Introduction

Reverse correlation is a method that is often used to characterize the response properties of neurons and to answer the simple question “What is the correlation between the response of a neuron and the stimulus used to elicit a response?” In any given sensory area within the brain, say, primary visual cortex, a given neuron will respond only to a restricted portion of that sensory domain, say, the upper left quadrant of the visual field. This phenomenon is captured by the notion of the receptive field, and our example neuron would be said to have a receptive field in the upper left quadrant of the visual field. In addition to spatial localization, receptive fields also describe the particular stimulus configuration to which the neuron is most responsive. Importantly, different neurons respond specifically to some stimuli and are indifferent or respond poorly to other stimuli. A visual neuron might be responsive to a small black spot on a white background, a line of a particular orientation, or a specific color, for example. Moreover, a given neuron may respond in opposite ways to two different stimuli, with different time courses. For example, a given cortical color cell might respond with excitation at a short latency and suppression at a long latency to one color and suppression at a short latency and excitation at a long latency to the opponent color. Thus, receptive fields have spatial and temporal structure. And a complete description of a receptive field includes both where in a sensory field a neuron is responsive to stimuli and to what stimuli the cell responds best. Although we will focus on visual neurons for this chapter, the concept of receptive field is useful in all sensory systems; for example, one can investigate the response properties of a neuron in somatosensory cortex to different kinds of stimuli applied to a specific region on the skin, the response properties of a neuron in auditory cortex to different frequencies, and the response properties of a neuron in olfactory cortex to different odors. In fact, reverse correlation techniques were originally developed for characterizing cells in the auditory cortex.^1,8

The problem of how to characterize a given neuron's receptive field is not trivial; in fact, it took several decades before neuroscientists appreciated that single retinal ganglion cells in the cat's eye respond only to small spots of light in a specific location of the retina.¹⁹ Previously, scientists had attempted without much success to elicit responses using full-field illumination, which they naively considered to be the “best” visual stimulus. We can sympathize. After all, what good would a visual neuron be if it could not report the difference between the room lights “on” and the room lights “off”? We now appreciate the sophistication of center-surround receptive fields and that these are a rather elegant way to encode the maximum amount of information (contrast borders) about the visual world with the smallest number of neurons—even if center-surround neurons do not respond very well to global illumination changes. In the primary visual cortex, the characterization of receptive fields was an even greater challenge. Having just established that retinal ganglion cells and cells in the lateral geniculate nucleus prefer small spots of light, scientists were rather discouraged to find that these stimuli were largely ineffective at driving cells in the next stage of visual processing, the primary visual cortex (V-1). It took an accident of experimental design to discover that most cells in V-1 actually respond best to oriented lines.^15,16 As David Hubel describes it in his Nobel lecture:¹³

For 3 or 4 hours [of recording from a single V-1 cell] we got absolutely nowhere. Then gradually we began to elicit some vague and inconsistent responses by stimulating somewhere in the midperiphery of the retina. We were inserting the glass slide with its black spot into the slot of the ophthalmoscope [used to stimulate the retina by projecting the spot onto a screen in front of the animal's eyes] when suddenly over the audiomointor the cell went off like a machine gun. After some fussing and fiddling we found out what was happening. The response had nothing to do with the black dot. As the glass slide was inserted its edge was casting onto the retina a faint but sharp shadow, a straight dark line on a light background. That was what the cell wanted, and it wanted it, moreover, in just one narrow range of orientations.

The experimental procedure employed by Kuffler and Hubel and Wiesel in their pioneering studies of the visual system can be characterized as follows:

Place an electrode into an Flash a Measure the neuron's

area of the visual system → stimulus on →? response and decide

and isolate a single cell the retina what stimulus to try next

As more and more different stimuli were tested, it became clear that cells are rather specialized, responding really well to some stimulus features (e.g., lines and edges) and not very well at all to others (e.g., different colors). Moreover, a given cell might respond best not just to lines and edges, but to lines and edges of a specific orientation. In fact, one might have one's electrode immediately adjacent to a cortical cell and not be able to elicit any response at all from the cell with a 45° line if the cell's preferred stimulus is an 80° line. Thus, in trying to fully characterize both the spatial and temporal response properties of a given neuron, one is faced with two problems: First, the particular stimuli that you choose to use will influence your decisions about what stimuli the cell is responsive to (you might mischaracterize an orientation-selective cell if you test it only with small spots), and second, it takes quite a long time to test a cell's response to every stimulus you can think of (as you would like to, so as not to miss the cell's actual stimulus preferences).

Reverse correlation methods overcome both of these problems, though they have limitations, which should be appreciated so as not to misinterpret the results. (We will discuss these at the end of the chapter.) Moreover, in addition to providing a full characterization of the spatial and temporal response properties of single neurons, reverse correlation methods have also proved useful in characterizing the receptive fields of multiple neurons simultaneously.⁹ We will first describe this powerful technique in general terms, as it is used to study the visual system, and will then use some specific examples to outline the limitations.