【作者】 顾本柏；

【导师】 李红；

【作者基本信息】 西南大学 ， 发展与教育心理学， 2011， 博士

【摘要】 人类在与环境发生作用时,总是遵循着简约、经济的原则,通过一定的方式把许多客体划为一类。通过分类(categorization)可以将数量庞杂的信息统一编码,有序存储,利于快捷提取,极大地提高了人类的认知效率。基于相似性的分类是人类最基本最常用的分类方式之一,也是其它分类方式(比如概念)产生和发展的基础。相似性是指人在感官上对事物内在联系的一致性的认识,是知觉目标与内在源类别表征之间包含相同的特征或结构。尽管不少归纳和分类方面的研究涉及到相似性表征,但是并没有具体探讨人类相似性表征的认知机制。即使一些研究专门探讨了视知觉相似性表征,但这些研究都是考察灵长目动物的相似性表征的脑机制。由于相似性表征在人类归纳和推理等高级认知中起着基础性作用,因此很有必要对人类的相似性表征的认知和发展机制进行深入系统地探讨。基于以前的研究,本研究通过6个实验比较系统地探讨了人类客体相似性表征的心理物理特征、电生理机制以及随年龄的发展变化特点。在实验1中,从互联网上图片库中选取430幅不同动物图片,进行黑白处理。根据形状相似、关键性特征部位在图片上的构图位置相似等原则对这些图片进行配对,最后确定60对图片作为候选图片。让40名(20男20女)在校大学生从熟悉度、清晰度、轮廓相似度以及关键特征在构图中位置的相似度共4个维度对这些配对图片的相似度进行评判打分,最终23对符合选择要求,最后从这23对图片中按照分数从高到低选取10对作为本研究后续实验的标准原型图片。在实验2中,对实验1中选定的10对配对图片进行Morph处理,每对配对图片生成与其相似度(依次为90%、80%、70%、60%)不同的对称图片8张。来自某高校的50名在校大学生(25男25女)参与了本实验研究。在每个trial中,被试的任务是快速准确地判断刺激图片和这2张标准原型图片中的哪一张最像,同时电脑自动记录被试的反应时和按键反应。整个实验包括10个block,每个block只有一对标准原型图片,刺激图片为2张原型图片及其Morph产生的8张图片。结果统计和分析表明,被试的反应时随着刺激图片与原型图片相似度的增高而降低,反应概率随着刺激图片与原型图片相似度的增高而增高。曲线回归分析表明,对客体相似性的主观反应与客体相似度之间大致呈二次函数曲线关系。在实验3和实验4中,采用ERP技术探讨了成人的客体相似性表征的时间进程和电生理机制。实验3采用实验1筛选出来的10对图片作为标准原型图片进行9级Morph处理。刺激图片为2张原型图片以及对其Morph处理产生的9张相似度依次为：90%、80%、70%、60%、50%的左右对称图片共11张。来自某高校的17名在读大学生(7男10女)参与了本实验。整个实验流程和任务与实验2一致,同时同步记录被试的脑电活动。实验结果表明,在行为指标上,随着刺激图片相似度水平的增高,被试进行客体识别判断的反应速度和正确概率都逐渐提高,表现出和实验2较为一致的结果；在ERP指标上,发现脑电成分N2和P300对客体相似敏感,相似度为50%、60%、70%条件下比80%、90%、100%条件下诱发的N2波幅明显大；更有意思的是,相似度为60%的刺激图片不但比50%的刺激图片诱发的P300波幅明显小,而且比相似度为70%、80%、90%、100%的刺激图片诱发的P300波幅也明显小。这似乎表明客体相似度为60%左右很可能为客体相似性识别的一个转折点。偶极子溯源定位分析结果表明,后扣带回和海马旁回很可能涉及到相似性表征加工。在实验3的基础上,实验4选取相似度变化跨度为2%且与60%接近的55%、57%、59%、61%、63%、65%作为实验中相似度的6种水平。整个实验流程和任务与实验2、3一致,同时同步记录被试的脑电活动。行为结果表明,在反应时上,随着刺激图片与原型图片相似度的增高,反应时也随着显著变小；在反应概率上,对相似度为55%、57%、59%的刺激图片的反应概率之间没有差异,但是它们显著小于对相似度为61%、63%、65%的刺激图片的反应概率,而相似度为61%与63%的刺激图片的反应概率之间差异不显著,63%与65%差异不显著,但61%与65%之间差异显著。更重要的是,在ERP结果上发现了与行为数据同样的结果。在N2成分上相似度为55%、57%、59%的刺激图片诱发的N2波幅显著小于61%的刺激图片诱发的N2波幅,同时61%与63%的刺激图片诱发的N2波幅没有显著性差异,但是都大于65%的刺激图片诱发的N2波幅；在P300成分上,与标准原型图片的相似度为55%、57%、59%的刺激图片诱发的P300波幅显著大于61%的刺激图片诱发的波幅,同时65%的刺激图片诱发的P300波幅也显著大于61%的刺激图片诱发的波幅,但是61%与63%的刺激图片诱发的P300波幅没有显著性差异。实验4的这些行为和脑电结果一致表明相似度大约61%～63%很可能为客体相似性识别过程中的过渡范围,刺激图片相似度低于61%,由于缺乏足够的信息,无法把这张图片识别为相对应的原型图片；相似度高于65%,所包含的相似性信息能足以对刺激图片进行正确表征识别。实验5和实验6探讨了4～5岁儿童的客体相似性表征的心理物理特征和认知机制。在实验5中,实验材料、程序和任务和实验3完全相同。42名4～5岁的儿童(22男20女)参与本次实验。结果表明,和实验2、3一样,儿童的反应时随着刺激图片与原型图片相似度的增高而降低,反应概率随着刺激图片与原型图片相似度的增高而增高；对客体相似度与儿童的主观相似度判断反应概率进行曲线回归分析发现,儿童对客体相似性的主观感觉与客体相似度之间也大致呈二次函数曲线关系。为了考察儿童与成人的客体相似度表征加工机制是否完全相同,对实验3中成人的行为数据和实验5中儿童的行为数据做了统计对比。结果表明,儿童的反应时在任意相似度水平条件下都比成人的显著大,这可能与儿童的按键反应速度较慢有关。更重要的是,儿童的反应概率仅在相似度为60%时比成人的显著高,而在其它相似度水平上二者之间没有差异。这说明客体相似性识别时相似度为50%时,无论儿童还是成人从客体中都未获取足够信息进行知觉辨别加工；当相似度为60%时,儿童从客体中获取的相似性信息比成人要多,因此正确反应的概率比成人要明显高；当相似度超过70%以后,成人和儿童在客体获取的相似性信息足够能进行客体识别加工,故二者之间的反应概率没有差异(反应概率都达到90%以上)。这个结论也说明儿童的客体相似性加工能力比成人要强,尤其是在客体之间包含的相似性信息有限时。这也从另一个侧面印证了儿童和成人在分类加工中采用的加工策略不同：儿童更多采用知觉相似性进行分类加工,而成人可能更多采用抽象概念进行分类加工,这和前人的研究一致。在实验5的基础上,实验6采用与实验4完全相同的实验材料、程序和任务,进一步探讨儿童的客体相似性表征的认知机制。42名4～5岁的儿童(16男26女)参与本次实验。结果表明,儿童反应概率和实验4中成人的不同,儿童的反应时随着相似度的增加而显著减小,其中在相似度为57%到59%之间反应时的减小值最大,在59%与61%之间反应时的减小值最小。同时,除了相似度为63%和65%之间反应概率不显著外,相似度的其余条件之间差异都显著,而且从55%到57%增加量最大,随后每相邻两个水平条件的增加量没有差异。最后,也对实验4中成人的行为数据和实验6中儿童的行为数据做了统计对比。结果表明,和实验3和实验5对比结果基本趋向一致,反应时在相似度各个水平上儿童的都比成人的明显大；反应概率在相似度为55%时儿童和成人之间没有差异,在57%、59%时儿童比成人显著高,在61%、63%、65%时二者之间差异边缘显著,儿童比成人稍微高。这个结果进一步证实和推进了实验5的结论：在相似度为55%以下,儿童对客体识别的能力与成人没有差异,也就是说儿童和成人都无法从刺激图片中获得足够的信息进行识别加工；而在相似度为57%～61%时儿童对客体识别的能力明显强于成人；在相似度为61%～65%之间二者的识别能力差别逐渐减小,当达到65%以上时儿童和成人客体识别的能力基本没有差异,这也从另一方面说明客体相似性表征时,我们能把正在识别加工的靶刺激比较恒定地知觉为某一目标刺激时,靶刺激与目标刺激的相似度至少在65%以上。总之,本研究发现：对客体进行相似性表征时主观反应概率与客体相似度大致呈二次函数关系；4～5儿童比成人的客体相似性识别能力要强,尤其是在相似度约为57%～59%时儿童客体相似性识别明显比成人强；无论成人还是儿童在相似度低于55%以下都未能对客体进行识别,而至少高于65%以上两者才能对客体比较正确地进行识别；脑电成分N2和P300对客体相似性表征敏感；偶极子溯源分析表明扣带后回和海马旁回很可能涉及到客体相似性表征加工。更多还原

