Surveys on English verbs

Jim Baker, University of Cambridge

Between May and July 2015 I ran a few different online surveys asking people to rate a lot of English sentences, some quite odd-sounding. This was done with a specific purpose in mind, and it seems only fair to try and explain that purpose and some of what I discovered to people who took part, particularly as I received an impressive number of respondents and several people seemed keen to discuss matters further. May I also take this opportunity to thank everybody who took part, and particularly those people who (without any encouragement from me) shared the surveys with their family, friends and colleagues, allowing me to collect even more useful data.

The point of the surveys was to consider different behaviours of different verbs. Specifically, I was interested in different intransitive verbs. Let me begin by explaining what I mean by that.

Transitive and intransitive

Most verbs are either transitive or intransitive. Transitive verbs have two "arguments", nouns or pronouns which play a role in the event or state the verb describes. For example:

(1) Lucy reads the book.
1 2

(2) Lucy loves you.
1 2

(3) He sees the cat.
1 2

Each of these verbs has two arguments, and so is transitive. Intransitive verbs, on the other hand, have only one argument:

(4) Lucy arrived.
1

(5) Lucy is working.
1

Classes of intransitive

My study is largely based on a paper by Prof. Antonella Sorace with the exciting title Gradients in Auxiliary Selection with Intransitive Verbs. In this paper, Sorace divides intransitive (one-argument) verbs up into various classes depending on their meaning. She suggests that these classes form an ordered hierarchy. The classes in the hierarchy, in order, are given below. Not all of the names for the classes are very transparent, so I've given some examples of verbs in each class as well. However, you needn't worry about understanding the exact details.

1. Change of location come, arrive, leave, fall
2. Change of state become, decay, die, happen, grow
3. Continuation of a pre-existing state stay, remain, last, survive, persist
4. Existence of state be, belong, sit, seem
5. Uncontrolled process tremble, skid, cough, rumble
6. Controlled process (motional) swim, run, walk
7. Controlled process (non-motional) work, play, talk

Across the world's languages, there are certain constructions that only occur with a subset of intransitive verbs. For example, in English, most people agree that you can say Lucy outplayed Chris but not Lucy outarrived Chris. The construction that prefixes out- to a start of a verb is allowed with a group of verbs that includes play but not arrive. We call these sorts of divisions between different verbs splits.

The idea behind Sorace's hierarchy given above is that it governs where such splits can and cannot occur. If intransitives are divided into two groups with respect to a certain construction (e.g. prefix out-), then (Sorace suggests) one group will consist of classes toward the top of the hierarchy and one of classes toward the bottom. The exact "cut-off point" may vary. For one construction it might be classes 1 and 2 that allow it, whereas all the other classes don't. Another might be allowed with classes 5–7 and not with classes 1–4. But in any case, it is predicted, only groups of neighbouring classes on the hierarchy behave in the same way. We don't expect to find a construction with classes 1 and 7 but not 2–6.

Sorace also suggests that there may be intermediate classes, towards the middle of the hierarchy, which show mixed behaviour: some verbs, or some speakers, allowing a given construction but not others.

However, Sorace only explores her ideas systematically with regard to a particular construction in some continental Western European languages. My surveys aimed to consider various splits affecting intransitive verbs in English in light of Sorace's predictions.

Results

What did I find? Well, in general, Sorace's predictions seem to hold up for English. In the surveys, people were asked to rate various words, phrases and sentences as "OK", "Not OK" or "Not sure". I assigned scores to these as follows: This enabled me to get an average score for each word, phrase or sentence. Each of these instantiated a particular verb in a particular construction that shows a split amongst different intransitive verbs. I was then able to get an average score for each of Sorace's classes for each construction. The smaller the score, the fewer speakers accepted the construction with that class.

Here is what the results looked like for prefix out- (outarrive, outplay etc.):

As you can see, there's a fairly good correspondence with the hierarchy, although not a perfect one. The classes towards the right hand side of the graph tend to receive lower scores than those toward the left hand side: forms like outarrive and outdecay (from classes 1 and 2) are less accepted than ones like outswim and outplay (from classes 6 and 7).

