Simultaneous recordings of one take at different exposures—underexposed, correct, overexposed. Maximum safety net for HDR or tricky lighting situations.
You're filming a scene with extreme dynamic range — backlight through windows next to shadowed faces, or you're unsure how the color grade will turn out later. Instead of risking a take and finding out afterward that the highlights are blown or the shadows are crushed, you shoot the same take multiple times: once underexposed, once correctly exposed, once overexposed. This is exposure splitting — and it's your safety net when the lighting situation is ambiguous or you need to remain flexible for HDR work.
The practice on set goes like this: You set up your camera on a tripod or gimbal, find the correct exposure for the most important information (usually the face or the main lighting situation), and shoot the scene or take. Then you repeat it with one or two stops darker — and then with one or two stops brighter. Or you vary the shutter speed if you're interested in motion blur. The three variants give you options later in the edit: You choose the best base during grading, or you blend them depending on the image area — for example, the sky from the overexposed version, the faces from the correct one, the shadows from the underexposed one.
This becomes particularly valuable for productions intended for HDR workflows (see also: HDR Grading, Log Curves). When you can't know the peak luminance of the monitor later or how aggressively the colorist will work, three exposures give you maximum freedom. Even with difficult natural light — a cloudy day that keeps getting brighter or darker — you save multiple exposure variations of the same take, instead of having to reshoot three days later.
The downside: You need more storage, more time for the same take, and in the edit, you have to organize the variations and, if necessary, register them pixel-perfectly. Some DoPs register the variations immediately with a clapper or timecode so the editor can sync them easily later. Work with your camera technicians and data management, not against them — who takes over the takes and how they are named determines your efficiency in post. With modern cameras that have a large dynamic range (Red, Arri, Sony), exposure splitting is less critical than with cameras with a narrower margin — but you should always have a safety net.