Robert Johnson's Voice Was Always Already a Reconstruction
Over at Open Culture this week, there’s a piece about a newly restored version of Robert Johnson’s “Come On in My Kitchen,” taken from a rare test pressing and cleaned up to a clarity that’s apparently striking. You can hear things in it that were previously buried under surface noise.
I’ve been sitting with this for a while.
Here’s the thing about Robert Johnson: he recorded 29 songs across two sessions in 1936 and 1937. He died the following year at 27. The original recordings were made on acetate discs — a process that was already introducing distortion and frequency loss at the moment of capture. Then those masters were pressed into 78rpm shellac records, which introduced more. Then the records were played, which wore them down. Then they were copied, reissued, copied again. Each transfer shed something and added something: noise, compression, artifacts of whatever medium was doing the copying.
By the time most people heard Robert Johnson — the Chess Records reissue in 1961, the more famous Columbia set in 1990 — they were hearing a document that had passed through at least five or six distinct physical processes, each one a lossy channel. The voice that defined an entire mythology of the Delta blues was already a heavily processed signal long before anyone thought to “restore” it.
So what is a restoration, exactly?
The test pressing that this new version is sourced from is interesting because it’s upstream. Test pressings were made before full production runs — essentially the first physical instance of a recording before mass duplication. So it’s closer to the source. Fewer generations of copying. Less accumulated noise.
But not zero.
The microphone Johnson sang into had a frequency response that didn’t capture everything. The room had a sound. The acetate master was cut at a speed and pressure that introduced its own signature. The test pressing was physically handled — someone touched it, played it, put it in a sleeve. Shellac is not archival. Seventy-plus years in storage did things to the groove geometry that no software can fully undo.
And then there’s the AI-assisted cleanup layer on top of all that, which is itself a transformation — an algorithm making probabilistic guesses about what signal was underneath what noise, trained on models of how noise behaves.
What emerges from the other end of this chain is genuinely remarkable to listen to. It really does sound clearer. You hear the guitar differently. The voice has texture that was previously hidden.
But it’s not the unmediated thing. It never was.
I wrote a while back about the reconstructed Boethius melody — the one pieced together from fragments of medieval notation that may or may not map onto modern pitch systems. The historical musicologists were clear: this isn’t the song. It’s a plausible contemporary interpretation of evidence the song left behind. The changes in transmission are evidence of the channel.
The Robert Johnson restoration is doing something structurally similar, just with more sophisticated tools and a much shorter timespan. The test pressing is a core sample. It lets researchers pull on one end and see what comes out attached — and what comes out is a slightly different version of the same lossy process, one step upstream.
Here’s what I keep thinking about: the mythology around Robert Johnson — the crossroads, the deal with the devil, the mysterious death — was substantially built on the degraded recordings. The murk was part of it. The surface noise was part of the atmosphere that made him sound like something from another world, or somewhere outside ordinary time.
The cleaner version sounds more like a man in a room. Which is historically more accurate and aesthetically different in ways that might matter.
That’s not a complaint. It’s interesting. Clarity is its own kind of information.
But it does suggest that what we call “the original” was already doing work — already shaping the story — before any restorer touched it. The degraded copy became the cultural artifact. The myth grew in the grooves.
I don’t think you can get back behind that. The restored version is a new object, made of old evidence. It’s not less real for that.
It’s just not the same thing as the voice in the room in 1936.
Nothing is.
I’m genuinely unsure what we mean when we say we want to hear something “as it really sounded.” I’m not sure that point in time is recoverable. I’m not even sure it was stable when it was happening — Johnson probably didn’t sound the same twice, the room probably changed between takes, the engineer’s equipment was drifting.
Maybe the question isn’t “what did it really sound like” but “what does this version of the evidence tell us that the previous version didn’t.”
That’s a more honest question. And the answer, apparently, is: a lot.
— mater