ebook img

PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES Perceptual Similarity Across Sixteen ... PDF

98 Pages·2016·1.94 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES Perceptual Similarity Across Sixteen ...

Running head: PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES Perceptual Similarity Across Sixteen Languages Esther Schott Department of Psychology McGill University, Montreal A thesis submitted to McGill University in partial fulfillment of the requirements for the degree of Master of Science June 2016 © Esther Schott, 2016 PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES Table of Contents Abstract ........................................................................................................................................... 3 Résumé ............................................................................................................................................ 4 Acknowledgements ......................................................................................................................... 5 Contribution of Authors .................................................................................................................. 6 Introduction ..................................................................................................................................... 7 Which Languages are Perceived as Similar? .............................................................................. 8 Properties of Languages Which May Explain Perceptual Language Similarity......................... 8 The Present Studies ................................................................................................................... 14 Languages investigated ............................................................................................................. 17 Study 1 – Implicit Language Similarity ........................................................................................ 17 Method ...................................................................................................................................... 18 Results ....................................................................................................................................... 25 Discussion ................................................................................................................................. 36 Study 2 – Explicit Language Similarity ........................................................................................ 37 Method ...................................................................................................................................... 38 Results ....................................................................................................................................... 41 Additional Analysis: Language Identification Questionnaire ................................................... 48 Discussion ................................................................................................................................. 50 General Discussion ....................................................................................................................... 52 Limitations and Future Directions ............................................................................................ 59 Novel Contributions to Knowledge .......................................................................................... 60 Conclusion .................................................................................................................................... 62 References ..................................................................................................................................... 63 Appendix A – Literature Search for Language Parameters .......................................................... 68 Appendix B – Language Parameters and Sources ........................................................................ 73 Appendix C – Details of Study Design ......................................................................................... 82 Appendix D – Sources for Audiobooks and Texts Recorded ....................................................... 84 Appendix E – Characteristics of Talkers Used in Studies 1 and 2 ............................................... 91 Appendix F – Perceptual Similarity Scores of Studies 1 and 2 .................................................... 92 Appendix G – Additional Analysis: Cluster Analysis in Study 1 ................................................. 96 Appendix H – Details on the Language Identification Questionnaire .......................................... 98 2 PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES Abstract Learning perceptually similar, and thus hard to separate, languages may be different from learning perceptually dissimilar, easy to separate, languages. To test this hypothesis, a measure of perceptual language similarity is needed. The purpose of this thesis is to establish, describe and explore such a measure. In Study 1, implicit language similarity across 16 languages was tested. Adult participants were presented simultaneously with two recordings that were either of the same, or of two different languages. They responded whether they heard 1 or 2 languages. Responding “1” when two languages were presented is an indication that these languages sound similar. To understand why some language combinations were more likely to be perceived as a single language, parameters relating to properties of the listener population, language family, geographic closeness, and properties of language