Control and null subjects are governed by morphosyntax in Finnish

I have done some work on Finnish control and null subjects, and now there is a first draft paper on that research that I will present in the forthcoming Uralic Syntax workshop in Budapest, Hungary (here is also a highly simplified 10-page presentation). The basic argument of that work is that control and null subjects are governed by morphosyntax in Finnish. Here I will explain the background, and at the end of this post I speculate a bit about what might be going on.

Few years ago, when we were working as a research group in Finland, we encountered data about the Finnish possessive suffix that led us to believe that the third person possessive suffix was accompanied by a null subject, little-pro. This research later become two papers (one here), but I’ve written about it here at Finnish Syntax Blog as well (see this, for example). Yet, all this research effort was hampered by the fact that very little was known about Finnish control in general. It is possible to work in such circumstances by making educated guesses, which is exactly what people have done thus far, but over time the process becomes frustrated.

I therefore spent one holiday season in Italy with pen and paper and tried to figure out how Finnish control works. The purpose was to provide an anchor that could be used to make further research a bit easier.

Well, what is control? Languages have ways of omitting linguistic material from speech, yet the semantic interpretation proceeds as if something were there.

(1) Pekka uskoi että __ voitta-a kilpailun helposti.
Pekka believed that (he) win-3sg competition easily

This sentence must mean that Pekka is both the believer and the prospective winner. It is not possible interpreted this clause as saying that Pekka is the believer and Merja is the winner. Control is about how we fix the semantic interpretation for constituents that are not uttered (marked by “__” in (1)). Something in our mind/brain tells us that the empty place must refer to the same thing as the matrix subject. There are many similar constructions in Finnish (2).

(2)
a. Pekka uskoi {__ voittavansa kilpailun}
b. Pekka voitti kilpailun {__ harjoiteltuaan koko kesän}
c. Pekka halusi {__ voittaa kilpailun}
d. Pekka voitti kilpailun {__ leikiten}
e. Pekka punastui {__ laulettaessa}
f. Pekka hävitti {__ voittamansa pokaalin}
g. {__ nukkuessa} virkistyy
h. Pekka rikkoi {__ pyöränsä}
i. Pekka näki Merjan {__ juoksemassa}

and so on and so forth. You can get a good grasp of the magnitude of this problem by checking out the table of contents of my paper, especially Section 4 that lists all the constructions I had to examine (well, here is a condensed list: finite clause, noun phrase, adposition phrase, TUA-adverbial, ESSA-adverbial, MA-participle, VA-infinitival, A-infinitivial, MA-infinitival, VA-participle, generics, KSE-adverbial).

There are three problems that one should be able to answer for all control constructions in one language (or possibly in all languages):

A. Licensing. When are controlled null subjects possible?
B. Control and interpretation. Where do they get their semantic interpretation?
C. Types of null subjects. What types of null subjects are there?

How you want to proceed to answer A-C depends on some of your background assumptions about language.

Let’s follow most of the professional linguists in Finland and assume that Finnish is a pidgin language, hence it has words that occur in a linear sequence such that they also match with meaning and thought. Some say that the relation between linear sequence of words and meaning is “isomorphic”, implying that Finnish is a kind of super-pidgin that has no ambiguity. Regardless, the idea is that Finnish has no syntax; there is meaning/thought and a list of words. This notion is supported among other things by the fact that the word order in Finnish is quite free. (Since Hungarian has syntax, a plausible scenario is that some 5,000 years ago the proto-Uralic language was a pidgin language, and then the Hungarian population made contact with real languages, while the Finnish group had no such contacts, and Finnish remains a retarded pidgin to this day.) At any rate, this is the most authoritative theory of Finnish we have to date. Accordingly, you will harvest a theory of meaning/thought to explain (1-2). Something in your thought compels you to omit words. For example, you might omit ‘Pekka’ in (1) because it occurs for the second time in your thought, and so you will also omit ‘Pekka’ (3).

