Stimulus generalization is obviously beneficial, but it can be carried too far. It’s sensible for you to feel fear when someone wearing an angry expression approaches you; and thanks to generalization, you’ll feel this fear even if the person’s face is a bit different from angry faces you’ve seen in the past. But if you generalize too much, you might feel fear in response to other facial expressions—and so end up being afraid in many social situations. What you need to do, therefore, is discriminate—respond in a way that’s guided by the stimuli in your view.
The phenomenon of discrimination is easy to demonstrate. An experiment might, for example, use a loud boat horn as the US; this stimulus reliably produces a startle response when it’s sounded. In the first part of the procedure, a red warning light is paired with the boat horn, and after a few pairings, this stimulus will reliably produce a conditioned response—you’ll tense your muscles whenever the light appears. Once this CR is established, we proceed to the next step: Now we intersperse trials pairing the red light + horn with other trials in which a new stimulus—say, an orange light—is presented with no US (no boat horn). In this setup, you’re likely to generalize at first by tensing up in response to both the red light (technically referred to as the CS+) and the orange light (the CS–). As the training continues, however—trials pairing the red light + horn, mixed with trials presenting the orange light alone—you’ll learn to discriminate and will cringe only when you see the red light.
In a discrimination procedure you learn, of course, that the CS+ signals the approach of the US. What about the CS–? In our example, we might think you’re learning nothing about the orange light because it’s never followed by the horn, and for that matter never followed by any kind of US. Even so, the CS– (again, the orange light) does provide information. When it arrives, this signals a period in which the US is likely not to arrive. If the US is the horn, then the CS– indicates that the loud noise is not coming soon. If the US is food, then the CS– signals the start of a period in which no food will be available.
Essentially, then, CS– takes on a meaning opposite to that of the CS+. It means “no noise,” or “no food,” or, in general, “no US.” Correspondingly, the animal’s response to the CS– tends to be the opposite of its response to the CS+. If the US is a noise blast, then the CS+ elicits fear and the CS– seems to inhibit fear—and so the animal is calmer in the presence of the CS– than it would be otherwise. If the US is food, then the CS+ elicits salivation and the CS– causes the animal to salivate less than it ordinarily would. Overall, the CS– takes on the role of inhibitor: Whatever the response produced by the CS+, the CS– makes that response less likely.