The Silence in the Recording Is Data

There are four seconds of nothing on the tape of Modesta Huayta's testimony, right after I asked what happened to her sister, and for years I marked that gap in my notes with a single word, pause, as if all silences were the same silence.

They are not. A four second gap before an answer about a missing sister is not the same event as a four second gap before an answer about the price of potatoes, even though a waveform editor will show you an identical stretch of flat line in both places. Treating silence as an absence of data, something to be trimmed for a cleaner recording or skipped past in a transcript, throws away one of the only pieces of a testimony that a person cannot fully control the way they can control their words.

Silence coding

I now assign every marked silence in a transcript one of five codes before I write another word of the surrounding text, a practice I call silence coding, and the code changes what, if anything, I am permitted to write into the gap on the page.

Code one is calculation: the speaker is retrieving a fact, a date or a number, and the silence ends the moment they have it. Code two is refusal: the speaker has decided, in that instant, not to say something, and the silence is followed by a change of subject rather than an answer. Code three is grief: the silence sits on a fact the speaker already knows perfectly well and does not need to retrieve, and what is taking the four seconds is not memory but the cost of saying it out loud again. Code four is translation lag: the speaker is composing the sentence in a second language and the pause is purely mechanical, carrying no emotional weight at all. Code five is listening: the silence belongs to me, not them, a pause I introduced by not filling the space with a follow up question, and the speaker used the room I left to keep going on their own.

The Modesta example

Modesta's four seconds, on relistening with the codes in hand, was grief, code three: her breathing before the gap was already unsteady, meaning she was not searching for the fact, she already had it, sitting on it. I wrote that silence into the page as a described pause, "she was quiet for a moment," which is honest and does almost no work. Compare a code one silence from earlier in the same session, when I asked what year her family moved to the valley: there her breathing was even, and the pause ended the instant a number surfaced, "1968." That silence needed no description at all, because a reader waiting for a date to arrive does not need to be told the speaker was thinking; the arrival of the number does that work on its own.

Getting the code wrong is not a small error. If I had written Modesta's grief silence as if it were calculation, "she paused to remember," the sentence would have implied she had forgotten something about her sister, when the opposite was true: she remembered everything, immediately, and the four seconds was the cost of choosing to say it to a stranger with a recorder running.

The pushback

The obvious pushback is that assigning a code to a silence is itself an interpretive act with no more authority than a guess, since I cannot hear grief the way I can hear a word, and dressing a guess up as a code with a name risks more false confidence, not less. I take this seriously, and my answer is that I do not code a silence from the silence alone. I code it from breath, from what came immediately before the question, and from what came immediately after the gap closed, three pieces of audible evidence, not intuition. I am also willing to leave a silence uncoded, and I have several in my files marked simply unclear, which I think is more honest than forcing every gap into one of five boxes because five boxes felt tidier than admitting I could not tell.

What silence coding gives me, imperfect as it is, is a reason to stop treating four seconds of nothing as four seconds of nothing. On a waveform they are identical. In a life, they were never the same four seconds twice.