were investigated. In a multidimensional scaling analysis, geography and properties of the languages explained which language combinations were more likely to be perceived as a single language. To test which aspects of the results from Study 1 were task-dependent or task-independent, an explicit measure of language similarity was used in Study 2. Participants listened to each language separately and were asked to arrange languages that sound more similar closer together. The results from a multi- dimensional scaling analysis of the explicit judgments of language similarity could be explained by properties of the listener population, language family, geography, and properties of the languages. In comparison to Study 1, the explicit task may be more affected by listeners’ meta- linguistic knowledge than by properties of the language. Finding out why some languages may be perceived as more similar can shed light on which properties of languages are salient to a naïve listener. The implicit and explicit measures of language similarity described in this thesis can be used to test hypotheses about the role of language similarity in infants’ and adults’ language learning. 3 PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES Résumé L'apprentissage de langues qui sont similaires sur le plan perceptif, et conséquemment plus difficile à dissocier, devrait être différent de l'apprentissage des langues qui, elles, sont différentes sur le plan perceptif, et plus facile à dissocier. Pour vérifier cette hypothèse, une mesure de la similarité perceptuelle des langues est nécessaire. Le but de ce mémoire est d'établir, de décrire et d'explorer cette mesure. Dans l'Étude 1, la similarité implicite de 16 langues différentes a été mesurée. Deux enregistrements, provenant ou non de la même langue, ont été présentés à des participants adultes. Les participaient devaient ensuite énoncer s'ils avaient entendu 1 ou 2 langues. Si les participaient répondaient «1», cela indiquait que les langues paraissaient semblables. Afin de mieux comprendre pourquoi certaines combinaisons de langues étaient plus souvent perçues comme étant une seule langue, des paramètres relatifs aux caractéristiques de la population du participant, la famille linguistique, la proximité géographique et les caractéristiques des langues ont été examinés. À l'aide d'une analyse de positionnement multidimensionnel, il a été possible de déterminer que la proximité géographie et les caractéristiques des langues permettaient d'expliquer quelles combinaisons de langues étaient plus propices à être perçu comme une seule langue. Pour vérifier quels aspects des résultats de l'Étude 1 étaient dépendants ou indépendants de la tâche, la similarité des langues a été mesurée à l’aide d’une tâche explicite dans l'Étude 2. Les participants ont écouté chaque langue séparément et devaient subséquemment regrouper les langues en fonction de leurs similarités sonores. Les résultats issus de l'analyse de positionnement multidimensionnel des jugements explicites de similarité des langues peuvent être expliqués par les caractéristiques de la population, la famille linguistique, la géographique et les caractéristiques des langues. En comparaison avec l'Étude 1, la tâche explicite de l'Étude 2 pourrait être plus influencée par les connaissances métalinguistiques de l'auditeur que par les caractéristiques des langues. Comprendre pourquoi certaines langues seraient perçues comme plus similaires permet de mieux saisir quelles sont les caractéristiques des langues auxquelles un auditeur naïf porte attention. Les mesures implicites et explicites de la similarité des langues présentées dans cette étude peuvent être utilisées afin de mieux comprendre le rôle de la similarité des langues dans l’apprentissage des langues. 4 PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES Acknowledgements I would like to thank Kristine Onishi for her support in pursuing this thesis. A big thank you also to Mark Baldwin. As members of my thesis committee, Debra Titone and Linda Polka have provided valuable feedback, I would like to thank them as well. This project has benefitted much from a dedicated undergraduate student, Kayla Baskin, thank you for all your hard work! I want to extend a huge thanks to Giovanna LoCascio, who has been an invaluable help throughout my time in the department. I am also grateful for support from a Richard H. Tomlinson Master’s Fellowship, as well as to the resources provided by the Centre for Brain, Language and Music. I would also like to thank Amélie, Julie, Camille, and the volunteers of the MIDC lab. Thanks also to Ramsey, AJ, Zhen, Alia, Chantelle, Nida, Jonas, Karim, Geneviève, Matteo, Jane and Sebastian. A heartfelt thanks also to my brother and parents who have supported me from afar. Finally, I would like to thank the participants, and the talkers of the 16 languages. 5 PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES Contribution of Authors I, Esther Schott, conducted the literature search that forms the basis of the investigation into the language parameters that relate to perceptual language similarity, designed the studies, found and prepared the audiobooks and text stimuli to be recorded, recorded the stimuli, supervised the research assistants who listened to the stimuli and judged whether they were usable, and produced the final stimuli and programmed the studies. I supervised research assistants conducting the studies, and analyzed and interpreted the data. I wrote the manuscript, including the preparation of tables and figures, and revised the manuscript based on comments from Kristine Onishi. Kayla Baskin helped with finding audiobooks and text stimuli to be recorded, as well as helping with recording stimuli, and conducted, analyzed and interpreted pilot studies (not included in this thesis). Kristine Onishi provided feedback and supervision in all aspects of this thesis, including the design, stimulus generation, data analysis and interpretation, and provided comments on drafts of this thesis. 