本文旨在檢視以英語為強勢語言的華裔繼承語使用者(heritage speakers)和成人後方學習華語為第二語言之學習者（下稱二語者）之三聲變調感知。當兩個三聲連讀，第一個三聲變為二聲；當三聲在非三聲之前，該三聲變成半三聲。42位華語母語人士、21位繼承語使用者和25位二語者參與了聲調辨識實驗，受試者在聽了兩個單音節之後，須從四個音檔中選出母語人士如何將剛聽過的兩個單音節連讀成一個雙音節。雖然有些統計結果未達顯著，整體趨勢顯示：（1）與二語者相比，繼承語使用者有些許優勢（與母語者更加相似）；（2）跟其他音調組合相較，繼承語使用者在辨識三聲變調和半三聲有些許困難，但二語者似乎對所有雙音節組合皆有困難。

This study investigates how heritage speakers (HSs) and late learners of a second language (L2ers) of Mandarin (both English-dominant) perceive Mandarin tone 3 (T3) sandhi. T3 sandhi occurs when two T3 syllables occur consecutively, and the first syllable becomes a T2. When a T3 syllable precedes a non-T3 syllable, the T3 syllable becomes a half-T3. Forty-two Mandarin native speakers (NSs), 21 HSs and 25 L2ers completed a Tone Identification Task. Participants listened to two monosyllabic sound files separately and were then asked to choose the disyllabic sound file (one out of four) that corresponded to NS production of the disyllabic sequences. While the differences are not always statistically significant, the trend is that (1) HSs performed better than L2ers overall(HSs being more native-like than L2ers); (2) while HSs have more difficulty identifying T3 sandhi and half-T3 conditions than other conditions, L2ers seems to show overall difficulty with disyllabic tones.