What Is an Accent

Contents

1. The picture under the word
2. Three cases the picture cannot hold
3. The neighbours
4. What the foundation holds
5. A departure, not a feature
6. The speaker hears it last
7. Clear and heavy
8. Changing an accent
9. What it costs
10. An ordinary Tuesday

Ruth grew up in Leeds and has worked in Houston for six years. On a Monday she tells her team that the shipment is at the dock, and Tom, who was born in Houston, asks her to say it again. She says it again. He nods and writes it down. At Christmas she goes home, and her sister Beth says, before the coats are off, “You’ve gone American.” Ruth says she sounds exactly as she always has. Both of them mean it.

The plain picture of an accent is a sensible one. An accent is a way of speaking that goes with a place or a first language. A person carries it in her mouth, the way she carries a height or an eye colour, and others can hear it. Some people have a strong one, some a light one, and a few have none, which is what people mean by speaking neutrally. The picture is good enough to sort a room, and it is the picture most of us use when we say that someone “has an accent” as though it were a thing she was holding.

Underneath it is a container picture. The accent sits in the sound, and a listener detects what is there. If that were so, two competent listeners given the same sentence would agree about how much accent it held, and a speaker would be able to hear her own with a little care. I want to see what the picture commits us to, and I will start where it fails.

Three cases the picture cannot hold

The seam runs through the same place in all three. The word accent is said of the voice, and what it reports changes when the listener changes, and changes again when the listener is the speaker. A feature of a sound would not behave like that. Something about the listener is in the report.

The neighbours

Several earlier pieces stand close, and a reader could think I am repeating one of them, so I will say what each owns and what differs.

What Is a Wink is the nearest, and it was published yesterday. It owns one mark and several readers, and it splits the readers on the host’s relation to the movement: made, or only occurring in its site. Here nothing of that kind is in doubt. A sentence is plainly made. The split lies entirely in what each listener already holds, and the new matter is the case where the quotient sits near one and the sound goes unremarked. What Is a Language owns a vocabulary as named conditions, and it names an accent in its care passage as a thing a person is judged by; it does not take it as its subject. What Is a Reputation owns the point that the value of a mark is set by the reader’s prior, for a person’s name. I use the same direction of reading, on a feature of a sound. What Is a Hug and What Is Revenge own the author’s smaller reading of an act. I use it in one section, to say why Ruth is the last in her room to hear herself, and I use it as the foundation states it.

What the foundation holds

Six statements do the work, and I use them as the foundation states them. First, the human experiences only the quotient of the actual over the expectation, and meets neither term alone. Second, the expectation comprises a subconscious prediction, which is on the real side, and ideas, which are on the imaginary side. Third, the foundation’s definition of surprise is the logarithm of the magnitude of the quotient, signed. Where the magnitude is one, that value is zero. The host is there to read it, so the value is defined, and this is not the undefined-at-the-boundary case.

Fourth, records are write-protected: a record updates expectation going forward and does not re-open. Fifth, from the foundation’s section on the host’s own act in its own denominator: prediction is retrained by the record and by nothing else, and the host’s own acts are among the records, so the prediction holds the host’s own making by kind. An act the host makes lands on a prediction trained on her own making of such acts, and the same act received lands on one trained on no such thing. The foundation gives a direction and no amount. Sixth, every artifact is a miss against its exemplar, and the evaluative question is fidelity and not success.

A caution about the fifth. The foundation says that the maker’s reading is smaller, and it says in so many words that this is structural and is not a bias, a distortion or a fault. I keep to that. Speech is an act, and what it leaves is a sound received by whoever is there, so the sentence applies. The foundation’s laboratory support concerns finger force and touch and does not mention voices, and I do not borrow it for this.

A departure, not a feature

Take the sentence Ruth said on Monday, “The shipment is at the dock.” It is one set of sounds. Tom’s prediction was trained on the speech around him in Houston from his earliest years. It holds, by kind, how that sentence goes. Ruth’s version departs from it, and he reads the quotient. Departure is the name for the size of that gap, and he reads it as a small surprise, which is why he asked her to say it again.

Now take Callum, a fortnight later, who has a record of Glasgow, a record of the film and television that carries Texas speech, and two weeks of the office. His prediction for that sentence is not Tom’s, and his quotient is not either. Beth in Leeds holds a record of Ruth, aged fifteen, saying nearly the same words, and she reads Ruth’s present speech against it. Each of them meets a quotient. No listener meets the sound by itself. Whether the sound is called an accent is settled on the side of the denominator, in how far this ear’s training lies from it.

Figure 01

One sentence, four listeners

Listener	What the prediction was trained on	What she or he reads
Tom, in Houston	A lifetime of speech around him there, and no Leeds.	A departure, small to moderate. He asks once.
Callum, two weeks in	Glasgow speech, screen Texas, ten working days of the office.	A smaller departure than Tom’s, in a different direction.
Beth, in Leeds	Ruth at fifteen, and the town around them both.	A departure from Ruth’s old speech: “gone American.”
Ruth, in either room	Her own making of these sounds, for decades.	Near one. Unremarked, in both rooms.

