hauteur des notes. Selon la théorie de l'attention dynamique (Jones, 1976), il y a une synchronisation entre des régularités temporelles des événements externes et des oscillateurs internes qui guide l'attention à travers le temps et aide à développer les attentes perceptives et temporelles. Notre hypothèse est que des structures métriques fortes pourront stimuler l'apprentissage implicite. Nous faisons l'hypothèse que le fait de présenter des hauteurs de notes avec des structures métriques fortes permet de développer des attentes temporelles par rapport à l'arrivée du prochain évènement. Ces attentes facilitent le traitement de hauteur des notes et ensuite «boostent» l'apprentissage implicite de la grammaire artificielle. Trois études ont été réalisées pendant cette thèse. L'étude 1 était une étude comportementale dans laquelle nous avons étudié l'influence d'une présentation temporelle régulière (avec une métrique forte) vs. irrégulière sur l'apprentissage implicite d'une grammaire artificielle basée sur des structures de hauteur des notes. Les résultats ont montré que la présentation temporelle influence l'apprentissage implicite et que la présentation temporelle régulière représente un avantage pour l'apprentissage implicite par rapport à la présentation temporelle irrégulière. Dans une étude électrophysiologique (L'étude 2) nous avons étudié quelle présentation temporelle de la grammaire artificielle, rythmique avec une métrique forte ou isochrone, serait plus efficace pour apprendre implicitement la grammaire des hauteurs des notes. Les résultats électrophysiologiques ont montré que les structures métriques apportent un bénéfice supplémentaire à l'apprentissage implicite. Dans l'étude 3 nous avons étudié comment des structures métriques fortes permettent d'améliorer les capacités d'apprentissage implicite chez des patients avec des lésions dans le cortex frontal inférieur qui ont été décrits comme déficitaires pour apprendre des structures artificielles. Les résultats comportementaux et électrophysiologiques ont montré que les patients atteints de lésions dans le cortex frontal inférieur sont capables d'apprendre une nouvelle grammaire artificielle malgré leurs lésions et leur déficit syntaxique. Il pourrait être utile d'exploiter cet avantage de la présentation métrique chez les patients, pour qui un déficit de l'apprentissage implicite a été montré avec des matériaux non-métriques et non musicaux / The thesis aims to investigate how temporal regularities can influence the implicit learning of artificial pitch structures. Implicit learning refers to the acquisition of structure knowledge by mere exposure. According to the Dynamic Attending Theory proposed by Jones (Jones, 1976), internal attentional oscillators synchronize with external temporal regularities, helping to guide attention over time and to develop temporal and perceptual expectations about future events. We made the hypothesis that strongly metrical structures might boost implicit learning, and in particular, that the strongly metrical presentation of pitch structures helps listeners to develop temporal expectations about the occurrence of the next event and thus benefits to the processing of the pitch dimension, leading to better learning of the artificial material. Three studies were realized during this PhD thesis. In Study 1, we used a behavioral approach to investigate how regular and irregular temporal presentations of an artificial pitch grammar influence implicit learning. The data revealed that both types of temporal presentations can influence implicit learning, but that the regular presentation leads to an advantage over the irregular presentation. In Study 2, we used behavioral and electrophysiological methods to investigate which type of regular temporal presentation of the artificial grammar, i.e. strongly metrical or isochronous, leads to better implicit learning of pitch structures. Electrophysiological results showed that the metrical framework provided an additional benefit for the pitch structure learning. In Study 3, we investigated whether the strongly metrical presentation allows patients with left inferior frontal lesions (with previously reported deficits for implicit learning) to learn the artificial pitch grammar. Behavioral and electrophysiological results showed that patients with left inferior frontal gyrus lesions acquired the new artificial grammar despite their lesions and despite previously reported deficits in implicit learning and syntax processing of natural language. It might be useful to exploit the potential benefit of the strongly metrical presentation further in patients for who impaired IL has been shown with non-musical and non-metrical materials

