Despite the popularity of RNN, LSTM and GRU, there was little evidence that confirmed their performance. We tried to investigate to reveal underlying processes and conditions to deal with such of them. The meanings of multi-layered RNNs might also be possible based on these findings as well.
The current study was designed to see what the representation unit of Korean Eojeols is, and to see whether the syntactic-categorically and semantically ambiguous Eojeol, which is the between-unit of words and phrases in Korean sentences, activates multiple meanings, even though the multi-meanings are associated with the different syntactic-category. Two primed-lexical decision experiments and two lexical decision experiments were performed. In Experiment 1, the semantic relatedness between the prime Eojeol and the target word was manipulated in two SOA conditions (150ms and 1000ms). The prime Eojeol was syntactic-categorically (e.g. noun or verb Eojeol) and semantically ambiguous. The results showed that all the noun and verb related meanings were activated by the syntactic-categorically and semantically ambiguous prime Eojeols at the short 150ms SOA condition, but only the noun-related meaning was activated at the long 1000ms SOA condition. In Experiment 2, the prime Eojeol, in which the fully conjugated Eojeol was not ambiguous but the embedded word or word stem in the Eojeol has many meanings, was presented as a related or unrelated with the target words. As in Experiment 1, two SOA conditions, 150ms and 1000ms, were used for the primed-lexical decision task. Differently from Experiment 1 results, only the verb-related meaning, consistent with the fully conjugated Eojeol, was activated but not in the noun-related meaning. Experiment 3 and 4 were designed to see whether the frequency of the Eojeol is the main factor to determine the lexical decision time presented in the isolated and sentence embedded form. In two experiments, the results showed the strongest effect of the Eojeol frequency among other variables to support the possibility of the Eojeol as the mental representation unit. The results of four experiments indicate that the fully conjugated form of Eojeols is the possibility of Eojeol as a lexical entry unit especially when used in the verb Eojeol.
With Korean-Chinese-English trilinguals living in north-east China as participants, three repetition priming experiments were conducted to investigate multilinguals' lexical representation and processing. In three experiments, primes were presented masked or visible. Participants were requested to perform the lexical decision task. Significant intra- and inter-language priming effects were obtained and found to be modulated by the proficiency of each language of the trilinguals. However, some of the findings were not fit with the predictions from major models of bilingualism. The predominant role of L2 revealed in this study poses a problem to the existing models of bilingualism which have given a superior role to L1. To explain the dominating effects of L2 in the triliguals, we propose that the status of L1 could be adjusted as a consequence of inter-language interactions.
Studies about cross-cultural psychology have demonstrated that East Asians make relatively broader and more complex causal attributions than Westerners. For instance, Korean participants indicated a greater number of causes could contribute to the event than Americans (Choi et al., 2003). Present study tested whether East Asians make broader cognition in reasoning about causal conditional “if A then B”. We hypothesized that East Asians may consider the possibility of precondition A (“does A happens or not”) as well as the possibility of “A causes B”. In the two on-line experiments, we recruited Americans and Japanese participants and presented eight causal conditionals, which manipulated the probability of A happens. Participants judged the probability at which a causal conditional is true or false. Japanese participants estimated conditional was true when the probability of Awas high, while Americans did not. These results indicated that Easterners might make broader cognition in causal conditional reasoning.
Previous studies revealed that internet addiction make human intelligence and cognitive abilities worse. In this study, we tried to find out how much worse using correlation analysis. K-Young test used for measuring the degree of addiction, and several behavioral experiments used for measuring intelligence and cognitive ability. Participants were a total of 300 people, and 29 people on the potential risk was included. As a results, there was a significant correlation between degree of internet addiction and working memory’s recall accuracy, visual attentional resource allocation ability, mental-attentional capacity, and mental rotation ability. And there was no significant correlation between addiction degree and motion control ability, emotional perception ability, and intelligence. This results shows that the internet addiction's harmfulness about inteligence and cognitive ability is still not certain.
The present study compared the effect of eye movements on postural control in 13 Parkinson’s disease (PD) patients (Hoehn-Yahr stage 2-3) and 13 age-matched healthy adults. Participants viewed, from a distance of 1 m, a computer display of an oscillating object that continuously changed shape and reported the number of times the shape changed over each 60 s trial. Four oscillation rates (0, 0.5, 0.8, and 1.2Hz) were presented. Postural sway data (mean position, standard deviation, and range of the excursions in the anterior-posterior and medio-lateral directions) were obtained using a wireless motion tracking system via sensors attached to each participant’s head, neck, and cervical spine. The effect of eye movement frequency on postural sway was minimal in the healthy adults, consistent with the view that postural control and suprapostural task (eye movement control) are functionally integrated to facilitate the performance of suprapostural control tasks. By contrast, PD patients’ showed greater and more variable postural sway, particularly in the anterior-posterior direction. It appears that PD patients’ impaired postural control systems are no longer integrated functionally with their eye movement control. This result suggests an additional factor to consider in PD patients who are susceptible to falls due to postural instability.
Previous studies demonstrated that some moist materials induce strong disgust. The present study investigated whether onomatopoeic words representing moisture modulate disgust evoked by visual images consisting of the moist materials. We used twenty images as stimuli, ten of which consisted of moist materials (moist images) and the rest consisted of dry materials (control images). Each picture was presented with each of three kinds of onomatopoeia (“betobeto”, “sarasara” and “tekuteku”), and participants evaluated the degree of perceived disgust for each image. The results showed that onomatopoeic words representing moisture modulate perceived disgust for both types of the images. The moist images with “betobeto” induced stronger disgust than the images with “tekuteku”. On the other hand, the dry images with “betobeto” induced stronger disgust than the images with “sarasara” and “tekuteku”. Our findings suggest that disgust is formed by integration of visual and verbal information.
画像の輝度はその画像の誘意性に相関する可能性があることが知られている(Lakens et al., 2013)。 本研究では，飲食店評価webサイトのユーザー投稿画像でも，こうした輝度と誘意性の関係が認められるかを調べることを目的とした。同webサイトからダウンロードした524枚の画像に対して，被験者は画像にある料理の食べたさと，写真としてのよさを評価(0-100)した。その結果，画像の赤色成分(R値)とwebサイトの飲食店評価との間に弱い正の相関が見られた(r =.11, p = .02)。さらに，写真のよさとwebサイトの飲食店評価との間にも弱い正の相関が見られた(r =.35, p < .001)。これらの結果は，画像の誘意性は単なる輝度のみでは決定できず，評価する対象(可食物，風景)の性質に依存する可能性を示す。また，こうした飲食店評価は掲載画像に影響を受けていることを示唆する。
静止画であるにもかかわらず動いて見える錯視が生じる知覚メカニズムについて，眼球運動が錯視量を増加させる要因であることは古くから指摘されていた。ところが，錯視運動と平行の方向に眼球運動を行うと，逆に錯視量が抑制されるというケースが近年報告された(Matsushita et al., 2013)。しかしながら，この錯視量抑制現象は着色されたグラデーションパッチを刺激として用いた実験でしか報告されていない。そこで本研究では色やグラデーションが錯視量抑制現象にとって必須であるかどうかを検討した。そのためにグレースケール化したグラデーションパッチ(実験1)および静止画が動いて見える錯視でありながらグラデーションを含まない矢印ドリフト錯視(実験2)を用いつつ，先行研究と同じパラダイムで実験を行った。錯視運動量はマグニチュード推定法により取得した。その結果，錯視量抑制現象にとって刺激の色やグラデーションが必須ではないことが示された。