Department of Psychiatry and Behavioral Sciences, University of Minnesota Health.
Department of Psychology, University of New Hampshire.
Neuroimage. 2021 Aug 1;236:118092. doi: 10.1016/j.neuroimage.2021.118092. Epub 2021 Apr 23.
Fractals are self-similar patterns that repeat at different scales, the complexity of which are expressed as a fractional Euclidean dimension D between 0 (a point) and 2 (a filled plane). The drip paintings of American painter Jackson Pollock (JP) are fractal in nature, and Pollock's most illustrious works are of the high-D (~1.7) category. This would imply that people prefer more complex fractal patterns, but some research has instead suggested people prefer lower-D fractals. Furthermore, research has suggested that parietal and frontal brain activity tracks the complexity of fractal patterns, but previous research has artificially binned fractals depending on fractal dimension, rather than treating fractal dimension as a parametrically varying value. We used white layers extracted from JP artwork as stimuli, and constructed statistically matched 2-dimensional random Cantor sets as control stimuli. We recorded the electroencephalogram (EEG) while participants viewed the JP and matched random Cantor fractal patterns. Participants then rated their subjective preference for each pattern. We used a single-trial analysis to construct within-subject models relating subjective preference to fractal dimension D, as well as relating D and subjective preference to single-trial EEG power spectra. Results indicated that participants preferred higher-D images for both JP and Cantor stimuli. Power spectral analysis showed that, for artistic fractal images, parietal alpha and beta power parametrically tracked complexity of fractal patterns, while for matched mathematical fractals, parietal power tracked complexity of patterns over a range of frequencies, but most prominently in alpha band. Furthermore, parietal alpha power parametrically tracked aesthetic preference for both artistic and matched Cantor patterns. Overall, our results suggest that perception of complexity for artistic and computer-generated fractal images is reflected in parietal-occipital alpha and beta activity, and neural substrates of preference for complex stimuli are reflected in parietal alpha band activity.
分形是在不同尺度上重复的自相似模式，其复杂性用分数欧几里得维数 D 表示，范围在 0（一个点）到 2（一个充满的平面）之间。美国画家杰克逊·波洛克（JP）的滴画具有分形性质，而波洛克最著名的作品属于高-D（~1.7）类别。这意味着人们更喜欢更复杂的分形模式，但一些研究表明，人们更喜欢低-D 分形。此外，研究表明顶叶和额叶的大脑活动可以跟踪分形模式的复杂性，但之前的研究是根据分形维数人为地对分形进行分类，而不是将分形维数视为参数变化的值。我们使用 JP 艺术作品中的白色层作为刺激，并构建了统计匹配的二维随机 Cantor 集作为对照刺激。当参与者观看 JP 和匹配的随机 Cantor 分形图案时，我们记录了脑电图（EEG）。然后，参与者对每个图案的主观偏好进行评分。我们使用单次试验分析构建了将主观偏好与分形维数 D 相关联的受试者内模型，以及将 D 和主观偏好与单次试验 EEG 功率谱相关联的模型。结果表明，参与者更喜欢 JP 和 Cantor 刺激的高-D 图像。功率谱分析表明，对于艺术分形图像，顶叶阿尔法和贝塔功率参数跟踪分形模式的复杂性，而对于匹配的数学分形，顶叶功率在一系列频率范围内跟踪模式的复杂性，但在阿尔法频段最为突出。此外，顶叶阿尔法功率参数跟踪了对艺术和匹配的 Cantor 图案的审美偏好。总的来说，我们的结果表明，对艺术和计算机生成的分形图像的复杂性的感知反映在顶叶-枕叶阿尔法和贝塔活动中，而对复杂刺激的偏好的神经基质反映在顶叶阿尔法波段活动中。
