Update on the Grammar Writing Process VIII

I wrote earlier this month that I had been revising the index of my Ayeri grammar. I also noticed that a discussion of verbs with predicative complements like tav- ‘become’, maya- ‘feel’, etc. has been missing so far. I completed and added those things yesterday, respectively. This also means that the manuscript is basically finished since it should now include everything I meant to discuss, and it should be rather presentable. However, hold your horses, it still needs another round of proofreading to weed out mistakes that have crept in due to adding and deleting index tags from the source files, as well as mistakes which are due to my not being a native English speaker, or just plain lapses.

Update on the Grammar Writing Process VII

Oh wow, it’s been a full hundred days since I last gave a report here about my progress. The project is still going on, in case anyone wondered.

After littering the LaTeX source files of the grammar with index tags in April, I’ve been working on clearing them up again for the last 3 weeks or so to make the keyword index actually useful. And while I’ve been at it, I’ve been fixing the one or the other issue I’ve come across as well—spelling, formatting, content. I had hoped to be done with this task by June 1st, but just as usual, everything is taking twice as long as expected. Let me tell you, it’s pretty annoying to go through all page references one by one, and to check whether they’re leading to actually relevant information.

I seriously want this off my desk as soon as possible now, even though I’ve learned a lot by writing this book. However, it’s been preoccupying me for long enough—on July 3rd, it’s going to be 2 years. July 1st is the deadline I gave myself, though knowing my perfectionist tendencies, it’s probably rather going to be August: I’ve been considering to ask some native English speakers I know for some additional proofreading.

I also feel a little guilty about spending so much time on writing this grammar instead of working as hard on my Ph.D. project for university: there are about 1,800 commits to the repository (about 1,000 in the last 12 months alone), and if we assume that each one equals about 45 minutes of work on average (reading takes a lot of time, which is balanced by correcting small things), this amounts to 1,350 hours. This, in turn, is about equal to 34 weeks on a full-time job. On the other hand, I suppose I should be fine if I’ll continue working on my thesis with as much zeal as on the grammar, once the grammar book is done for the time being. Blood, sweat, and tears, etc. Anyway, I’ve come so far with this book project, I don’t want to put it on hold indefinitely, especially now that it looks like the end is only weeks away. And then I can hopefully move on.

Update on the Grammar Writing Process VI

Incidentally, it was the 600th day since I started working on a refined version of the Ayeri Grammar just yesterday. I’ve spent the last 2 weeks proofreading the chapter on syntax and I’m now ready to go back to the start and proofread the whole book again. You can see the very raw manuscript containing the first round of corrections right after writing the chapters and under it the batch waiting for the second round of proofreading in the first photo. The second photo shows that it’s not an inconsiderable amount of paper—and I printed it as 4 pages per sheet.

I’ve said before that I definitely wanted to have a bound copy of the book myself once it’s done (maybe in summer?), so I’ve looked a little at print-on-demand services recently since I figured that other people might want to have one as well. Apparently, whether you use Amazon’s CreateSpace or Lulu.com, a paperback the size I’ve been considering—15.6 × 23.4 cm with approx. 480 pages—can be produced for about $16, so about €13, which sounds pretty reasonable. You even get an ISBN and distribution included, though I have no idea whether a very small profit margin for the author is already factored in as well (we’re likely talking quarters per copy here). I’m certainly not expecting to sell many copies, since conlangs are a pretty specific thing, but if I made the one or the other buck this way which I could then reinvest in running this site, I’d not be entirely unhappy. Since I like to use print and digital in parallel especially with textbooks, I’m planning to keep up the PDF version of the grammar for free. This is essentially what Language Science Press does, minus the peer review they also provide. If anyone were to want a print copy, they could order it in addition and pay a reasonable price for getting a proper book. I think this is a pretty fair offer.

Update on the Grammar Writing Process V

Happy new year, everyone! I suppose it’s time again to provide a brief update on my progress with writing my grammar of Ayeri. The whole last year I’ve been trying to figure out describing its syntax formally. This will continue to preoccupy me for the time being also in the new year because verbs are still not fully described, and complementizer phrases (used for complement clauses, relative clauses and such) are lining up to be next. Then, I will also have to work on correcting some things in the sections on raising and control with regards to syntactic typology (I should have figured out constituent structure first), and also describe pronominal binding. And after this, I will have to go back to the beginning of the chapter and fix things for consistency and do proofreading.

The compiled PDF is now close to 400 pages (in A4 format, but with generous margins because LaTeX) without frontmatter, appendices and backmatter, and 400 pages is what I had wanted the main part to be at most once everything is done. The section on the syntax of verbs alone is already almost 100 pages long currently, though granted, verbs are probably the most complex part of the language (or any language?), and all those diagrams take up an awful lot of space. I will definitely have to shave some pages off after writing will be done hopefully some time later this year, though, and especially the argumentative parts are probably predestined for some literal cutting to the chase in spite of my trying not to ramble unnecessarily. The description of Ayeri’s alphabet might also rather go in the appendix. Years at university have taught me that good writing can’t be produced on the spot, anyway.

Honestly, sometimes I wish I had an editor to look over my writing to guide me with it. With the syntax chapter especially, I wish someone could check the plausibility of my hypotheses and analyses once writing is done, too. And then, there’s still proofreading of the whole grammar to do. My English may be pretty good overall, but I’m always somewhat distrusting my abilities as a non-native speaker. Proofreading one’s own writing is generally hard in my experience, though, even in one’s native language.

