June 29, 2006

Diagnosing soup label syntax

This evening, Thomas Norman contacted the Language Log Help Line with a problem:

My question concerns a pressing grammatical issue: the Campbell's Chunky soup label.  Every can of soup proudly proclaims itself the "soup that eats like a meal."  The meaning of the slogan is obvious, but the soup itself is certainly not eating anything.  While discussing the matter at dinner, my father and I also came up with the expression "she takes a good picture."  The intended meaning is clear--she looks good when you take a picture of her--but the structure of the sentence seems to imply that she is the one taking the picture.  Is there a name (other than "idiom") and explanation for constructions like these, or are they just grammatical curiosities?  (Or perhaps more accurately, ungrammatical curiosities.)

Though I'm a phonetician, specializing in the sounds of language, we linguists are sworn to provide aid and comfort in the event of any language-related emergency. Also, I seem to be the only one on duty today. So I'll give it a shot.

Thomas' question is clearly a matter of grammatical voice. As the Wikipedia explains,

In grammar, the voice of a verb describes the relationship between the action (or state) that the verb expresses and the participants identified by its arguments (subject, object, etc.). When the subject is the agent or actor of the verb, the verb is said to be in the active voice. When the subject is the patient, target or undergoer of the action, it is said to be in the passive voice.

Now, Thomas understandably expects the passive form to be something like "the soup that is eaten like a meal", and he's worried about what has happened to the is and the -en. Well, in the grammatical tradition of the classical languages (or what I remember about it from high school), the Campbell's Soup slogan would be called not "active", not "passive", but "middle voice". And supporting this memory (Mr. Mansur would be proud), a message contributed by Carl Conrad to the B-Greek mailing list ("Kemmer's Middle-voice categories", 10/18/2001) lists 19 "middle-voice categories (to which [he has] added notes on some Greek and Latin verbs falling into these categories) compiled from Suzanne Kemmer's The Middle Voice, Amsterdam/Philadelphia: J. Benjamins Publishing Co., 1993", with category #18 in this list given as:

18. Facilitative: inherent characteristic of patient allows action to take place: "soup eats like a meal."

Bingo! However, there seems to be some variation here in more recent terminology. SIL LinguaLinks says that

Middle voice is a voice that indicates that the subject is the actor and acts

* upon himself or herself reflexively, or
* for his or her own benefit.

(which would cover some of Kemmer's middle-voice categories, but not #18). LinguaLinks offers an alternative category that fits better:

Mediopassive voice is a passive voice in which the

* verb has stative meaning, and
* actor is not expressed.

where "stative" is a category of verbal aspect, such that a "stative" verb, as Wikipedia explains,

is one which asserts that one of its arguments has a particular property (possibly in relation to its other arguments). Statives differ from other aspectual classes of verbs in that they are static; they have no duration and no distinguished endpoint.

The Wikipedia article on the mediopassive voice gives the English examples

The book reads well. The trousers wash easily. Ripe oranges peel well.

We've clearly identified the syndrome and assigned a diagnosis to the symptoms: this pattern is known as middle voice, Kemmer category #18, or in some jurisdictions, mediopassive voice.

As so often in traditional forms of learning such as medicine and grammar, there is a bit of hocus-pocus about this. We owe more of an explanation than simply the assignment of a terminological category, however comforting that may be. I recall taking an infant with an uncomfortable and persistent rash to the doctor, one dark December afternoon, and being told that it was hibernal eczema. I felt at the time that I wanted more than a translation of the season and the symptom into Latin and Greek, respectively; but I suppose that what I was getting for the insurance company's money was the reassurance that this rash was a familiar condition, and that nothing much could or should be done about it.

So, Thomas, you can relax. The condition is nothing to be alarmed about: it happens in all the best languages. Indeed, as the Wikipedia entry for mediopassive tells us,

Proto-Indo-European ... had two voices, active and mediopassive, where the middle-voice element in the mediopassive voice was dominant. Ancient Greek also had a mediopassive voice in the present, imperfect, future, perfect, and pluperfect tenses, but in the aorist and future tenses the mediopassive voice was replaced by two voices, one middle and one passive.

Eat your soup in peace; and if you come back during normal clinic hours, one of our syntactic specialists will tell you about the latest research on the hows and whys of the many voices of grammar.

[And why voice is not tense.]

[Update: some additional insight is provided by CGEL (the Physicians' Desk Reference of syntax), which discusses examples like this one on p. 307 under the scare-quoted heading 'Middle' intransitives:

In this case the transitive use is primary and the intransitive is interpreted as having an unexpressed causer. Cross-linguistically, the primary use of the general term middle is for a term in a system of voice -- it applies to a voice that is in some sense intermediate between active and passive. The term is certainly not applicable to English in this sense: there are just two categories in the syntactic system of voice in English, active and passive. She doesn't frighten easily is active in form, but it has some semantic affinity with the passive, and it is in this semantic sense that it can be thought of as intermediate between ordinary actives and passives: we put scare quotes around the term to signal that it is being used in an extended sense and is not to be interpreted as denoting a formal category in the voice system.

Intransitives like She doesn't frighten easily characteristically have the following properties:

[40] i A causer (normally human) is implied but can't be expressed in a by phrase.
       ii The clause is concerned with whether and how (especially how readily) the subject-referent undergoes the process expressed in the verb.
       iii The clause is negative, or is headed by a modal auxiliary (especially will), or contains an adjunct of manner (such as well or easily).
       iv The clause expresses a general state, not a particular event.

The implication of a causer ([40i]) is what makes such clauses semantically similar to passives: compare She isn't easily frightened. But the causer cannot be expressed: *She doesn't frighten easily by noises in the dark. Property [ii] shows that these intransitive actives are by no means identical in meaning to passives. Compare The shirt irons well, whcih says something about the quality of the shirt, with The shirt was ironed well, which tells of the skill of the ironer. Properties [iii] and [iv] exclude such examples as *She frightens and *There was a sudden noise outside and she frightened immediately.


Posted by Mark Liberman at June 29, 2006 11:18 PM