Pre employment testing is much akin to drawing blood for a medical test -- you don't need to exsanguinate a patient in order to check their platelet count, and you don't need to drown an applicant in extensive pre employment tests in order to establish that they're the right applicant for you. Some companies out there are subjecting every applicant to multiple hours of psychometric testing, which is not only expensive for the company, but has been proven to not work all that well. One study looked at the number of actual correlations there were between job testing and subsequent performance, and found that a perfectly valid assessment could be captured in only one to one and a half hours.
Of course, all of that can change fairly dramatically depending on the specific job role and the anticipated number of applications. If you're looking to narrow a field of dozens (or in this economy even hundreds) down to a few that you want to look more deeply at, giving them a quick Big 5 personality test and a brief cognitive test can take about a half hour per candidate and will rapidly tell you which are the ones you might want to test more extensively.
On the other hand, if you're looking at the top 4 or 5 candidates for an executive position, you'll want to have them fill out an in-depth series of assessments before the interview, so you can go with premeditated questions and a clear picture of what else you need to know about this particular applicant.
The basic question, then, is where to draw the line. In order to answer that, it should be pointed out that no amount of psychometric testing is going to suffice unless it's properly paired with a strong behavior based interviewing technique. These tests essentially serve, to return to our medical analogy, as a nurse that comes in and gets your vital statistics and basic information to turn over to the doctor -- or interviewer -- who then makes the final decision.
Understanding that, the line should be drawn at the point of diminishing returns: the point at which each successive test becomes noticeably less useful to the interviewer. Experts seem to generally concur that this happens somewhere between the second and the fifth test depending on the specifics of the job and the pay grade.