A Question of Alignment XII: Conclusion

In this series of blog articles—taken (more or less) straight from the current working draft of chapter 5.4 of the new grammar for better visibility and as a direct update of an old article (“Flicking Switches: Ayeri and the Austronesian Alignment”, 2012-06-27)—I will finally reconsider the way verbs operate with regards to syntactic alignment.

All articles in this series: Typological Considerations · ‘Trigger Languages’ · Definition of Terms · Some General Observations · Verb agreement · Syntactic Pivot · Quantifier Float · Relativization · Control of Secondary Predicates · Raising · Control · Conclusion


Now that a few tests have been conducted, let us collect the results. As Table 1 shows, Tagalog and Ayeri are not really similar in syntax despite superficial similarities in morphology. According to Kroeger (1991)’s thesis—which essentially seeks to critically review and update Schachter (1976)’s survey by leaning on LFG theory—Tagalog prefers what Kroeger analyzes as the nominative argument for most of the traits usually associated with subjects listed below. That is, in his analysis, the nominative argument is the NP in a clause which is marked on the verb, which corresponds to Schachter (1976)’s ‘topic’, or Schachter (2015)’s ‘trigger’—’trigger’ is also the term often seen in descriptions of constructed languages in this respect. Kroeger (1991) finds in his survey that the nominative argument is largely independent from the actor, so that the logical subject is not necessarily the syntactic subject; what Schachter (1976) calls ‘topic’ also does not behave like a pragmatic topic in terms of statistics.

Essentially, what Tagalog does according to Kroeger (1991)’s analysis, is to generalize voice marking beyond passive voice, so that any argument of the verb can be the subject. However, unlike passives in English, higher-ranking roles (for passives, the agent) appear not to be suppressed or to be demoted to adverbials like it happens in English with the periphrasis of the agent with by in passive clauses. Linguists have been grappling for a long time with this observation, and constraint-based approaches, such as LFG (recently, Bresnan et al. 2016) or HPSG (Pollard and Sag 1994) pursue, may be able to explain things more succinctly than structuralist ones due to greater flexibility. In any case, Kroeger (1991) avoids the terms ‘active’ or ‘passive’ possibly for this reason, and instead uses ‘actor voice’ (AV), ‘objective voice’ (OV), ‘dative/locative voice’ (DV) etc. (14–15).

Ayeri, in contrast to Tagalog, very much prefers the actor argument (called agent here for consistency) for traits usually associated with subjects, independent of whether the agent is also the topic of the clause—in Ayeri it is the topic which is marked on the verb, not the nominative argument. In spite of a few irregularities like patient agreement in agentless clauses and using topicalization as a way to disambiguate the syntactic pivot in ambiguous cases, Ayeri is remarkably consistent with a NOMACC language. The fact that there is a subject in the classic, structural sense is also evidence for the hypothesis that Ayeri is configurational. Since it clearly prefers agent NPs over other NPs, not all arguments of a verb are on equal footing. Tagalog, on the other hand, treats the arguments of verbs in a much more equal manner.

Table 1: Comparison between Tagalog (Kroeger 1991) and Ayeri
CriterionTagalogAyeri
Marked on the verbnominative argument (NOM)topic argument (TOP)
Verb agreementoptional; if present with NOM, independent of being Arequired; typically with A, independent of being TOP
Syntactic pivotdetermined by NOM, independent of being Ausually with A, but determined by TOP in ambiguous cases
Quantifier floatreferring to NOM, independent of being Areferring to A, independent of being TOP
Relativizationonly of NOM, independent of being A(all NPs may be relativized)
Control of secondary predicatesreferring to NOM, independent of being Areferring to A or P depending on semantics, but independent of being TOP
Raisingusually of NOM; A possible but marked for someonly of A, independent of being TOP; no ECM
ControlA deletion target, independent of being NOM (with exceptions)A deletion target, independent of being TOP

It was pointed out before that in Tagalog, the syntactic pivot depends on what is marked as a subject (Kroeger 1991: 30–31). This and other examples from Kroeger (1991) may make it seem like Tagalog is not fixed with regards to the distinction between NOMACC and ERGABS alignment. However, Kroeger (1991) also points out that there is a statistically significant preference to select patient arguments as subjects, and that OV forms of verbs are “morphologically more ‘basic’” (53) than their respective AV counterparts. These observations point towards an interpretation of Tagalog as syntactically ergative, though Kroeger (1991) deems such an interpretation problematic due to non-nominative agents keeping their status as arguments of the verb—which also distinguishes Tagalog from an ergative languages like Dyirbal, where “ergative (or instrumental) marked agents are relatively inert, playing almost no role in the syntax, and have been analyzed as oblique arguments” (54).

In conclusion, is Ayeri a so-called ‘trigger language’? Yes and no. It seems to me that what conlangers call ‘trigger language’ mostly refers to just the distinct morphological characteristic of languages like Tagalog by which a certain NP is marked on the verb with a vague notion that this NP is in some way important in terms of information structure.1 Ayeri incorporates this morphological feature and may thus be counted among ‘trigger languages’ by this very broad definition. However, the real-world Austronesian alignment as a syntactic phenomenon goes much deeper than that and is much more intriguing, as I have tried to show in this series of blog articles, and I did not even cover all of the effects Kroeger (1991) describes in his survey. Ayeri, in syntactically behaving rather consistently like a NOMACC language, (somewhat sadly, in retrospect) misses the point completely if ‘trigger language’ is understood to also entail syntactic characteristics of Philippine languages.

