by Suzette Haden Elgin, Ph.D.
(linguistics; UCSD)



In Lesson Two we looked at this small set of English words...

sign, resignation, malignant, signature, resign, benign, malignancy, malign, significant

...and tried to figure out why native speakers sometimes do — and sometimes don't — pronounce the G that appears in them. We considered every one of the nine words in the set, looking for patterns. We extrapolated from the set to other English words of the same kind, looking for evidence that the patterns we had found also applied to them. And we came up with this rule as a working hypothesis:

Rule 1: When an English word ends with GN, the G has to be silent.

In this lesson we're going to look at that rule again, and talk about it for a while. There are 4 things that we need to go over.

1.It's important to understand that the rule says only what it says. It says that when you pronounce English words ending with GN, you don't pronounce the G. It says nothing more. It doesn't say, for example, that when a GN sequence isn't word-final you do pronounce the G. That would be a different hypothesis, separate from the rule we've proposed, and it would have to be tested. We'd have to look at sets of words like these:

gnarl, gnash, gnat, gnaw, gnome, gnu

And like these:

designer, benignly, resignedly, signing

We'd quickly discover that there are other positions for a GN sequence in English words where the G has to be silent. Rule 1 is an accurate statement for the data we examined in Lesson Two, but it clearly isn't the last word on the matter.

2. It's important to understand that (as many of you suggested in your e-mails) there are many other ways to word Rule 1. As long as all the proposed wordings accurately describe what native speakers of English do when they say English words ending in GN, they're all acceptable. Choosing among them is a matter of personal preference. You might choose a particular wording for "stylistic reasons" — that is, because you liked the way it sounded. You might choose a wording because you felt that it was more elegant than the others. You might choose a wording because it fit better into your preferred model of linguistic theory. (We'll come back later in this course to what "model of linguistic theory" means; for now, trust me.) All of that is acceptable.

3. It's important to understand that in linguistics it's almost never safe to call a rule "the right answer." In math, two and two are always going to be four, and you can count on that. Any linguist who proposes a rule about a human language knows that some other linguist might propose a better one the following day. It's not unusual for a linguist to be giving a paper about a rule at a conference and have some other linguist in the audience break in and say, "Oh, that's not right! Here's what's really going on!", followed by a different proposed rule. That's a hazard of doing business, if you're a linguist. When you propose a rule, you're saying that — based on the information you have available at that moment — it's your best and most carefully-reasoned hypothesis about the language behavior in question.

4. Finally, it's important to understand that the rule we're working with isn't a rule about letters of English, but a rule about English sounds. Suppose we were talking about the words "telephone", "telegraph," and "pharmacy," and we decided that the relevant rule is something like "In those (and similar) words, PH is pronounced as F." That would be a rule about letters of English, and about the English writing system, and about English spelling. Our rule about word-final GN isn't like that; it's a phonological rule — a rule about sounds.

Moving Right Along

With all that specified, we can move on. It's certainly possible that English would have a phonological rule applying only to word-final GN. That could happen. But it's a bit suspicious. When you come across a linguistic rule that limited, you always want to ask yourself whether it might really be part of some other, bigger rule. We already know from the brief look at words like "gnaw" and "designer" that restricting the rule to word-final GNs is an error; we can see that it applies to GNs at the beginning of words, and to GNs at the end of some English syllables. (That is, the G in "designer," at the end of its second syllable, is silent; the G in "designate," also at the end of a second syllable, is pronounced.) That's enough to make us think that something more must be going on here. And in that situation, what a linguist does first is look at other English sounds that are the same type of sound as G and N. When linguists refer to the meaningful sounds of a language — its phonemes — they put them between slashes. So G and N, as English phonemes, are written as /g/ and /n/; in phonemic notation, "gnu" is written as /nu/ and "gnome" is written as /nom/. The phoneme /g/ is a consonant of the kind that's called a STOP because when you say it by itself it completely stops the flow of air through your throat and mouth; you can't say /g/ unless you put a vowel with it. The phoneme /n/ is a consonant of the kind that's called a NASAL because it involves the nose in an intimate fashion. Here are the complete sets of English stops and nasals:

Stops: /p/ /b/ /t/ /d/ /k/ /g/
Nasals: /n/ /m/ /ng/

[There's a fancier symbol for /ng/, but not all of your browsers would let me use it here, so we'll stay with /ng/; it's the sound — two letters, but just one sound — at the end of "swing" and "long."]

Suppose we look for some English words that have not just GN but other examples of a stop followed immediately by a nasal. Look in a dictionary, and you'll find a few. You'll find words like "knee" and "pneumonia"; you won't find any words like "bmag" or "dnak" or "tngoverly" or "lobn" or "matn" ... and so on. If I tried to convince you that there could be English words like "bmag" or "matn" or "bnadn," you'd flatly refuse to believe it. You know better, because the specifications for pronounceable English words are in your internal grammar, and those words don't fit the specs.

You have this all figured out by now, I suspect. [Often, when my own linguistics profs said that in class, I didn't yet have a clue, and I had to live with the awful idea that I might be the only person in the class who was still clueless. If you don't have this all figured out yet, you and I have something in common.] Clearly, the rule we need isn't a rule about G and N or about /g/ and /n/; it's a rule about stops and nasals. And one way to word it would be this:

Rule 2: No English word can have a stop followed immediately by a nasal.

This is a rule about sounds, remember. Certainly English words can have sequences of letters that put a nasal right after a stop — as in "gnaw, knot, sign," and many others we've been looking at. But when that happens, something has to give. "Something has to give" isn't elegant, but it accurately states the facts. Whenever a word of English would otherwise turn out to have a forbidden sequence of sounds, for whatever reason, something has to give.

One mechanism for that kind of giving is to do what we do with a stop followed by a nasal — we delete one of the two sounds from our pronunciation of the word. (Whether we then go on to spell the word in a way that matches the pronunciation is a separate decision and depends on other factors.)

Another way is to do what we do with lots of English plurals. English won't allow a nasal to occur immediately after a stop; it also won't allow any two of its hissing-and-buzzing sounds to occur one right after the other. If you want to talk about more than one beach (/bich/) you have to add an S to it to mark it as a plural, which would leave you with /bichs/ to pronounce. That's not allowed. Something has to give. There could be a rule that made either the /ch/ or the /s/ silent, and if you were analyzing another language you might find that option being used; English doesn't do it that way. Instead, it inserts a vowel — the vowel that sounds like "uh" — between the /ch/ and the /s/ to break up the forbidden cluster. The result — "beaches" — is a fine English word.

That's enough; we're not through, but it's enough for now. (Notice that you knew how to do all these things, even though you didn't know that you knew and you couldn't recite the rules.)

My thanks to all of you who've been sending messages about these lessons; special thanks to Douglas Dee, Aya Katz, Richard Kennaway, Ken Rolph, Sheri Wells-Jensen, Jim White, Elizabeth Barrette, and Sue Surova. I appreciate your help, and I look forward to your input about Lesson Three. Please e-mail me directly at

— Suzette Haden Elgin

Updated by webspinner