December 28, 2006

Two ways to look at the passive


Since we last looked at the injunction Avoid Passive in any detail (in a posting by Mark Liberman that has links back to a pile of earlier postings), I've looked at some more treatments of the passive in books of advice.  Here I'm going to report on two extremes: at the low end, Toni Boyle and K.D. Sullivan, The Gremlins of Grammar: A Guide to Conquering the Mischievous Myths That Plague American English (2006), and at the very high end, Virginia Tufte, Artful Sentences: Syntax as Style (also 2006).  The two books share a semantic characterization of the passive, but otherwise they could scarcely be more different.


I'm going to follow both of the books in talking about the "passive voice", though older writers on English grammar (especially in the 19th century) regularly object to this term, on the grounds that voice and tense and mood and aspect and the like are names of grammatical categories realized in inflectional morphology.  Latin has a passive voice, these writers explain, because it has a system of inflected verb forms that are primarily devoted to use in constructions of a certain sort.  English, on the other hand, has constructions of this sort, but no verb form primarily devoted to use in them; the English passive (as in This book was written by a friend of mine) uses the "past participle" (to give it its traditional, and very opaque, name), which is also used in perfect-aspect clauses (like Kim has written many books) and adjectivals (like Written instructions are better than oral ones and When we arrived, at 5, the door was closed and locked).  To put things another way, the work that's done by inflectional morphology in Latin is done in English "analytically", or "periphrastically", that is, by syntactic constructions.

What is this work?  Simplifying a lot, the passive provides a way to treat what is normally the direct object of a verb (or, occasionally, the object of a preposition) as a subject.  Note that my characterization is framed entirely in syntactic terms: the syntactic category verb and the syntactic functions subject, direct object, and prepositional object.   There is no talk of actions, actors, agents, performers of actions, or recipients of actions.  (These semantic notions are not irrelevant, because they're tied, in very complex ways, to the syntactic notions of subject, direct object, etc.  They're just not identical to them.)

English has a number of constructions that do this work.  Among them are several that use the past participle form of the verb and optionally allow the expression of the normal subject in a prepositional phrase with by (in actual writing and speech, the by-phrase is much more often omitted than not).  I will refer to all of these as "passive constructions".  Among them are the "BE-passive" in (1) and the "GET-passive" in (2):

(1) Kim was attacked by wolves.

(2) Kim got attacked by wolves.

There's now a problem in using the technical term "active (voice)".  It contrasts with "passive (voice)", but how?  Is it narrowly contrasted, so that active VPs are only the ones that can have passive counterparts?  If so, then enormous numbers of verbs are neither active nor passive:  in particular, intransitives of several types, as in (3), and unpassivizable transitives of several types, as in (4). 

(3a) Kim slept.
(3b) Sandy disappeared.
(3c) Terry seemed unhappy.
(3d) Chris became a detective.
(3e) Three hours elapsed.
(3f) Terry screamed.

(4a) Kim resembles Sandy. (*Sandy is resembled by Terry.)
(4b) The play concerns poverty. (*Poverty is concerned by the play.)
(4c) I realized the answer. (*The answer was realized by me.)
(4d) These movies star Freddy the Pig. (*Freddy the Pig is starred by these movies.)
(4e) I have two houses. (*Two houses are had by me.)

Or is "active" broadly contrasted to "passive", so that anything that's not passive is active?  If so, then intransitives and unpassivizable transitives are all active.  In this case, we might as well abandon the misleading technical term "active" completely: VPs are either passive or not; the non-passive VPs don't necessarily have anything in common with one another, beyond not being passive. 

This is all background.  Now a word about the attitude we take at Language Log to the passive, which is that passive constructions have their uses and that a blanket injunction to avoid them, or even to avoid them as much as possible, is silly.  Good writers, including Strunk and White themselves, use them with some frequency, as we have pointed out many times here on Language Log.  In fact, most of Tufte's discussion of the passive (pp. 78-89) is devoted to its virtues, with many well-chosen examples.

(Quite often, people have written me to say that in their experience active clauses are usually, or even almost always, clearer than their passive counterparts.  These are, of course, impressions, not the results of systematic studies of passive use; they are subject to the effects of selective attention and confirmation bias.  When people have looked at polished writing to count passive clauses -- not an easy task, and subject to some judgment calls -- they find that 10-20% of the clauses are passive.  And when you look at specific examples, very few of them would be improved by conversion to actives, and many would be changed for the worse.)

