The research reported in this thesis examines the relative contributions of top-down and bottom-up information during lexical access. I evaluate the Cohort Model of lexical access (Marslen-Wilson and Welsh, 1978; Marslen-Wilson and Tyler, 1980; and Marslen-Wilson, 1987) which specifies that the first stage in lexical access is fully autonomous and that during this stage all processing proceeds in terms of analysis of the acoustic-phonetic input. Implicit in this model is the assumption that bottom-up processing is immune to any effects of contextual or top-down information. I examine the extent to which listeners ever rely exclusively on bottom-up information during lexical access and investigate this issue empirically, by measuring effects of context on both the production and the perception of words in various contexts. I test the hypothesis that a word uttered in a constraining context will be acoustically indistinguishable from its competitors by, first, measuring one acoustic parameter (VOT) across constraining and non-constraining contexts and, then, examining the intelligibility of tokens of that parameter taken from the varyingly constraining contexts. The data from these experiments suggest that the realization of VOT is not an aspect of bottom-up information which would create problems for a bottom-up processor in terms of providing ambiguous acoustic-phonetic information. I then investigate whether bottom-up processing during lexical access is immune to effects of context. Following Grosjean (1980) and Tyler (1984), I utilize the Gating Paradigm. Using incongruous contexts, I argue that direct assessment of the contributions made by different information sources during lexical access can be made. By presenting bottom-up information which is inappropriate to the contextual (topdown) information, I evaluate the extent to which one information source is given priority over the other. I vary both the contextual constraints available to the listener and the acoustic clarity of bottom-up information. The observed pattern of listeners' identifications of the words suggested that whilst bottomup information was given priority, top-down information was available and was utilized during lexical access. I present data which support the working structure of the Cohort Model of lexical access. I conclude, however, that the model places disproportionate emphasis on initial bottom-up processing. It appears that top-down information is not prohibited from contributing to processing during the initial stage of lexical access.

The main goal of the present thesis is to contribute to the study of overt visual attention under natural conditions. All publications included in the thesis employ eye-tracking methodology and local image feature analysis. The particular scientific question posed, if all parts of the thesis are considered together, can be formulated as: What experimental (i.e. experimenter controlled) and pseudo-experimental (e.g. demographic characteristics such as age) variables define, quantify and set limits for bottom-up determinants of fixation election? In this summary, I will bring together the answers to this question obtained during the course of the thesis and try to elucidate their significance in relation to the puzzles and riddles of eye movement research. After presenting an overview of the field, I will show how and to what extent fixation selection is altered as a function of age and as certain stimulus components are modified. Finally, I will argue that visual attention involves varying degrees of reliance on bottom-up cues according to the attending agent’s knowledge that is relevant in a given context.

The ability to locate symbols on a visual display forms an integral part of the effective use of AAC systems. Characteristics of display design and perceptual features of symbols have been shown to influence rate and accuracy of symbol location (Thistle&Wilkinson, 2009; Wilkinson, Carlin,&Jagaroo, 2006). The current study endeavoured to compare the use of two colour-coded organisational strategies (alphabetical order and categorisation) for their effectiveness in symbol location and to investigate if some bottom-up features influenced the performance of the participants in these tasks. 114 learners in Grade 1 to 3 in a mainstream school were randomly divided into two groups. Both of the groups were exposed to two visual search tests in alternating order. The tests involved searching for 36 visual targets amongst 81 coloured Picture Communication Symbols on a computer screen in one of two colour-coded organizational methods, namely alphabetical order or categorisation. The data from the research task was collected through computer logging of all mouse selections. Findings showed that locating symbols on a computer screen with a categorisation strategy was significantly faster and more accurate than with an alphabetical strategy for the Grade 1 to 3 participants. The rate and accuracy of target symbol location in both the strategies decreased significantly as grade increased, as did the differences between rate and accuracy of target location when using the two strategies. It was also found that although the tests in this study placed heavy top-down processing demands on the participants, there was still evidence of bottom-up factors influencing their performance. Implications for display design in AAC clinical practice were discussed. / Thesis (PhD)--University of Pretoria, 2012. / Centre for Augmentative and Alternative Communication / unrestricted

