Australian Linguistic Society

Towards a Comparative Analysis of Mandarin De Constructions

Michelle Fu, Asian Studies,University of Manitoba, Winnipeg, Canada/University of Sydney, Sydney,

1. Background

De subordinate constructions such as wo pao de hen kuai 'I run very fast' have been a controversial issue in Chinese language research. The central debate has been on whether it is the first predicate wo pao de 'I run' or the second predicate hen kuai 'very fast' that qualifies as the main predicate. The Primary Predicate Hypothesis (PPH) adherents (cf. Chao 1968, Tai 1973, Tang 1977, Li & Thompson 1978, 1981, Chu 1983, C-R Huang and Mangione 1985, C-R Huang 1990) claim that V1 wo pao de is an adverbial adjunct, while V2 hen kuai is the matrix predicate. The Secondary Predicate Hypothesis (SPH) supporters (cf. Mei 1972, 1978, Henne, Rongen & Hansen 1977, Paris 1979, Zhu 1982, C-T Huang 1982, 1988, Ross 1984, Li 1985) hold the view that V1 is the main predicate, while V2 is an adverbial adjunct.

This paper provides an analysis of de constructions in comparison with non-de subordinate constructions in Chinese within the theoretical framework of the Role and Reference Grammar (RRG) (Foley & Van Valin 1984, Van Valin 1993). It will be argued that a distinction should be made between the type of subordination represented by de constructions and that exemplified by non-de sentences. It will be shown that when the structural properties of these two sentence types are examined, it becomes clear that it is the second predicate in a de construction which qualifies for the main predicate. The findings thus support the PPH position.

This paper is organized as follows. First, the specific issue investigated in this study is discussed. Then a brief summary of RRG concepts is provided. Next, an analysis of both de and non-de constructions is conducted, followed by a conclusion.

2. The Issue

One major piece of evidence that both PPH and SPH adherents have used to support their respective hypotheses concerns the A-not-A question formation. PPH followers have pointed out that when a de construction is in an A-not-A question form, it is V2 that occurs in such a form, but not V1, as shown in (1):

     (1) a. Wo pao de  kuai-bu-kuai.
            I  run DE  fast-not-fast
            'Did I run very fast?'

         b. *Wo pao-bu-pao  de kuai?
             I  run-not-run DE kuai

They have argued that, since in Mandarin only a main verb can appear in an A-not-A form, V2 must be a main predicate while V1 a subordinate predicate.

SPH adherents, on the other hand, have argued that an A-not-A form can also occur with subordinate verbs; therefore it is not a legitimate test for main predicates. For example, C-T Huang (1988: 279) claimed that example (2) is a clear counterexample to PPH's argument, where the A-not-A form occurs in the lower clause:

     (2) Ni  renwei tamen hui-bu-hui    lai?
         You think  they  will-not-will come
         'Do you think that they will come or
                                  do you think they won't?'

Certain questions arise: why do (1) and (2) behave differently with regard to the A-not-A form? Is there anything that is more fundamental underlining the constructions of (1-2) other than their different behaviour in the application of the A-not-A form? These are the questions that this paper attempts to answer.

3. Relevant Concepts of RRG

Following Foley and Van Valin (1984) and Van Valin (1993), it is assumed that a clause consists of three levels: the nucleus, the core and the periphery. The nucleus contains nothing but a predicate. The core consists of core arguments and the nucleus. The periphery contains the nucleus, the core, and peripheral arguments, such as temporal and locative phrases.

When two clauses combine to form a subordinate construction, they can be linked at any of those three levels. Again following RRG, it is assumed that the crucial property of subordination is syntactic embedding. Subordination formed at the core level has two distinct construction types, the unmarked and marked cases. The former refers to the construction type in which a bare core is embedded under another core as a core argument, i.e. as either a subject or an object. The latter refers to the construction type in which a periphery is embedded under a core as a core argument.

