Part of the research for this paper was made possible by Social Sciences and Humanities Research Council of Canada Grants 410-810949, 410-830354, and 410-850540 and by National Science Foundation Research Grant BNS-9011190. I am indebted to Meihan Low for assistance with the sources written in Chinese.
Word order, both at the clause level and even more at the phrase level, varies among Sino-Tibetan languages. In this Chapter, I describe some of this variation and examine it in the light of word order tendencies found among the languages of the world as a whole. In Section 1, I briefly summarize some of the variation in word order within Tibeto-Burman (TB) languages, and discuss what features of word order in these languages are typical and atypical. In Section 2, I discuss word order in Chinese, identifying some typological unusual features and discussing possible explanations for them. An overall theme shared by the two sections is that word order in Sino-Tibetan is best understood in an areal context.
The discussion in this section summarizes briefly what I discuss in much greater depth in Dryer (forthcoming). It is based on examination of descriptions of ninety-three TB languages.
The distribution of the two orders of object and verb in TB is straightforward: all TB languages are ov, except for Bai and the Karen languages, which are vo (and more specifically svo). Although available data varies in the descriptions, the ov languages within TB generally share a variety of other word order characteristics typical of ov languages, in employing postpositions rather than prepositions, in placing genitive modifiers before the possessed noun, in placing relative clauses (if they are externally headed) before the head noun, in placing postpositional phrases before the verb, in employing clause-final markers for subordinate clauses, in placing markers of polar questions (if they employ them) at the end of sentences, and in placing auxiliary verbs after the main verb. An example of an exceptional feature found in a few TB languages is the placement of manner adverbs. While most TB languages more commonly place manner adverbs before the verb, a few TB languages, all of them Kuki-Chin-Naga languages, commonly if not preferentially, place manner adverbs after the verb. This is described as the preferred position in Tiddim Chin (Henderson 1965) and Angami (Giridhar 1980) and is illustrated for Tiddim Chin in (1).
|‘He called out loudly, as if answering someone.’ (Henderson 1965: 4, sentence 5)|
In some cases, the vo TB languages exhibit the mirror image of the characteristics mentioned above for ov TB languages; however, in many other cases, they do not. Both Bai and the Karen languages employ genitive-noun order. In fact this order is the one word order feature that is apparently shared by all Sino-Tibetan languages. In employing this order, Bai and the Karen languages are actually not atypical: as discussed in Dryer (1991), the two orders of genitive and noun are about equally common among svo languages.
The Karen languages exhibit a number of features that are atypical of vo languages. I illustrate this with features from Bwe Karen. While Bwe Karen does employ some clause-initial markers of subordinate clauses, there are also some clause-final subordinators, illustrated by kha lέ ‘if’ in (2).
|‘If you stay, I will stay.’ (Henderson 1997: 78)|
Another feature of Bwe Karen that is atypical for a vo language is the placement of a word meaning ‘able’ after the main verb, as in (3).
|‘We can’t build our granaries close to the village.’ (Henderson 1997: 142)|
While this is rather unusual among vo languages, it is something found in a number of vo languages in other families in southeast Asia, including Tai-Kadai (e.g. Nung: Saul and Wilson 1980: 47–8, 55), Mon-Khmer (e.g. Chrau: Thomas 1971: 97), and Hmong-Mienic (Hmong Njua; Harriehausen 1990: 179–80).
Bai also exhibits features atypical of VO languages. Foremost among these is the placement of relative clauses before the modified noun, as in (4).
|‘Words that are written tidily are easy to read.’ (Xu and Zhao 1984: 73)|
As discussed in greater detail in Section 2 below on Chinese, this order is extremely rare among vo languages. There is at least one word meaning ‘able’ that follows the main verb, as in (5).
|‘I can go anywhere.’ (Xu and Zhao 1984: 22)|
And although Bai has some prepositions, it also has some postpositions, like no 33 marking the indirect object in (6).