6 PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES Perceptual similarity across sixteen languages Introduction Bilingual infants, that is, infants growing up in a bilingual environment are faced with the task of acquiring two languages simultaneously. Although bilingual infants learn two languages while monolinguals learn only one, bilingual infants show a similar time course of language acquisition: for example, they can discriminate their two native languages a few days after birth (e.g., Byers-Heinlein, Burns, & Werker, 2010), show discrimination of phonetic contrasts (e.g., Albreda-Castellot, Pons, & Sebastián-Gallés, 2011; Sundara & Scutellaro, 2011) and are able to learn novel word-object associations at roughly the same time as monolinguals (e.g., Byers- Heinlein, Fennell, & Werker, 2012; Mattock, Polka, Rvachew, & Krehm, 2010). While this seems to indicate that bilingual infants reach their language milestones at around the same time as monolinguals, little is known about how bilinguals acquire the sounds, words and structure of each language to which they are exposed. One strategy that bilingual infants may use to acquire each of their languages is to separate the two languages in their language input (e.g., Mehler et al., 1988; Nazzi, Bertoncini, & Mehler, 1998). If the ability to separate languages is a critical process for language acquisition in bilingual infants, this may lead to differences between groups of bilingual infants. Learning two languages that are perceptually similar, and thus hard to separate, should be different from learning two languages that are perceptually dissimilar, and thus easy to separate. There is evidence that some language pairs are easier to separate than others: American infants can discriminate some languages at birth (e.g., English and Spanish; Moon, Cooper, & Fifer, 1993) and others only at 5 months (e.g., English and Dutch; Nazzi, Jusczyk, & Johnson, 2000). Presumably, languages that are discriminated later in development are harder to separate. By comparing infants who are exposed to language pairs that are easy versus hard to separate, we can test whether the ability to separate languages is important in bilingual language acquisition. To investigate the effect of language similarity, a measure of perceptual language similarity is needed. This measure of language similarity should reflect perceived similarity, rather than the structural similarity of language pairs: Infants are naïve listeners, meaning that infants are more likely to rely on the language input in their environment rather than their limited knowledge about the structure of language. 7 PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES Which Languages are Perceived as Similar? There is some indication of which languages are perceived as similar from the studies of infants’ language discrimination at birth (Byers-Heinlein et al., 2010; Dehaene-Lambertz & Houston, 1997; Mehler et al., 1988; Moon et al., 1993; Nazzi, Bertoncini, et al., 1998; Ramus, 2002b) and later in development (Bahrick & Pickens, 1988; Bosch & Sebastián-Gallés, 2001; Christophe & Morton, 1998; Dehaene-Lambertz & Houston, 1997; Molnar, Gervain, & Carreiras, 2013; Nazzi et al., 2000). However, these studies have only investigated few pairs of languages, and many of these studies are designed to specifically test the role of rhythmic class in language discrimination (e.g., Christophe & Morton, 1998; Mehler et al., 1988; Nazzi, Bertoncini, et al., 1998; Nazzi et al., 2000). Thus, the languages investigated in these studies may have been selected as maximally dissimilar on their rhythmic properties. Based on a rhythmic class account, English and Dutch both employ stress-timed rhythm, so separating English and Dutch should be harder. English and Japanese, a stress- and a mora-timed language, should be easier to separate. There are two problems with relying only on rhythmic class to explain perceptual similarity: First, some languages cannot be accurately classified as syllable-, stress- or mora-timed (Grabe & Low, 2002; Ramus, 2002a), which excludes bilingual populations exposed to these languages from being studied. Second, while rhythmic class has received much attention in infants’ perception of languages, rhythmic class is just one of many properties of a language. In adults, other factors, such as language family, geographical origin, sounds and prosody have been studied, and may also affect infants’ perception of language similarity. Investigating the role of these factors in perceptual language similarity may help understand why some languages are perceived as more similar than others. Properties of Languages Which May Explain Perceptual Language Similarity Language family. Languages may be perceived as more similar if they are from the same language family. Linguists classify languages with an assumed common language ancestor as part of the same language family (McMahon & McMahon, 2005). Presumably, languages from the same language family are more perceptually similar as a result from their common language ancestor, from which languages in the same family may have inherited the same or similar sounds, words or prosody. Some studies found an effect of language family on the perception of languages (Bradlow, Clopper, Smiljanic, & Walter, 2010; Maddieson & Vasilescu, 8 PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES 2002). For example, French native speakers were trained to identify a small set of languages and then listened to new talkers of these languages intermixed with new languages. Participants were then asked to indicate whether the presented language sounded like one of the languages they had previously been trained on. When hearing Hausa, an Afro-Asiatic language, participants often responded that the language sounded like Amharic, also an Afro-Asiatic language (Maddieson & Vasilescu, 2002). Yet in that same study, other combinations that were often confused, such as Korean and Cantonese are not from the same language family, showing that that languages that are perceived as similar are not necessarily from the same language family. When the authors tested a group of American listeners with a background in linguistics, the pairs of languages that these listeners confused could be better explained by language family. This indicates that the effect of language family on perceptual similarity may be mediated by meta- linguistic knowledge rather than perceived similarity. In a different study, that uses a task that is potentially less affected by meta-linguistic knowledge, no effect of language family was found (Meyer, Pellegrino, Barkat-Defradas, & Meunier, 2003). Taken together, these findings suggest that being from the same language family is not associated with languages being more perceptually similar. Yet, linguistically trained listeners’ confusions of languages indicate that these listeners were likely to confuse languages from the same language family, providing weak evidence for languages from the same language family being more perceptually similar. Geographical closeness. Languages from geographical close areas may also be more perceptually similar. Geographical closeness may have allowed contact between the language populations which in turn could lead to borrowing of words or other linguistic structures (Enfield, 2005), and may ultimately lead to geographically close languages sounding more similar than geographically distant languages. There is evidence that languages from the same geographic region are more perceptually similar (Bradlow et al., 2010; Maddieson & Vasilescu, 2002; Muthusamy, Jain, & Cole, 1994). In one study, English native speakers listened to recordings of unfamiliar languages and were asked to group these languages by how similar they sound. The resulting language groupings could be explained by geographic origin, with Asian languages grouped closer to each other, as well as European languages grouped closer together (Bradlow et al., 2010). These findings suggest that geographic closeness plays a role in how perceptually similar adults judge languages. However it is unclear how adults’ explicit judgments of languages relate to infants’ perception of language. Adults have meta-linguistic 9 PERCEPTUAL SIMILARITY ACROSS 16 LANGUAGES knowledge about languages, which infants do not have, and adults’ knowledge may affect their answers when they are explicitly asked to make judgments about language similarity. Thus, an explicit task may not be adequate for testing perceptual language similarity, especially to assess language similarity as perceived by infants. In summary, there is some evidence that listeners do perceive languages that are geographically closer as more similar, at least in explicit tests of perceptual language similarity. Language family and geographic closeness may each be a reason why languages may be perceived similar, but they likely have an indirect relationship with language similarity. If languages that are from the same family or geographic area are more perceptually similar, this may actually be caused by similarities in the acoustic signal of the language, for example similarities in sounds or prosody. Sounds. Languages may be perceived as similar because they use the same sounds, or use sounds in parallel ways. Listeners have been found to pay attention to the sounds of unknown languages on several levels (Bond, Stockmal, & Muljani, 1998; Bond & Stockmal, 2002; Bradlow et al., 2010; Frota, Vigário, & Martins, 2002; Lorch & Meara, 1989; Muthusamy et al., 1994; Ramus, 2002b). Languages that share sounds have been shown to be confused which each other (Muthusamy et al., 1994), suggesting that similar sounds are associated with languages being more perceptually similar. Listeners are not just sensitive to particular sounds, but also broader classes of sounds such as the presence of front rounded vowels (Bradlow et al., 2010). Listeners are also sensitive to sequences of sounds (Meyer et al., 2003). In the sequential discrimination task in that study, naïve listeners heard two talkers, one after the other, speaking the same or different languages and were asked to say whether the talkers spoke the same or different languages. Which pairs of languages were found to be harder to discriminate could in part be explained by whether the language allowed consonant clusters, that is, sequences of consonants without intervening vowels. One caveat of using the sequential discrimination task as a measure of naïve listeners’ perception of languages is that by presenting the languages one after the other listeners have to rely on their memory when judging whether the two languages are the same. Remembering 10

Description:
Appendix F – Perceptual Similarity Scores of Studies 1 and 2 . 92 In Study 1, implicit language similarity across 16 languages was tested. Fairy tale : it jumps , gold shoes, white shoes http://Edu.Pcbaby aimenu-ete-aimete-lo-nmu-oran-ba-ereke-the-tortoise-and-the-.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.