【Abstract】 When interacting with the environment, human beings categorize them by simple and economic principles by. The categorization process can remarkably increase the efficiency of cognition as it lends itself to orderly and quickly encoding, storing and retrieving immense and complicated information. The similarity-based categorization is one of the most common ways to classify information and lays the basis for many other categorizations, such as concepts. Similarity refers to common features or structures between perceived objects and the representation of its prototype categorization, based on which humans can obtain the essential relationship of the world. Few studies so far have been done to explore the mechanism of the similarity-based representation, even though a couple of researches about categorization mentioned this issue. Also there exist several researches which specialize in the similarity-based representation, but the research objects are the brain mechanism on primates. The core and basic role of the similarity-based representation in induction and reasoning cognition motivate us to examine its mechanism deeply and systematically. On the basis of prior studies, six experiments have been done in the present research to explore the mechanism of the similarity representation of objects in human beings in a developmental perspective with using psychophysical, electrophysiological methods.In experiment 1,430 animal pictures were selected from the internet and were decolorized so that to make them look black and white.60 pairs of pictures met the requirement of candidate pictures by matching the similarity of their shapes and the position of the key features.40 college students were asked to evaluate the similarity of the candidate pictures according to the familiarity, similarity of the shapes, similarity of the profiles and the position of the key features.23 pairs of pictures met the final requirements and 10 pairs which ranked higher scores were chosen as the standard pictures for the following experimental materials.In experiment 2,8 symmetrical pictures were generated as stimulus pictures based on the different levels of similarity (90%,80%,70% and 60%) to the standard pictures for each of the 10 pairs using Morph technique.50 college students (25 males and 25 females) participated in the experiment and were asked in each trial to judge which of the two standard pictures was more similar to a stimulus picture as quickly and accurately as possible. Reaction time and responses were recorded.2 standard pictures and the corresponding 8 morphed pictures were contained in each of the whole 10 blocks. The results showed that the reaction time while decreased and the accuracy rate increased as the similarity between the stimulus pictures and the standard pictures increased. Curvilinear regression analyzing indicated that quadratic function can explain the relation between the perceptual similarity of objects and the physical similarity of objects.In experiment3, ERP techniques were employed to explore the time course and the electrophysiological mechanism of the similarity representation of objects in adults.2 standard pictures and 9 symmetrical pictures which were generated based on the 5 levels of similarity (90%,80%,70%,60% and 50%) to the standard pictures were contained in the experiment materials.17 college students (7 males and 10 