|‘We gave you a flower.’ (Xu and Zhao 1984: 51)|
As discussed in detail in Dryer (1992), there are a number of word order characteristics which, contrary to widespread belief, do not correlate with the order of object and verb. These include the order of adjective, numeral, and demonstrative with respect to a modified noun and the order of degree words with respect to a modified adjective. Among the vast majority of ov languages in Asia that are not TB, these pairs of elements occur in the order modifier-modified, and this has led some linguists to the mistaken belief that these features are to be expected of ov languages. However, as shown in Dryer (1992), it is not the case that these features are typical of ov languages. For example, with respect to the order of adjective and noun, it is actually somewhat more common for these to follow the noun in ov languages outside of Asia. The ov TB languages are in many respects atypical among ov languages in Asia, but normal for ov languages in the world as a whole, in that in most ov TB languages, some of these modifiers normally follow the modified word. The distribution of these word order characteristics among TB languages is also interesting in that there is considerable variation in their distribution and it is often the case that even within a given subgroup of TB, some languages will employ one order while others employ the opposite order. It is possible to describe this variation only very briefly here; I discuss it in much greater detail in Dryer (forthcoming).
Both orders of adjective and noun are well-attested as preferred orders among TB languages. Among the eighty-five TB languages for which I was able to obtain information on this, the preferred order is adjn in twenty-eight languages and nadj in forty languages, and in seventeen languages, both orders occur without any indication in my source that one order is preferred. Furthermore, assuming for the purposes of discussion the classification of TB languages proposed by Bradley (1997), in three of Bradley’s six highest-level subgroups of TB (Bodic, North-Eastern India, and Central), there are some languages in which adjn is the preferred order and other languages in which nadj is the preferred order. In the other three subgroups, all of the languages are either nadj or allow both orders, with neither order dominant. Even within a number of groups at the next level down in Bradley’s classification, there are four groups containing languages of each of the two types (Bodish, Bodo-Garo, Tani, and Digarish ‘Mishmi’). For example, within Bodo-Garo, Deuri (Brown 1895) is adjn, while Kokborok (Karapurkar 1976) and a few others are nadj.
The geographical distribution of the two orders of adjective and noun is shown in Map 3.1. This map makes clear the extent to which adjn order is found in the western part of the area in which TB languages are spoken while nadj order is found more to the east.
This distribution can be understood in terms of the distribution of the two orders in non-TB languages in the surrounding area. The languages to the west and southwest of TB, especially the Indic languages within Indo-European, are consistently adjn, while the languages to the east of TB other than Chinese, namely, Tai-Kadai and Mon-Khmer languages, are consistently nadj. Map 3.2 shows the distribution of the two orders of adjective and noun in a larger area of Asia that surrounds TB languages. The overall impression given by Map 3.2 is two clear areas in South and Southeast Asia, one to the west which is adjn and one to the east which is nadj. But these two areas split TB down the middle.
Map 3.1 Order of Adjective and noun
Map 3.2 Order of Adjective and noun in Asia
Note that the clearest exceptions to the tendency for adjn order in the west are a number of nadj TB languages in the extreme western side of TB. These languages are all languages that are closely related to Tibetan, however, and they thus represent either languages which have moved into that area relatively recently, and thus have been less subject to influence from Indic languages, or languages whose less accessible location in the Himalayas has also meant that they have been less subject to influence from Indic languages.
The overall pattern of the distribution of the two orders of demonstrative and noun among TB languages is somewhat similar to the distribution of the two orders of adjective and noun: Demn order is more common to the west, while ndem order is more common to the east. However, demn order is more common than adjn order and there are a number of languages which are demn but nadj. Among the seventy-nine tb languages for which I was able to obtain information on the order of demonstrative and noun, fifty-one are demn, twenty are ndem, three allow both orders without there being evidence for one order being dominant, and five normally have demonstrative words simultaneously preceding and following noun, as in the example in (7) from Nishi.
|‘this woman’ (Hamilton 1900: 20)|
As with the order of adjective and noun, we find both orders represented within the same subgroup. In four of Bradley’s highest-level subgroups there are some languages which are demn and others which are ndem (Bodic, Kuki-Chin, North-Eastern, and South-Eastern). Note that this set of subgroups is very different from the analogous set with the order of adjective and noun: Bodic is the only subgroup in both sets. In other words all six subgroups are inconsistent, either in the order of adjective and noun or in the order of demonstrative and noun. Again we find languages of each type even within lower-level subgroups. For example, among the Burmish languages, Maru (Clerk 1911) is demn while Achang (Dai 1985) is ndem.
