基于对北京市居民空间认知的3500份调查问卷数据，采用了ARCGIS的空间插值和Geoda095i软件的空间自相关的数理统计方法，分析北京市居民的认知空间结构。研究得到四点结论：①认知地点的类型并不是影响认知率的因素。②居民对北京市的空间认知呈现出西高东低的“>”形格局，在西北—东南方向隆起动物园和北京火车站两个认知高峰，鞍部为低认知率的东方新天地和sogo崇光百货。③居民认知呈现明显的空间相关性和尺度依赖性。地名认知在较小的空间尺度上为高—高、低—低集聚的趋同性，在较大的空间尺度上表现为高-低集聚的趋异性，两者的转换尺度为城市居民日常生活空间的半径6 km左右。④北京城市的认知空间可细分为北海公园、中友百货、动物园和城乡贸易中心四个次级认知区域，次级认知区域为同质的高认知区，内部包含多个认知高峰。

City spatial cognition research included cognition of place name, city image, cognitive distance and cognitive distortion. The national research of spatial cognition began in the 1980s, focusing mainly on the city image. Research of place name is more focused on the origin and meaning, less involved in the form of space. The purpose of this paper is to analyze the pattern of cognitive space based on a large number of questionnaires, then explored it’s genesis deep. The main methods are spatial interpolation of ARCGIS and spatial autocorrelation of Geoda095i. Four conclusions were obtained as follows: Firstly, the type of cognitive point is not the influencing factor of cognitive rate. Secondly, the spatial cognitive pattern showed a ">-shaped pattern". The two peaks of Beijing Railway Station and Peking Zoo uplifted in the northwest - southeast direction, and saddle lied in lower Sogo Department Store in low recognition rate area. Thirdly, the cognition of residents had clear spatial relevance and scale-dependent. Cognition among sites is high-high, low-low comparability in less spatial level, but it is high-low otherness in larger spatial level. 6km is their boundary. Fourthly, cognitive space can be divided into four sub-regions: Beihai Park, Zhongyou Department Store, Peking Zoo, and Chengxiang Entrepot. The sub-region is homogeneously high cognition area, and contains many cognition peaks.