Channels Panel

The Channels panel (Window → Channels) breaks the active document into its colorimetric components. At the top is a composite row showing the whole image, followed by one row per channel — R/G/B, C/M/Y/K, or L/a/b depending on the document's color mode — plus any alpha channels you've added.

Channels are data, not pictures

This is the idea the whole panel is built on. A channel is a single number per pixel: the red channel is "how much red light," the cyan plate is "how much cyan ink," the L channel is "how light." A channel on its own is not a color — it's a quantity. You only get a color you can show on screen when you combine all the channels and interpret them through the document's color profile.

So the panel really has two kinds of view: the composite is the picture (all channels, run through the profile, converted for your display), and a single channel is the raw data. Every rendering choice below follows from keeping those two jobs separate.

The composite row

The composite row always shows the full image, regardless of which channels you've toggled on or off below it. Toggling channel visibility affects the canvas and the individual channel rows — not this thumbnail. It is rendered in the document's own color space and converted to your display through the active color profile (or OCIO display), exactly like the main canvas.

"Show Channels in Color"

This setting changes how the individual channel rows (and the canvas, when a single channel is isolated) are drawn. It is not cosmetic — it is a real difference in what the pixels mean. In Photoshop it lives in Preferences; here it is a control on the panel itself, because that's where it's actually used.

Setting	What you see	Profile applied?
Off (default)	The channel's raw value mapped straight to gray: 0 is black, maximum is white. The number is the gray level.	No — plain grayscale, no color management at all.
On	The channel tinted in the color it represents — the red channel in red, the magenta plate in magenta — isolated through the profile so an ink shows in its reproduced color.	Yes — that one component, run through the profile.

Why the grayscale view uses no profile

The obvious question is: why show the raw channel as gray, instead of running it through the profile and then converting that to grayscale? The latter sounds more "correct," but it's the wrong thing for what a single channel is. Three reasons:

1. A lone channel has no color to convert. A profile is defined on complete (R,G,B) or (C,M,Y,K) values. A single channel is one axis of that. To "run it through the profile" you'd first have to invent a color like (R,0,0) — an arbitrary choice the profile was never meant to interpret.
2. It would break "equal data looks equal." If you profiled (R,0,0) and then desaturated, the gray you'd see would be the perceptual lightness of pure red. Red and blue aren't equally light, so a value of 200 in the red channel and 200 in the blue channel would show as different grays. You could no longer compare two channels, judge which region is darker in the actual data, or trust that what you see is what you're editing.
3. Editing acts on the numbers. Curves on a channel, building a mask, per-channel noise reduction, frequency separation, hunting compression artifacts in one channel — these all operate on the stored magnitudes. The display has to be a faithful, undistorted map of those magnitudes. Color tinting actively hurts this, which is exactly why retouchers inspect and edit channels in grayscale.

So which view is "right"?

There isn't one correct view — there are two different jobs, and the setting picks which one:

View	It's the…	Use it for
Grayscale (no profile)	editing / measurement truth	Working a channel: masks, curves, comparing, measuring. (Almost always.)
In color (through profile)	intuition / soft-proof view	Understanding a channel's contribution, or eyeballing how an ink reproduces.

It defaults to raw grayscale because the dominant use of a single channel is as data to be shaped, not as a thumbnail to admire.

Channels as a workspace for masks

Because channels (and alpha channels) are raw, profile-free grayscale signals, you can treat them as generic image data and run ordinary processing on them to build selections and masks: copy a color channel into an alpha channel, run blur / curves / HSL / high-pass / levels on it, then load the result as a selection — the classic recipe for luminosity masks, color-range masks, and edge masks. This whole family of techniques only works because the channel is the raw signal; if the view (or the data you filtered) were color-managed first, the mask you're sculpting would be subtly wrong. The profile-free grayscale view isn't a limitation — it's what makes channels useful as a computation substrate.