This study attempts to investigate the transformation of the central place system in Miura Peninsula, located on the fringe of the Tokyo metropolitan area. The study area includes Kamakura-shi, Zushishi, Yokosuka-shi, Miura-shi and Hayama-mdchi in Miura Peninsula (Fig. 1). Three hundred eightythree districts are selected as analytical units in order to clarify the distribution of central functions.In this study, retail activities are defined as central functions.<br> Data employed in this paper are in the form of central function scores. The formula for the central function score is<br> E_ij{nR_i/n∑j=1R_j}<br> (j=1, 2, 3, …n), <br> where C_ij is the central function score of function j in central place i, E_ij is the number of the establishments serving function j, R_j; is the annual amount of sale of function j per establishment, and n is the number of the central functions. The central functions are the 53 activities shown in Table 1. E_ij is derived from the yellow pages, and R; from the Census of Commerce. In this paper, an attempt is made to elucidate the transformation process of the intra-urban central place system from 1972 to 1985.<br> The major results of this study are summarized as follows:<br> (1) Central functions are classified by similarity of the distribution pattern. For 1972, central functions are classified into five function groups (Table 2): the high (Fa₁), middle (Fa₂), middle-low (Fa₃), low (Fa₄) and non-order functions (Fa₅) Fa₅ is made up of motor vehicle and motorcycle stores. For 1985, central functions are also classified into five functional groups (Table 3): the high (Fb₁), middle (Fb₂), low (Fb₃) and two non-order functions (Fb₄ and Fb₅). Group Fb₄ is constituted of motor vehicle stores, motorcycles stores and gasoline stations. Group Fb₅ includes dress materials, secondhand and accessories stores. The six high-order functions in 1972 moved to the middle in 1985, and the seven middle-low order functions in 1972 moved into the low category in 1985 (Fig. 4).<br> (2) To investigate the central places, cluster analysis (Ward's method) is applied to the standardized central function scores calculated for each functional group. For 1972, analytical unit districts are classified into eight groups (Fig. 2): the highest (Ca₁), high (Ca₂), middle (Ca₃), middle-low (Ca₄), low (Ca₅), and non-hierarchical central place groups (Ca₆ and Ca₇) and non-center group (Ca₈). For 1985, analytical unit districts are classified into seven groups (Fig. 3) the highest (Cb₁) high (Cb₂), middle (Cb₃), low (Cb₄) and non-hierarchical central place groups (Cb₅ and Cb₆), and noncenter groups (Cb₇). In terms of Ca₆ and Cb₅, there are also self-sufficient high-order functions such as department stores and large supermarkets. Groups Ca₇ and Cb₆ contain many motor vehicle stores.In both 1972 and 1985, the highest and high-level central places are located near railroad stations. This means that the central place system in Miura Peninsla is formed around the railroad station as a core. For 1972, the highest and high-level central places are surrounded with middle, middle-low and low-level central places in order of proximity. In 1985, by contrast, the low-level central places are dispersed along main roads, suggesting that central functions were accumulated in the districts located along the main roads.<br> (3) New functional groups (Fc₁-Fc₅) are established in order to examine changes in the distribution of central functions (Fig. 5).