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

The notorious term ‘trigger language’ comes up in discussions on Conlang-L as early as 1995, where it may well have originated as an established term in the fictional-language community for what will be described below in brief. That is, I have not been able to find any earlier mentions of the term ‘trigger’ as referring to an alignment system in the archives; other mainstays of the fictional-language community, such as the ZBB, were established only about a decade later. In a message dated December 16, 1995, John Cowan writes that he wants “to propose a reform of Radilu, to make it use the Tagalog concept of a trigger” (Cowan 1995). By his definition, this entails that

each clause contains one noun phrase which is not marked for case, but rather has a distinct marking called the “trigger marker”. […] The verb carries a marking (which of course looks nothing like the noun case markers) that tells the true case of the trigger. […] This involves changing the name of “nominative” and “accusative” to “actor” and “patent” [sic], since there is no longer a “subject” or “object” as such. Of course, word order is free (Cowan 1995)

He also notes that “Usually the trigger is definite (Tagalog doesn’t have articles)” (Cowan 1995). Essentially, it seems that the motivation for Cowan’s system is that the ‘trigger’ indicates that a certain NP is definite. As we will see further on, this is similar to how Tagalog marks one of its relations on the verb, with that relation being definite. Things are more complicated in reality, though. Especially the claim that Tagalog lacks subjects and objects is problematic. However, the term ‘trigger’ seems to have currency in that, for instance, Schachter (2015) chooses it explicitly to refer to the “non-case-marked argument” (1659). In a parenthetical remark he adds that some

previous treatments have referred to the argument in question as the topic and some as the subject. However, as will become clear below, each of these labels appears to carry some inappropriate connotation, making a netural term like Trigger seem preferable […] There also seems to be good reason to reject the term focus. (Schachter 2015: 1659)

It may be noted that term ‘focus’ is used in Schachter and Otanes (1972), the main reference grammar of Tagalog. What is interesting in comparing Schachter (2015)’s and Kroeger (1991)’s respective analyses of Tagalog’s syntactic alignment is that both make the same observation in spite of coming to opposite conclusions: Tagalog is ambiguous as to whether the subject notion is vested in the NP whose role is marked on the verb or the actor, since certain syntactic constructions typically associated with subjects apply to either or both. While this ambiguity leads Schachter (1976, 2015) to ultimately conclude that Tagalog lacks a single unified relation which can be analyzed as a syntactic subject,[1. Cowan (1995)’s sketch may be based on Schachter (1976). Curiously, Schachter (2015) does not acknowledge Kroeger (1991) at all, nor does he refer to any other research more recent than 1985. The reason may be that Schachter retired in the early 1990s, as the UCLA linguistics department’s Department history suggests. It may also be noted that Schachter (2015) was apparently published posthumously.] Kroeger (1991) reaches the opposite conclusion by performing further tests and taking a functionalist rather than purely structuralist perspective. Thus, he concludes:

• “Tagalog has a well-defined grammatical subject” (225). What Schachter (1976) lists as evidence against are special cases which can be explained by the high semantic and pragmatic prominence of actors more generally (Kroeger 1991: 225). Tagalog basically applies the the notion of a logical subject distinct from the syntactic subject to some constructions, though the syntactic subject is more important overall (36).
• “grammatical relations are defined independently of phrase structure” (225);[1. This point especially may be a problem for generative theories of syntax.]
• “patients can become subjects even when the agent is expressed as a direct (non-oblique) argument of the verb” (225).
• “Subject selection in Tagalog does not work by demotion or suppression of thematically more prominent arguments. Rather, all arguments seem to be equally eligible for mapping onto the subject relation” (226).

Kroeger (1991) also provides evidence based on statistics and examples that the marked-for relation, which he classifies as being in the nominative case according to his hypothesis that it is the syntactic subject, is neither especially salient in terms of pragmatic topichood, nor does it show signs of carrying pragmatic focus specifically. He finds that rather, nominative marking works independent of discourse functions (56 ff.). All things considered, the term ‘trigger language’ is probably ill-fitting, not just for Ayeri.

The tests for typical properties associated with grammatical subjects which Kroeger (1991) performs partially extend those presented in Schachter (1976). Moreover, his conclusions build on a more modern, functionally oriented approach than Schachter’s. For this reason, I will follow Kroeger rather than Schachter. Either way, in order to compare what is going on in Ayeri, we will have to test verb agreement, syntactic pivot, relativization, control of secondary predicates, raising, and control.[1. The tests which Kroeger (1991) dismisses as irrelevant to determining subjecthood in Tagalog have been omitted here if they were also not profitable to answering this question for Ayeri. The same goes for a number of tests which are specific to the grammar of Tagalog and thus have no application in Ayeri.] First of all, it will be helpful, however, to define some terms which will be used in the discussion further on.

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

Verbs govern the relations of the various phrase types to each other and they are thus central to the formation of clauses. Just from looking at the numerous examples given both on this website and in the grammar, it should be clear that Ayeri’s preferred word order is verb-first, which opens up a few typological questions—first and foremost, whether Ayeri actually has a verb phrase, or in terms of generative grammar: whether it is configurational in this regard. Ayeri definitely has a constituent structure as far as NPs, APs, PPs, etc. are concerned. However, due to VSO word order, it is not obvious whether verb and object actually form a VP constituent together, since V and O are not adjacent to each other. Since Ayeri marks topics in terms of morphology, it will also be necessary to discuss how this mechanism works and how it relates to the notion of the subject.