females) participated in the experiment. The procedure and the task were same as those in experiment 2. Consistent behavioral results with those in experiment 2 were obtained: The reaction speed and accuracy increased as the level of the similarity increased. The ERP data showed that N2 and P300 were sensitive to the similarity information:the levels of 50%,60%,70% elicited larger amplitude of N2 than the levels of 80%,90%, 100%. Interestingly, the level of 60% elicited smaller amplitude of P300 not only than the level of 50%, but also compared to the levels of 70%,80%,90%,100%, which seemingly indicated that the similarity level of 60% might be a turning point in the perception of similarity for humans. The results of dipole source analysis indicated that the posterior cingutate and parahippocampal gyrus might involve in the processing of similarity representation.Based on experiment 3,6 levels of similarity which were all closed to the level of 60% and differed from each other in the amount of 2% were employed as 6 conditions in experiment 4. The experiment procedure and the task were as same as those in experiment 2 and 3. Behavioral data showed that reaction time decreased significantly as the similarity level increased. In terms of the accuracy rate, no difference were found among the levels of 55%,57%,59%, but they were all lower than the levels of 61%, 63%,65%; No difference was observed between the level of 61%and 63%,63% and 65%, but the only difference was found between 61% and 65%. More importantly, interesting ERP results were observed:for the N2 component, the condition of 55%, 57%,59% induced smaller amplitude than the condition of 61%, at the same time, no difference were observed between the conditions of 61% and 63% both of which induced larger amplitude than the condition of 65%. As for the P300 amplitude, the conditions of 55%、57%、59% elicited larger amplitude than the condition of 61% did, at the same time, the condition of 65% elicited larger p300 amplitude than the condition of 61%. No difference was observed between 61% and 63%. Both the behavioral and the ERP results suggested that similarity levels around 61%-63% were likely to be the transition bound in similarity perception. When the similarity level was lower than 61%, people hardly recognized the picture as the prototype picture for a lack of sufficient information. While those levels which were higher than 65% contained enough information for people to correctly represent and recognize the morphed pictures.In Experiment 5 and 6, we explored the cognitive and electrophysiological mechanism of similarity representation of objects for children aged around 4-5 years old. The materials, the procedure and the task in experiment 5 were same as those in experiment 3. Forty-two 4-5 years old children (22 males and 20 females) participated in the experiment. The results showed the same behavioral pattern as that in experiment 3:the reaction time decreased and the accuracy rate increased as the similarity between the stimulus pictures and the standard pictures increased. Also, curvilinear regression analysis indicated that the relation between the perceptual similarity of objects and the physical similarity of objects conformed to quadratic function. To further examine the patterns between the children and the adults, we compared the behavioral results of adults in experiment 3 and those of children in the experiment 5. As a result, it was found that children responded slower than adults did, which might be due to the fact that it took