本文採用意象圖式 (image schema) 與原型範疇理論 (prototype theory) 為分析的觀點，旨在探討包括 babaw, baraw, daya（上）、truma, lhbun（下）、 brah（前）、 bukuy（後）、iril（左）、narat（右）、ruwan（裡面）、ngangut（外面）、kska/ska（中間）等十二個太魯閣語方位詞彙的空間概念用法。基於意象圖式的觀點進行分析，可以清楚地呈現每個方位詞所表達的概念，尤其是「上」和「下」，有兩種以上的說法，但是它們皆可用不同的意象圖式來呈現與區分。從原型範疇理論的觀點來看，這些方位詞大多僅存在最核心的用法，沒有明顯的語意擴展。有趣的是，筆者發現 daya、lhbun、brah、 bukuy這四個方位詞彙可能是由實詞衍生出方位詞的用法。

This paper adopts image schema (Johnson 1987; Lakoff 1987; Langacker 1987) and prototype theory (Wittgenstein 1958; Berlin and Kay 1969; Rosch 1973, 1975) and investigates the spatial conceptualization of several locative expressions, including babaw, baraw, daya ‘over/above/upon,’ truma, lhbun ‘under/beneath,’ brah ‘before/in front of,’ bukuy ‘back/behind,’ iril ‘on the left,’ narat ‘on the right,’ ruwan ‘inside,’ ngangut ‘outside,’ and kska/ska ‘among/between/in the midst of.’ Based on image schema theory, the meaning of these locative expressions can be clearly represented; in particular, the multiple locative expressions related to ‘over/above/upon’ and ‘under/beneath,’ which now can be distinguished by adopting different image schemas. According to prototype theory, most of these locative words have core usages only, and do not have semantic extensions. Interestingly, we observe that four of these locative expressions, daya, lhbun, brah, and bukuy, may be derived from content words.