4. Analysis
4.1. Syntactic Position Test

Since examples (1-2) are subordinate constructions, the syntactic position test will be used to show the embedding situation in them. It tells us whether the embedded clause can function as a core argument or not. The application shows that the embedded clauses in both de and non-de constructions occupy the syntactic position normally reserved for a core argument. Note (3-6):

     De constructions
     (3) a. Na   ge ren    pao de hen  kuai.
            that CL person run DE very fast
            'That person ran very fast.'
            Literally: 'That person's running was very fast.'

         b. Na   ge ren    hen  kuai.
            that CL person very fast
            'That person was fast.'

     (4) a. Ta  chang de tiao-le    qilai.
            3sg sing  DE dance-PERF INC
            'S/he sang to the point that s/he started to dance.'

         b. Ta  chang de ge.
            3sg sing  DE song
            'the song that s/he sang'

     Non-de Constructions
     (5) a. Zhang laoshi  neng       lai  tai  hao  le.
            Zhang teacher be-able-to come very good RTM
            'That teacher Zhang can come is really great.'

         b. Zhang laoshi  tai  hao  le.
            Zhang teacher very good RTM
            'Teacher Zhang was really great.'

     (6) a. Wo zhidao ta  zou   le.
            I  know   3sg leave PERF/RTM
            'I knew that he had left.'

         b. Wo zhidao zhe  jian shi.
            I  know   this CL   matter
            'I knew this matter.'

Comparing (a) with (b) examples in (3), (5) and (6), it is seen that both the highlighted clauses and the highlighted noun phrases hold the same syntactic position. This suggests that the highlighted clauses in these sentences are embedded and function as the core arguments of the complex constructions of which they are a part. Examples in (4), however, are different. When V2 is replaced by a noun, as in (4b), the result is a phrase rather than a sentence. Therefore V2 in (4a) and the noun phrase in (4b) do not share the same syntactic position.

These facts point to a crucial difference between de and non-de constructions in term of their respective embedding situation. That is: de constructions only allow the embedded clause to function as a logical subject but not a logical object, as shown in (3-4). Non-de constructions, on the other hand, allow the embedded clause to function as either a logical subject or object, as shown in (5-6). To conclude, V2 in de constructions is not an embedded predicate and cannot be subordinate to V1; V1, on the other hand, is an embedded constituent and is subordinate to V2. Such a finding supports the PPH position.

4.2 Unequal Information Structure

Following Foley and Van Valin and many other scholars (Haiman 1978, Thompson 1985, Givon 1987, Ramsay 1987, Lehmann 1988, Lambrecht 1994), it is assumed that the embedded clause typically encodes background information, while the matrix clause provides foregrounded information. Haiman (1985) has stated that the clause that contains presupposed information is not subject to challenge. The negation test, adopted from Haiman (1985), shows that the content of the embedded clauses in examples like (1-2) cannot be questioned.

     De Construction
     (7) a. Na   ge ren    pao de hen  kuai.
            that CL person run DE very fast
            'That person ran very fast.'

         b. Ta  pao de bu  kuai.
            3sg run DE NEG fast
            'S/he did not run fast.'

         c. *Ta  mei pao.
             3sg NEG pao
             'S/he did not run.'

     Non-de Construction
     (8) a. Ming zhidao ta  zou   le.
            Ming know   3sg leave PERF/RTM
           'Ming knew that he had left.'

         b. Ta  bu  zhidao.
            3sg NEG know
            'S/he did not know.'

         c. *Ta  mei zou.
             3sg NEG leave
             'S/he did not leave.'

The denial in (7b) where the higher predicate is challenged is acceptable, while the denial in (7c), where the embedded predicate is challenged, is not. Similarly, the response denying the matrix clause in (8b) is acceptable, whereas the response denying the subordinate clause in (8c) is not. Therefore, it can be concluded that the embedded clauses in (7-8) contain presupposed information known to the speaker and the hearer when those sentences are uttered. The matrix clauses in (7-8), on the other hand, contain the foregrounded information which can be challenged.

Apart from the above observations, the application of the negation test has also shown that it is V1 in a de construction that contains background information. V2, whose content can be denied, contains foregrounded information. Based on these facts, a conclusion can be drawn that V1 is the subordinate clause and V2 the matrix clause. Such a finding again supports the PPH stand.