There is less variability in the order of numeral and noun among TB languages. The only languages in which numn is dominant are Bodic, although both types are about equally common in Bodic. Within Bodic, the distribution is largely predictable from lower level subgroups in Bradley’s classification: West Himalayish and Kiranti are NumN while Central Bodish (Tibetan), Western Bodish (Tamangic), and Eastern Bodish (e.g. Monpa) are nnum. Within what Bradley classifies as Central Himalayan, all are numn except for Newari; however, the classification of Newari is notoriously problematic.
I was able to obtain data on this characteristic for fewer languages, and I did not include affixes expressing degree. Again, there is a split, with twenty-five languages in which degadj is preferred, eleven languages in which adjdeg is preferred, and four in which both orders occur with no evidence that one is preferred. The subgroup with the most adjdeg languages is actually geographically central, namely, Kuki-Chin. And again, we find inconsistencies within lower level groups. Among the western languages of Central Bodish, Jad (Sharma 1989) is adjdeg while Nyamkad (Sharma 1992) is degadj.
I use the expression ‘Chinese languages’ to apply to what are traditionally called ‘dialects’ of Chinese, following the use of these terms within linguistics. I will, however, largely restrict discussion to Mandarin. There are some differences in word order among the different Chinese languages, but I will generally ignore these here.
This section has two primary purposes. The first is to document the extent to which Mandarin is unusual in its word order in a number of ways. The second is to argue that these unusual characteristics are best understood areally and reflect the geographically intermediate position of Mandarin between the languages of Southeast Asia and the languages of Northeast Asia. In Section 3.1, I describe the word order characteristics of Mandarin, discussing ways in which it is unusual, and in Section 3.2, I discuss possible areal accounts of these unusual characteristics.
The dominant order at the clause level in transitive clauses in Mandarin is svo, as in (8).
|‘I like Susan.’|
|‘Susan likes me.’|
sov word order is also possible, with the object marked by the object marker ba, as in (9), though this word order is less common (see Sun and Givón 1985 for evidence from narrative and oral text).
|‘I bought the book.’ (Li and Thompson 1981: 21)|
Since the dominant order in Mandarin is vo, we might expect it to have prepositions rather than postpositions. There are in fact a couple of different sorts of words that can be classified as adpositions. There is a set of words that can be described as prepositions, illustrated by the words ba ‘object marker’ in (9), cóng ‘from’ in (10a), and dào ‘to’ in (10b).
These words are called coverbs by Li and Thompson (1974), and different coverbs seem to be in different stages of grammaticization from their original status as verbs to their current status as more preposition-like elements. I will assume here that this grammaticization has proceeded far enough for some of them to justify calling themselves prepositions.
Mandarin also has postpositions, as illustrated by qǐ ‘from’ (< ‘start’) in (11a) and shide ‘like’ in (11b) (these examples kindly provided by Liu Danqing).
|‘He will go to work from tomorrow on.’|
|‘He is sly like a fox.’|
Note how both of these postpositions optionally co-occur with a preposition; however, the postposition in these constructions is obligatory. The number and frequency of both prepositions and postpositions in Mandarin suggests that it is best not to say that the language is primarily prepositional or primarily postpositional. Nevertheless, the frequency of postpositions is somewhat unexpected of a vo language.
These adpositions combine with noun phrases to form adpositional phrases. But these adpositional phrases more often precede the verb, as in (9), (10a), and (11), though they sometimes follow the verb, as in (10b). Mandarin is thus an instance of a language in which the object normally follows the verb but adpositional phrases more often precede the verb. This makes Mandarin highly unusual from a cross-linguistic perspective: among 199 vo languages in my database that I code for order of adpositional phrase and verb, only three are pp-v, while the other 196 are v-pp. The three that are pp-v are all Chinese languages: Mandarin, Cantonese, and Hakka. The Chinese languages are thus the only known instances of languages of this sort. It is not clear to me, however, how old the pp-v order in Chinese is. The claims in the literature regarding the position of pps in Classical Chinese are somewhat contradictory. On the one hand, Li and Thompson (1974) claim that classical Chinese was predominantly v-pp and that pp-v is a development since the twelfth century ad. On the other hand, Sun (1987) argues that both orders of pp and v were common in Classical Chinese.
