From Mary Lynn Garcia, author of Design and Evaluation of Physical Protection Systems:
"A complex question and I will try and explain as clearly as I can via written message!
So, I think the question is really how many monitors can an operator effectively observe; in a well-designed video assessment (not surveillance) system, there can be as many cameras as needed. What is important is how much video is presented to the operator to make a decision. When Sandia designs a system, we use 4 monitors, only 1 of which present video of the current highest priority alarm. The other 3 are used to present a live image of the highest priority alarm sector, the next highest priority alarm sector and a free monitor for operator choice. Note that an assessment system integrates sensors and video and uses a human operator to assess the causes of an alarm in real time.
As to a surveillance system (the operator is monitoring a number of cameras and sensors are not used), where an operator is observing live video from all cameras, and the images are usually scanning all cameras, it is harder to say. We do know that humans are bad at monitoring video over an extended period, as you note using the 20 minute number. This is because we have a limited ability to stay alert for a long time, especially when doing something as tedious as looking at a bunch of video images being presented all at the same time. By the way, that number is not the result of Sandia research but by a group of people from the UK (I think); recent research shows that the period of time is actually less, more like 15 minutes! This same group demonstrated that humans could successfully observe up to 9 large monitors BUT monitor size, distance of incident from camera (which is sort of resolution), duration of the incident, and disruptions such as telephone calls or other activity around operator (think a large control room) reduce operator effectiveness. And this is only for the 20 minutes noted. Other research has shown that the complexity of the video scene also contributes to operator ineffectiveness. This means that if there is a lot going on in the scene (maybe an exterior view of a bust street or a parking lot at shift change), and bad night lighting, it will be very hard for the operator to pick out anything of importance.
Think about it this way--suppose you are the video operator, sitting in a typical control room. There will be others in the room, most likely, doing their tasks (answering the phone, dispatching security personnel via radio, supervisors checking on operators, etc). the operator is generally seated looking at a bank of monitors, often above their head and in front of them by some medium at best distance. Then, the large monitors are broken into smaller, 16/32 or whatever, "screens" and video is canning on each screen at some rate, say 5 seconds per view. If there are 50 cameras and 16 screens displayed on a monitor, that means that it will take 3 times through to get all 50 cameras sequentially scanned onto the large monitor with the 16 "screens". if there is something happening in the scene, it will be hard for the operator to distinguish this due to all the factors noted above. if we combine all the factors (distance, resolution, complexity of scene, lighting, duration of event, disruptions and a 20 minute period of highest alertness), you can see it adds up to something not very effective. If we also consider that the operator is often told to "look for something suspicious" with little explanation as to what this means, it makes it very tough for the operator. Since as security people, we are primarily concerned with malevolent human acts of theft or sabotage, surveillance systems give us little "bang for the buck" in terms of preventing successful adversary attacks. There can always be some value in the recorded video for an after-the-fact investigation, but it is too late to protect the asset at that point. This is acceptable as long as the consequence of loss is acceptable.
Sorry this took so long, but the short answer is not many, and it will not be effective for a long time or for many events. If anyone in your class has a copy of the Design book (either edition), all of this is explained in better detail and with pictures in Chapters 8 and 9. Chapter 9 in particular gives details about how monitor layout can affect operators and gives some info on how many things humans can deal with at the same time."