Listening comprehension has been the forgotten skill in second language acquisition. However, in recent years, more and more studies have focused on listening comprehension and now acknowledge its importance in language acquisition. Empirical studies have explored how listeners use the two main listening processes (top-down processing and bottom-up processing). In this study, 31 low-proficiency level Mandarin Chinese as a foreign language (CFL) learners from Brigham Young University took the Chinese Computer Adaptive Language Test (CCALT) and listened to four listening passages, measured by idea unit analysis and local and global question types. The data from these measurements suggest that low-proficiency level CFL participants in this study used both top-down and bottom-up processing while they listened to short listening passages. The results suggest listening comprehension at various proficiency levels needs to be studied further in Chinese and with different types of listening passages.

L2 listening research has moved toward a focus on understanding the process of listening. However, there are still few detailed studies of L2 listening that reveal learners’ comprehension processes when listening to scripted and unscripted listening texts. Studies in which such processing has been discussed have lacked detailed diagnoses of how bottom-up and top-down processing interactively affect listeners’ comprehension. This study was designed to show how listeners’ process and comprehend texts, with a focus on how their bottom-up and top-down processing either assist or impede their comprehension. In this study, a group of 30 L1 Japanese university English language learners’ listening abilities were diagnosed. The 30 participants were at three listening proficiency levels—high, mid, and low—based on TOEIC listening proficiency scores. The diagnostic procedure involved participants listening to two scripted and two unscripted listening texts and then reporting what they comprehended through three tasks—L1 oral recalls, L2 repetitions, and verbal reports. Other data was also collected in the study to relate the comprehension of listening texts to other important listening-related variables including listening proficiency, lexical knowledge, listening anxiety, study abroad experience, short-term phonological memory, and working memory. The main finding of the study was that miscomprehension of listening texts was invariably multi-causal, with a combination of both bottom-up and top-down factors leading to comprehension difficulty. Although not a new finding, the study offered more detail than current research about how bottom-up and top-down processing occur interactively. Regarding the overall difficulty of the listening texts, unscripted texts were more difficult to comprehend than scripted texts, and high-proficiency participants had fewer listening difficulties overall than mid- and low-proficiency participants. Quantitative and qualitative results revealed common processing difficulties among all participants due to L1-related phonological decoding issues (e.g., /l/ vs. /r/), connected speech, unknown lexis, and a lack of familiarity with unscripted speech hesitation phenomena (e.g., um, like). Qualitative transcript examples showed how top-down knowledge influenced misinterpretations of words and phrases interactively with bottom-up information, making inaccurate understandings of listening difficult to overcome. In addition to revealing participants’ difficulties and the severity of their comprehension difficulties, the diagnostic procedure showed common strengths—key words and phrases understood well by participants. High-frequency vocabulary and shorter utterances were both shown to be comprehended well. Finally, quantitative results in the study revealed relationships of participants’ listening comprehension with other important listening related variables. Listening proficiency and listening anxiety had strong relationships with listening comprehension of the listening texts. Working memory and short-term phonological memory had no relationship with listening text comprehension. Finally, study abroad experience showed a relationship with comprehension, but with many caveats, and listening vocabulary knowledge was not related with comprehension, but again, with numerous caveats to consider. Based on the results, theoretical and pedagogical implications were posed. Theoretical implications from the study relate to the understanding of four concerns in L2 listening research. Mainly, data in the study will aid researchers’ understanding of how L2 English listeners process speech interactively (i.e., with bottom-up and top-down information) for comprehension, how L2 English listeners experience connected speech, how L2 listeners deal with unknown lexis, and how L2 listeners experience difficulties with features of unscripted speech. Pedagogical implications of the study include the need for increased teacher and learner awareness of the complexity of L2 listening, the need to have learners to track their own listening development, and the need for teachers to expose learners to unscripted listening texts and make them familiar with features of unscripted speech. Finally, suggestions for further research are posed, including conducting diagnostics assessments of L2 listening with listeners of different L1s and with more varied proficiency levels, using different diagnostic procedures to examine L2 listening comprehension, and using more instruments to understand listening-related variables’ relationships with L2 listening comprehension. / Teaching & Learning

