ABSTRACT
Los compuestos verbo-nombre del español presentan la particularidad de carecer de un núcleo que aporte los rasgos morfológicos, de categoría o semántico-referenciales. El acceso al significado depende entonces de la relación que mantienen los constituyentes entre sí y no a partir de la identificación de un elemento nuclear, como ocurre en compuestos que poseen un lexema referencial (telaraña, bocacalle). Dentro del compuesto verbo-nombre, la relación semántica que se establece entre los constituyentes no es única. Si bien la estructura semántica prototípica responde al patrón agente-paciente (cortacésped), existen otras posibilidades argumentales, como las relaciones locativas (pasacalle). El estudio realizado tuvo como objetivo determinar hasta qué punto la estructura argumental proyectada por el verbo tiene una influencia en el procesamiento cognitivo de estas unidades. Primeramente, se llevó a cabo un juicio de aceptabilidad para asegurar que los estímulos se correspondieran con la realidad lingüística de los participantes del estudio experimental. A continuación, se administró una prueba de decisión léxica con compuestos que poseían distintos tipos de estructura argumental: (1) agente / paciente (algo que V a N, abrelatas), (2) agente / paciente menos prototípica (procesos metafóricos, chupasangre) y (3) locativos (lugar donde x hace V a N, guardamuebles). Los resultados muestran que los tiempos de decisión (respuesta) ante compuestos locativos fueron significativamente mayores que ante los prototípicos. Este resultado no puede ser explicado por diferencias en longitud o frecuencia de los compuestos o sus constituyentes, por lo que parecen apoyar la hipótesis de que la estructura argumental juega un rol central en el procesamiento de estas palabras.
Research on the processing of compound words offers important insights on how the mental lexicon is organized. It is a current topic in psycholinguistics if compound words are represented and processed as unitary lexical units (full-listing models) or only as individual constituents analyzed via acombinatorial mechanism (full-parsing models). There is enough experimental evidence that both mechanisms are involved (dual-route models). Several characteristics of the stimuli, like length, morphological family size, frequency of compounds and their constituents are important factors to determine how they are processed. Compound words are meaningful units that contain smaller meaningful units. Therefore, in the domain of compounds' studies, several hypotheses have been proposed to explain how only one interpretation is achieved from two independent meanings. Models that describe the construction of lexical semantic features in compound words, like APPLE - Automatic Progressive Parsing and Lexical Excitation (Libben 1994,1998) or CARIN - Competition Among Relations in Nominals (Gagné, 2000), are based on the notion of a morphological head. According to these theories the recognition of the head would trigger an interpretation of the whole word. In noun-noun compounds (pez espada, 'swordfish', telaraña 'spiderweb'), in which the head has the referential features, the identification of this head and the posterior clarification of the relationship with the non-head lexeme is the way to interpret the whole compound (pez 'fish' and tela 'web', respectively). However, not every compound has a head with the referential attributes inside. Verb-noun compounding is an extremely productive word-formation process in Romance languages. Spanish verb-noun compounds have the particularity of being exocentric: these constructions do not present a categorical, morphological or semantic head. Therefore, access to the meaning depends on the relationship between both of their constituent lexemes. For these units, the traditional distinction between semantically transparent and semantically opaque compounds is not suitable, because it doesn't take under consideration the projection of the argument structure by the verbal constituent. The semantic relationship established between the lexemes within a compound is not unique. While the prototypical semantic structure responds to the agent-patient pattern (cortacésped), there are other semantic possibilities, such as locative relationships (pasacalle). The present study addresses the issue of the comprehension of Spanish verb-noun compounds in order to provide evidence about the role of the argument structure projected by the first lexeme in the whole-word meaning. It is proposed that the argument structure of the verbal constituent has a cognitive influence on the processing and comprehension of these units. Firstly, an acceptability judgment test was administrated in order to identify a group of verb-noun compounds that were adequate for the Argentinean Spanish lexicon. Secondly, a lexical decision task was conducted with the stimuli selected as acceptable. Thirty native speakers (20 females), ranged in age from 19 to 34 years old, with at least 12 years of schooling, participated in the experiment. The lexical decision task included three types of compounds according to their argument structure: (1) Agent / patient (abrelatas), (2) Agent / patient with less prototypical features or metaphorical processes (chupasangre), and (3) locatives (guardamuebles). Stimuli were matched according to the whole-word and constituent frequency and length. For the statistical analysis, ANOVAs were calculated for error rates and response times (RTs) for each condition. Results show that reaction times (answers) to locative compounds were significantly higher than to agent-patient compounds. This contrast cannot be explained by differences in frequency or length, of the compounds or their constituents. Consequently, the present results seem to support the hypothesis that argument structure plays a central role in the processing of these words.