The comparative construction in Mandarin also uses a coverb construction, as in (12).
|‘Zhangsan is fatter than him/her.’ (Li and Thompson: 142)|
The construction in (13) is Marker-Standard-Adjective (bǐ + tā + pàng). Again, this order is quite rare crosslinguistically; only Mandarin and Hakka in my database have this order, the typical order in vo languages being Adjective-Marker-Standard, as in English taller than Mary. In fact, this is the normal order in Cantonese, as in (13).
|‘Today is hotter than yesterday.’ (Matthews and Yip 1994: 166)|
Both orders are possible in both Mandarin and Cantonese, the difference being which is the dominant construction.
Although manner adverbs can either precede or follow the verb in Mandarin, their normal position is preverbal, as in (14).
|‘She/he walked quickly.’ (Li and Thompson 1981: 323)|
This is also not the usual order for vo languages, it being more common for manner adverbs to follow the verb, though there are a number of other vo languages outside Chinese which more commonly place manner adverbs before the verb (such as Latvian).
Mandarin is also unusual in being a vo language that places relative clauses before the noun, as in (15).
|‘the book [that I gave you]’ (Li and Thompson 1981: 117)|
Among 254 vo languages in my database for which I code the order of relative clause and noun, all are nrel, except for the three Chinese languages (Mandarin, Hakka, and Cantonese) and Bai, as illustrated above in Section 2.1.
There are three characteristics where svo languages are intermediate between verb-initial and verb-final (Dryer 1991), where some svo languages exhibit the order associated with verb-initial and others exhibit the order associated with verb-final. For all three of these, Mandarin exhibits the order associated with verb-final languages. One of these is the use of genn order in genitive constructions, as in (16).
|‘the rabbit’s ear’ (Li and Thompson: 113)|
A second is the use of sentence-final question particles, as in (17).
|‘Can you write Chinese characters?’ (Li and Thompson: 547)|
A third is the fact that interrogative phrases in content questions do not need to occur at the beginning of the clause, typically occurring in situ, in the position that corresponding non-interrogative phrases would occur in, as in (18).
|‘What are you doing?’ (Li and Thompson 1981: 522)|
The last two of these characteristics are also ones shared by other vo languages of Southeast Asia, in Hmong-Mienic, Tai-Kadai, and Mon-Khmer. The genn order contrasts, however, with the ng order of Tai-Kadai and Mon-Khmer.
There are other respects in which Mandarin behaves more like a typical vo language. The predicate follows the copula, as in (19).
|‘I am Susan.’|
Words meaning ‘able’ precede the other verb, as in (17) above. Verbs meaning ‘want’ precede the verb denoting what is wanted, as in (20).
|‘I want to go to China.’|
To a large extent, the inconsistencies in Mandarin word order can be characterized in terms of the distinction between two types of dependents, what in different grammatical traditions have been called complements or arguments on the one hand vs adjuncts or modifiers on the other. Mandarin typically places the head before a complement, a dependent which is required grammatically and semantically to complete the meaning of the phrase: verb before object, adposition before object, copula before predicate, verbs with meanings like ‘want’ or ‘able’ before their verbal complements. On the other hand, Mandarin typically places heads after adjuncts (i.e. modifiers), dependents which are not required grammatically or semantically but which optionally elaborate on the meaning of their phrases; this is reflected in placing nouns after adjectives, relative clauses and other modifiers of nouns, adjectives after intensifiers or standards of comparison, and verbs after manner adverbs and adpositional phrases. Whether this pattern is more than a coincidence, either synchronically or diachronically, is not clear.
We have seen in the preceding section that Mandarin (and other Chinese languages) has a number of characteristics that are highly atypical of vo languages. Can we offer anything to explain why Chinese might have these unusual characteristics? I will start with the assumption that Proto-Sino-Tibetan was ov, reln, and pp-v, and that the reln and pp-v orders are at least partly a retention of these features from Proto-Sino-Tibetan (LaPolla 1994; Liu 1999). Let me focus on two of these characteristics, vo&reln (vo with prenominal relative clauses) and vo&pp-v (vo with preverbal adpositional phrases). The cross-linguistic rarity of these types implies that there are some causal factors discouraging such languages from arising in the first place and possibly also encouraging such languages to change to some other type if they do arise. Languages elsewhere in the world which were ov&reln or ov&pp-v and which became vo have apparently also become nrel and v-pp, either simultaneously or shortly after becoming vo. Chinese, however, has apparently retained these characteristics for a long time. reln order is also the dominant order across TB, suggesting that the reln goes back to Proto-Sino-Tibetan. Since vo order is also apparently fairly old in Chinese, dating back to Proto-Sinitic or close to that, this means that Chinese has been vo and reln for a long time. The situation with pp-v order is less clear, as noted above: either it too dates all the way back to Proto-Sino-Tibetan, or Chinese has moved from pp-v towards v-pp and back towards pp-v.
