认知与发展心理学研究室知识库

为了研究Ebbinghaus 错觉的产生机制，我们结合心理物理法、事件相关电位、功能性磁共振成像和近红外光谱成像等方法，具体考察了四个问题：（1）Ebbinghaus 错觉的产生是否需要意识参与；（2）三维空间下的Ebbinghaus 错觉；（3）运动相关的Ebbinghaus 错觉；（4）基因与环境在Ebbinghaus 错觉加工中的作用。
研究一采用持续性闪烁抑制范式（continuous flash suppression paradigm，CFS）及后向掩蔽范式（backward masking paradigm）考察了意识下的Ebbinghaus错觉加工。在实验1 中，我们首先测量了Ebbinghaus 错觉效应量，发现当外周诱导刺激是圆时，无论测验刺激是圆还是不规则图形错觉效应均显著，但是前者的效应量显著大于后者，此结果重复了前人的研究，说明所采用的实验刺激有效。在实验2 中，我们采用CFS 来掩蔽测验刺激，但是外周的诱导刺激仍然可见，结果发现当测验刺激是圆时，看起来更大的圆（被小诱导圆围绕）更容易突破抑制进而达到意识水平，但是当测验刺激是不规则图形时则没有观察到这一现象。这说明Ebbinghaus 错觉可以在意识下得到表征，并且这一表征可能发生在比较早的加工阶段。实验3 进一步排除了反应偏向这一因素在实验2 中的作用。在实验4 中，我们采用后向掩蔽来使外周的诱导刺激不可见，但测验刺激可见，结果仍然可以观察到显著的错觉效应。这一结果进一步验证了Ebbinghaus 错觉可以在意识下得到表征。
研究二采用EEG 考察了三维空间内的Ebbinghaus 错觉。我们引入双眼视差这一深度线索，使用立体镜呈现刺激，通过变化外周四个诱导圆的水平视差来使测验圆看起来在诱导圆的前面、后面，或者与诱导圆在一个平面上，而实际上测验圆在屏幕上的物理位置并没有发生改变。我们假设当测验圆与外周诱导圆在一个平面时既包含低水平成分（lateral inhibition）也包含高水平成分（sizecontrast）；当测验圆与外周诱导圆看起来不在一个平面时，Ebbinghaus 错觉中的高水平成分将被大大减弱或者消失，但低水平成分不受影响。实验5 的结果表明，无论测验圆是否与诱导圆在同一平面上，错觉效应量均显著，但是当测验圆与诱导圆不在同一平面上（有深度线索）时的错觉效应量显著小于它们在同一平面（无深度线索）时。为了验证有深度线索时错觉效应量的减少不是由于外周诱导圆在屏幕上的水平位移导致的，实验6 不使用立体镜，而是把实验5 中单眼内的刺激呈现在屏幕中央，这样被试的双眼同时看见相同的刺激，结果发现与实验5 中的三种条件相对应的条件下均观察到显著的错觉效应，并且与实验5 中有深度线索相对应的条件下的错觉效应量显著增加，这说明实验5 中有深度线索时错觉效应量的减少不是由于外周诱导圆的水平位移导致的。实验7 的EEG 结果表明，在错觉图形呈现的整个时间窗内，无论是否引入深度线索，看起来更大的测验圆都在顶枕叶区域引发更大的alpha 频段（8-13 Hz）的振幅,并且右侧顶枕叶区域内的alpha 振幅与三种条件的错觉效应量的均值呈显著的负相关（与有深度线索时的错觉效应量也有负相关的趋势）；在错觉图形呈现的晚期时间窗内，在无深度线索时，看起来更大的测验圆在中央顶区引发了更大的beta 频段（14-30Hz）的振幅。另外，beta 与alpha 振幅比值的绝对值与无深度线索和有深度线索时的错觉效应量之差呈显著的正相关。并且alpha 振幅与beta 和alpha 振幅比值的绝对值这两个指标能够解释无深度线索时错觉效应量中53%的变异量。综上所述，研究二发现了Ebbinghaus 错觉中的低水平成分与高水平成分在神经生理学上的证据，即alpha 活动与低水平成分有关，beta 活动与高水平成分有关，并且beta 活动会受到alpha 活动的调节。
研究三采用fMRI 技术考察了当测验刺激具有某种操作属性时错觉效应量的变化。测验刺激包括两种：西瓜与篮球。与西瓜相比，篮球具有更确定的操作属性。实验8 的行为结果表明，西瓜与篮球的错觉效应均显著，但是前者显著大于后者。我们猜测篮球错觉效应量的减少可能是由于它的操作属性引起的。fMRI结果首先发现功能定义的初级视皮层（primary visual cortex, V1）的表面积与Ebbinghaus 错觉效应量之间存在显著的负相关，并具有左脑偏侧化的特点，这与前人研究一致。另外，我们还发现结构定义的V1 的皮层厚度与错觉效应量之间存在显著的正相关。全脑分析表明，当测验刺激被小诱导圆包围时，篮球显著激活了左侧内顶叶（left inferior parietal lobule, lIPL），这是加工客体操作属性的核心脑区。以V1 作为种子的PPI 分析表明，当测验刺激是篮球时，我们发现了与错觉效应有关的V1 和lIPL 之间的显著连接。DCM 结果表明，lIPL到V1 的反馈连接强度与篮球和西瓜的错觉效应之差存在显著的负相关。综上所述，研究三的结果表明，Ebbinghaus 错觉可以在V1 上得到表征；客体的操作属性能够调节Ebbinghaus 错觉加工，并且这种调节依赖于lIPL 到V1 的反馈连接。
研究四结合近红外光谱成像技术和双生子考察了基因与环境在Ebbinghaus错觉加工中的作用。实验9 的行为结果表明，遗传因素可以解释Ebbinghaus 错觉中61.27%的变异量。任务前的静息态数据表明三个感兴趣区域（顶枕区、左\右侧顶颞区）之间的时间序列连接强度均显著，并且顶枕区与左侧顶颞区间的连接强度具有遗传性（29.44%），另外，这两个区域间的连接强度与行为上的错觉效应量呈显著的负相关。对任务态的数据分析发现错觉效应在左侧顶颞区具有遗传性（26.22%），并且低估成分在顶枕区具有遗传性（40.06%），高估成分在左侧顶颞区具有遗传性（42.79%）。研究四的结果表明Ebbinghaus 错觉具有很高的遗传性，并且负责加工Ebbinghaus 错觉的大脑活动也具有一定的遗传性。