In this paper, top-down and bottom-up processing were studied regarding their effect on speech-in-noise. Three cognitive functions were also studied (divided attention, executive functioning, and semantic comprehension), and the effect they have on the speech processing and on each other. The research questions asked were if a difference in speech-in noise perception can be observed regarding the different levels of top-down and bottom-up support, if speech-in-noise is related to any of the researched cognitive abilities, and if there exists any correlation between these abilities. The method is a within-subject experimental design, consisting of four different tests: PASAT, to measure attention, LIT, to measure semantic comprehension, TMT, to measure executive functioning and SIN, to measure speech-in noise. The results showed a significant difference between top-down and bottom-up processing, a significant difference between top-down processing in decreasing and increasing conditions could also be seen. A negative correlation between the benefit of top-down support and the semantic comprehension task was found. Regarding the cognitive abilities a few correlations were found; the semantic comprehension task had a positive correlation to both the central executive task and the attentional task, the attentional task had a negative correlation to the central executive task, and both of the central executive subtasks had a positive correlation to each other. Most of the findings were expected, built on earlier cognitive hearing theories and studies.

THIS THESIS IS IN THE EXAMINATION PROCESS The ability to locate symbols on a visual display forms an integral part of the effective use of AAC systems. Characteristics of display design and perceptual features of symbols have been shown to influence rate and accuracy of symbol location (Thistle&Wilkinson, 2009; Wilkinson, Carlin,&Jagaroo, 2006). The current study endeavoured to compare the use of two colour-coded organisational strategies (alphabetical order and subcategorisation) for their effectiveness in symbol location and to investigate if some bottom-up features influenced the performance of the participants in these tasks. 114 learners in Grade 1 to 3 in a mainstream school were randomly divided into two groups. Both of the groups were exposed to two visual search tests in alternating order. The tests involved searching for 36 visual targets amongst 81 coloured Picture Communication Symbols on a computer screen in one of two colour-coded organizational methods, namely alphabetical order or subcategorisation. The data from the research task was collected through computer logging of all mouse selections. Findings showed that locating symbols on a computer screen with a subcategorisation strategy was significantly faster and more accurate than with an alphabetical strategy for the Grade 1 to 3 participants. The differences between the rates and accuracy of target symbol location using the two strategies decreased significantly as grade increased. It was also found that although the tests in this study placed heavy top-down processing demands on the participants, there was still evidence of bottom-up factors influencing their performance. Implications for display design in AAC clinical practice were discussed. / Thesis (PhD)--University of Pretoria, 2012. / Centre for Augmentative and Alternative Communication (CAAC) / Unrestricted

本研究旨在探討不同聽力教學法對台灣國小學童聽力表現之影響，並試圖了解國小學童處理聽力之過程。本研究將七十六位國小五年級學童，其中包括五十二位男性與二十四位女性，分為「自下而上教學(bottom-up)」與「相互作用教學(interactive)」兩組，進行十五週之聽力教學。學童於聽力教學前後，均以劍橋兒童英檢初級(Starter Level)之聽力測驗進行前測與後測；同時，學童於每週聽力教學後，均須完成一份聽力小考，共計十五份。受試學童之前測、後測與十五次聽力小考皆以統計計量法加以計算，作為探討兩種聽力教學法成效之依據。此外，後測結束後，「相互作用教學(interactive)」組之學童接受問卷調查，以了解國小學童處理聽力之過程。 本研究主要發現「自下而上教學(bottom-up)」與「相互作用教學(interactive)」兩組之聽力表現並無重大差異。但整體而言，「相互作用教學(interactive)」組之平均分數較優於「自下而上教學(bottom-up)」組。問卷結果顯示，受訪之國小學童較常使用「由上而下」之訊息處理方式(top-down processing)，但大多數之受訪者表示，了解每一個單字的意義是聽力理解之關鍵因素。本研究最後提出若干相關建議以供參考。 / The purposes of this study were to identify effective teaching activities that improve elementary EFL students’ listening comprehension and to investigate when learners rely on top-down processing, bottom-up processing, and interactive processing. Participants in this study were 76 fifth graders from two classes in an elementary private school in Taipei County, including 52 males and 24 females. One of the two classes was treated with bottom-up pre-listening activities, while the other class was treated with interactive pre-listening activities. A pre-test and a post-test, adopted from the Starter level in the Cambridge Young Learners’ English Test, were implemented before and after a 15-week listening instruction. During the 15-week listening teaching, participants were also required to take a listening quiz after each session of teaching. Scores of the pre-test, post-test and the 15 listening quizzes were calculated by employing statistical measures to investigate the effectiveness of the two types of pre-listening activities, namely interactive and bottom-up pre-listening activities. Further, a questionnaire was administered for the interactive group after the post-test to investigate when learners relied on top-down processing, bottom-up processing and interactive processing. The major findings of the study are: first, results from participants’ listening scores revealed that there was no significant improvement between pre-test and post-test of the two groups although the mean scores of the interactive group were higher than that of the bottom-up group; second, results from the questionnaire indicated that participants employed top-down processing more frequently than bottom-up processing in their listening, while they viewed the understanding of each word (bottom-up processing) as essential to listening comprehension. Suggestions and implications are made at the end of the study.

