The purpose of this paper is to conduct a qualitative study to investigate the L2 acquisition of spatial words in Chinese time expressions by English-speaking learners and to analyze their error patterns. Three subjects were selected from different proficiency levels: beginning, low-intermediate, and high-intermediate. Each subject was observed and assessed monthly for six months. A grammatical judgment task in multiple choice format and a story-telling task with given pictures were employed to examine the subjects' comprehension and production of Chinese time expressions and their responses were further compared with the performance of 12 native controls. The overall results are as follows: First, for the English-speaking learners, the shang-xia 'up-down' type of time expressions was easier than the qian-hou 'before-after' type, indicating that the 'split' concept (the latter type) was more challenging than the 'zero-contrast' expression (the former type). Second, a hierarchical sequence of L2 time expressions obtained from the findings was xia 'down'＞ hou 'after'＞ shang 'up'＞ qian 'before,' showing that future time expressions were less difficult than past time expressions. Third, it was found the subjects' common errors included ＂wrong omissions,＂ ＂wrong additions,＂ ＂wrong substitutions,＂ and ＂wrong word orders,＂ where ＂wrong-substitutions＂ were more frequently found than ＂wrong word orders＂. Finally, there was a significant difference between the subjects' comprehension and production. It was found that the subjects performed better on the comprehension task than on the production task.