4.3 Further Comparison

It is now argued that de and non-de constructions can be further distinguished from one another in several other aspects with regard to operator scope and insertion of peripheral arguments, aspect markers, a negator and an optional pause. It is argued that distinctions in these aspects result from the fact that the embedded clause in a de construction is a bare core embedded in another core. However, the embedded clause in a non-de construction is a periphery embedded in another core. In other words, (1) represents the unmarked case of core subordination, while (2) illustrates the marked case. It is argued that although the embedded bare core in the de construction is a clause, it has been stripped of all the properties that a full clause normally has. The embedded periphery in the non-de construction, on the other hand, possesses almost all the qualities of a full clause.

4.3.1 Contrast in Bare vs Full-sentence Properties

As mentioned earlier the previous studies by PPH followers have pointed out that the A-not-A question marker may only occur in the matrix clause of a de construction. Such a marker, however, may appear in both the embedded clause (Huang 1988) and the matrix clause of a non-de construction.

Now the question is: why do de and non-de sentences behave differently with regard to the use of the A-not-A illocutionary force marker? It is argued here that failure of the occurrence of the A-not-A question marker in the embedded clause of a de construction is due to the fact that the embedded clause is a bare core and does not possess full-clause properties. On the other hand, the possible presence of the A-not-A question form in the embedded clause of a non-de construction is due to its full-clause properties. To illustrate, the fact that the embedded clause of a non-de sentence may carry the A-not-A question marker suggests that the clause is virtually a full sentence; on the contrary, the embedded clause of a de sentence does not allow such a marker. Such a contrast in bare vs full-sentence properties exhibited by the embedded clauses of de and non-de constructions respectively can be further substantiated by other evidence. Insertion of Modality Markers

When a modality marker appears in the embedded clause of a de construction, it has scope over the whole sentence, as shown in (9). However, when it occurs in the embedded clause of a non-de construction, it only qualifies the embedded clause where the marker appears, as in (10).

     De Construction
     (9) Na   ge ren    keyi pao de hen  kuai.
         that CL people may  run DE very fast
         'That person may run very fast.'
         '*That person may run and is very fast.'

     Non-de Construction
     (10) Ni  renwei tamen keyi lai  ma?
          you think  they  may  come Q
          'Do you think that they may come?'
          '*May you think that they may come?'

Note that keyi 'may' appears in the embedded clause in both examples. However, in (9) it affects the whole construction, while in (10) it affects only the clause in which it occurs. The freedom or lack of freedom to have its own modality marker indicates that the embedded clause of a non-de construction possesses full-sentence properties, while that of the de construction does not. Insertion of Peripheral Arguments and Aspect Markers

The freedom to specify independent peripheral arguments, as well as aspect markers in the embedded clause of a non-de construction reinforces its full-sentence properties. Compare (11) with (12):

     De Construction 
     (11) a. Na   ge ren    zuotian   pao de hen  lei.
             that CL person yesterday run DE very tired
             'Yesterday that person was tired from running.'

          b. *Na  ge ren    pao le/guo   de hen  lei.
             that CL person run PERF/EPX DE very tired

     Non-de Construction
     (12) a. Wo zhidao ta  zuotian   zou   le.
             I  know   3sg yesterday leave PERF
             'I know that s/he left yesterday.'

          b. Wo zhidao ta  zuotian   lai  guo.
             I  know   3sg yesterday come EXP
             'I know that s/he was here yesterday.'

Example (11a) shows that the peripheral argument zuotian 'yesterday', although appearing in the embedded clause, must be shared by both clauses. Example (11b) demonstrates that aspect markers le and guo cannot appear in the embedded clause of a de construction. In (12), however, the peripheral argument zuotian 'yesterday' exhibits semantic scope over the embedded clause only, without affecting the matrix clause. It can also be seen that the embedded clause allows occurrence of the aspect markers le and guo. These facts indicate that the embedded clause in (12) assume the status of a nearly full sentence, while the embedded clause in (11) does not. Insertion of a Negator

Additional support for such full-sentence properties of the embedded clause of a non-de construction may also be provided by the freedom to insert a negator into the embedded clause. See (13-14):

     De Construction
     (13) Na   ge ren    *bu pao de bu  hen  kuai.
          that CL person NEG run DE NEG very fast
          'That person did not run very fast.'

