The Curious Phenomenon of Stochastic Resonance

Here’s a curious little phenomenon. It’s called “stochastic resonance.” And it’s curious because we usually think of random noise as a bad thing. Noise is the nasty stuff that gets in the way of the signal, the stuff we want. But sometimes, under specific conditions, random noise helps boost the signal.

This phenomenon occurs in a few domains, from biology to ice ages. But it works in the visual domain as well and it’s definitely one of those “Oh…that’s kinda cool,” things when you see it. Here’s a mundane photo of some clouds through an airplane window:

A mundane photo of some clouds

We’ll clip the image with a threshold filter set to 128. Any pixel with a value less than 128 (down to 0) will be black, any pixel greater than 128 (up to 255) will be white.

Here’s the same image with the thresholding:

Original cloud image with thresholding

It’s tough to make out what’s going on in the image, isn’t it? In fact, if you didn’t know what the original picture was, you’d have a hard time making sense of its thresholded counterpart. Blobs of black and white. Most details that could clearly tell you you’re looking out an airplane window are lost.

Now let’s go back to the original image. This time let’s add some random black and white noise to it. This ought to do:

Original image with generous random noise applied

It doesn’t look that great, but now let’s put it through the same threshold filter. Once again, every pixel with a value less than 128 will be black, and every pixel with a value above 128 will be white.

Noisy cloud image with threshold process applied

That’s quite a difference isn’t it? The random noise before the thresholding has helped preserve details in the image that make it pretty clear we’re looking at some clouds through an airplane window. Neat!

The effect gets even more pronounced when you shrink the images and put them side by side.

The two thresholded images shrunk and put side by side. The difference is dramatic.

Stochastic resonance is well known in fields like biology, neuroscience, climate change, and of course signals. But I wonder what, if anything, it can teach us in the world of visual design?

At the very least it can help us ask some old questions through a new lens…

The recent design trend away from skeumoprhism and towards flat, minimalist design led by companies like Apple and Google certainly comes to mind. There is no direct evidence for it, at least not yet, but it’s worth asking whether the (non-random) visual noise imparted by skeumorphic visuals might help boost the signal?

Or maybe Apple and Google should just apply a random noise filter over their GUIs? I’m joking…


I’m writing a book about design in marketing. If you’re interested in the book, subscribe to my email list to receive the latest updates.