Une hypothèse dominant actuellement les théories sur l’évolution des capacités syntaxiques est celle d’une spécificité humaine pour le traitement des grammaires supra-régulières. Cette hypothèse est supportée par les données comparatives actuellement disponibles, qui ne fournissent pas de démonstration non ambiguë de cette capacité chez une autre espèce. Dans cette thèse, nous avons adopté une nouvelle approche consistant à examiner si ces échecs pourraient découler de la difficulté que représente l'extraction de régularités non-adjacentes. Pour tester cette hypothèse, nous avons mené une série de quatre études chez le babouin de guinée (Papio papio) et l’humain. La première étude montre que les babouins requièrent une quantité d’exposition beaucoup plus importante que l’humain pour apprendre des associations non-adjacentes. Dans une seconde étude, les babouins ont pu généraliser des patterns basés sur une répétition adjacente ou non-adjacente d’un élément, mais ils se sont montrés davantage sensibles à ces premiers. Une troisième étude, corrélationnelle, révèle que les babouins se montrant sensibles aux régularités non-adjacentes ne sont pas ceux obtenant les meilleures performances pour l’apprentissage de dépendances adjacentes. Une dernière étude suggère que les babouins sont sensibles à une structure en miroir (impliquant des dépendances centrées-emboitées), mais pas à une structure en copie (à dépendances croisées). Ces résultats mettent au jour une importante continuité des capacités syntaxiques au sein de la lignée des primates, mais révèlent également des différences inter-spécifiques importantes dans les contraintes mnésiques pesant sur celles-ci. / A current dominant hypothesis on the evolution of syntactic abilities propose that the processing of supra-regular grammars is a unique human capacity. In support of this hypothesis, artificial grammar learning studies conducted so far do not provide unambiguous demonstration of this capacity in a non-human species. In this thesis, we adopted a new approach by studying cognitive prerequisites for supra-regular grammar processing. Our hypothesis was that these previous failures could be attributed to a bias in these species towards the exploitation of local regularities and difficulties for processing more distant relationships, rather than an inability to master supra-regular grammars. We conducted a series of experiments in Guinea baboons (Papio papio) and humans to assess this hypothesis. In a first experiment, we show that baboons need much more exposure than humans to learn non-adjacent associations. In a second study, we show that baboons can generalize patterns involving an adjacent or a non-adjacent repetition of an element, but that they are more sensitive to the former. A third, correlational, study reveal that baboons succeeding to extract non-adjacent regularities are not those showing the best performance in learning local ones. A last study suggest that baboons are sensitive to a mirror structure (involving center-embedded dependencies), but not to a copy structure (crossed dependencies). Overall, our results reveal a stronger continuity in grammar processing capacities within the primate order than previously thought, but also highlight important species differences in memory constraints.

<p>The artificial grammar learning (AGL) paradigm has been intensively researched since the 60-s. In general, these investigations attempt to study the implicit acquisition of structural regularities. Among other things, it has been suggested that the AGL paradigm can serve as a model for the process of acquiring a natural language. Thus it can serve as a well-controlled laboratory task that might be used to understand certain aspects of the process of language acquisition. For example the AGL paradigm has been used in an attempt to isolate the acquisition of syntactic aspects of language. Several experimental studies show that the participants acquire knowledge of the underlying rule system since they are able to differentiate grammatical strings from non-grammatical ones. It has been argued that the traditionally conducted AGL paradigm with grammaticality instructions might make the task explicit, at least during the test phase. In order to imitate the language learning process as close as possible, to rule out the possibility of an explicit component during the testing phase (i.e., keeping the retrieval process implicit) and to rule out explicit rule conformity or rule following, we modified the classical AGL paradigm. In a behavioural study we combined the AGL paradigm with an altered mere exposure paradigm in an attempt to better model aspects of language acquisition. We were able to show that subjects, classifying under mere exposure instructions, categorize grammatical and non-grammatical strings just as well as those solving the classification task with the grammaticality instructions. This indicates that the mere exposure version might serve as a more appropriate model for language acquisition.</p>

The artificial grammar learning (AGL) paradigm has been intensively researched since the 60-s. In general, these investigations attempt to study the implicit acquisition of structural regularities. Among other things, it has been suggested that the AGL paradigm can serve as a model for the process of acquiring a natural language. Thus it can serve as a well-controlled laboratory task that might be used to understand certain aspects of the process of language acquisition. For example the AGL paradigm has been used in an attempt to isolate the acquisition of syntactic aspects of language. Several experimental studies show that the participants acquire knowledge of the underlying rule system since they are able to differentiate grammatical strings from non-grammatical ones. It has been argued that the traditionally conducted AGL paradigm with grammaticality instructions might make the task explicit, at least during the test phase. In order to imitate the language learning process as close as possible, to rule out the possibility of an explicit component during the testing phase (i.e., keeping the retrieval process implicit) and to rule out explicit rule conformity or rule following, we modified the classical AGL paradigm. In a behavioural study we combined the AGL paradigm with an altered mere exposure paradigm in an attempt to better model aspects of language acquisition. We were able to show that subjects, classifying under mere exposure instructions, categorize grammatical and non-grammatical strings just as well as those solving the classification task with the grammaticality instructions. This indicates that the mere exposure version might serve as a more appropriate model for language acquisition.