The truth about infer

« previous post | next post »

The other day, I dropped a passing reference to the misuse of infer to mean "imply". The facts, as John Cowan reminded me in a comment, are more complicated. A few minutes of research reveals that the truth about infer is even more complex — and more interesting — than I suspected.

Let's start with the simplest version. We have a person P, an audience A, some evidence E, and a conclusion C. We put these ingredients together in three ways:

(A) The evidence E leads to the conclusion C: "E implies C".
(B) The person P deduces the conclusion C [from the evidence E]: "P infers C [from E]".
(C) The person P indirectly communicates C [to the audience A]: "P implies C [to A]".

This roughly describes how I use infer and imply, and what most usage authorities prescribe for these words.

All of these uses have been around in English since the 16th century; and all of them are in all the dictionaries. But there's a serious problem with this simple story: infer has also been used since the 16th century in meaning (A) — and this sense is also in the standard dictionaries.

The earliest example in the OED of infer in usage (A), "Evidence infers Conclusion" comes from Sir Thomas More in about 1530:

4. To lead to (something) as a conclusion; to involve as a consequence; to imply. (Said of a fact or statement; sometimes, of the person who makes the statement.)

c1530 MORE Answ. Frith Wks. 840/2 The fyrste parte is not the proofe of the second, but rather contrary wyse, the seconde inferreth well ye fyrst.

MWCDEU, apparently writing in reference to an earlier edition of the OED that dates the quotation differently, calls this usage "More 1533":

More 1533, the OED shows, was used continuously up to the time the book was edited; Milton and James Mill are among those quoted. Here are some examples not in the OED:

He used Metcalf as an agent in all proceedings which did concern that foundation: which will infer him to be both a wise and an honest man. [Thomas Fuller, The Holy State and the Profane State, 1642]
However, as I have often heard Dr. Johnson observe as to the Universities, bad practice does not infer that the constitution is bad. [James Boswell, Life of Samuel Johnson, 1791]
Lucy … reseated herself with an alacrity and cheerfulness which seemed to infer that she could taste no greater delight. [Jane Austen, Sense and Sensibility, 1811]
… to be a literary man infers a certain amount of — well, even formal education. [William Faulkner, 25 Feb. 1957, in Faulkner in the University, 1959]

This sense of infer — our meaning (A), "Evidence infers Conclusion" — is not only in the OED, but was also in other authoritative dictionaries, such as Webster's 2nd and the Century Dictionary:

As MWCDEU explains:

More 1533 is not itself really the subject of much controversy, but it is recognized in dictionaries, and that recognition leads the commentators not to trust the dictionaries.

We may call the real disputed sense "personal infer" because — in distinct contrast with More 1533 — it usually has a personal subject.

In other words, the "real disputed sense" is our case (C), "Person infers Conclusion [to Audience]".

In understanding how this tangle arose, it may help to indulge briefly in the etymological fallacy. Originally infer basically meant "to bring in" — the OED's etymology is:

[ad. L. inferre to bear, bring, or carry in, to inflict, make (war), to cause, to occasion, to introduce; in med. L., to infer; f. in- + ferre to bear. Cf. F. inferer to allege, show, infer (16th c.).]

As you'd expect given this background, the word came into English in the 16th century with a wide variety of meanings, most of which are now obsolete:

1.a. trans. To bring on, bring about, induce, occasion, cause, procure; to bring upon (a person, etc.), to inflict; to wage (war) upon. Obs.
To confer, bestow. Obs.
with compl. To cause to be; to make, render. Obs. rare.
To bring in, introduce (in discourse or writing); to mention, report, relate, tell; to bring forward (as an argument, etc.), adduce, allege. (With simple obj., or more rarely obj. clause.) Obs.

To carry in, insert; to figure as inserted or projecting into. Obs. rare.

Note that the OED's obsolete sense 2. "to bring forward …, adduce, allege" is just a more direct form of the OED's sense 4. "to lead to (something) as a conclusion; … to imply".  Citations given for sense 2. start in 1526 and include this from Shakespeare:

1593 SHAKES. 3 Hen. VI, II. ii. 44 Full well hath Clifford plaid the Orator, Inferring arguments of mighty force.

The OED now adds to its sense 4. ("To lead to (something) as a conclusion; to involve as a consequence; to imply") a usage note : "This use is widely considered to be incorrect, esp. with a person as the subject." This note was not in the edition that the MWCDEU was based on, but was apparently added in recognition of the near-universal hue and cry about "personal infer" among usage mavens from 1917 onwards. As MWCDEU explains with respect to a similar history at Merriam-Webster:

The OED defined More 1533 without comment, as did Webster 1909 and undoubtedly other dictionaries. But in 1932 a member of the philosophy department at Boston University wrote to the Merriam-Webster editorial department questioning the definition of More 1933 in Webster 1909. In his opinion the sense was no longer in current good use. After some preliminaries about logic, he got to the point:

… no cultured person has in my hearing ever confused the two words. It is, however, the constant practice of the uneducated and the half-educated, to use infer for imply.