longer time for them to press the keys. More importantly, the accuracy rate for the children was higher than that for the adults only when the similarity level was 60%, which suggested that neither children nor adults could discriminate objects in the perceptual level when the similarity level were around 50%; It seems children could perceive more similarity information than adults when the level was 60%. However, the advantage disappeared when the level was higher than 70% as both of the groups could easily obtain enough information for the similarity perception. The results suggested that children were better than adults to perceive similarity information especially when limited information was available. The reason might be that the two groups adopted different strategies:children were more likely to categorize objects based on perceptual similarity while adults were more likely to categorize objects based on abstract concepts.In experiment 6, we further explored the cognitive mechanism of the similarity representation of objects for children by applying the same experiment materials, procedure and task. In addition, forty-two 4 or 5-year-old children participated in the experiment (16 males and 26 females). The results showed that the reaction time decreased as the similarity level increased. The maximal and minimal decrease amounts were found respectively between the level of 59% and 61%. Meanwhile, the difference of the accuracy rate were observed among all the levels except between the level of 63% and 65%, among them, the increase amount reached highest between the levels of 55% and 57%. No difference was found for the increase amount between any other following two consecutive levels. Finally, we statistically compared the behavioral data in experiment 6 and the one in experiment 4. The results agreed with the comparison results betweent the experiment 3 and 5:children reacted faster than adults in all levels of similarity. As for the accuracy rate, no difference was observed when the rate was 55%. And children perceived more accurately than adults when the level of similarity was 57%. A marginally significant difference was found between the two groups when the level was 61%,63% and 65%. The pattern further confirmed the results obtained in experiment 5:When the level of similarity was lower than 55%, hardly could any people recognize the objects and when the level was around 57%-61%, children demonstrated better capability to perceive similarity than adults; And when the level reached 61%-65%, the difference between the two groups disappeared. It is possible that the similarity level should at least reach 65% for people to stably perceive the target-object as some prototype object when representing the similarity of objects.To sum up, in the present research, we find that the relation between the perceptual similarity of objects and the physical similarity of objects conform to quadratic function; 4-5-year old children are better to perceive the similarity of objects than adults do, especially when the similarity is 57%-59%. When the level of similarity is lower than 55%, hardly can any people recognize the objects and when the level is higher than 65%, both children and adults are equally able to perceive objects correctly; The ERP component N2 and P300 are sensitive to the similarity representation. Furthermore, a hypothesis needs to be inspected in the future research in which the posterior cingutate and parahippocampal gyrus might be involved in the processing of similarity perception.更多还原