Recently we've had several discussions about how tones in Sinitic languages aren't as uncomplicated or inflexible as one might imagine or as is often claimed:
"Mandarin by the numbers"
In these posts and in the comments to them, we have seen how stress and musical tune / melody often override or distort the canonical tones for given morphosyllables in sung or spoken context. This is a completely different matter than tone sandhi, where tones are modified according to their position within a sequence of syllables (I believe that most instances of tone sandhi occur for simple physiological reasons, e.g., in normal speech it is virtually impossible to pronounce two full third tones in a row because they both dip so low in an individual's register that one needs a means for readying oneself for the onset of the utterance of the second third tone and does this by changing the first third tone utterance to a rising second tone).
Recently (2012), University of Oslo student Øystein Krogh Visted has written a very interesting M.A thesis entitled "Nuances of Pronunciation in Chinese: Lexical Stress in Beijing Mandarin." Here's a brief description of the thesis:
The pronunciation of Beijing Mandarin, which is the basis for Modern Standard Mandarin, is in reality not as straightforward as it is usually presented. General books on the language and common textbooks in English on the subject usually only give very basic, prescriptive (though supposedly descriptive) analyses of the basic features of pronunciation. Finer points are generally not discussed in any detail. The treatment of amongst other things the aspect of word stress (the parts of words that are emphasized in speech) in mastering and indeed properly understanding Chinese is thus neglected. It has not yet acquired the position in Chinese language-teaching it arguably needs, so that the language may begin to be taught and indeed learned in a more comprehensive manner. This book will take a basic analytical approach to the phenomenon of word stress in Beijing Mandarin. It compares and discusses available meta-information on the topic, as well as its theoretical underpinnings and practical applications, and from a pedagogical starting point aims to bring attention to these important nuances in the Chinese language.
Because of the above described phenomena and increasing polysyllabicization, some scholars have even mentioned to me that they think northern Mandarin, at least around Beijing, is evolving into a stress language rather than a tonal language, though I haven't seen any papers or books that document this development.
This has prompted me to reflect that the "older" Sinitic languages (e.g., Cantonese, Min) tend to have more tones, the "younger" ones (e.g., the Mandarin branch) fewer tones, and to pose the following questions:
1. How to account for the greater number of tones in the older languages?
2. How to account for the lesser number of tones in the younger languages?
3. How many tones did Middle Sinitic have?
4. How many tones did Old Sinitic have?
I ask these questions because it appears that northern Chinese is gradually shedding its tones and evolving into a stress language, at least in the opinion of some colleagues. Of course, others have pointed out that stress has always played a role in Chinese, but it just hasn't been described and studied in detail. Still, it does seem that stress and melodic contours are assuming greater weight now than in the past.
From W. South Coblin
Questions 1 and 2 are really different sides of the same coin. The diminution in number of tones in the younger languages is due to phonological mergers, merger and split being two fundamental processes in sound change. The medieval lexica of the Qieyun family of texts [VHM: medieval rhyme books] indicate four tones, i.e., ping, shang, qu,and ru [VHM: "even, rising, leaving / falling, and entering"]. It is assumed that the first three differed in level and/or contour. The forth was characterized by syllable final stops, i.e., checked finals.
It is widely assumed that this four-tone system had two phonetic registers, upper and lower, conditioned by the presence or absence of initial voicing/murmur on the syllables. This would have been a phonetic rather than phonemic difference, and many native speakers would presumably not have been aware of it. Thus, the lexica indicate only four tones. Later loss of voicing/murmur, i.e., loss of the conditioning factor, would have brought the register distinction into full phonemic prominence. This process is thought to have begun sometime in the Tang period in at least some dialect families. The result would then have been an eight tone system consisting of paired upper and lower (i.e., yin and yang) ping, shang, qu, and ru tones. This eight-tone system, which is actually rather rare today, would then have been reduced in various ways, due to mergers in most dialect groups. It is these mergers that have reduced the number of tones. Standard Cantonese is an exception, in that the upper yinru tone underwent a phonetic split, conditioned by vowel quality. That yielded nine phonetic (as opposed to phonemic) tones.