综上所述，Ebbinghaus 错觉既可以在较早的加工阶段（如意识下）和较早的视觉加工通路（如V1）内表征，也可以接受来自背侧通路（如lIPL）的调节，这进一步验证了前人关于Ebbinghaus 错觉效应包含两个成分的假设，并且我们为这一假设提供了直接证据，即低水平成分与alpha 活动有关，高水平成分与beta 活动有关。这提示我们Ebbinghaus 错觉是一种灵活的知觉现象，我们既可以任由这一现象的发生，让测验刺激更容易从背景中脱颖而出，进而得到更快的知觉加工，也可以在必要的时候（如对测验刺激进行手动操作时）减弱背景刺激的干扰，进而更好地完成任务。因此，Ebbinghaus 错觉是千百万年来生物进化并且不断适应环境的产物，具有很高的遗传性，对个体生存具有重要意义。

 
By combining the methods of psychophysics, event-related potentials (ERPs), functional MRI and functional Near Infrared Spectroscopy (fNIRS), we investigated the mechanisms of the Ebbinghaus illusion. Specifically, we probed four questions: (1) whether consciousness is necessary in the formation of the Ebbinghaus illusion; (2) the Ebbinghaus illusion in three-dimensional space; (3) action-related Ebbinghaus illusion; (4) the respective contributions of gene and environment in the formation of the Ebbinghaus illusion.
In study 1, combined with continuous flash suppression paradigm (CFS) and backward masking paradigm, we investigated the processing of the Ebbinghaus illusion without awareness. In Experiment 1, we measured the strength of the Ebbinghaus illusion effect, and found that when the surrounding inducers were circles, the illusion effect was significant, no matter the central target was a circle or an irregular figure, and the illusion effect of the former was larger than that of the latter. The results were consistent with previous studies. In Experiment 2, we used a dynamic color noise to suppress the central target but leaving the surrounding inducers visible, and found that when the central target was a circle, the circle with large perceived size (surrounded by small inducers) broke from suppression faster than the same circle with small perceived size, but there was no such difference for the irregular figure. The results suggested that the Ebbinghaus illusion could be represented without awareness, and this representation occurred in relatively early visual processing stage. Further, Experiment 3 excluded the possibility of the role of response bias in Experiment 2. In Experiment 4, we used backward masking to make the surrounding inducers invisible but leaving the central target visible, and we still observed significant illusion effect, further confirming the fact that the Ebbinghaus illusion could be represented without awareness.