A discussion of subject, topic, and configurationality is interesting also in that Ayeri’s syntactic alignment was originally inspired by the Austronesian or Philippine alignment system, though then under the term ‘trigger language’ which is itself not unproblematic. Tagalog, an Austronesian language of the Malayo-Polynesian branch, spoken mainly in the Philippines (Hammarström et al. 2017: Tagalog; Schachter and Otanes 1972), usually serves as the academic poster child in descriptions of Austronesian alignment. Ayeri departs from Tagalog’s system in a number of ways, though, and probably towards the more conventional. Austronesian alignment is not necessarily the best model to liken Ayeri’s syntax to. It will nonetheless be informative to compare both systems based on the work of Kroeger (1991, 1993), who provides an analysis of Tagalog’s syntactic alignment roughly in terms of the LFG framework and describes some heuristics which may be helpful in establishing what is actually going on in Ayeri. As mentioned in a previous blog article (“Happy 10th Anniversary, Ayeri”, 2013-12-01), I started Ayeri in late 2003—then still in high school and not knowing much about linguistics. Of course, I had to go and pick as a model the one alignment system which has long been “a notorious problem for both descriptive grammarians and theoretical syntacticians” to the point where it “sometimes seems as if Austronesian specialists can talk (and write) of nothing else” (Kroeger 2007: 41).

As mentioned above, Ayeri’s unmarked word order gives the verb first, and then, in decreasing order of bondedness to the verb, the phrases which make up the verb’s arguments: subject (agent), direct object (patient), indirect object (dative), followed by adverbials in the genitive, locative, instrumental, and causative case. Ayeri’s basic word order is thus VSO, a trait it has in common with about 7 % of the world’s natural languages according to Dryer (2013). Regarding word order typology, we can declare the generalization in (1), which is consistent also with word order in other areas of the language, where the head precedes the modifier. The head is here represented by the verb, the modifier by the object—like English, Ayeri is a VO language, thus. In addition to this, however, Ayeri regularly puts the verb as the head of the clause itself first.

It is commonly assumed that languages have a subject which occupies a certain position in the constituent structure which commands a constituent jointly formed by the verb and its dependents—the predicate. An SVO sentence in English thus very generally looks like in (2) (compare the examples in Bresnan et al. 2016: 101–111).

However, Ayeri is a VSO language, so the question arises how the basic constituent structure should be diagrammed in tree form, since V and O are not adjacent. As an initial hypothesis one might assume that they cannot form a unit together, since S somehow stands in between the constituents it is supposed to command. A very first stab at diagramming would probably be to come up with a flat, non-configurational structure, all but lacking a VP, as shown in (3).

Such a structure, though, does not do Ayeri justice in that, for instance, right-node-raising of a subject and object NP together is possible, so there is evidence that they form a constituent subordinate to the verb. NP–XP constructions where XP is not a maximal projection of a verb also exist in isolation, so NP and XP are probably contained in a small-clause constituent S separate from the verb. The verb in the initial position furthermore shows inflection, so one might rather construe it as an I⁰, projecting an IP, which frees up VP for other purposes while we can use IP to govern both Iʹ and S. In fact, such a structure is basically the conclusion Chung and McCloskey (1987) come to for Irish, which is also a VSO language (4a). Bresnan et al. (2016) give the chart in (4b) for Welsh, equally a VSO language (also compare Dalrymple 2001: 66, sourcing Sadler 1997). Kroeger (1991) suggests the two structures depicted in (4c) for Tagalog, based on the suggested constituent structure for Celtic languages.

What all of these c-structures have in common is that the inflected verb appears in I⁰, which is a sister of S. S, in turn, is a small clause containing the arguments of the verb. In the cases of Irish and Welsh, however, there is a VP sister of the subject NP which itself does not have a head, but contains the object NP as a complement. In the case of Tagalog, S is non-configurational, that is, while XP may contain a non-finite verb, the subject and object NPs are on equal footing.

Bresnan et al. (2016: 129–138) inform that the phenomenon of the verb ending up in a different head position (V⁰ apparently moves to I⁰) in (4b) is commonly known as ‘head movement’, except that LFG is built specifically without any movement. Since LFG is based on the assumption that all nodes in a syntactic structure are base-generated, that is, that there are no transformational rules generating the surface structure from a deeper layer of representation underneath it, there cannot be a trace of V left behind in VP. LFG avoids empty categories, as there is no information contained in an empty node. The functional information provided by the verb is not lost, however, it is merely now provided by the verb in I⁰. Essentially, the Welsh example does not violate endocentricity, since the finite verb in I⁰ still forms the verbal head in the functional structure representation of the clause. With regards to constituent structure, V⁰, if present, c-commands its NP sister; both V⁰ and NP are dominated by VP:

The AVM in (4b) shows that the contents normally found in V⁰ are provided by the head of its equivalent functional category, I⁰. I⁰ and VP are said to map into the same f-structure (Bresnan et al. 2016: 136). Endocentricity still holds in that IP dominates all nodes below it, thus also I⁰ and the object NP. In addition, I⁰ c-commands its sister node and all of its children, hence also the object NP. As Bresnan et al. (2016) put it: “X is an extended head of Y if X is the Xʹ categorial head of Y […], or if Y lacks a categorial head but X is the closest element higher up in the tree that functions like the f-structure head of Y” (136). For our example, replace X with I⁰ and Y with VP in the second half of the quote: I⁰ is the closest element higher up in the tree that functions like the f-structure head of VP, which itself lacks a categorial head.

The analysis of the sentence structure of Celtic languages shows that VSO languages do not automatically need to be considered ‘non-configurational’ and lacking a VP if the notion of extended heads is accepted. In any case, tests need to be performed to see whether one of the analyses presented in (4) holds true for Ayeri as well. However, this will not be in the scope of this series of blog articles.