     Non-de Construction
     (14) Wo zhidao ta  bu  zou   le.
          I  know   3sg NEG leave PERF/RTM
          'I know that s/he will not leave.'

It is not possible for the embedded clause of a de construction to be independently negated, as is shown in (13), while this is not the case with the embedded clause of a non-de construction, as is shown in (14). Phonological Evidence

Apart from the syntactic evidence mentioned above, phonological evidence exhibited by non-de constructions also appears to point to a full-sentence analysis for their embedded clause. For example, there can exist a very short pause between the clauses of the non-de constructions collected. The pause is not long enough to be registered by a comma, but it is noticeable occasionally. When it appears, it seems to function as a "silent" complementizer, linking related clauses. This situation contrasts with what is presented by clauses of de constructions, where no intonational break between them is possible.

5. Conclusion

This paper has analyzed the de construction by comparing it with the non-de subordinate construction. It has been argued that although those two sentence types are both subordinate constructions, they differ from one another syntactically and phonologically. In a de construction, the embedded clause can only function as a logical subject, but not a logical object. In a non-de construction, on the other hand, the embedded clause can function either as a logical subject or object. The analysis of the de construction indicates that V1 is the subordinate clause, while V2 the matrix clause, thus supporting the PPH stand. This paper has also argued that de and non-de constructions should be treated separately. This is because the former represents the unmarked case of core subordination in which the embedded clause is a bare core lacking full-sentence properties. The latter represents the marked case in which the embedded clause is a periphery possessing nearly all the properties of a full-fledged sentence.

This study has offered a solution to a long-standing problem in research on de constructions. It has an advantage over the previous accounts inasmuch as the present analysis has addressed both the similarities and differences between de and non-de subordinate constructions, while the previous studies have only dealt with the similarities by lumping the two sentence types into the same category, subordination, but have failed to examine the crucial differences between them.

Selected References

  • Chao, Yuen Ren. 1968. A grammar of spoken Chinese. Berkeley: University of California Press.
  • Chu, Chauncey Cheng-hsi. 1983. A reference grammar of Mandarin Chinese for English speakers. NewYork: Peter Lang Publishing Inc..
  • Givon, T. 1987. Beyond foreground and background. In R. Tomlin (ed), Coherence and grounding in discourse. Amsterdam: John Benjamins Publishing Company. 175-188.
  • Foley, William A. & Robert D. Van Valin, Jr. 1984. Functional syntax and universal grammar. Cambridge: Cambridge University Press.
  • Haiman, John. 1985. Natural syntax: iconicity and erosion. Cambridge: Cambridge University Press.
  • Henne, Henry, Rongen, Ole B. & Hansen Lars J. 1977. A handbook on Chinese language structure. Bergen: Universitetsforlaget.
  • Huang, C-R. 1990. Review of Functionalism and Chinese Grammar edited by James H-Y. Journal of Chinese Linguistics. 18 (2): 318-334.
  • Huang, C-T. James. 1988. Wo pao de kuai and Chinese phrase structure. Language. 64 (2): 274-311.
  • Li, Y-H. Audrey. 1985. Abstract case in Chinese. University of Southern California dissertation.
  • Ross, Claudia. 1984. Adverbial modification in Mandarin. Journal of Chinese linguistics. 12: 207-234.
  • Tai, James H-Y. 1973. A derivational constraint on adverbial placement in Mandarin Chinese. Journal of Chinese linguistics. 1: 397-413.
  • Van Valin, Robert D., Jr. 1993. A synopsis of role and reference grammar. In Robert D. Van Valin, Jr. (ed), Advances in role and reference grammar. Amsterdam: John Benjamins Publishing Company. 1-166.