Unclogging the Sink: Avoiding Center-Embedding with Subordinating Verbs

While working on the chapter on verb morphology in the revised Ayeri grammar I’ve been working on all the way since July, I had to stop the other day and consider how to deal with participles. With regards to Ayeri, these are the infinite subordinated verbs which appear together with verbs that may take another verb – or even a whole phrase headed by such an infinite verb – as a complement:

    1. Cunyo
      Cun-yo
      begin-3SG.N
      makayam
      maka-yam
      shine-PTCP
      perinang.
      perin-ang
      sun-A

      ‘The sun began to shine.’
    2. Manangyeng
      Manang=yeng
      avoid=3SG.F.A
      pengalyam
      pengal-yam
      meet-PTCP
      badanas
      badan-as
      father-P
      saha
      saha
      in.law
      yena.
      yena
      3SG.F.GEN

      ‘She avoids to meet her father-in-law.’

Example (1a) shows two intransitive verbs being combined, cun- ‘begin’ (subordinating) and maka- ‘shine’ (complement); example (1b) illustrates an intransitive verb which takes a transitive verb as a complement, manang- ‘avoid’ (subordinating) and pengal- ‘meet’ (complement). We can also observe word order differences in that the agent NP in (1a) follows the initial verb complex while the agent NP is an enclitic pronoun directly following its governing verb in (1b), which is not surprising, since agent pronouns take verbs as clitic hosts. That is, the agent pronoun -yeng ‘she’ (3SG.F.A) bundles together with its governing verb manang- ‘avoid’ rather than cliticizing to the whole verb phrase,1 i.e. appending -yeng ‘she’ (3SG.F.A) to pengalyam ‘meeting’. As far as describing the morphology of participles goes, we are basically done here, and this is also where I stopped thinking previously.

However, if we actually continue from here and take a morphosyntactic point of view, an interesting question arises: what happens to the constituents’ linearization if we (a) use full NPs instead of cliticized agent pronouns, and (b) not only combine an intransitive verb with a transitive one, but try all possible combinations–intransitive, transitive, ditransitive, and any of these with adverbial adjuncts? I think that for the sake of sketching out my thoughts here, it won’t be necessessary to give examples of all 18 possible combinations, since the more complex cases should all be treated alike anyway.

Assuming that she in (1b) is Maha, who just doesn’t get along with her husband’s father, where does the agent NP go? As noted above, the position of the agent clitic is somewhat special, so I take it that (1a) should be the basic word order. Thus, for (1b), we get:

  1. Manangye
    Manang-ye
    avoid-3SG.F
    pengalyam
    pengal-yam
    meet-PTCP
    badanas
    badan-as
    father-P
    saha
    saha
    in.law
    yena
    yena
    3SG.F.GEN
    ang
    ang=
    A=
    Maha.
    Maha
    Maha

    ‘Maha avoids to meet her father-in-law.’

Besides the fact that the agent winds up at the very end of the clause instead of at its otherwise preferred position after the verb, this still seems to be comprehensible. What happens, however, if we use a transitive subordinating verb? The dictionary, for instance, lists pinya- ‘ask (s.o. to do sth.)’ as a candidate. With regards to the normal constituent order of Ayeri, we can assume that the patient NP will follow the agent one:

  1. Ang
    Ang
    AT
    pinyaya
    pinya-ya
    ask-3SG.M
    sahayam
    saha-yam
    go-PTCP
     
    Ø=
    TOP=
    Yan
    Yan
    Yan
    sa
    sa=
    P=
    Pila.
    Pila
    Pila

    ‘Yan asks Pila to go.’

This also still looks very harmless with regards to parsability. However, things become more complicated if we increase the complexity of the embedded phrase by making the subordinate verb transitive (4a) or even ditransitive (4b):

    1. Ang
      Ang
      AT
      pinyaya
      pinya-ya
      ask-3SG.M
      konjam
      kond-yam
      eat-PTCP
      inunas
      inun-as
      fish-P
       
      Ø=
      TOP=
      Yan
      Yan
      Yan
      sa
      sa=
      P=
      Pila.
      Pila
      Pila

      ‘Yan asks Pila to eat the fish.’
    2. Ang
      Ang
      AT
      pinyaya
      pinya-ya
      ask-3SG.M
      ilyam
      il-yam
      give-PTCP
      koyaley
      koya-ley
      book-P.INAN
      ledanyam
      ledan-yam
      friend-DAT
      yana
      yana
      3SG.M.GEN
       
      Ø=
      TOP=
      Yan
      Yan
      Yan
      sa
      sa=
      P=
      Pila.
      Pila
      Pila

      ‘Yan asks Pila to give the book to his friend.’

The distance between the subordinating verb and its arguments grows by the increasing number of constituents in the embedded phrase, which in turn becomes increasingly “deep” in terms of underlying syntactic structure or “heavy” with regards to syntactic weight. The parser in both the speaker’s and the listener’s brain thus has to keep track of more and more relations in parallel. In terms of information flow, this does not strike me as beneficial or intuitive either to a speaker constructing the phrase, or to a listener having to decode the utterance. For the same reason, there is already a rule that relative clauses constitute “heavy” elements which pull their referent NP all the way to the back of the clause to keep the upper right field free of clutter. While (4a) is a little awkward since there are two patients to keep track of at the same time and at the time the first patient NP occurs we don’t know yet whether -ya is just for agreement or a cliticized agent-topic pronoun, example (4b) seems even more impenetrable with its three referents in the embedded clause and two in the matrix clause, placed at the end. Putting less important information before important information also goes against information-flow intuition.2 Restructuring seems advisable here, thus, and there are two possibilities:

    1. Ang
      Ang
      AT
      pinyaya
      pinya-ya
      ask-3SG.M
      _​​i​
       
       
       
      Ø=
      TOP=
      Yan
      Yan
      Yan
      sa
      sa=
      P=
      Pila
      Pila
      Pila
      [ilyam
      il-yam
      give-PTCP
      koyaley
      koya-ley
      book-P.INAN
      ledanyam
      ledan-yam
      friend-DAT
      yana]​​i​​.
      yana
      3SG.M.GEN

      ‘Yan asks Pila to give the book to his friend.’
    2. Ang
      Ang
      AT
      (da​​i​-)pinyaya
      (da=)pinya-ya
      (so=)ask-3SG.M
       
      Ø=
      TOP=
      Yan
      Yan
      Yan
      sa
      sa=
      P=
      Pila
      Pila
      Pila
      [, ang
      ang
      AT
      ilye
      il=ye.Ø
      give=3SG.F.TOP
      koyaley
      koya-ley
      book-P.INAN
      ledanyam
      ledan-yam
      friend-DAT
      yana]​​i​​.
      yana
      3SG.M.GEN

      ‘Yan asks Pila to give the book to his friend.’

In the past I might have preferred (5a) as a solution, but since the analogy to relative clauses suggests itself and complement clauses are useful, I find that I tend towards the sentence in (5b) currently. To indicate that the sentence structure has been remodeled and information will be following, it might also be useful to add the da- particle to the verb: a more literal translation of (5b) could be ‘Yan asks Pila such that she give the book to his friend.’ Either way, however, the normal VSO word order is restored this way, important information is present early on, and the heavy constituent is banned to the back. Information should now be able to flow easily again.

  1. If Ayeri has such a thing, which is kind of a problem for me right now. I need to read up on VSO-language syntax before I can make conclusive claims.
  2. Hat-tip to Oliver Schallert (blog article in German).