It is true that some writers seem to be overfond of the passive, and can use some encouragement to re-word.  My impression, from working with students, is that the problem is rarely a simple fondness for passives, but usually involves a more complex set of difficulties in organizing discourses for an audience.  The ineffective passives are just a symptom of a larger problem.

Now to the two books.  Gremlins has a very brief treatment, less than a page (pp. 77-8).  The section, titled "Verbs Have Voices", starts with an explanation of the voices of English:

Verbs have two voices to choose from, active and passive.  If a verb is in the active voice, the subject is doing the action.

The matador confronted the bull, stared him in the eye, flicked his cape, then ran back to the side of the arena.

(The usual confusion between expressions and the things they denote, between words and the world.  Subjects of sentences are linguistic expressions, and expressions don't do actions; denotations of subjects might sometimes do actions, however.  This usage is so widespread that it might seen churlish to complain about it.  But I think it's useful for students to keep the distinction between form and meaning in mind; remember that this is a book for ordinary people, not professional linguists or philosophers.  Once that's well established, there's no problem in using the looser locution, since things will be clear in context.)

Ok, Gremlins talks about "verbs", period, suggesting that they hold to the view that all verbs are either passive or not, and they use "active" to refer to the non-passive ones.  That's just a terminological choice.  Then they give a version of the standard semantic characterization, in which verbs denote actions and subjects of active verbs denote the agents in those actions, and one example.  The example has four active VPs in coordination, sharing the subject the matador.  The first two of these can be seen as denoting actions only by stretching the notion of "action" considerably; confronting something and staring something in the eye are not caused changes of state.  It is, of course, easy to find much more extreme examples, of active VPs that transparently do not denote caused changes of state: many of those in (3) and (4) above, plus things like:

(5a) Kuwait lies to the south of Iraq.
(5b) The tank holds 14 gallons.
(5c) Everyone appreciates fine wines.
(5d) Fine wines please everyone.
(5e) Picnics attract ants.

When you look at polished writing and ask how many clauses have verbs denoting actions and subjects denoting the agent of those actions -- again, not an easy task and subject to judgment calls -- the figures are once more in the 10-20% range.  Action verbs with agentive subjects are certainly not in the majority.

I'm dwelling on these very familiar points because the characterization and the example appear in a book of advice; they're SUPPOSED TO BE HELPFUL to writers.  I can't imagine how they could be.  The semantic characterization is no more than recitation of a piece of a catechism, reproduced without understanding; a reader who takes it to be a claim about English (or languages in general) and tries to test it will quickly come upon examples like those above and conclude  that the claim is false, while everyone else will just memorize it as a definition and pass on, no wiser.  But why do semantic characterizations persist, in the face of such abundant counterevidence?

I suspect that the answer is in fact that they are treated as dogma.  They are seen as being so fundamentally true that action and doer of the action have come to be understood as 'meaning of a verb' and 'meaning of a subject in an active clause', respectively.  Plenty of people have responded to examples like those in (4) and (5) by patiently explaining to me that  they do indeed describe actions, in some extended or metaphorical interpretation of the word action.  For them, the semantic characterizations couldn't possibly be false.  If so, then including them in an advice book is nothing more than instruction in the catechism.

In any case, Gremlins passes immediately to the passive, leading with:

The passive verb always uses some tense of to be.

The book that was written in four weeks was made into a movie in four years. [(I)]

A quibble: "some tense" should be "some form".   Is written and was written have tensed forms of BE (present and past, respectively), but be written (base form), being written (present participle), and been written (past participle) do not, yet all of them are passive.  A small point, true, but also another instance of the often shocking laxness in the use of standard grammatical terminology in popular writing ABOUT GRAMMAR.

More important, we've already seen that passive verbs don't always use some form of BE; there's also the GET-passive, as in (2).  In fact, there's a whole lot more -- in particular, BE-less passives in various verb-complement constructions, as in (6), and in various free adjunct constructions, as in (7).

(6a) The fiends had Kim attacked by wolves.
(6b) We saw Kim attacked by wolves.

(7a) Attacked by wolves, Kim fled.
(7b) With Kim attacked by wolves, everyone was terrified.
(7c) Once attacked by wolves, you'll never feel the same about the forest.

And there are many constructions with the verb BE in them that are also not passives -- the progressive, in (8), and an assortment of copular constructions, sampled in (9).

(8) Wolves are attacking Kim.

(9a) Terry is unhappy.
(9b) Superman is Clark Kent in disguise.
(9c) There are penguins on the porch.

I mention all this because Gremlins has, for some reason, taken the occurrence of a form of BE as criterial for passives, when in fact it is neither necessary nor sufficient.