  • Bresnan, Joan et al. Lexical-Functional Syntax. 2nd ed. Chichester: Wiley Blackwell, 2016. Print. Blackwell Textbooks in Linguistics 16.
  • Kroeger, Paul R. Phrase Structure and Grammatical Relations in Tagalog. Diss. Stanford University, 1991. Web. 17 Dec. 2016. ‹http://www.gial.edu/wp-content/uploads/paul_kroeger/PK-thesis-revised-all-chapters-readonly.pdf›.
  • Pollard, Carl and Ivan A. Sag. Head-Driven Phrase Structure Grammar. Chicago, IL: U of Chicago P, 1994. Print. Studies in Contemporary Linguistics.
  • Schachter, Paul. “The Subject in Philippine Languages: Topic, Actor, Actor-Topic, or None of the Above?” Subject and Topic. Ed. Charles N. Li. New York: Academic P, 1976. 493–518. Print.
  • ———. “Tagalog.” Syntax—Theory and Analysis: An International Handbook. Ed. Tibor Kiss and Artemis Alexiadou. Vol. 3. Berlin: De Gruyter Mouton, 2015. 1658–1676. Print. Handbooks of Linguistics and Communication Science 42. DOI: 10.1515/9783110363685-007.
  1. I want to encourage everyone to actually do some reading of the professional literature on a given topic instead of only relying on the second-hand knowledge of other people in the conlanging community. It’s hard but you’ll learn from it. With the internet, finding articles and books is as easy as ever. This is one of the reasons why I give citations under the more serious blog articles, and make sure to link literature that is legally available online.

A Question of Alignment XI: Control

In this series of blog articles—taken (more or less) straight from the current working draft of chapter 5.4 of the new grammar for better visibility and as a direct update of an old article (“Flicking Switches: Ayeri and the Austronesian Alignment”, 2012-06-27)—I will finally reconsider the way verbs operate with regards to syntactic alignment.

All articles in this series: Typological Considerations · ‘Trigger Languages’ · Definition of Terms · Some General Observations · Verb agreement · Syntactic Pivot · Quantifier Float · Relativization · Control of Secondary Predicates · Raising · Control · Conclusion


Control verbs behave basically in the opposite way of raising verbs: the subject of the subordinate verb is also an argument of the verb in the matrix clause—subject or object—and this argument acts as a controller for the subject of the subordinate verb. The main clause predicate thus is thought to assign two thematic roles. In GG it is assumed that the subject of the lower clause is a silent PRO element which is coindexed with the controller (Carnie 2013: 442–445, 451).

    1. Subject control:

      Johni tries [that Johni gets a job]
      = Johni tries [PROi to tPRO get a job]

    2. Object control:

      The officer ordered Maryi [that Maryi turn back]
      = The officer ordered Maryi [PROi to tPRO turn back]

Kroeger (1991) refers to subject control as ‘Equi’ and reports that according to Schachter (1976: 505), it is typically the actor of the subordinate verb that is the target of deletion. At first sight this would be a strong argument in favor of defining the actor NP as the subject, however, he notes that under certain circumstances, “the controllee in a transitive complement clause [is allowed] to be either the Actor (regardless of case marking) or the argument which bears nominative case” (Kroeger 1991: 37). This is the case, for instance, with himukin ‘persuade’ and magpilit ‘insist on’. Subordinate verbs marked for non-volitive mood form an exception as well (36–37, also 96–97). Kroeger (1991) illustrates the main pattern of control in Tagalog with the following set of example sentences in (2).

  1. Tagalog (adapted from Kroeger 1991: 37):

While the nominative argument of the subordinate verb changes between the actor in (2a), the theme in (2b), and the recipient in (2c), it is always the actor which is dropped as the coreferential argument. Why the example sentences in (2) use balak ‘plan, intend’ in its object-voice form is not explained. However, Kroeger (1991) mentions that “alternation in the voice category of the matrix verb and the case marking of the controller does not affect the control relation” (37). In other words: whether the actor in the matrix clause is the subject or not does not matter; for Tagalog’s equivalent of subject-control verbs the control relationship always finds its origin in the actor argument, although there are a few exceptions, as mentioned above. The set in (3) presents an interesting example of (‘obligatory’) control based on the patient/theme argument in Tagalog’s equivalent of object-control verbs.

  1. Tagalog (adapted from Kroeger 1991: 93–94):

Regarding (3ab), Kroeger (1991) explains that “when the complement verb appears in its volitive (unmarked) form, the controllee must be the Actor of the embedded clause” (93). Thus, Maria cannot be the patient subject in (3b), since she is still the controllee. If the verb of the embedded clause is marked for non-volitive mood as in (3c), however, the sentence becomes grammatical: “When the embedded verb is marked for non-volitive mood, the pattern is reversed: the controllee must be the subject, and not the Actor. Actor gaps cannot be controlled in non-volitive complements” (94). The difference between obligatory and non-obligatory control adds a further complication to acceptability, but these details do not need to preoccupy us for the purpose of comparison to Ayeri, which lacks these distinctions.

As previously with raising verbs, it is possible in Ayeri to combine a subordinating verb with a full complement clause (4a), an embedded IP complement (4b), or complete incorporation (4c). In both (4a) and (4b) cases, it is necessarily the actor which is coreferened, as the bottom arrow shows. In (4c), the bottom arrow does not show coreference, but the relation of verb agreement. The arrow on top, as before, shows what the respective verb picks as the clause’s topic for all example sentences.