Here's another example, the [verb] one's way into construction (sentences like Lucy worked her way into the upper echelons of university administration):

Again, there's quite a good correspondence. But there are complications, too, for example many class 2 verbs allow this construction quite freely, e.g. quite a lot of (though not all) speakers accept ✓?The vine grew its way into the house. This isn't what we expect, and something it's down to me to try and find an explanation for.

Sometimes things go the other way. For example, what we call "prenominal past participles" (a good example of alliteration) are most likely to be accepted in classes 1 and 2:

Prenominal past participles are found in phrases like the frozen ice (which means "the ice which has frozen") and the recently arrived recuits (which means "the recruits who have recently arrived")—although for some reason we can't just say the arrived recruits. We also can't generally use them with verbs from classes 3–7, e.g. the talked man [= "the man who has talked"].

Subjects and objects

Sorace's work and mine form part of a much larger body of work on something known as the Unaccusative Hypothesis. The idea behind this hypothesis is that some subjects are actually really objects at some level. Consider the following (transitive) sentence:

(6) Lucy froze the ice solid.
SUBJ OBJ

Lucy is the subject of this sentence (as most of my sentences!) and the ice is the object. The adjective solid must describe the object, it can't describe the subject (Lucy may or may not actually be solid, but that's beside the point). Consider also Lucy hammered the metal flat, where flat describes final state of the object ("the metal"), not the subject ("Lucy", again), or Lucy burned the toast black, where black describes "the toast".

But what about the following?

(6) The ice froze solid.
SUBJ

Here the ice is the subject, but we've seen that adjectives can only describe objects in this type of sentence. The Unaccusative Hypothesis tries to explain this by suggesting that the ice is actually an object at some "deeper" level, but surfaces as the "subject" for other reasons.

There is also other evidence that some intransitive subjects are really objects. Recall prenominal past participles. When these are used with transitives, they describe the objects of the equivalent sentence. So, we can take a sentence like Hannibal destroyed the city and produce the destroyed city [= "the city which has been destroyed"], but not the destroyed Hannibal [= "Hannibal who has destroyed something"]. But with some intransitive we can create prenominal past participles which describe the equivalent of the subject, like The leaves have fallen => the fallen leaves. This suggests that these subjects, too, are really objects.

Or compare Lucy froze the ice and The ice froze. Exactly the same thing happens to the ice in both cases (it freezes). The fact that it's a object in the one case suggests further that it might also be an object in the second.

Not all intransitive verbs have this kind of "deep" object though. The subject of a verb like play or talk only ever behaves like a subject, and never like an object.

How does this relate to the survey?

The idea behind Sorace's hierarchy is basically that verbs belonging to the smaller-numbered classes are more likely to have "subjects that are really objects" than verbs belonging to the larger-numbered classes are. The various different types of construction I tested are precisely those which people have suggested tell you whether a subject is really an object or not. For example, prenominal past participles are supposed only to be found with object-verbs, and the out- prefix and one's way into only with true-subject-verbs.

Most of the time, as we saw with some of the examples above, things line up quite well with the hierarchy. This suggests Sorace's prediction—verbs with objects coming closer to one end, those with "true" subjects coming closer to the other—holds up pretty well.

But there are still a few problems. Some verbs don't always behave exactly as predicted. Furthermore, the various diagnostic constructions which I tested in the survey don't tend to identify exactly the same split. This suggests a finer-grained analysis is needed, but that's pretty complex so I won't go into details.

Finally, some constructions don't seem to line up with the hierarchy at all. For example, some people have suggested that sentences of the type There arrived a man, There came a guest etc. are only possible with verbs that have underlying objects. Actually, though, there isn't any correlation with the hierarchy in this direction (and indeed very little correlation with the hierarchy at all):

Respondents to the survey were as likely to accept sentences with verbs from class 7 like ?There worked a man as those with verbs from class 1 like ?There arrived a man. This is not what the mainstream theory about this type of sentence predicts! My conclusion from this is that the acceptability of there sentences isn't actually sensitive to whether the subject is in any way object-like or not, but rather to some other factor. This is in line with what certain other researchers have suggested might, in fact, be the case.

Conclusion

All in all this survey has been very illuminating. It has demonstrated a lot of things about English verbs, supporting some ideas and suggesting others require alternation. I must say again that I am very grateful to all the people who took part.