Uma das tarefas básicas dos mecanismos atencionais é decidir qual a localização dentro do campo visual, em que devemos prestar atenção primeiro. Um objeto que contenha características distintas, tais como orientação, forma, cor, tamanho, brilho, textura, etc. diferentes, pode atrair a atenção de uma maneira \"bottom-up\". A informação \"top-down\" baseia-se no conhecimento prévio e tem uma grande influência nas localizações atendidas. Inspirado nos mecanismos da Atenção Visual Humana, embora sem a pretensão de simulá-la, este trabalho prevê o desenvolvimento de uma nova metodologia que integra os dois tipos de informações: \"bottom-up\" e \"top-down\". Características \"bottom-up\" são geradas a partir de Momentos e essas informações são utilizadas em mapas de saliência, enquanto que um conhecimento prévio é utilizado para gerar pistas \"top-down\". Neste trabalho, desenvolveu-se uma metodologia específica para a busca e o reconhecimento visual em cenas com múltiplos objetos, utilizando para isso uma rede \"fuzzy\" contendo três subsistemas \"fuzzy\". Dada uma imagem de entrada, o objetivo consiste em se detectar regiões que possam conter informações mais significativas, a fim de que se possa guiar e restringir processamentos mais complexos. A inclusão de mecanismos de atenção (seleção de uma região de interesse dentro da imagem) é de fundamental importância pois os resultados obtidos pelo método podem ser usados para controlar a aquisição da imagem de uma maneira dinâmica. O modelo proposto está estruturado em três estágios principais: O primeiro estágio consiste em se segmentar os objetos e extrair características globais dos mesmos baseadas principalmente na teoria dos momentos, tais como tamanho, orientação, formato e distância e também média de nível de cinza. Por intermédio da comparação de um objeto com os outros presentes na cena, características \"bottom-up\" de conspicuidade são usadas para guiar a atenção ao objeto mais diferente. Por intermédio do uso da lógica \"fuzzy\" é possível inferir com grande flexibilidade algumas regras de decisão baseadas nos princípios de percepção visual tais como as leis Gestalt. O segundo estágio consiste de um subsistema \"fuzzy top-down\" que combina diferentes características de acordo com a relevância das mesmas em diferentes tarefas. Finalmente, o terceiro estágio consiste de um subsistema \"fuzzy\" que integra as informações obtidas dos subsistemas anteriores e fornece um índice geral de saliência, e indica a provável localização do objeto a ser reconhecido. A nova abordagem foi testada com objetos geométricos levando-se em consideração as características que atraem a atenção dos serem humanos / One of the basic tasks assigned to the human attentional mechanisms is to decide which location in the visual field we must pay attention first. An object containing distinctive features (such as different orientation, shape, color, size, shine, texture, etc.) can attract attention in a bottom-up way. Top-down information is based on the previous knowledge and has a large influence on the attended locations. Inspired on human visual attention mechanisms, although it doesn\'t want simulate it, this work presents a new methodology to integrate two different kind of information: bottom-up and top-down. Bottom-up features are obtained from Moment Theory and this information is used in salience maps, while a previous knowledge is used to create top-down hints. In this work, an specific methodology to visual search and recognition was developed to be applied to scenes containing multiple objects by a fuzzy net with three fuzzy subsystems. The aim of this methodology is to detect regions that may contain the most significant information, in order to guide and to restrict most complex processing. The inclusion of attentional mechanisms (the selection of a region of interest in the image) is fundamental and can be used to control the image acquisition in a dynamic way. The proposed model is structured in three main stages. The first stage segments the objects and extracts global features of them, based on the Moment Theory such as size, orientation, shape and distance and gray level average. By comparing one object with the other ones present in the scene, bottom-up features of conspicuity are used to guide the attention to the most different object. The Fuzzy Logic allows us to infer with great flexibility some of decision rules based on the visual perception principles such as the Gestalt Laws. The second stage is a top-down fuzzy subsystem that combines different features according to the relevance of them in different tasks. Finally, the third stage is a fuzzy subsystem that integrates the information obtained from the previous sub-systems and gives us a general salience index. The new methodology was tested in geometrical objects considering the feature that attracts attention to human beings

