Saturday, August 13, 2011

On the acceptability of "on accident."

Over on the microblogging site Tumblr (where I don't always post about language, but you're welcome to look me up if you'd like), the following image has been making the rounds. As you might imagine, I do not particularly agree with its claims:


It is hardly the only message of its kind. Indeed, there is a great abundance of such misinformation on language and grammar in the world today, which was one of my main impetuses for creating this blog. Language Hippie is an attempt to share solid facts, the results of linguists' careful scientific inquiry into language, in a world where statements like the above image spread rapidly from person to person without much analysis of their claims. I'd like to focus on this particular image because of the way its author has framed and phrased its claims as authoratative. By that I mean it is a professional-looking image, and one that covertly claims authority on language with its reference to the findings of linguists (uncited and seemingly irrelevent as they are). It also refers in passing to language descriptivism, which is fairly rare in such a message. Usually, prescriptivist claims come from individuals who have never been introduced to the merits of descriptivism as a scientific approach to language. This author has heard of descriptivim but is dismissive, and all of this merits a closer look.

First, however, it is important to separate three possible arguments against "on accident", which may appear similar at a glance but are actually distinct claims. The first is the image's primary message, that the use of words "by accident" is absolutely correct and the use of words "on accident" is completely wrong. This claim represents the prescriptivist intuition that there is a single correct form of language, just like there is a single correct answer to the math question "What is 4 times 5?" or a single correct formula for the chemical carbon dioxide. The above image puts forward this argument via matter-of-fact statements like "Things do not happen on accident" as well as the claim that to use the word "on" in this phrase would be to "forget education" and spread the "rot" of the English language. The overall suggestion is that the phrase "by accident" is inherently correct, and "on accident" is inherently wrong. In this view, a speaker who uses the latter instead of the former is akin to a child who says that the first president of America was Thomas Jefferson, for in each case there is a particular correct fact about the world that has not yet entered his or her knowledge. (And, like the ignorant child, we should not tell someone who says "on accident" that their way of things is legitimate or equal to the truth of the matter.) This is the standard prescriptivist claim, that there is a single correct form of language that is being ignored, forgotten, or unlearned in some fashion.

Note that this argument, like the above picture, provides no reasoning for why one particular form of language is correct and another is wrong. "Things do not happen on accident" is presented as a simple factual claim like "4 times 5 equals 20", with no explanation for why this is the case. However, there are certainly arguments that could be made in favor of a form like "by accident" that go beyond simply asserting its inherent correctness. Although the above image does not do so, it is possible to argue from either a historical or a logical perspective for one linguistic form over another. Speaking from history, someone could plausibly say, "We have never used this form before, and deviation is unacceptable", while from logic, another might assert that the one form of language simply doesn't make sense. This type of reasoned argument against "on accident" is more nuanced in its approach, but now that we have defined the arguments, we can begin to examine and critique their qualities.

As I mentioned earlier, my goal on this blog is to share what linguists have discovered through applying the scientific method to language. And the primary, most certain finding of the field of linguistics has been that language is not fixed. It changes from person to person, from group to group, and most certainly over time -- so that even the same individual might not speak the exact same language at age 50 as he or she did at 30. So when linguists investigate a particular language, their conclusions about its grammar are not irrevocable facts the way discoveries in other fields might be, but are instead closer in spirit to the results of a census. They represent a split-second snapshop of the state of the language at that particular moment, and should always be interpreted as temporary. If a linguist declares something to be ungrammatical, he or she is merely stating that at the time of writing, the language does not appear to allow such a form.

It follows from the perpetual flux of language that there is no such thing as an inherently correct linguistic form. That is, there is no word, phrase, or grammatical construction that is necessarily grammatical, or necessarily ungrammatical, in a language. Although speakers might currently favor some particular construction, that is no indication that it used to be said in the language, or that it will be used by speakers in the future. Chaucer's Canturbury Tales, for instance, were written in the 14th century and contain the word "thanne." That word would have been ungrammatical in the 10th century, when English speakers said "├ża," and it is certainly ungrammatical today, when we use the word "then" instead. Yet in Chaucer's time, "thanne" was a perfectly acceptable word. And similarly, a construction that is today not considered acceptable might very well enter the language in the future -- which is to say, there is no word or phrase so inherently incorrect that it could never be part of a given language. Indeed, the fact of language change dictates that there is really no such thing as inherent grammaticality or ungrammaticality. The first argument against "on accident", that it is simply an inherently incorrect form of language, can therefore be discarded.

The fact of language change also serves as a counterargument to the potential argument from history, that the phrase "on accident" has traditionally never been used in English. It may or may not be true that this phrase is a new addition to the language, but from a linguistic perspective, an item's longevity should have no bearing on its grammaticality or acceptability to speakers. Phrases that have been popular in a language for centuries may suddenly fall out of favor, and new coinages may suddenly be heard on every speaker's lips. This is simply how language functions. A construction should not be villified for its newness, for nearly every word we say today was at some point a new entry to the English language. The claim that deviation is unacceptable in language is therefore absurd, and the argument from history can be discarded as well.

Finally, then, we reach the argument from logic. This argument against "on accident" claims that the phrase is compositionally nonsensical -- that is, that the meanings of its pieces do not make sense together. It could be argued, after all, that the preposition "on" designates placement at the top of something, as in the phrase "sitting on the table", while the preposition "by" indicates means, like in the phrase "going by train." According to this argument, "on accident" would mean placement at the top of an accident rather than happening through the means of an accident, and so the phrase cannot be an acceptable substitute for "by accident." The phrase simply doesn't make sense.

Another finding of linguistics, however, is that language is not always logically compositional: the meanings of an utterance's pieces do not always add up to the meaning of its whole. This is readily evident in English, where the phrase "I'm not unhappy" is most definitely not synonymous with "I'm happy." Or consider the word "inflammable", whose individual pieces appear to mean "not able to be set on fire" and yet whose definition is the exact opposite of that. And logically, the verb "am" indicates a first-person subject, so the pronoun "I" should be redundant and unnecessary in the sentence, "I am a lawyer" -- and yet dropping it to say "Am a lawyer" sounds curiously ungrammatical.

The fact is, language meaning does not always follow the rules of logic, and it is a futile endeavor to apply logic to judgments of grammaticality. If this is still not clear, simply consider the phrase "on purpose", whose meaning is roughly opposite to "on/by accident." If we follow the definition of the preposition "on" given above, we would have to conclude that "on purpose" is nonsensical and ungrammatical, and should properly be stated as "by purpose" instead. And yet, as a linguist looking at English, I have never come across a speaker saying something happened "by purpose." Despite the logic of its composition, it is simply not a phrase that speakers are saying. And "on accident", despite its composition, is.

Ultimately, this last fact is the truest judge of grammaticality: that the phrase "on accident" is something that people are saying. The job of a linguist is to examine the language that people speak and describe it as best we can, and from this perspective, "on accident" is wholly grammatical. In fact, since people rarely complain about linguistic forms that no one is actually using -- "It's not correct to say 'sdog' as the plural for 'dog'! The 's' has to go at the end of the word!" -- claims of ungrammaticality are almost always misguided. If someone has taken the time to complain about a certain language form being unacceptable, that is usually a pretty good indication that someone in the language is saying it. And if it's being said, it's a part of the language -- no exceptions.

Despite the above image's disdain for descriptivism as asking people to "forget education", we are actually sharing the real scientific findings of how language works. It is not necessarily "how you feel about English" that matters, but it is true that everything you say is grammatical. And for a scientist studying language, it would be sheer folly to discount that data due to a misguided preconception of what we are looking for when we study grammar.