The tonal situation in Early (now usually called "Old") Chinese has been a source of controversy. Karlgren and Li Fang-kuei, among others, assumed that the four "classical" tones were already present at that stage. Li definitely believed that tones were present at the time of the composition of Shijing odes, but he was noncommittal about their ultimate origins. The most common newer view is that tones are of secondary origin in Chinese and arose through a process called "tonogenesis", a term coined by James A. Matisoff. Tonogenesis became a hot topic in the sixties and seventies of the last century, when it began to be thought that most or perhaps all tonal systems in East Asian languages were ultimately secondary. Briefly, East Asian tonogenesis is conceived of as a process in which phonemic tones arose when certain early syllable final consonants were lost. In Chinese, as well as in various other languages, it is widely thought that the lost consonants in question were the glottal stop, laryngeals such as -h, and sibilants such as -s.
The classic view for Chinese is that the loss of a final glottal stop yielded the medieval shang tone, while loss of final -s and/or -h yielded the qu tone. The ping tone would have arisen when neither consonant was present, and ru would simply comprise syllables having various obstruent finals, usually posited as -p, -t, or -k, though some authorities envisage other such stops. This view, with individual variations, is rather widely held by people who work on Old Chinese today.
You have asked what I think. The answer is that I do not work on Chinese of that period at all and do not concern myself with the question, except in those rare cases where it may have some bearing on the historical and comparative dialectology I do nowadays. For pre-medieval Chinese, I use Jerry Norman's Early Chinese system, which is specifically designed for use in the historical study of dialects and is much simpler than the Old Chinese systems developed by people who specialize in that field. In this system, ping tones are unmarked, shang tones are marked with an -x, and qu tones are indicated by an -h. These symbols are purely algebraic and are borrowed from Li. They are non-committal regarding tonal origins, since no modern Chinese dialect is atonal. Thus, this system of tonal marking is quite adequate for all historical work involving the modern vernaculars. For the medieval period, I use Norman's Common Dialectal Chinese. There, tones are indicated by the 1-8 numbering system of Y. R. Chao that is conventionally used by Chinese dialectologists today.
From Axel Schuessler:
Old Chinese (OC) is assumed to have had no tones. During (before?) Hàn an assumed final glottal stop (still surviving in some southern dialects) got lost and resulted in Middle Chinese (MC) shangsheng 上聲 [VHM: "rising tone"] (i.e., words with a glottal stop presumably already had a certain short-stopped intonation which survived after ʔ was lost, hence the tone). An OC final -s ( or h as well as -s) was also lost by MC, so that only the earlier concomitant intonation survived, hence qusheng 去聲 [VHM: "leaving / falling tone"]. Where there were no OC *ʔ and *s, the syllable either ended in a vowel (> pingsheng 平聲 [VHM: "even tone"]), or a stop consonant (rùsheng 入聲 [VHM: "entering tone"]). This adds up to four MC tones.
After MC, voiced initial consonants became voiceless (b>p, d>t, etc). To maintain a distinction, MC tones split according to voicing of initial consonants. In Mandarin, MC pingsheng split: voiceless consonant became today's tone 1 (gē < kâ 'song'), voiced initial became tone 2 (with aspiration:) táo < dâu. MC shangsheng became Mandarin tone 3, after voiced initial it merged with qusheng > Mandarin tone 4 (hence 受 OC *duʔ > MC shangsheng 'to receive' and *dus > MC qusheng 'to give' are both Mandarin shòu). All MC qusheng syllables have become Mandarin tone 4. Because of these splits in pingsheng, syllables beginning with today's l- (e.g., lí-kai ["leave"]) can only end up in Mandarin tone 2, there is no tone 1 beginning with l- (except for a few rare, odd exceptions, perhaps). MC rùsheng words became very irregular and unpredictable in Mandarin.
As you see, in northern Chinese MC tones reshuffled themselves, merged and split to wind up with 4 tones again; some Shandong dialects of Mandarin are said to have three tones. Because of the splits of the MC four tones, Sinitic languages can have up to 8 tones (2×4) when no mergers occurred. This is what you find in the south.
I seem to recall that Mantaro Hashimoto once wrote a paper (and perhaps someone else did so later as well) showing that Sinitic languages in the north near toneless Altaic languages tend to have few tones, languages in the south (Cantonese, Min, etc.) near Tai languages with their multitudinous tones tend to have preserved more tones. Perhaps Tai substrates may have something to do with this too (Anne Yue-Hashimoto would know more about it). So there is some north-south gradation due to neighboring languages. There is a Sinitic language, I think a dialect of northern Chinese somewhere in Central Asia, that has lost all tones.
It is evident from the accounts of historical phonologists such as Coblin and Schuessler that the number of tones in Sinitic languages has waxed and waned, that they are not immutable, and that they are indeed in a state of flux at the current time. How many there will be a century from now is impossible to predict.