As with raising verbs, the embedded and incorporated verbs appear in a non-finite form, the participle. For (4b) the reason may be that there is no overt agent in the clause with which to agree, and agreement with the patient does not make sense here because the clause does not express a passive either. In (4c) the reason may be that the main verb already carries person features. If the topic marking on the finite verb is altered as in (5), the meaning of the sentences does not change with regards to grammatical relations and voice, giving us yet more reason to assume that the agent is the grammatical subject, and that topic marking has no influence on these matters. Ayeri thus has actual subject-control verbs in the way English has them.

In object-control constructions, the object of the matrix clause’s verb is an actual argument of it, as shown in (6). This argument becomes the subject of the embedded clause, and there is no change in the meaning of the verb between both versions of sentences. We have seen above that Ayeri does not allow to-object raising, since it is not possible to assign patient case to an external agent because Ayeri’s case marking is not purely based on grammatical functions, but there is still also some semantic motivation. Ayeri does, however, allow object control, so it seems to be possible at least to implicitly convert the matrix clause’s patient or theme to the agent of the embedded clause, while the opposite is apparently not possible. Whether syntactic precedence or some kind of accessibility hierarchy is involved here still needs to be investigated.

    1. John asked [Mary to give Peter the book]
      = John asked Mary
    2. The teacher instructs [the students to calculate parables]
      = The teacher instructs the students
    3. I persuaded [my friend to come along]
      = I persuaded my friend

The example sentences in (7) follow the format of those above. Again, it is generally possible to use a complement clause as in (7a) or (7b), as well as complementing the verb in the matrix clause with a non-finite clause with object control (7c). However, the incorporation strategy is not possible here because this would cause a doubling of case roles (7d). As we will see below, however, this is not an issue for intransitive complement clauses.

Strictly speaking, it does not matter in (7a) and (7b) whether the coreferenced argument is the topic in both clauses or not; it is simply not unlikely that it is. Again, topicalization does not have an effect on grammatical relations—although it was shown above that Tagalog, in the canonical case, deviates from its normal behavior as well with regards to control verbs to the point where this construction has been used as an argument in favor of the actor argument being the subject. As for Ayeri, unlike in coordinated main clauses, topicalization is not a strategy for disambiguation of several possible controllers for the pronominal agent of the complement/embedded clause here. Due to the semantics of the verb in the matrix clause, it is clear that the patient argument is to be understood as the agent of the subordinate verb. Thus, there is no ambiguity in anaphoric reference in the complement clause.

  • Carnie, Andrew. Syntax: A Generative Introduction. 3rd ed. Chichester: Wiley Blackwell, 2013. Print. Introducing Linguistics 4.
  • Kroeger, Paul R. Phrase Structure and Grammatical Relations in Tagalog. Diss. Stanford University, 1991. Web. 17 Dec. 2016. ‹http://www.gial.edu/wp-content/uploads/paul_kroeger/PK-thesis-revised-all-chapters-readonly.pdf›.
  • Schachter, Paul. “The Subject in Philippine Languages: Topic, Actor, Actor-Topic, or None of the Above?” Subject and Topic. Ed. Charles N. Li. New York: Academic P, 1976. 493–518. Print.

A Question of Alignment X: Raising

  • Example (6) in this article is likely wrong, structurally. I’ll update it when I’ve figured some more things out.
  • It’s internally more consistent to maybe analyze it as a VP complement getting its verb extracted and adjoined as a complement to the main verb. There is no “incorporation” of the subordinate object into the matrix clause as such. [Figure 1] [Figure 2]

In this series of blog articles—taken (more or less) straight from the current working draft of chapter 5.4 of the new grammar for better visibility and as a direct update of an old article (“Flicking Switches: Ayeri and the Austronesian Alignment”, 2012-06-27)—I will finally reconsider the way verbs operate with regards to syntactic alignment.

All articles in this series: Typological Considerations · ‘Trigger Languages’ · Definition of Terms · Some General Observations · Verb agreement · Syntactic Pivot · Quantifier Float · Relativization · Control of Secondary Predicates · Raising · Control · Conclusion


Raising verbs involve the sharing of the subject of an embedded clause with the structural subject or object position of its matrix clause; the complement clause’s subject appears as a gap in English. The raised subject is not semantically an argument of the matrix clause’s verb. The matrix clause’s subject may also be a dummy ‘it’ or ‘there’ in English.

    1. It seemed that Johni knows the answer.
    2. Johni seemed _i to know the answer.
    3. *Johni seemed it.

    1. I expected that Lindai sings the national anthem.
    2. I expected Linda _i to sing the national anthem.
    3. !I expected Linda.

Kroeger (1991: 27–28) states that, as expected, raising is restricted to nominative arguments in Tagalog. Non-nominative actors may be raised into the matrix clause as well, however, but at least for some speakers there needs to be a resumptive pronoun—basically, an overt pronominal ‘trace’ in terms of GG—in the complement clause, as shown in (4). Example (3) shows a case of raising of the nominative argument of the complement clause to the patient of a transitive verb; the nominative argument of the complement clause subsequently is realized as a gap coindexed with the patient of the matrix clause, that is, the raised argument. In English, one would speak of to-object raising, though here the patient of gusto, sila, is in its nominative form, so syntactically, ng Nanay ‘mother’, the actor, is the object in this clause. In (4a), the verb of the complement clause, lutuin ‘cooks’, marks its patient argument as the subject. Yet, the non-subject agent, Charlie, is raised to occupy the patient role in the matrix clause. The position of the non-subject agent in the complement clause is subsequently realized as a resumptive pronoun, niya, coindexed with the raised NP. Example (4b) shows that it would be ungrammatical to have a gap in its stead.

