徽州方言的古濁聲母如同今天絕多數的漢語方言一般都已化成不帶音的清聲母，不同的是它們呈現出兩類不同的層次:送氣與不送氣。這兩類唸法在徽州方言的核心區休寧、屯溪一天的方言裡尋階著分化條件，相當混亂地分佈在濁聲母字中，我們以為這正表示徽州方言濁聲母的變化並不是單純的清化而已。從明清時期的文獻中我們發現徽州方言的濁聲母在當時是不送氣的讀法，因此我們將不送氣清音的層次假設為較早的語音表現，送氣清音的出現則是與四世紀以來客贛移民有關。 這個不送氣的唸法其實是吳由表現的反映，今吳、湘方言都保留濁聲母帶音的讀法，卻也都有清化不送氣的例子，我們以為徽州方言古濁聲圇清化送氣正是吳語層的表現。而客?方言所影響的送氣層出現的時間可能相當晚，應在客?移民在南方落方之後才逐漸展開，同時徽州各地所受的影猖因地而異，多數點的濁聲母都唸送氣清音，唯有核心地區尚存不送氣的讀法。換句話說，送氣清音是以覆蓋的方式代替了原不送九的唸法，未被覆蓋的字仍讀不送氣，因此出現了混亂的分佈。 江灣的濁聲母清化類型不同於徽州各地，舒聲韻的從、邪二母都是送氣音，其他舒聲字仍不送氣;入聲韻絕多數字都是送氣音，唯少數字讀作不送氣:大體上我們可以舒、促韻母為分化條件判斷送氣與否，期間並不混亂。我們以逼江灣絕多數舒聲字清化不送氣正是徽州較早的語言表現，與其他徽州方言一致;但入聲字清化多送氣的現象卻需要加以說明，因為若將清化送氣的例子視為早期語音是其他徽州方言完全相反的分析，我們以為並不合理。江灣方言入體字清化送氣也是客?方言的影響，只是這個影響最早出現在入聲韻中，舒聲韻並不明顯。 於是我們以下圖表示江灣方言古濁聲母的變化過程，括弧中代表外來方言的影響。 舒聲 *b →P (←Ph) 入聲 *b →P (←Ph) 綜合說來，從徽州方言的表現可窺出有三支方言曾在此發揮影響力: ?吳語 ?客贛方言 ?官話 吳語層的時間最早，事實上我們認為徽語可視為一麵吳語，這種吳語今天之所以表現得與今吳地之吳語大不相同正適足以說明它所受到的外來方言影響遠超過吳語，這裡所稱的外來方言影響指的就是客贛方言與官話。

　　　　　 The voiced initials in Huizhou dialects are devoiced completely just like most of other Chinese dialects. There seem to be two different strata found in Huizhou dialects: aspirates and nonaspirates. In this paper we have stated the majority of the reflexs of the voiced stops and affricates in most Huizhou dialects are aspirates. However, we find both in Xiuning (休寧) and Tunxi (屯溪), which are said to be the core of Huizhou regin, without clear conditions. This division into aspirates and nonaspirtates does not correlate with any known distinction in Anclent or in Archaic Chinese. It is caused by respective contact of two different types of dialects in different periods: ?Wu dialects, and?Hakka-Gan dialects. We assume the nonaspirates to be the earlier stage in Huizhou dialects, basing our argument on the descriptions we found in the literature three hundred years ago. The Hakka-Gan immigrations into southern China brought aspirates down and exerted an influence on many southern dialects such as Xiang, and Wu. That Huizhou dialects originate from Wu dialects is proposed in this paper. That is to say, Huizhou dialects and Wu dialects were of the same type before Hakka-Gan immigrants came to southern China. This is why the majority of the reflexs of the voiced stops and affricates in most Huizhou dialects are aspirates and both types can be fund unpredictably in 休寧 and 屯溪 dialects. The behavior of voiced stops and affricates in Jiangwan (江灣) is quite different from that in Wu, Hakka-Gan or other Huizhou dialects, having aspirates as reflexs in most entering (or ju-) tone and nonapirates in other tones (從母 and 邪母 are exceptions)。That the distributions of aspirates and nonaspirates can be roughly divided by tones indicates that this might be a conditioned-change at the first sight, which is a sharp contrast to other Huizhou dialects disussed above. That the nonaspirates in nonetering tones are treated as the earlier stage in this dialect is obviously true, just like in other Huizhou dialects; what is the story about those aspirates in entering tone? We view the nonaspirates as the earlier, and the aspirates then become the later stage resulting from dialects contact. This hypothesis coincides with that we made for other Huizhou dialects: the nonaspirates are earlier than aspirates.