We cannot answer the puzzle by simply saying that reln and pp-v order are simply retentions from Proto-Sino-Tibetan, since other instances of ov languages changing to vo have apparently invariably ceased to be reln and pp-v. What is striking, however is the extent to which Chinese languages resemble languages to the north, including Japanese, Korean, Mongolian, Tungus, and Turkic. These languages also place relative clauses before the noun and adpositional phrases before the verb. Of course, since these are characteristics associated with ov languages in general, the fact that Chinese resembles languages to the north in these respects is no different from saying that it resembles ov languages elsewhere in the world.
But there are other ways, however, in which Chinese resembles ov languages to the north far more than it resembles ov languages elsewhere in the world. As shown in Dryer (1992), the two orders of relative clause and noun are about equally common in ov languages. The reln order in Chinese is an ov characteristic only in the sense that reln languages are usually ov. The opposite is not the case: it is not the case that ov languages are generally reln. nrel order is in fact as common as reln order among ov languages. The cross-linguistic distribution of the two orders of relative clause and noun among ov languages partly follows an areal pattern: reln order is more common in Eurasia, while nrel order is more common elsewhere in the world. Hence the reln order in Chinese cannot be simply viewed as an ov characteristic; rather it is a characteristic associated with ov languages in Asia, both in TB and those north of Chinese. Map 3.3 shows the cross-linguistic distribution of the two orders of relative clause and noun among ov languages. Map 3.3 shows clearly how reln order is more common in Eurasia, particularly eastern Asia, in the area surrounding Chinese (except to the south, where the languages are not ov).
Map 3.3 Worldwide Distribution of two Orders of Relative Clause and Noun Among OV Languages
The reln order of Chinese resembles the common reln order to the north and to the west, in TB. But there are other respects, however, in which Chinese word order resembles word order in languages to the north more closely than word order in TB languages. Consider, for example, the order of adjective and noun. We saw above in Section 2.2.1 how both orders of adjective and noun are found in TB, though nadj order is somewhat more common, particularly in the east, towards Chinese. On the other hand, languages to the north are consistently adjn. When we look at the distribution of adjn and nadj order in Asia in Map 3.2 above, we see that Chinese is situated between a large set of nadj languages to the south (Tai-Kadai and Mon-Khmer) and southwest (eastern TB languages) and a large set of adjn languages to the north. The order of adjective and noun in Mandarin thus more closely resembles the languages to the north than many TB languages, especially those that are situated more closely to Chinese.
The situation regarding the position of demonstratives and numerals relative to the noun and the position of intensifiers relative to the adjective is similar: Chinese languages consistently place the modifiers before the modified element, like almost all languages in northeast Asia and unlike the majority of TB languages. The only TB languages like this are a subset of Bodic languages in Nepal and northwest India and these are the TB languages that are most distant geographically from Chinese. The tendency to consistently place modifying elements before the modified element is a property of ov languages of northern Asia. As discussed above, it is not a property of most ov languages outside Asia. In this way, therefore, Chinese resembles languages of northern Asia far more than it resembles TB languages or other ov languages, suggesting that these characteristics are best understood in terms of areal influence from languages of northeast Asia.
Someone wishing to deny the claim of areal influence from the north could take one of two approaches. One might try to argue that these characteristics reflect word order from an earlier time, perhaps going back to Proto-Sino-Tibetan. Since some of these characteristics (prepositional phrase and manner adverb before verb, standard of comparison before adjective, relative clause before noun) are ones generally found only in ov languages, this hypothesis would have to claim that Proto-Sino-Tibetan was ov and that Chinese has retained these characteristics, despite changing to vo order. Since TB languages are also largely ov, and share these characteristics, this is not an implausible scenario for these characteristics. But this leaves two things unexplained. First since these characteristics are so rare in vo languages other than Chinese, why has Chinese maintained them, when languages elsewhere in the world changing from ov to vo order apparently seldom if ever retain these ov characteristics? Second, while this explanation may make sense for the characteristics of Chinese that are generally associated with ov order, it does not explain why Chinese resembles languages to the north in ways that are not associated with ov order, namely, in placing adjectives, demonstratives, and numerals before nouns and intensifiers before adjectives. We would have to say that Proto-Sino-Tibetan also had these characteristics, coincidentally similar to languages to the north. And since the majority of TB languages do not have these characteristics, we would have to say that all these TB languages have lost these characteristics, except in the subset of Tibetic languages that are like Chinese in these ways. However, there is a more obvious explanation for the fact that these Tibetic languages place these modifiers before the modified element: they are also adjacent to languages which consistently place modifiers before the modified element, namely, Indic languages in Indo-European. The Indic languages (and also Dravidian) belong to a large arm of consistently premodifying languages that connects with the area in northern Asia where this is found via Pakistan, Afghanistan, Turkmenistan, Uzbekistan, and Kazakhstan.
