華語文教材與前人之研究顯示，「捲起」與「拱起」舌位皆可做為華語捲舌音發音方式，而此兩種方式在聽覺上對非母語人士而言是否不同仍待解答。本研究探討此問題，採用ABX區分實驗，以華語母語人士作控制組，並招募以華語為二語的學習者作為實驗組（英語、法語、日語共三組），隨後統計分析各組之實驗數據。有兩項結果值得注意：一、在各自與對應的齒齦音比較時，除母語組以外，兩種發音方式的捲舌音對非母與人士在聽辨上幾無差別；二、日語組的反應正確率顯著低於其他兩組。當假設日語母語人士能察覺華語齒齦後音與日語齦齶音之間的語音差異，則可解釋為何該組表現較差。而此假設更暗示，如非母語人士將注意力放在語音細節上時，其可能能夠分辨兩種捲舌音發音方式之間的不同。最後，研究結果為華語教學提供些許建議。

This study investigates two articulatory variants of the Mandarin retroflex as perceived by non-native speakers who learned Mandarin Chinese as a second/foreign language (L2). A review of Mandarin teaching materials and previous studies reveals that the following two gestures produce Mandarin retroflexes: ＂curling up＂ of the tongue and ＂bunching up＂ of the tongue. The focal question is whether these two gestures affect non-native speakers’ perceptions differently when they are distinguished from their alveolar (non-retroflex) counterparts. To address this question, an ABX discrimination task involving Taiwan Mandarin speakers (as a control group) and L2 Mandarin learners (as experimental groups) was implemented. Data were statistically analyzed for Taiwan Mandarin, Japanese, English, and French participants, resulting in thefollowing two notable findings: i) the two types of retroflexes were not perceived differently by the non-native speakers when paired with their alveolar counterparts, and ii) the mean accuracy rate of the Japanese group was significantly lower than those of the English and French groups. One explanation for the Japanese participants’ poor performance is that they are able to detect the phonetic differences between Mandarin post-alveolars and Japanese alveolopalatals. This explanation implies that non-native speakers may be able to distinguish the two retroflex variants if their attention is drawn to the phonetic details. These findings have several implications for Mandarin pedagogy.