Meanwhile, there are several constructions involving subjects that are understood as objects of verbs, but are in fact NOT passive constructions, for example the four illustrated in (10), in which the subjects are understood as object of the verbs read, skim, lift, and wash, respectively.

(10a) This book reads easily.
(10b) This book is easy to skim.
(10c) This box is too heavy to lift.
(10d) My shirt needs washing.

Two of these -- in (10b) and (10c) -- do have a form of BE in them, and they all have non-agentive subjects, so at least two of them are problematic for the way Gremlins delineates passives.

But all of these details are as nothing in the face of the fact that this section of the book is the first place in it where passives are mentioned, and the fact that the two short passages above (three sentences of text in all) are the whole of the book's treatment of the nature of active and passive voice.  Obviously, no one could make any sense of this if they didn't already know how to recognize actives and passives, at least in the easy cases, so what is this section for?

The point is to trumpet Avoid Passive (which is what comes next); the stuff about be is there, I think, just as a demonstration that serious grammatical issues are somehow involved.  The tactic here is one I've seen in a number of popular advice books (I hope to post on some other examples eventually): the goal of a section of the book is to proscribe some usage, but first there are some ornamental technicalities, which serve to suggest that the proscription is somehow grounded in Real Grammar and therefore should be taken seriously.  The ornamental technicalities are, typically, one or more of the following: truncated (Gremlins on the passive might be a new record here); therefore desperately incomplete; inaccurate on factual details; illustrated by flawed examples; discussed with technical terms used inappropriately; and not entirely relevant to the proscription.  Oh yes, and the examples are almost always invented and almost always given without context.

In any case, the punch line is:

The subject of a passive verb never acts--which gets pretty boring.  It's like listening to music that's always in a minor key.  Dreary.  So writing or talking in the active voice is best.

The truly remarkable part of this is its framing as an objection to ALWAYS using the passive (where it's available), something no one has ever even come close to suggesting.  (Even in advice to use the passive in describing the design of experiments and tests in the scientific literature, the manuals don't tell you to use the passive everywhere.  But, anyway, Gremlins isn't addressed to people writing scientific journal articles.) 

Then there's the bad-mouthing of music in minor keys.  Undeniably, minor scales and chords are popularly associated with melancholy, but there's plenty of minor music with other emotional tones (Beethoven's Fifth Symphony is in C minor, a key that many have seen as characteristically "stormy" and "heroic" for Beethoven), and most music of any length modulates between minor and major (sometimes shorter compositions do too; as Daniel Levitin notes in This Is Your Brain on Music, p.38, "Light My Fire" by the Doors has the verses in minor chords, but the chorus in major chords).

And then the analogy between passive syntax and minor music, which seems to turn on perceived associations between passivity, in the real world, and, on the one hand, passive syntax, and, on the other, minor keys -- in combination with a celebration of activity, energy, control, etc. in the real world, which are associated with active syntax and major keys.  There's a lot to be said on the topic -- why, for example, is the contrast not between restiveness (bad) and placidity (good)? -- but, as far as I'm concerned, none of it belongs in a book like Gremlins.  There might indeed be some metaphorical associations, between grammatical voice and extralinguistic matters, that have some psychological reality for at least some speakers, but they're likely to be subtle in their effects, much more subtle than other factors that I'll take up below.

Finally, a comment on "the subject of a passive voice never acts".  For passivizable verbs with non-agentive subjects, the passives are just as (metaphorically) "active" as the corresponding actives, as is the case for (5c) and (5d) and their passives:

(5c - active) Everyone appreciates fine wines.
(5c - passive) Fine wines are appreciated by everyone.
(5d - active) Fine wines please everyone.
(5d - passive) Everyone is pleased by fine wines.

As far as I know, there are no verbs with agentive direct objects -- there is, after all, SOME significant connection between the syntactic functions in sentences and the participant roles in situations -- but you can concoct passives in which the subject denotes an agent, just not the agent of the verb that is passivized.  What I have in mind are things like:

(11) I was moved/impelled/inspired to sing the national anthem.

Here the impulse or inspiration is internal to the speaker of (11).  The effect of the sentence is to assert that the speaker sang the national anthem -- performed an action -- and did so as a result of this internal impulse or inspiration.

Back to the Gremlins text.  The activity connection is pursued further in its final part:

[So writing or talking in the active voice is best.]  To see why, let's take the last example [(I)] and turn it around.

