Signal and Noise: Stimuli


Before we can develop an understanding of Signal Detection Theory, we need a signal to detect!

What’s a signal? Well, it could be your friend’s voice at a party, or a distant boat on the sea, or the smell of truffles under the earth. Any stimulus that you are trying to perceive can be thought of as a signal.

Our signal will be coherent motion in a random-dot kinematogram (RDK). Go ahead and see what it looks like:

Push the Run button to display the signal. Push the Pause button to temporarily freeze the display. Or push the Reset button to reinitialize the demo with a new direction of motion.

This is coherent motion because the dots are moving together in the same direction. Another way to describe this is to say that the signal is present.

Perhaps you are beginning to wonder why certain words are highlighted with color. Throughout this site, color is used to represent key concepts. We’ve just met our first: yellow used to represent the concept of a signal being present. As you move on through the site, you’ll see yellow recurring in various contexts that relate to a signal being present. Other colors will follow. Head to the Legend to see a listing of how color is used.


Now, when there is only signal, detecting it is not much of a challenge, but we typically also have noise. Noise is the music and other voices at the party, the swirling clouds of fog on the sea, the mélange of other smells wafting up from the soil. All of the other stimuli you are not interested in, but that are interfering with your ability to detect the signal.

Our noise will be random motion of the dots. Give it a try:

Noise is incoherent or random motion. Every dot is moving in a different direction. When there is only noise, and no signal, we say that the signal is absent.

Signal and noise

In many situations, the signal we are trying to detect is buried in noise: your friend’s voice among the music and other voices at the party, the distant boat among the clouds of fog on the sea, the smell of truffles among the other smells of soil. The stimulus you are trying to detect, the signal, is buried amongst many other stimuli, the noise.

Our signal, coherent dot motion, is mixed with noise, random dot motion:

Here, half the dots are moving coherently in the same direction — the signal, and half the dots are moving in different, random directions — the noise.

Variable signal strength

How strong is the signal compared to the noise? That can vary continuously from “pure” signal at one extreme to “pure” noise at the other. And this brings us to one of the motivations for using RDKs as stimuli: by varying the proportion of dots that are moving coherently between one (all the dots) and zero (none of the dots), we can elegantly vary our stimulus from “pure” signal to “pure” noise, or anywhere in between:

Move the Coherence slider up and down or edit the number to adjust the proportion of coherence — even while the dots are moving!

Moving forward, we’ll describe stimuli with any signal at all as having signal present, and stimuli with only noise as having signal absent.

Random-dot kinematograms, also called random-dot motion displays, are widely used to study motion perception (e.g. Britten, Shadlen, Newsome, & Movshon, 1992) and there are many variations on the basic concept (Pilly & Seitz, 2009). For us, they provide a nice parametrically variable stimulus!