The people are composites and the readings are the argument. No measurement of anyone is reported here, and the sizes are directions and not amounts, since the foundation states no scale for how far two listeners’ denominators lie apart.

A feature of the sound would not distribute like that. A feature is the same for each listener, and a departure is a relation, so it differs with the listener’s record. That is what the case of Tom answers. His report, “I have no accent,” is true of his quotient with the speech around him, which is near one, and surprise at one is zero. A listener from Dublin hears something in the same sentence because her record is different. The mouth is the same and both are right.

One thing I do not say. A set of listeners in one town may hold very alike predictions, and I would like to write “the town’s expectation” and be done. The foundation has no expression for an overlap between two hosts’ expectations, so I speak of each listener’s record and let the likeness be a fact about many records. The gap is recorded in the foundation’s own findings, and I leave it where it is.

The speaker hears it last

Now Ruth. She makes the sounds. By the foundation’s sentence, an act the host makes lands on a prediction trained on her own making of such acts, and the maker’s reading is the smaller. So her own speech reads small to her. This is not a failing. She has made these vowels for decades, and her prediction holds them.

Put the two factors together. When Ruth is in Leeds, she is among listeners whose records are near hers, and both factors are small. When she is in Houston, the room’s records are far from hers and the first factor is large, while the second stays small. She is in the one position where the room hears her most and she hears herself least. That is the answer to the second case. The small surprise she feels when she is asked where she is from is a mark of her own landing as news.

How does she learn of it at all? By marks. A repeated question is a mark in her record, and so is a smile, a guess at her origin, a “say that again.” Each is the record of a listener’s quotient, arriving in the only form that could reach her. She reads them through her own prediction, and by the foundation’s sentence on records, they train it forward and do not rewrite how she sounded.

Clear and heavy

The two interviewers can now be read. The one who writes “very clear” has a prediction close to the candidate’s speech, and the quotient is near one. The one who writes “heavy” has a prediction further off, and the quotient departs. Each word reports the writer’s quotient. Neither reports how the candidate’s speech stands to anything outside the room.

That matters because the second interviewer has a cost: she lost a sentence. The loss is hers, and it is a loss in the plain sense that she could not follow. I leave that intact. But a lost sentence is a failure of the reading and carries no verdict on the candidate’s fidelity to anything. Fidelity, in the foundation, is the nearness of an artifact to the exemplar of a condition, and the question worth asking of a made thing is how near it came. An utterance is an artifact, so it is a miss of the exemplar of what it says, in any accent. The foundation names an exemplar for a condition. As I read it, it names none for a sound kind, and so no sound kind sits nearer than another.

Changing an accent

People do change how they sound, and the foundation says how it can go and how it cannot. It cannot go by editing the past, because the sounds Ruth made at fifteen are in the record and stay there. It can go by new marks. Each month in Houston adds speech to the records of the people around her, and adds speech from them to hers. Her own prediction retrains by her own making, and so slowly drifts toward the room’s. Tom’s retrains on her. Both move, and neither controls the other’s. She has no hand on his denominator, and she can add a mark to his record and nothing more.

I cannot say how fast. The foundation supplies no rate for the update to the expectation, and the piece does not pretend it does. Studies of listeners adapting to an accent report change within a short exposure, and I report that and do not lean on it. The only certain thing is the direction, and the other certain thing is that she will be the last to find out that she has drifted, since her own reading moves with her.

What it costs

The first cost is that a speaker’s own ear is the weakest instrument for the question. A friend can report what she hears, and what the friend reports is her quotient. There is no listener whose report is the accent itself, and so there is no standard to pass or fail. The word “neutral” holds a position, usually the position of whoever is speaking it.

The second falls on the speaker. Ruth cannot set the room’s expectation before she speaks, and the foundation says plainly that no host has a hand on her own denominator. She can repeat herself, write the figure down, speak more slowly. Each is a new mark. Whether it is worth making is a decision she is the site of (R13), and the cost of being asked again is hers to bear regardless of whose quotient it was.

The third is that this gives nothing to anyone who is judged. A listener who decides, from a sound, that a speaker is less able has made an act, and the act is the listener’s and lands as a mark in the speaker’s record. What the argument removes is an inference, from how a person sounds to what she knows. It does not undo being passed over.

An ordinary Tuesday

1. When you lose a sentence, say that you lost it. “I missed that, once more” reports your quotient, which is where the departure was. “Your accent is hard” reports it as a fact about her mouth.
2. Give a new voice a little time before you decide how hard it is. The next hearing is read against a prediction that has already moved. This is reported, with no rate from the foundation.
3. If you are the one who is asked to repeat, do not read it as a verdict on what you said. A repeat is a listener’s quotient and a sentence you can offer again.
4. For anything that matters, add a mark that does not pass through the ear. Write the number, the name, the address. A mark on a page is read by listeners at their own pace, and it stays.
5. Do not wait to hear your own. You are the worst placed to. If you want to change it, ask someone you trust what they hear, and take the answer as theirs.
6. When you give feedback, write the sentence you lost. “Heavy accent” says where your ear was. “I lost the third answer” says something that can be fixed or checked.