She wrote the book in four weeks, but it took four years to make the movie. [(II)]

You want others to remember what you say and write, so keep it active.  The exercise will do you good.

Taking it from the end:  the activity connection is there in the pun on exercise; and the preceding sentence introduces a new (and unsubstantiated) claim, that active sentences are easier to remember than passive sentences.  Now consider the passive example, (I).  It is indeed awkward, but that's at least in part because the book that was written in four weeks is hard to contextualize.  (It's only too easy to invent awkward examples, especially out of context.)  If the referent of the book is given in the context (as it must be in (II)), then the following (which also makes the contrast explicit) is something of an improvement:

(III) The book was written in four weeks but was made into a movie in four years.

In (III), the restrictive relative clause (modifying the book) that makes the original hard to contextualize has been turned into the first conjunct of a coordination; the Gremlins rewriting, (II), does the same.  That is, (II) is not a simple "turning around" of passives into actives; that would produce something like

(IV) The person who wrote the book in four weeks made it into a movie in four years.

in which, as in (I), the contrast between four weeks and four years is poorly expressed, because four weeks is inside a relative clause and four years is in the main clause.  A minimal fix would put the two NPs in parallel positions:

(V) X wrote the book in four weeks and made it into a movie in four years.

(where X is some subject NP).  Converting a passive with no by-phrase into an active requires supplying material not in the original; in this case, Gremlins supplies, without comment, a subject she.

But (V) implies (almost surely incorrectly) that the person who wrote the book also made the movie of it.  Version (IV) shares this defect, but there's no such problem with (III), since (III) contains no NPs denoting the writer of the book or the maker of the movie.  That's one of the virtues of the passive: it allows you to omit any expression of the subject of its active counterpart.  In any case, fixing the problem with (V) requires you to supply different subjects for wrote and made:

(VI) X wrote the book in four weeks and Y made it into a movie in four years.

The first lesson here is that rewriting to avoid some proscribed usage often requires rewording other parts of the sentence, sometimes substantially.  Advice manuals almost always do this subsidiary rewriting without comment, though if readers need advice on using actives and passives they almost surely need help in the rewriting process.

Now look at the first conjuncts in (III) and (VI): the book was written in four weeks (passive) vs. X wrote the book in four weeks (active).  These clauses are not interchangeable in discourse, because the passive version is about the book, while the active version is likely to be understood as being about X; in general, a subject is likely to be understood as denoting something that is both topical in the sentence (what the sentence is about) and topical in the discourse (what the discourse is about at this point).  That's another of the virtues of the passive: it allows you convey that a certain discourse referent (denoted by the subject of the passive) is topical.  Gremlins, like most advice on the choice between active and passive, fails to even hint at the enormous importance of topicality in this choice.

Next, look at the second conjunct in (VI) -- and Y made it into a movie in four years -- and compare it to the active Gremlins version, (II), and the improved passive version, (III).  As I've already pointed out, (II) and (III) bring out the contrast between four weeks and four years by using but instead of and.  This is another way in which rewriting can introduce material not explicit in the original.  That's the second lesson here: advice manuals very often make alterations in the original that are not required by a straightforward undoing of the proscribed usage; they "improve" the original in other ways as well and so heighten the contrast between the "bad" original and its rewriting (almost always without comment or explanation, of course).

In fact, the second conjunct of the Gremlins version, (II) -- but it took four years to make the movie -- goes way beyond the minimal rewriting in (VI).  Strikingly, (VI) has an action verb and an agentive subject in this conjunct, but (II) does not!  The verb in (II), took, is indeed active voice, but in the sense here TAKE belongs with the verbs in (5) above, which don't even come close to denoting actions.  In fact, in this sense, TAKE is unpassivizable (with either of the two available verbs):

(12a) It took four years to make the movie.
(12b) *Four years were taken (by it) to make the movie.
(12c) *The movie was taken (by it) four years to make.

As for the subject, it's a "dummy" it, a place-holder with no denotation of its own (certainly not as an agent in an action); instead, in this construction to make the movie 'making the movie' is interpreted as the subject of took four years.  The verb TAKE in related constructions, as in (13) and (14), is equally unpassivizable:

(13a) Making the movie took four years.
(13b) *Four years were taken (by making the movie).
(13c) Making the movie took Allen four years.
(13d) *Allen was taken (by making the movie) four years.

(14a) The movie took four years to make.
(14b) *Four years were taken (by the movie) to make.
(14c) The movie took Allen four years to make.
(14d) *Allen was taken (by the movie) four years to make.