In study 2, we used electroencephalograph to investigate the processing of the Ebbinghaus illusion in three-dimensional space. We introduced depth cue with binocular disparity, and presented stimuli with stereoscope. We varied the horizontal disparity of the four surrounding inducers to make the central target appear in front of, behind, or in the same plane of the surrounding inducers, though the physical location of the central target on the screen was invariant. We speculated that the illusion effect consisted of both low-level component (lateral inhibition) and high-level component (size contrast) when the target and its inducers were in the same plane, but the high-level component was largely diminished or even eliminated when they were perceived in different planes. Experiment 5 demonstrated that the strength of the Ebbinghaus illusion effect was significant, no matter the central target and its surrounding inducers were in the same or different planes, but the former was significantly larger than the latter. In Experiment 6, in order to verify that the reduction of illusion effect when the central target and its inducers were in different planes was not caused by the physical displacement of the inducers, we presented the stimuli to both eyes without stereoscope, and these same stimuli were presented to one eye in Experiment 5. We found that the illusion effect in all the three conditions corresponding to Experiment 5 was significant, and the illusion effect in the conditions corresponding to the conditions with depth cue in Experiment 5 was significantly increased, suggesting that the reduction of illusion effect when the target and inducers in different planes in Experiment 5 was not due to the physical displacement of the inducers. In Experiment 7, the results of EEG demonstrated that, in the whole time window when the illusory figure was presented, no matter the target and the inducers were perceived in the same plane or not, the target with large perceived size induced larger alpha-band power (8-13 Hz) in occipitoparietal region, and the alpha power in right occipitoparietal region was negatively correlated with the average illusion effect of the three conditions (there was marginally significant correlation with the illusion effect in the conditions with depth cue). In the late time window, when the target and the inducers were presented in the same plane, the target with large perceived size induced larger beta-band power (14-30 Hz) in the central parietal region. In addition, the log-transformed absolute ratio of beta and alpha power was positively correlated with the difference of illusion effect between the conditions of the same and different planes. Moreover, the alpha power in the whole time window and the log-transformed absolute ratio of beta and alpha power in the late time window could explain 53% variance in the behavioral illusion effect with no depth cue. In sum, we found the neurophysiological evidence of the two components in the Ebbinghaus illusion effect, i.e., alpha activity was correlated with low-level component, and beta activity was correlated with high-level component, and the beta activity was mediated by alpha activity.
In study 3, we investigated the Ebbinghaus illusion effect when the target contained manipulation information with fMRI. There were two types of targets, watermelon and basketball. The basketball was supposed to contain more specific manipulation information in contrast to watermelon. The behavioral results in Experiment 8 demonstrated that, the illusion effect of both of the watermelon and basketball was significant, but the former was larger than the latter. We conjectured that the reduction of the illusion effect for the basketball might be caused by the manipulation information contained in it. The fMRI results demonstrated that the functional defined surface area of the primary visual cortex (V1) was negatively correlated with the strength of illusion effect and this correlation was left lateralized, consistent with previous studies.
Moreover, we also found that the structure defined cortical thickness of V1 was positively correlated with the strength of illusion effect. The voxel-wise whole-brain analysis showed that when the surrounding inducers were small, comparing to watermelon, the basketball activated left inferior parietal lobule (lIPL), which was the central area processing manipulation information. The PPI analysis with the seed of V1 showed significant connections between V1 and lIPL when the target was basketball. Moreover, the DCM results demonstrated that the strength of feedback connection from lIPL to V1 was negatively correlated with the difference of illusion effect between basketball and watermelon. Above all, the Ebbinghaus illusion could be represented in V1; manipulation information could modulate the processing of the Ebbinghaus illusion, and such modulation relied on feedback projections from lIPL to V1.
In study 4, combined with fNIRS and twins, we investigated the respective contributions of gene and environment in the formation of the Ebbinghaus illusion. The behavioral results of Experiment 9 showed that, the genetic variance was 61.27% in the Ebbinghaus illusion. The analysis of resting state data before task demonstrated that the time course correlation between the three regions of interest (parietoccipital, left/right parietotemporal) reached significance, and the genetic variance in the connection between parietoccipital and left parietotemporal was 29.44%. Moreover, the connection strength between these two areas was negatively correlated with the behavioral illusion effect. The analysis of task-related fNIRS data found that the genetic variance of illusion effect was 26.22% in the left parietotemporal region, and the genetic variance was 40.06% in the parietoccipital region when the target was surrounded by large inducers, and the genetic variance was 42.79% in the left parietotemporal region when the target was surrounded by small inducers. Taken together, study 4 found that the Ebbinghaus illusion is highly hereditary, and the brain activity responsible for the Ebbinghaus illusion is partially hereditary.
In summary, the processing of the Ebbinghaus illusion could occur both in the early processing stage (such as under awareness) and in the early visual pathway (such as V1), and could receive modulation from dorsal pathway (such as lIPL), thus further verifying the two-component hypothesis in the Ebbinghaus illusion. Moreover, we provided direct neurophysiological evidence supporting this hypothesis, i.e., the low-level component was correlated with alpha activity, and the high-level component was correlated with beta activity. This indicates that the Ebbinghaus illusion is a flexible perceptual phenomenon, it can occur without restriction, allowing the target popping out from the context and receiving fast perceptual processing, or be reduced or eliminated when it is necessary to perform specific task better (such as manual manipulation of the target). Therefore, the Ebbinghaus illusion is the product of evolution and adaptation in million years, is highly hereditary, and is critical for survival.