  1. Tagalog (adapted from Kroeger 1991: 26):

  1. Tagalog (adapted from Kroeger 1991: 28):
    1. *gusto ko si Charliei na lutuin _i ang suman

Kroeger (1991) presumably switches to labeling the raised NP as ABS in (4) because it is the patient-subject of gusto ‘want’ (note the actor ko occurs in genitive case); the patient of the embedded clause, suman ‘rice cake’, is also marked as a subject with the verb indicating this by object-voice marking. This is basically consistent with how an ABSERG language would mark subjects. Unfortunately, Kroeger (1991) only gives examples of ‘to-patient’ raising, but not of ‘to-actor’ raising (Carnie 2013: 430). As we will see below, Ayeri has no problem with the former (as to-subject raising), however, it cannot do the latter (as to-object raising), probably for semantic reasons. First of all, let us look at to-subject raising, however.

    1. *

      Surpye {ang Pada.}

      surp-ye ang=Pada

      seem-3SG.F A=Pada

      ‘Pada seems.’

In (5), Pada is both the topic and the subject of koron- ‘know’, but not of surp- ‘seem’, as (5d) shows. However, Pada can be made the subject of the matrix clause, as shown in (5b). Raising results in an intransitive matrix clause, which means that topicalizing the only argument of the verb is blocked, as illustrated by the ungrammaticality of (5c). The verb in (5b) also becomes non-finite, like in English. Unlike in Tagalog, it cannot carry any marking for grammatical relations. Furthermore, it is possible in Ayeri to form a complex predicate like surp- koronyam in (6), literally ‘seems knowing’, with all of the arguments of the embedded clause becoming arguments of the matrix clause, that is, the matrix verb is interpreted as a transitive clause and may carry topic marking for any of its syntactic (rather than semantic) arguments.

If the topic is actually the subject, it should be possible in Ayeri to raise non-actor topics into the matrix clause easily. Of course, this is possible in Tagalog. In (7a), thus, Manuel is the one arrested, so he is the patient of the subordinate clause which acts as the subject of the matrix clause. The fact that Manuel is a patient-subject of the subordinate verb, hulihin ‘be caught’, is reflected in its being marked for objective voice. The English translation is consequently given with the subordinate clause phrased in the passive voice. Similarly, in (7b), the subordinate verb, sinuhulan ‘be bribed’, is marked for directional voice. According to this, ang pangulo ‘the president’ is a non-actor subject of the subordinate verb here as well. It also is in the matrix clause, since the matrix verb, napagbintangan ‘be accused of’, is marked for directional voice.

  1. Tagalog (adapted from Kroeger 1991: 26):

As we have seen above, the marking of the privileged NP on the verb in Ayeri has no effect on grammatical relations; making a transitive verb agree with an NP other than the agent NP was also judged questionable. Thus, we would expect Ayeri to not allow for the same flexibility as Tagalog. The next two sets of example sentences, (8) and (9), thus feature non-actor topics in the complement clause in the (a) examples which we attempt to raise into the subject position of the matrix clause in the (b) examples.

Comparing (8) and (9) with (7a) and (7b), it becomes apparent that Ayeri is very dissimilar to Tagalog with regards to the promotion of a non-actor NP to the subject of the matrix clause in that it is not possible to produce a grammatical result this way. Besides yet more evidence for the disconnect between the marking on the verb and subject assignment and also evidence in favor of an interpretation of the actor NP as the subject, it is possibly the fact that the subordinate verb appears in a non-finite form when raising occurs that prevents some of the flexibility of Tagalog observed above. Even if Ayeri were to work like Tagalog large and by, since finiteness in Ayeri also includes topic marking, it would not be possible for the non-finite verb to mark the assignment of grammatical roles to its complements, overt or covert.

The examples (3) and (4) from Tagalog quoted initially both feature to-object raising: the subject of the complement clause becomes an object of the matrix clause’s verb. This phenomenon is also known as exceptional case marking (ECM) or accusative and infinitive (AcI) and entails that the matrix verb assigns accusative/objective case to the raised subject (Carnie 2013: 439–442). The raised subject is not semantically an object of the matrix verb, however, but an external agent:

    1. Mother wants them to study tonightMother wants them
    2. Mary expects him to tidy the roomMary expects him
    3. John hears people sing in the streetJohn hears people

Ayeri avoids this kind of construction. The reason for this is probably that even though it treats agent and patient as semantic metaroles rather permissively, case marking is nonetheless based on semantic roles rather than purely based on syntactic function. Due to the uniqueness condition, a verb in Ayeri cannot have two agent arguments, yet the raised object is an agent, albeit an external one. It is still salient enough as an agent to preclude assigning it patient case, though.

    1. Galamye {ang Sipra,} {ang sibunja} Ijān sangalas.

      galam-ye ang=Sipra ang=sibund-ya Ø=Ijān sangal-as

      expect-3SG.F A=Sipra AT=tidy-3SG.M TOP=Ijān room-P

      ‘Sipra expects that Ijān tidy up the room.’

    2. *

      {Ang galamye} Sipra {ang/sa Ijān} sibunjam sangalas.

      ang=galam-ye Ø=Sipra ang=/sa=Ijān sibund-yam sangal-as

      AT=expect-3SG.F TOP=Sipra A=/P=Ijān tidy-PTCP room-P

      Intended: ‘Sipra expects Ijān to tidy up the room.’

    3. *

      {Ang galamye} sibunjam Sipra {sa Ijān} sangalas.

      ang=galam-ye sibund-yam Ø=Sipra sa=Ijān sangal-as

      AT=expect-3SG.F tidy-PTCP TOP=Sipra P=Ijān room-P

      Intended: ‘Sipra expects Ijān to tidy up the room.’

