青年和老年人群认知联系网络的局部结构特征分析*

目的：从词汇联想网络角度比较青年和老年人认知联系结构的局部特征。方法：健康青年和老年人各50名。刺激词为Kent-Rosanoff中文对等词。反应词之间通过共享刺激词而构建“反应词-反应词网络”（RRN）：青年组RR网络（YRRN）和老年组RR网络（ORRN）。采用Pajek1.14和Matlab7.0进行网络分析和可视化。局部变量包括点度中心性、中间中心性、接近中心性、首层云集系数（CC1）和次层云集系数（CC2）。结果：在不同年龄组之间,各局部变量相关系数均<0.6。各组点度中心性与中间中心性、接近中心性和CC2线性正相关,而与CC1之间缺乏明显线性关系。点度中心性或CC1数值排位前十位的词汇在两组间均重复很少。对YRRN和ORRN的交集和差集的比较显示在各年龄组均存在各自特有的重要词汇及其联系。结论：同一个词汇在不同年龄人群认知联系结构中的重要性不同;不同的局部变量可以反映词汇间联系的特征;点度和首层云集系数可以为认知康复研究提供方法学依据。

Local structure properties of Chinese cognitive associative networks in young and old people

Objective: To compare the local cognitive associative structures in young and old people by analyses of word associative networks. Method: One hundred native healthy subjects including fifty young and fifty old people were in this study. The stimulus words were 100 Chinese words corresponding to Kent-Rosanof list. Two "response-response" networks(RR networks) were constructed by defining response words as nodes and the relationships of sharing same stimulus words amongst response words as linkages, including RR network of young (YRRN) and old (ORRN). Pajek1.14 and Matlab7.0 were used for network analysis and visualization. Local variables included degree centrality, betweenness centrality, closeness centrality and 1st clustering coefficient (CC1) and 2nd clustering coefficient (CC2). Result: Pearson correlation coefficients of local variables in YRRN and ORRN were all smaller than 0.6. In both RR networks, multi-dimensional scaling diagrams showed positive linear correlations of degree centrality with betweenness centrality, closeness centrality and CC2, but nonlinear correlations of degree centrality with CC1. By the 10 most valued words of degree centrality or CC1, there were few overlaps between each age group. For selection of word associative relationships, the differences and intersections of YRRN and ORRN showed that there were important words and relationships. Conclusion: There are different significances of a same word in different age groups; different local variables should represent characteristics of inter-concept relationships; degree centrality and CC1 should provide supports for researches on cognitive rehabilitation.