从中共中央确立公务员考试制度至今已有21年。在这21年中，公务员考试制度在不断地探索、发展和完善。行政能力测验作为国家公务员考试的重要组成部分，其质量影响到公务员队伍候选人的整体素质。为了能够建设题库，更有效地推进行政能力测验，首先我们要了解各类题型都考了什么，是如何考的。本研究针对行政能力测验中的图形推理分测验，展开了一系列研究，回答了图形推理“考什么”、“怎么考”等问题，希望对今后的命题工作，特别是题库建设的工作提供一些有价值的理论建议和切实可行的实践指导意见。研究一通过采访从事公务员行政职业能力测验命题行业专家和综述前人文献，构建图形推理能力初步框架、图形推理认知过程初步框架以及图形推理答题策略初步框架。通过对历年行测图形推理分测验的数据分析，计算出历年图形推理题目的难度、区分度、各选项的选答率以及鉴别力，为研究二选择试题，编制《图形推理测验筛选被试版》以及《图形推理测验口语报告版》。研究二使用《图形推理测验筛选被试版》，选取图形推理能力高、低两组被试，以研究一编制的《图形推理测验口语报告版》为实验材料，通过即时性口语报告以及追溯性口语报告了解被试在解题过程中表现出来的能力、认知加工过程以及答题策略。使用质性编码的方式对口语报告文本编码，得出《图形推理能力测评框架》、《图形推理认知过程测评框架》以及《图形推理答题策略测评框架》。图形推理能力测评框架包括4种能力，演绎推理能力、归纳推理能力、类比推理能力以及空间能力。图形推理认知过程框架包括4个阶段，目标管理、认知分析、概念分析以及概括选择回答。图形推理的答题策略包括规则建构、选项建构、边规则建构边选项剔除以及先选项归类再规则建构4种主要策略。研究三对历年行测中图形推理题目进行质量分析，综合各研究结果，提出了图形推理题目的命题指导意见。

It has been more than twenty-one years since the central government established the principle that every civil servant must pass the examination before admission. During these years, civil service examination system has constantly explored, developed and improved. As an important component of National Examination of Civil Service, Administrative Aptitude Test (AAT), to some degree, can influence the overall competence of civil servant. In order to lay foundation for the construction work of AAT item bank and promote AAT, firstly and most importantly, closely examine what could be tested in each subtests and how these subtests test. In this study, aimed at figural reasoning subtests, try to answer the question what figural reasoning subtest should test andhow figural reasoning subset test, thus this study provides valuable theoretical and practical suggestion. In study 1, construct the 1st version of competence-based framework of figural reasoning subtest, the 1st version of cognitive process framework of figural reasoning subtest and the 1st version of answer strategy framework of figural reasoning subtest after interviewing experts of figural reasoning subtests and summarizing previous literatures in this related area. Compute difficulty, discriminability and perception and discriminability of each option by analyzing figural reasoning data of previous years, and select typical test questions to make a figural reasoning test designed to select participants and another figural reasoning test designed for oral report. In study 2, using the figural reasoning test designed to select participants, choose participants with high ability or low ability to do the oral report experiment using thinking aloud method with figural reasoning test designed for oral report in order to complete the final version of competence-based framework of figural reasoning subtest, cognitive process framework of figural reasoning subtest and answer strategy framework of figural reasoning subtest. Figural reasoning subtest examines 4 key competencies, which are inductive ability, deductive ability, analogical ability and special ability. Cognitive process framework contains 4 phases, objective management, perceptual analysis, conceptual analysis and response generation and selection. The answer strategy framework is consisted of 4 strategy, rule construction, option construction, construction with rule and option, rule construction after option classification. In study 3, analyze the quality of figural reasoning questions in the previous years, and summarize the results of all studies aiming at providing practical advice about how to design highly qualified figural reasoning subtest.