The logician is, of course, talking not about More 1533 but personal infer.

The editors gave serious consideration to the complaint … One editor suggested adding a note to the definition of More 1533, but he was overruled by another who observed that the sense in question was not quite the same as More 1533. Instead they added a new definition, "5. Loosely and erroneously, to imply," which appeared in Webster's Second (1934). Thus the dispute was established, although most of the usage-book comment came after World War II.

OK, what's the history of "personal infer"? MWCDEU thinks it's a recent development:

The personal infer is relatively recent. We have found no examples earlier than this one:

I should think you did miss my letters. I know it! but … you missed them in another way than you infer, you little minx! [Ellen Terry, letter, 3 Oct. 1896]

But through the magic of internet search, I believe that we can move that date back a bit, and also help to explain how the usage arose — see below! Continuing with MWCDEU,

This sense is clearly an oral use at the beginning — we have no examples of it in print, except in usage books, until the middle of this century. But it is obscure in origin — was it a theater usage? Or possibly an Americanism Ellen Terry picked up on one of her American tours?

The development of infer-imply as a usage issue is curious, too. The ealiest mention of the subject is in MacCracken & Sandison 1917, an English handbook apparently originally intended for for use at Vassar College. Here is what it says:

A speaker or his statement implies (suggests, expresses, though not explicitly) something which a hearer infers (draws or deduces) from the statements. Infer is constantly used where imply is intended. CORRECT: Do you mean to imply [not infer] that I am deceiving you?

You will note that the example shown has a personal subject and that the authors say "infer is constantly used." Since we know of no printed evidence existing in 1917, we infer the usage to be spoken.

Following the hint that this might be a "theater usage", I decided to check 19th-century plays in Literature Online. What I found is that it's often very difficult in practice to distinguish the meanings given at the start of this post as (B) "Person infers (deduces) Conclusion from Evidence" and (C) "Person infers (communicates) Conclusion to Audience".

Here are the first three examples that I found, in alphabetical order:

In the anonymous play Wilhemina, from 1826, we have this sequence:

Dofrestom: Dost thou?–Then is thy doom sealed.
Count Griffel: Sealed, boy? What darest thou infer by that?
Dofrestom: Thy death, and punishment hereafter.

Does Count Griffel mean "What do you dare to deduce" or "What do you dare to suggest"? It could be either one. The use of "by" is consistent with either one — thus we might say "by lemma 7.13 we infer that …", or "by 'watery' we imply insipidness".

In Joanna Baillie's Basil (1851), we find:

Count Rosinberg: He's form'd for great occasions, thou for small.
Valtomer: But small occasions in the path of life
Lie thickly sown, while great are rarely scatter'd.
Count Rosinberg: By which you would infer that men like Fred'ric
Should on the whole a better figure make,
Than men of higher parts. It is not so …

Does Count Rosinberg mean "by which you would deduce that …" or "by which you would suggest that …"? Again, it's hard to tell.

From Nathaniel Harrington Bannister's England's Iron Days (1837):

Elgina: There is an awful hour which all do wish
Bright and joyous—quiet and undisturb'd—
No care no dejected passions to cloud
The mind—And that is when we die. The clock
Of death to the wicked man, is awful
In its sound, for it rings the knell of a
Soul, gone to answer to that tribunal,
For deeds done in its tenement of clay.
To the righteous man, the peal is like a
Mother's lullaby to her dear infant
Gliding into the sweet arms of slumber.
Let virtue halo thy closing days
Lord Edgar: Elgina, what mean'st thou? what would'st thou infer?
Elgina: If thou dost wish to live in good men's
Thoughts, or live beyond the grave; confess thy
Fault; and, Alaster meet not in the lists.

Same deal here — is Lord Edgar asking what Elgina would deduce, or what she would suggest? And what's with all the Counts and Lords talking about inferring, by the way? Not that we have a statistically significant sample yet, but still, three out of three…

OK, fourth time lucky. From John Brougham (1810-1880), Flies in the Web:

Alice: Traitor! But my consent he shall never have.
Foxglove: Why, in the name of all that's consistent, you don't mean to infer that you love this fellow?
Alice: I do, I do! I dared not whisper to myself how much! But now the barrier of pride is overthrown, and I am humbled—humbled to the very dust.

Here at last we get past the nobility — and the thees and thous — and find a case that seems unambiguous. It doesn't make sense for Foxglove to ask Alice whether she is deducing her amorous attachment; he must be asking about what she's suggesting or implying.

The play was first performed at the Theatre Royal, Manchester, in 1860; and the source of this text appears to be French's Acting Edition no. 1021, published ca. 1860 — so we can antedate Ellen Terry's 1896 letter by 36 years, and with a print citation.

For quite a long time before that, probably all the way back to the 16th century, there were examples that were ambiguous between the now-standard sense of infer — our case (B), "person deduces conclusion" — and the deprecated usage that MWCDEU calls "personal infer" — our case (C), "person communicates conclusion".

