Resumen de Constraints on early word segmentation and mapping
Acquiring a language is a vast, complex process where many factors intervene. However, infants are able to acquire a basic language competence to communicate effectively in a relatively short time. Words are the smallest units with semantic content, or a meaning. Before infants start saying –or acquiring– their first words, they must have discovered them from the input they receive. However, words are not usually uttered in isolation, nor are there reliable pauses between them; but rather, speech is an ongoing sound stream from which these units must be extracted. Speech segmentation is, thus, fundamental, not only because it is a prerequisite necessary to build the receptive vocabulary, but also because it has been found to predict later vocabulary.
To build this vocabulary, language segmentation is not the only ability required. Segmented word-forms will become words with referential value. Thus, learning words requires identifying the sound sequence from the speech stream that corresponds to a word, identifying the target object from the visual world, and mapping word and label, which implies cognitive associative processes and retention in memory of this link. Traditionally, research has focused on these processes independently, with studies tapping on only word segmentation, and word-object association, but it is obvious that these processes are interconnected. Recent research has begun to adopt an integrative approach of these processes.
This thesis tries to extend the previous research on word segmentation in a natural language in two syllable-timed languages, namely Catalan and Spanish, both as an independent process, or segmentation and mapping, as a joint process. To both explore the emergence and constraints of these processes, different factors or constraints are going to be analyzed. Thus, the aims of this dissertation, described in Chapter 2, addresses three research areas, namely, (1) the segmentation ability in Catalan- and Spanish-learning infants using CVC.cv disyllabic words, and how this ability can be modulated due complexity of sentential contexts, age and bilingualism; (2) the contribution of visual information in a segmentation task; and (3) the emergence of simultaneous segmentation and mapping as a learning mechanism.
The results of the different studies have shown that monolingual and bilingual Catalan- and Spanish-acquiring infants are able to segment disyllabic words around 6.5 months of age with a familiar pattern of response. When infants around the same age were tested with a monosyllabic word aligned with visual information, they elicited a novelty pattern of response, which could be interpreted as a facilitation effect of the visual information. These results seem to indicate that segmentation in these two languages, when stimuli favor it, is an already well-established skill at a very early age. However, these results differ from those obtained in French (Nishibayashi, et al., 2015). This could probably be because segmentation cues in Catalan and Spanish differ from French, as besides the syllable, infants in the tested languages can use lexical stress, enabling to extract words earlier. Although this was expected for monosyllabic words, the consistency of syllabic structure and lexical pattern of the CVC.CV words used in our test might have favored an earlier segmentation. Moreover, when infants were tested at different ages (6.5, 8 and 9.5 months of age), a developmental change, from familiarity to novelty in the response pattern, has been found in disyllabic word segmentation from 6.5 to 9.5 months of age. Results from the developmental experiments at three ages, as well from the audiovisual experiment, support Hunter and Ames’ (1988) model and the interpretation of the authors concerning familiar and novel preference, with novel preferences indicating easiness in carrying out the task.
It should be noted that results of the studies when testing two groups of monolingual infants (Spanish and Catalan), as well as testing bilingual population acquiring these two languages, have found no delays in the segmentation ability, as probably infants acquiring rhythmically close languages such as Catalan and Spanish use the same segmentation strategies. This can be seen in the experiments assessing disyllabic segmentation at 8 months of age, as well as the experiments exploring the developmental pattern of this ability at 6.5, 8 and 9 months of age.
Results of the dual segmentation and mapping studies suggested that, even when segmentation skills could be improved by using visual information at 6 months of age, infants were only able to perform a dual segmentation and mapping task by 9 months of age, and use this ability as a word learning mechanism. However, results with preterm population indicated that the task is cognitively demanding. Although infants at younger ages or infants with neurocognitive difficulties still might carry out both abilities simultaneously when conditions are favored, it is most likely that younger infants use sequential or single segmentation and mapping, which imply less cognitive demands, and can be used as a learning mechanism, as well.
However, results of this research also have pointed out at certain limitations that infants might experience in segmentation (and mapping) ability. For example, unsuccessful results were found at 8 months of age when linguistic context in which the words are embedded exceeded infants’ capacities, suggesting an important processing constraint when using a short exposure to the material. Although positive results of dual segmentation and mapping can be found at a young age (Shukla, et al., 2011), demands of the task, when using natural speech and more than one word, exceed cognitive capacity of young and at-risk infants. Finally, at 4.5 months of age, even when segmentation skills could be improved by using visual cues, young infants with very limited language experience and cognitive maturity cannot obtain positive evidence of segmentation.