In fact, even if the premodifying characteristics of Chinese are retained from Proto-Sino-Tibetan, it is likely that there has still been areal influence from the north in contributing to the Chinese retaining these characteristics. In general, it is probably the case that areal influence more often has an effect in encouraging languages to retain characteristics than in causing changes. In other words, even if Proto-Sino-Tibetan was consistently premodifying, like Modern Chinese, it is unlikely to be a coincidence that the languages that have retained these characteristics are precisely those (Chinese, southern and western Tibetic) that are adjacent to non-Sino-Tibetan languages that have exactly those characteristics, and that the languages which have lost these characteristics are ones that are not adjacent to such languages.
The idea that Chinese word order has been influenced from the north has been suggested by others, especially by Hashimoto 1986. Hashimoto provides a further type of argument for this influence, the fact that syntactic and phonological differences among Chinese languages follow a north–south pattern in that where one finds differences among Chinese languages, the languages to the north tend to be more similar to non-Sino-Tibetan languages (Tungus, Mongolian) to the north of Chinese. However, Hashimoto’s discussion assumes (following views shown to be incorrect by Dryer 1992) that the premodifying order within noun phrases is an ov characteristic. But the fact that this is not an ov characteristic, the fact that adjectives, demonstratives, and numerals do not tend to precede the noun in ov languages actually provides further support for Hashimoto’s position, since one cannot attribute these characteristics to Chinese being ov in the past (or moving towards ov).
An alternative hypothesis is that these characteristics of Chinese reflect internal changes that coincidentally led to characteristics that resemble languages to the north. In most cases, I do not think that this possibility should be ruled out, or even viewed as unlikely. When one examines the geographical distribution of typological characteristics, there are bound to be many instances of adjacent languages being similar by accident. However, the fact that Chinese is so unusual in some of these characteristics lowers the likelihood of coincidental resemblances, since there is a need to explain why Chinese has these characteristics when they are not found elsewhere in the world.
The most salient overall generalization about word order within Sino-Tibetan is that where one finds differences among languages, the different languages tend to be more similar in word order to adjacent non-Sino-Tibetan languages. In the last section, I have dwelt on the resemblances of Chinese to languages to the north, and have pointed out the resemblances of western and southern Tibetic languages to Indic languages, but I have also pointed out, in Section 2, the fact that the more eastern TB languages more closely resemble Tai-Kadai and Mon-Khmer languages to the east. We see this in its strongest form with the Karen languages, which are vo, like languages to the east. We also see it in the overall tendency for postmodifying order for various sorts of modifiers to be more common towards the east of TB, in Lolo, Bai, Qiang, and Pumi. However, even towards the east, we find genn order everywhere, even in Karen, as well as reln order, except in Karen.
On the other hand, the details are much more complex than these overall patterns might suggest. We have seen that for a number of modifiers, such as adjectives modifying nouns, there is considerable diversity, even within subgroups of TB. In addition, Tibetan, and the Bodic languages most closely related to it, do not fit the overall east–west pattern within TB, since they are towards the west, yet they tend to place modifiers after the noun.
Nor, surely, should all the geographical patterns be understood in terms of non-Sino-Tibetan languages influencing Sino-Tibetan languages rather than the other way round. It is precisely because we find such variation within Sino-Tibetan, compared to most adjacent families, that it is possible to see how the variation within Sino-Tibetan can be understood in terms of languages within Sino-Tibetan resembling adjacent groups of languages. In some cases, it may be that the direction of influence may have gone from Sino-Tibetan to non-Sino-Tibetan, but where that might be the case is not clear.