The example sentences in (11) show that to-object raising is not possible with verbs of wanting—here using galam- ‘expect’ by way of example. That is, the subject of the complement clause in (11a), Ijān, cannot take the object position of the matrix clause in (11b), nor is it possible to form a complex predicate with the arguments of the subordinate verb, sibund- ‘tidy’, becoming arguments of the matrix clause’s verb, galam- ‘expect’, in the way of (6) in (11c).

Other verbs which allow to-object raising in English include verbs of wanting like need or want, or verbs of perception like see or hear. English also permits this construction for verbs of cognition like believe, consider, know, and think, and for verbs expressing a causative relationship like make or let. Verbs like make or let do not have direct counterparts in Ayeri, as Ayeri uses a morphosyntactic strategy rather than a lexical one to express causative relationships. However, as (12) shows, Ayeri does not allow to-object raising with verbs of perception and verbs of cognition either.

    1. *

      {Ang tangya} Yan keynamas malyyam kirinya.

      ang=tang-ya Ø=Yan keynam-as maly-yam kirin-ya

      A=hear-3SG.M TOP=Yan people-P sing-PTCP street-LOC

      ‘Yan hears people sing in the street.’

    2. *

      Paronyeng {sa Avan} tesayam.

      paron=yeng sa=Avan tesa-yam

      believe=3SG.F.A P=Avan lie-PTCP

      ‘She believes Avan to lie.’

A Question of Alignment IX: Control of Secondary Predicates

In this series of blog articles—taken (more or less) straight from the current working draft of chapter 5.4 of the new grammar for better visibility and as a direct update of an old article (“Flicking Switches: Ayeri and the Austronesian Alignment”, 2012-06-27)—I will finally reconsider the way verbs operate with regards to syntactic alignment.

All articles in this series: Typological Considerations · ‘Trigger Languages’ · Definition of Terms · Some General Observations · Verb agreement · Syntactic Pivot · Quantifier Float · Relativization · Control of Secondary Predicates · Raising · Control · Conclusion


Secondary predicates in Tagalog are interesting insofar as depictive adjectives which occur after the verb always modify the nominative argument:

  1. Tagalog (adapted from Kroeger 1991: 29–30):

Kroeger (1991: 30) explains that (1c) is anomalous, since the subject is indicated as ang isda ‘the fish’, however, lasing ‘drunk’ is not a property usually associated with fish—it would fit better with ‘Maria’. However, this interpretation would be ungrammatical since ‘Maria’ is not the subject of the clause.

Secondary predicates in Ayeri also follow the finite verb, and they refer to the agent. If what was identified as the topic would be the subject like in Tagalog, thus, the reference of the adjective should change in the way shown in (1). However, as we will see below, this is not the case.

In (2a), the topic NP, Migray, happens to be the same NP that is modified by the secondary predicate, gino ‘drunk’: Migray is drunk. However, (2b) generates the same reading even though this time, sangal ‘the room’ is marked as the topic of the clause. A reading in which the room is drunk cannot be forced by morphological means, although it needs to be pointed out that predicative adjectives relating to the object inhabit the same postverbal position. Considering structure alone, the sentence in (2b) is ambiguous, though context certainly favors the reading provided in the translation of (2b), since ‘drunk’ is not typically a property of rooms.

Different than in (2), the adjective in (3), sati ‘cold’, refers to the object of the clause, kangaley ‘milk’, even though kangaley is not the topic of the clause. By structure alone, Niyas could also be the one who is cold, rather than the milk, however, this would be unlikely considering context and extralinguistic experience. Equally unlikely is the possible interpretation of the milk becoming cold by Niyas’ drinking it.

Different than in Tagalog, thus, it is not morphology but the meaning of the verb which determines whether the postverbal predicative adjective refers to the agent or the patient.1 However, since in Ayeri, the predicative adjective following the verb can refer to either the agent or the patient depending on context, this test does not have a very clear outcome. At least we could establish here that alternations in the morphological marking of the privileged NP—tentatively, the topic—has no impact on the relation between adjective and noun. The marking on the verb is thus not used for manipulating grammatical relations in this context, unlike in Tagalog.

  1. Unfortunately, Kroeger (1991) does not provide any examples of object predicatives in Tagalog, and neither does Schachter and Otanes (1972) readily contain information on these.

A Question of Alignment VIII: Relativization

In this series of blog articles—taken (more or less) straight from the current working draft of chapter 5.4 of the new grammar for better visibility and as a direct update of an old article (“Flicking Switches: Ayeri and the Austronesian Alignment”, 2012-06-27)—I will finally reconsider the way verbs operate with regards to syntactic alignment.

All articles in this series: Typological Considerations · ‘Trigger Languages’ · Definition of Terms · Some General Observations · Verb agreement · Syntactic Pivot · Quantifier Float · Relativization · Control of Secondary Predicates · Raising · Control · Conclusion


Kroeger (1991) observes that in Tagalog, only nominative arguments may be relativized. He refers to Keenan and Comrie (1977)’s accessibility hierarchy of NPs, according to which, he reports, “if only a single argument of any clause can be relativized, that argument must be the subject” (Kroeger 1991: 24). That is, the argument in the main clause which is modified by a relative clause must be the nominative argument, and additionally, Tagalog requires that there must not appear an overt nominative argument in the relative clause itself. The verb in the relative clause carries inflection for the role of the relativized argument in the relative clause, which is itself gapped. Thus, (1a) is grammatical, while (1b) is not.

