Yeah most workstation towers (i.e. the Mac Pro) have multiple graphics cards slots because a lot of pro apps use what's called nodes and hardware acceleration. Graphics cards are very good at rendering things such as HD video, 3D animation, etc. They're MUCH faster than the CPU which is what these apps would use in a more consumer orientated computer.
Hardware acceleration and nodes are able to use multiple GPU's (or in some cases, multiple computers altogether) in tandem with one another to display an image by nodeing out the multitude of tasks needed to display a fully rendered 3D character in Maya for example across multiple graphics cards so that one alone won't bear the burden by itself. The result is much faster render times. Although multiple graphics cards are bringing that 3D character to life inside the computer, only one card is needed to display the result of it on the monitor - the one your monitor is physically connected to.
Problem is that while games obviously support hardware acceleration, they don't however support nodes to divide it across multipile pieces of hardware so whatever card your monitor is connected to is the one that you're gonna see the performance of on the screen. Nodes is a feature only really seen in the professional applications field (Photoshop, Final Cut, Logic, Maya, etc).
When your hear about people gaming on multiple graphics cards, they're talking about something like this:
Radeon HD 7990 Dual GPU - Graphic solutions GeForce & Radeon
That is two graphics cards rolled into one. It uses only 1 PCI slot on the mother/logicboard and one monitor can get the net result of both cards.