過去的研究指出，具有威脅訊息的情緒臉孔會比正向情緒臉孔快速地吸引個體的注意力。視覺搜尋研究發現情緒臉孔的突顯程度會引發由下而上的注意力攫取，而由下而上的注意力攫取是否能夠被由上而下的因素所凌駕，一直是一個很重要的論爭，因此，本研究同時呈現兩種情緒臉孔，並以文字線索引發由上而下注意力選擇，以探討由上而下的因素是否能夠凌駕情緒臉孔之注意力攫取，並比較不同情緒臉孔攫取注意力的強度。實驗一探討文字線索是否能夠幫助注意力搜尋目標情緒臉孔。實驗二使用一致效果當作注意力攫取指標，探討當臉孔情緒與作業相關時，由上而下因素是否能夠凌駕干擾情緒臉孔為生氣臉孔或快樂臉孔之注意力攫取，以及比較生氣與快樂臉孔攫取注意力的強度。實驗三延續實驗二的研究，將臉孔情緒改為與作業無關，比較生氣臉孔與快樂臉孔攫取注意力的強度。實驗四與實驗二實驗程序相同，但增加倒立中性臉孔與生氣臉孔或快樂臉孔同時呈現情況，如此在以倒立中性臉孔為目標或干擾臉孔情況，可以比較生氣臉孔與快樂臉孔之由上而下因素注意力選擇與由下而上注意力攫取強度。本研究結果在大部分情況下並未獲得一致效果，顯示由上而下之注意力選擇會影響由下而上之情緒臉孔攫取注意力，使得干擾臉孔未能吸引注意力於其上。然而由下而上注意力攫取仍影響搜尋情緒臉孔的反應時間，結果顯示生氣臉孔與快樂臉孔競爭下，生氣臉孔的突顯程度大於快樂臉孔，說明生氣臉孔較快樂臉孔吸引注意力。 / Previous studies using visual search task showed that the saliency of the emotional faces was sufficient to induce the bottom-up attentional capture. It is a critical debate that whether top-down factors can override bottom-up attentional capture. To address this issue, we presented two emotional faces at the same display while using word cues to induce top-down attentional selection. Experiment 1 investigated the effect of word cues on the target face searching. Using congruency effect as an index of attentional capture, Experiment 2 investigated whether top-down factors could override attentional capture of the emotional distractors when emotion is task-relevant and the degree of attentional capture of angry faces was compared with that of happy faces. In Experiment 3, face emotion was changed to be task-irrelevant rather than task-relevant, and the degree of attentional capture of angry and happy faces was also contrasted. In Experiment 4, the invert neutral face was included additionally as another target or distractor. We can use the invert neutral face as the common basis to dissociate the components of top-down attentional selection and bottom-up attentional capture, and contrast the attention effects between angry and happy faces in these two aspects independently. The results showed that congruency effect could not be found in most conditions. This implies that top-down attentional selection can influence the bottom-up attentional capture of emotional faces. But reaction time for searching target was still found to be influenced by bottom-up process. When an angry face competes with a happy face, the saliency of the angry face is higher than the happy face.