  1. Tagalog (Kroeger 1991:24, from Foley and Van Valin 1984: 141–142):
    1. bata=ng b-in-igy-an ng=lalake ng=isda

      child=LNK PFV-give-DV GEN=man GEN=fish

      ‘the child which was given fish by the man’

    2. *isda=ng nag-bigay ang=lalake sa=bata

      fish=LNK AV-PFV-give NOM=man DAT=child

Ayeri, however, has no such restrictions. Non-topic NPs may be relativized, and relative clauses not uncommonly contain their own agent NP. The relativized NP may even be referred to in the relative clause by a resumptive pronoun or pronominal clitic, since verbs must not go uninflected. Since all NPs are accessible for relativization, it is not a suitable criterion for testing the subjecthood of what we so far identified as the topic NP.

  1. {Ang ilya} inunley ganyam inunaya si gumasayāng edaya.

    ang=il-ya inun-ley gan-yam inunaya-Ø si gum-asa=yāng edaya

    AT=give-3SG.M fish-P.INAN child-DAT fisherman-TOP REL work-HAB=3SG.M.A here

    ‘The fisherman who used to work here, he gave fish to the child.’

In (2), inunaya ‘the fisherman’, is both the topic of the clause and modified by a relative clause. He is referenced anaphorically by the 3SG.M.A suffix -yāng on the verb in the relative clause, since he is the actor in both. However, as the next examples show, these circumstances are not requirements for grammatical statements.

    1. {Ang ilya} inunaya inunley ganyam si {ang pyabasaye} benanya-hen.

      ang=il-ya inunaya-Ø inun-ley gan-yam si ang=pyab-asa=ye.Ø benan-ya=hen

      AT=give-3SG.M fisherman-TOP fish-P.INAN child-DAT REL AT=pass.by-HAB=3SG.F.TOP morning-LOC=every

      ‘The fisherman, he gave fish to the child which passes by every morning.’

    2. {Ang ilya} inunaya ganyam inunley si petigayāng hiro.

      ang=il-ya inunaya-Ø gan-yam inun-ley si petiga=yāng hiro

      AT=give-3SG.M fisherman-TOP child-DAT fish-P.INAN REL catch=3SG.M.A freshly

      ‘The fisherman, he gave fish which he caught freshly to the child.’

In (3a), the recipient NP ganyam ‘to the child’ is not the topic of the clause, but it is modified by a relative clause anyway. The relativized NP is again represented within the relative clause by means of verb morphology. The topic marker on the verb identifies the person suffix on the verb as the clause’s topic. In (3b), it is likewise not the topic NP which is relativized, but the patient NP inunley ‘fish’. This NP, however, is not represented in the relative clause because the verb does not inflect for the role of the patient, which the relativized NP carries in the relative clause as well. There is no morphology to alter the voice of the verb in such a way that the matrix clause’s patient NP becomes the subject of the relative clause.

  1. {Ang ilya} inunaya ganyam inunley si hiro nay lepan.

    ang=il-ya inunaya-Ø gan-yam inun-ley si hiro nay lepan

    AT=give-3SG.M fisherman-TOP child-DAT fish-P.INAN REL fresh and tasty

    ‘The fisherman, he gave fish which is fresh and tasty to the child.’

Relative clauses in Ayeri may even just consist of a predicative adjective, as (4) illustrates. In these cases, there is no case-marked noun or topic contained in the relative clause.

  • Foley, William A. and Robert D. Van Valin, Jr. Functional Syntax and Universal Grammar. Cambridge: Cambridge UP, 1984. Print.
  • Keenan, Edward L. and Bernard Comrie. “Noun phrase accessibility and uni­versal grammar.” Linguistic Inquiry 8 (1977): 63–99. Print.
  • Kroeger, Paul R. Phrase Structure and Grammatical Relations in Tagalog. Diss. Stanford University, 1991. Web. 17 Dec. 2016. ‹http://www.gial.edu/wp-content/uploads/paul_kroeger/PK-thesis-revised-all-chapters-readonly.pdf›.

A Question of Alignment VII: Quantifier Float

In this series of blog articles—taken (more or less) straight from the current working draft of chapter 5.4 of the new grammar for better visibility and as a direct update of an old article (“Flicking Switches: Ayeri and the Austronesian Alignment”, 2012-06-27)—I will finally reconsider the way verbs operate with regards to syntactic alignment.

All articles in this series: Typological Considerations · ‘Trigger Languages’ · Definition of Terms · Some General Observations · Verb agreement · Syntactic Pivot · Quantifier Float · Relativization · Control of Secondary Predicates · Raising · Control · Conclusion


Another property usually associated with subjects is the ability of quantifiers referring to the subject NP to ‘float’ into the VP. This is possible also in English, consider, for instance:

    1. All the children are writing letters.
    2. The children are all writing letters.

Both of these sentences are equal in meaning: for all children in the set, every child is writing an unspecified amount of letters. It is not the case in (1b) that for an unspecified amount of children, together they write the total amount of letters. All refers to the children in both cases, even though all is not placed in the subject NP, the children, in (1b). Kroeger (1991) mentions an example from Schachter and Otanes (1972) concerning lahat ‘all’, which is also able to float into a position right after the sentence-initial verb from the NP it normally modifies and which it would normally occur in:

  1. Tagalog (adapted from Kroeger 1991: 22, from Schachter and Otanes 1972: 501):
    1. sumusulat lahat ang=mga=bata ng=mga=liham

      AV.IPFV-write all NOM=PL=child GEN=PL=letter

      ‘All the children are writing letters.’
      Not: *‘The children are writing all the letters.’