ABSTRACT
La formación de clases de equivalencia entre estímulos ha sido propuesta en el campo del Análisis Experimental del Comportamiento como un prerrequisito conductual para el lenguaje. Adicionalmente, existe evidencia de que la transferencia de función entre estímulos equivalentes podría explicar la adquisición de estructuras sintácticas simples. No obstante, la simplicidad de las funciones sintácticas estudiadas no capturó la complejidad combinatoria de la gramática natural. Si la transferencia de funciones en clases de equivalencia es un modelo posible del desarrollo de estructuras gramaticales, debería ser verificado en contextos más válidos para el estudio del lenguaje. Los objetivos del trabajo realizado fueron los siguientes: (1) analizar la transferencia de funciones sintácticas en clases de equivalencia en un contexto válido para el estudio de la adquisición de reglas gramaticales, utilizando para ello el paradigma de gramáticas artificiales y (2) analizar los potenciales cerebrales relacionados con el procesamiento de esta transferencia de función, en secuencias gramaticales y no gramaticales. Se encontró evidencia comportamental de transferencia de función en un subgrupo de los sujetos experimentales. El potencial P600, típicamente asociado al costo de integración sintáctica en contextos lingüísticos, fue observado en estos sujetos ante: violaciones gramaticales con estímulos originales de la gramática artificial y secuencias con estímulos relacionados por equivalencia (gramaticales y no gramaticales). Se interpretó que el procesamiento de las secuencias artificiales implicó mecanismos neurobiológicos similares a los asociados a la sintaxis del lenguaje y que el patrón de actividad P600 observado puede ser explicado por el aumento del costo de integración de los estímulos al contexto previo.
Stimulus equivalence class formation has been proposed as a behavioral prerequisite for language within the field of experimental analysis of behavior. Additionally, there is evidence that transfer of function among equivalent stimuli may explain acquisition of simple syntactic structures. However these experiments analyzed sequence functions that did not capture the complexity and versatility of natural grammar. If transfer of function between stimuli that belong to the same equivalence classes is indeed a useful model for the development of grammatical structures, then we should be able to verify it in a more valid context for the study of language. Artificial grammar learning tasks have been applied to the study of several aspects of language acquisition, from word segmentation to phrase structure and syntax rules. Furthermore, it has been shown that patterns of brain activity during processing of artificial grammars resemble those observed in language syntax processing. In particular, structural violations of language sentences and artificial grammar sequences both activate Broca's area. Therefore, artificial grammars provide a valid paradigm to study the learning of syntactic functions. The main objectives of the current work were: (1) to analyze transfer of function within equivalence classes in a valid context for the study of syntax acquisition, applying the artificial grammar paradigm and (2) to analyze brain potentials related to the transfer of function in grammatical and ungrammatical sequences. Fifteen subjects were trained to form two three-stimulus equivalence classes and then performed an artificial grammar learning task. One stimulus from each equivalence class was included as an item in the artificial grammar categories. During a test stage, subjects were asked to classify new artificial grammar sequences as grammatical or ungrammatical, while their EEG activity was registered. Half of these new sequences were built using the original training items and the other half contained equivalence-related stimulus. Subjects were assigned to two groups according to their performance in this test stage. Those participants whose percentage of correct responses was above 50 % were considered to pass, while those below were assigned to the fail group. We found behavioral evidence of transfer of function in the pass subgroup. These participants were able to correctly discriminate grammatical from un grammatical sequences that were built using original or equivalence-related stimulus. Event-Related potential Analysis of the EEG signal indicated a posteriorly distributed positivity with a topography and time-course similar to the P600 potential. Within linguistic contexts, P600 is interpreted as the neural correlate of prediction and integration costs during syntax processing. It has been proposed that sentence comprehension depends on predictive mechanisms that combine lexical, semantic and syntactic information from linguistic input to anticipate future words. Processing of incoming stimuli is facilitated by pre- activation, allowing rapid integration to previous context. However, when the input does not match predictions, this integration becomes slower and more difficult, requiring additional neural resources. The P600 has been considered and index of increased integration costs, generated by unfulfilled predictions of word category and morphology based on previous context. In the present experiment, the P600 was observed after: grammar violations with the original artificial grammar lexicon and artificial sequences containing equivalence-related stimulus (both grammatical and ungrammatical). Results showed that artificial grammar processing involved neurobiological mechanisms that are similar to those associated in natural grammar processing. We interpreted the observed P600 pattern in terms of an increased stimulus integration cost, both in the case of grammatical and ungrammatical equivalence-related stimulus. Even though we consider that transference of function and equivalence class formations are by themselves insufficient to explain the complexity of natural grammar, we propose that this processes might be relevant to its acquisition and evolution, constituting a behavioral prerequisite for language development.
ABSTRACT
El objetivo del estudio fue comparar el rendimiento y potenciales cerebrales registrados durante el procesamiento de relaciones de equivalencia y gramáticas artificiales. El experimento se realizó en una muestra de adultos diestros sanos de entre 20 y 30 años. Se entrenó a los sujetos en dos tareas, una de aprendizaje de relaciones de equivalencia y otra de gramáticas artificiales, y se comparó el porcentaje de aciertos y los potenciales cerebrales registrados durante la fase de test de ambas. No se encontraron correlaciones en el desempeño en ambas tareas, pero sí negatividades similares en el rango de 400 a 600 ms, que fueron interpretadas como instancias del componente N400 (asociado al procesamiento lexical y semántico). Adicionalmente, se observó en la gramática artificial un componente P600 (vinculado al procesamiento sintáctico). Los resultados sugieren un solapamiento parcial en los mecanismos neurales del procesamiento de gramáticas artificiales y relaciones de equivalencia.