Perhaps "personal infer" survived from the start as a residue of the OED's "obsolete" sense 2. "To bring in, introduce (in discourse or writing); to mention, report, relate, tell; to bring forward (as an argument, etc.), adduce, allege". But even if it's not simply a survival, this sense would have been invented over and over again in response to those contextually ambiguous uses of the now-standard OED sense 3.a. "To bring in or ‘draw’ as a conclusion; … to deduce, conclude". So I wouldn't be surprised to find that with more time to search, we could find still earlier examples.

Thus the story of infer as "imply" is not just a recent mistake in matching similar words to a complex of related concepts. There's a long history of erratic specialization (from the original sense "bring in" to the much more limited meaning "deduce") and sporadic generalization (from contexts where "deduce" might be taken to mean "suggest").


  1. Rob Gunningham said,

    August 11, 2008 @ 9:36 am

    You're right it is interesting, but I'll bet you're not going to start using non-personal infer instead of 'imply' yourself, are you? There's no reason to do so.

  2. goofy said,

    August 11, 2008 @ 10:26 am

    There is something else that's worth noting, from MWDEU: "it seems likely that the repeated injunction not to use 'infer' for 'imply' has diminished the literary use of More 1533, and if anything is in real danger of being lost, it is that long-standard use."

  3. E W Gilman said,

    August 11, 2008 @ 10:37 am

    Very fine exposition. I have made a note of the antedate, just in case I ever get a chance to get back and make some revisions in the book.

  4. Ryan Denzer-King said,

    August 11, 2008 @ 1:45 pm

    Could it be that 'imply' and 'infer' were in free variation in the 16th century? And that some people naturally settled on using one in one situation and the other in another situation? I do this all the time when confronted with words which are similar in meaning: I tend to specialize them further because my neurons apparently don't like free variation.

  5. Sili said,

    August 11, 2008 @ 3:14 pm

    Glad I wasn't the only one to ponder if this was some sort of U/non-U issue by the third peer.

    I think I need to start taking a paper, just so I can infer it every morning (morningly?).

  6. Noetica said,

    August 11, 2008 @ 8:47 pm

    There are two distinct relevant senses of imply, and they may not be adequately drawn apart in the exposition above:

    The evidence E leads to the conclusion C: "E implies C".
    The person P indirectly communicates C [to the audience A]: "P implies C [to A]".

    In "E implies C", imply seems to be used in SOED's sense 2:

    Involve as a necessary consequence; involve the truth or existence of (a thing not expressly asserted). E16.

    But in "P implies C [to A]", imply seems to be used in SOED's sense 4:

    4 Express indirectly; insinuate, suggest, hint (at). L16.

    Whether this equivocation affects anything in what follows is another matter; but the difference is worth making explicit in the present context. In the jargon of logic, imply is always used in SOED's sense 2, as pretty well equivalent to entail.

  7. Stephen Jones said,

    August 12, 2008 @ 12:55 am

    I do this all the time when confronted with words which are similar in meaning: I tend to specialize them further because my neurons apparently don't like free variation.

    One thing I always noted is that where there are various words for something English tends to specialize whereas Spanish keeps the free variation. As I find it difficult to believe that English- and Spanish-speakers neurons are arranged differently I presume this is because Spanish free variations tend to be dialectical.

  8. Me said,

    August 14, 2008 @ 8:12 pm

    Also check out, 8th paragraph from the bottom. You could make it work if it weren't for the parallel use of "suggest" in the next sentence.

  9. Jonathan said,

    August 17, 2008 @ 2:10 pm

    As this hasn't been discussed explicitly, I presume I must either be wide of the mark or stating the obvious here, but … rather like 'lie' and 'lay' , the (standard) meanings of 'imply' and 'infer' stand in a causative relation to each other, and the syntactic relationships appear parallel too:

    X lies on the bed. / A causes X to lie on the bed. / A lays X on the bed.X is inferred. / A causes X to be inferred. / A implies X.and hence: B infers X. / A causes B to infer X. / A implies X to B.

    There is also of course a zero-change way of forming causatives (which I think remains productive) – could this help to explain non-standard 'infer'?

    The tune changes. / A causes the tune to change. / A changes the tune.X is inferred. / A causes X to be inferred. / A infers X.

    The alternation "X implies Y." / "A implies Y." also seems to fit the zero-change process (X and Y being abstract, A a volitional agent):

    The tune changes. / A causes the tune to change. / A changes the tune.X implies-Y. / A causes X to imply-Y. / A implies-Y by X.

    [(myl) This general idea was discussed in an earlier set of comments here. The trouble is that imply can't be a causative, at least not of the usual type, since then you'd say "I implied Kim that conclusion", meaning "I caused Kim to infer that conclusion".]

  10. Jonathan said,

    August 17, 2008 @ 2:12 pm

    (Sorry, the line breaks got messed up – there should be breaks where a full stop isn't followed by a space or line break. Trying to be too clever…)

RSS feed for comments on this post