    2. sinusulat lahat ng=mga=bata ang=mga=liham

      IPFV-write-OV all GEN=PL=child NOM=PL=letter

      ‘The/some children write all the letters.’
      Not: *‘All the children are writing letters.’

In (2a), lahat ‘all’ refers to the children, which constitute the subject NP according to voice and case marking, while we get the opposite case in (2b), where it refers to the letters, which are marked as the subject this time. Of course, it is equally possible in English to say The letters are all written by the children, where the letters is the subject that the floated all refers to.

A lot of clitic quantifiers in Ayeri have a double meaning as intensifiers. For instance, -ikan can refer to both quantities and qualities, meaning ‘much, many’ or ‘very’ depending on context. Thus, many of the suffixed quantifiers, if appended to the VP, are understood to modify the verb as an intensifier and are thus unsuitable for floating. The only exception is -aril ‘some’, which only pertains to NPs as a quantifier. However, since the floating of suffixed quantifiers would produce readings which are ambiguous at best, floating of -aril is avoided as well. Example (3) shows an attempt to float -hen ‘all’ into the IP, resulting in a meaning different from the sentence with the unfloated particle for the reasons just stated above.

    1. {Ang tahanyan} ganye-hen tamanyeley.

      ang=tahan-yan gan-ye-Ø=hen taman-ye-ley

      AT=write-3PL.M child-PL-TOP=all letter-PL-P.INAN

      ‘The children, all of them are writing letters.’

    2. {Ang tahanyan-hen} ganye tamanyeley.

      ang=tahan-yan=hen gan-ye-Ø taman-ye-ley

      AT=write-3PL.M=completely child-PL-TOP letter-PL-P.INAN

      ‘The children, they are completely writing letters.’
      Intended: ‘The children, they are all writing letters.’

Besides suffixed quantifiers, Ayeri also has free quantifiers like sano ‘both’ or diring ‘several’, however. These free morphemes only have a quantifying reading, not an intensifying one, and are thus suitable for floating, since they do not produce ambiguities with regards to what is being modified.

    1. {Ang apayan} yan sano layjya.

      ang=apa-yan yan-Ø sano lay-ye-ya

      AT=laugh-3PL.M boy-TOP both girl-PL-LOC

      ‘The boys, both of them are laughing at the girls.’

    2. {Ang apayan} sano yan layjya.

      ang=apa-yan sano yan-Ø lay-ye-ya

      AT=laugh-3PL.M both boy-TOP girl-PL-LOC

      ‘The boys, they are both laughing at the girls.’

Since, as described above, topicalization has no impact on what constitutes the subject, meaning does not significantly change when the topic of a sentence like (4b) is switched to the patient in example (5a). Unlike in Tagalog in (2b) above, yanang ‘boy(s)’ as the agent NP remains as the subject, and the floated sano still refers to this NP rather than the locative NP, layye ‘(at) the girls’. This fact is also reflected in the lack of plural marking on yanang, since sano indicates the NP’s plurality. We would expect the forms yanjang and lay if sano were to refer to ‘the girls’ rather than ‘the boys’, as in (5b).

    1. {Ya apayan} sano yanang layye.

      ya=apa-yan sano yan-ang lay-ye-Ø

      LOCT=laugh-3PL.M both boy-A girl-PL-TOP

      ‘The girls, the boys are both laughing at them.’

    2. {Ya apayan} yanjang lay sano.

      ya=apa-yan yan-ye-ang lay-Ø sano

      LOCT=laugh-3PL.M boy-PL-A girl-TOP both

      ‘The girls, the boys are laughing at both of them.’

As we have seen above, the modification of subject pronouns with clitic quantifiers is avoided due to many of them serving a double role as intensifiers with related meanings which could be readily understood as referring to the verb instead of the pronoun. With free quantifiers, this problem does not arise, however, so that there is no problem in placing them right after the finite verb. Ambiguity may be in the phrase structure of the clause here, but not at a functional level, as it is clear that the quantifier modifies the subject pronoun from semantic coherence.

  1. {Ang girenjan} sano bahalanya.

    ang=girend=yan.Ø sano bahalan-ya

    AT=arrive=3PL.M.A both finish-LOC

    ‘They arrived both at the finish.’

It is possible in Ayeri for pronouns to be modified by enclitic intensifiers indirectly by using sitang ‘self’ as an indeclinable dummy pronoun to carry the clitic so as to avoid ambiguity created by floating the clitic right after the finite verb. This is also possible for the purpose of quantification of pronouns with clitic quantifiers.

    1. {Ang girenjan} panca sitang-hen bahalanya.

      ang=girend=yan.Ø panca sitang=hen bahalan-ya

      AT=arrive=3PL.M.A finally self=all finish-LOC

      ‘All of them finally arrived at the finish.’

    2. {Ya girendtang} panca sitang-hen bahalan.

      ya=girend=tang panca sitang=hen bahalan-Ø

      LOCT=arrive=3PL.M.A finally self=all finish-TOP

      ‘The finish, all of them finally arrived there.’

Since sitang is indeclinable, it is the pronominal clitic which carries inflection for case, as (7b) shows. An analysis of sitang-hen as ‘self.TOP=all’ is therefore not possible. Moreover, -tang sitang-hen does not constitute a clitic cluster, because it is possible to place word material between the verb and sitang-hen, as (7) shows. The question remains, however, whether sitang-hen is stranded at the end of the verb phrase or located in the position a full agent NP would normally occupy.