(3)
a. *Pekka uskoi että Merja rakastaa __ (__ = Pekkaa).
b. *{Pekan(1) veli} uskoi että __(1) voittaa helposti kilpailun (__ = Pekka voittaa)
c. *{Pekan(1) veli} rikkoi {__ pyörän}. (__ = Pekan pyörän)
d. *Pekka rakastaa __. (__ = Pekkaa/itseään)
e. *Merja kertoi Pekalle __ olevansa mukavaa seuraa (__ = Pekka on).
f. *Merja kertoi Pekalle että __ soittaa huomenna (__ = Pekka soittaa).
g. *Pekka uskoi {__ voittavan kilpailun}

It does not work like that, actually, but this is just one possibility. The pidgin theorist is free to consider alternative hypotheses. For example, we know that deeply drunken people omit words that occur in their thoughts; sometimes they repeat some one thought endlessly, and so on–there are many possible hypotheses to disrupt the connection between thoughts and words!

But if you think, like I do, that Finnish is a real language, then it has syntax in addition to words and meanings. In fact, I report in my paper that the phenomenon is a mixture of semantic and syntactic factors.

Let’s return to A-C. There are two ways to proceed with this type of research. One is, adopt a tight syntactic framework or theory, for example the pidgin hypothesis above, and examine one relevant hypothesis within that theory, as I did in (3). Or one could adopt some minimalist theory. Another possibility is to explore the data without testing any hypothesis and find regularities and patterns. The former is natural scientific method, the latter is the traditional Aristotelian scheme. Aristotle popularized the idea that one should collect data and then classify/derive it from a theory. The difference is that in the natural scientific systems, the facts are allowed to change or test the theory; in the Aristotelian scheme, the theory never changes (indeed it remained unchanged for some thousand years).

I used the latter method here, because so little was known. Both are useful methods. Tycho Brahe used Aristotle’s framework in collecting astronomical data but he never changed his theory; Kepler rotated that data over and over and examined it in an amazingly open minded way, and thus found out that Plato was wrong about the planetary orbits; and Newton finally worked it all out in a natural scientific way. Kepler could not have formulated his laws without Tycho’s careful observations, and Newton could not have worked out a theory of gravitation without Kepler’s laws.

These two approaches have different constraints and outcomes. If you work within the Aristotelian scheme, I expect that you consider all and every construction in your language, or a sample of languages. If something doesn’t fit, it should still be listed and mentioned. The idea is that you provide a good overview of the facts, much like Tycho Brahe did with astronomical observations, so that other people can then rotate your data for other purposes. You will also have to provide negative facts; what type of sentences are not possible (e.g. 3a-g).

If you work within the natural scientific system, on the other hand, you only have to consider all the data that is relevant to your hypothesis, but then also nothing else. One of the reasons why there is so little studies on Finnish control is because most prior studies have been done within the natural scientific framework that requires one to be very selective, and because almost everybody inside Finland follows the pidgin theory that has produced nothing to date.

Those who have been in this field long enough will recognize that what I propose for Finnish is very close what was part of the GB-theory during the 1980s. There is a system in our brain that computes syntax (this is a radical and extreme hypothesis, but correct I think). That system is connected to three extrasyntactic systems: the lexicon, spellout (speaking, gesturing) and meaning/thought (thinking, decision making). The point of contact between syntax and speaking is PF; and the point of contact between syntax and thought is LF. Null elements emerge when the syntax plays with constituents (bare phi-sets in my theory) that cannot be interpreted by the extrasyntactic systems, and that there is some process related to morphosyntax (Case assignment, phi-agreement) that makes constituents visible at/for the PF- and LF-interfaces.

This is the old “Visiblity Hypothesis” that was proposed 30 years ago. But I think that the Finnish strongly supports this notion; from hereon you have to look at the facts and decide what to think.

Lack of PF-interpretation leads into silence; lack of LF-interpretation leads into a frantic search for meaning, hence to control. It is clear why speech/gesturing can do without content, while meaning cannot; there are no holes in our thought. Or is there? Frege thought there is.