((14a) and (14c) illustrate further constructions, like those in (10), which have a subject understood as the object of a verb but which are nevertheless not passive.)

What's happened here is that the Gremlins version of the second conjunct introduced an entirely new construction, not  in the original (again, without comment or explanation).  On top of that, the construction totally fails to fit the Gremlins characterization of active clauses, and indeed suppresses any mention of the maker(s) of the movie, just the way an agentless passive does.  Goodness knows what readers are supposed to make of all this for practical purposes.

Now I'm not claiming that there's something wrong with the second conjunct of (II).  In fact, I think it's pretty good.  There are several variants or expansions of it that might also do:

(15a) ... it took four years to make the movie. [in (II)]
(15b) ... it took four years for Y to make the movie.  [with mention of the maker(s)]
(15c) ... it took four years to make the movie of it. [with explicit reference to the book]
(15d) ... it took four years for Y to make the movie of it. [combo of (b) and (c)]

(16a) ... making the movie took four years.
(16b) ... making the movie took Y four years.
(16c) ... making the movie of it took four years.
(16d) ... making the movie of it took Y four years.

No doubt you can imagine still other possibilities.  What's good about all of these is that they bring out the two relevant contrasts, between the movie and the book and between four years and four months.

The problem with (II) is its first conjunct, specifically the subject of this clause.  Version (II) treats the writer of the book as topical, and that's possible (if so, then (II) conveys a topic shift, away from the writer of the book to the book itself, in contrast to the movie), but it's likely, especially when the sentence is viewed out of context, that the book is topical, in which case we want the book to be the subject of this clause -- that is, we want a passive.  The book's writer can then be downgraded in its discourse status (by being mentioned in a by-phrase), or you can suppress mention of the writer entirely, depending on your wider aims in the discourse:

(VII) The book was written (by X) in four weeks, but it took four years to make the movie.

(or with any of the other variants for the second clause, or with one of the constructions in (14) in the first clause).  And if you want to treat X as topical, then there are further possibilities, with active verbs in both clauses, for instance:

(VIII) X wrote the book in four weeks, but Y took four years to make the movie.

To sum up: the Gremlins treatment of the passive is appalling, but in detailing just what is appalling about it I've tried to bring out some important points.  What's especially disheartening, though, is that The Elements of Style (right back to the Strunk 1918 original) -- cited approvingly in the Gremlins reading list, by the way -- gets some of this right.  In particular, Strunk appreciates the significance of topicality in choosing between active and passive.  Here's his summary:

The need of making a particular word the subject of the sentence will often, as in these examples [given just before this], determine which voice is to be used.

(Note the nominal of making a particular word... instead of the verbal to make a particular word... and the passive is to be used instead of the shorter and active to use.  Strunk wasn't very good at following his own advice.)

Now we get to Virginia Tufte.  Tufte just assumes her readers are acquainted with the concepts and terminology of traditional grammar; her aim is to show you what you can do with the resources of English.  The section on "the passive verb" begins with the usual semantic characterization:

Otters eat clams.  The verb is in the active voice: the subject performs the action.  Clams are eaten by otters.  The verb is in the passive voice: the subject receives the action.

Oh dear.  But then she jumps right into a discussion of discourse organization:

Which form you use depends on whether you have previously been writing about otters or clams.  One of the uses of the passive is to shift the topic or the emphasis.  Another is to move the noun phrase that was the subject of discussion to a new location in the sentence, usually toward the end...

She also notes that "for good or ill" the passive allows you to omit this noun phrase entirely.

More generally,

There are, as with other inversions, many reasons for turning to the passive, including the need for special emphasis or rhythm, for strategic rearrangements of different kinds to aid modfication or to increase cohesion, for adjustments in a parallel series, and for certain more thematic effects, often providing a contrast with the active verbs.

These points are illustrated with pages of examples, extensively and sensitively discussed.  Real examples, with contexts.  On occasion, passives are rewritten into less effective actives.  In the middle of the section she turns to the sorts of passives that critics most often complain about, especially in chilly officialese and impersonal reporting, though she notes that the coldness and impersonality of her examples might well have been intended by those who wrote them; it's not at all clear that the syntax is the problem.

There is no hymn to the energetic activity of the active, no castigation of the boring submissiveness of the passive.  It's all about what you can do with the two voices.

If you don't know how to recognize a passive (at least in the easy cases), then you'll need some background before you can tackle Tufte.  Don't, however, try to get it from Gremlins.

zwicky at-sign csli period stanford period edu

Posted by Arnold Zwicky at December 28, 2006 10:42 PM