There are several schools of thought when it comes to race performance and pacing:
Some running enthusiasts and coaches consider it desirable to run a race with evenly paced splits. This school of thought most likely originated in the world of track and field, where feedback may exist every 400 meters (or more frequently, depending on whether the track is indoors or out) of the race distance. Races with "rabbits" hired or invited to set the goal pace for the leading runners can serve as a benefit or a drawback, depending on the racing conditions...or the condition of the athletes. But the goal more often than not is to bring the lead runners through an intermediate point of the race at a consistent and fast pace, conserving enough energy to finish in the desired time.
The tactical (slow!) nature of the first laps of most championship-level races (N.B. In the last forty Olympic Summer Games athletics distance event finals the world or Olympic record has been broken in the 800, 5,000 or 10,000 meters eight times; three times in the women's events. A single runner has broken an Olympic record at two different Games!) is evidence of a second school of thought when it comes to pacing, that of racing with descending splits.
Watch the first laps of a college championship 5,000-meter track race and the pace appears pedestrian, even to the casual jogger. Some of that mis-perception may have much to do with camera angles - unless it's one of those rail-mounted ones which roll at speed it is impossible to tell how fast the runners are truly moving, but when you do the math, the difference in splits between the first 800-to-1,200 meters and the last are the difference between citizen-athletes pushing the envelope at the start and greyhounds letting it all hang out for the tape.
The runners at "the pointy end of the spear" going for the overall win at a race engage in pace tactics which by nature need to be flexible: How fast are you and your fellow competitors going out? Who's making a move at what point? Is there enough real estate (and heart, lungs, leg strength, energy, etc.) between where you are now and the finish line to catch anyone who might be ahead, or (as in the case of Meb Keflezighi at the Boston Marathon) to stay ahead of anyone who might be coming from behind?
For the majority of citizen-athletes I believe the race distance serves to determine the pacing tactic which ought to be used. Shorter distances, such as the 5,000 meters, are served well by either strategy. When it comes to the marathon I strongly recommend the negative split focus; a citizen who decides to push the pace too hard (even as little as five seconds per mile too quickly) at the front end of a marathon may be in for a very uncomfortable - and very long - last fifteen kilometers.
There are few substitutes to running many miles worth of workouts at varying paces to learn how a particular pace feels. After ten-plus years of track workouts a six-minute-mile pace feels a particular way; a seven-minute-mile pace feels a different way. Treadmill running, for all of its perceived disadvantages, encourages (enforces?) pace discipline. A runner who is not neuromuscularly prepared to run a particular pace on a treadmill runs the risk of being ejected off the back of the treadmill deck (a source of comedy entertainment on the Internet), injuring themselves (a source of income for emergency room physicians), or both.
There are less painful ways to encourage pace discipline, however.
Eighteen months after the initial achilles' tendon issues, I run on the average two out of every three training miles on a treadmill. Another benefit of treadmills (especially for brittle coaches/runners) is the ability to control variables. If something feels wrong on a run the runner can choose to end the workout rather than try to push through pain (after almost 20 years one tends to know when it's discomfort) and shut everything down.
A quarter mile into last weekend's Fiesta of Five Flags 5K, it was pretty cute to hear a young lady say to her companion, 'oh, my, our pace is 6:18...this could be trouble...' When I decided to return to short distance racing, I retained the "control the variables" concept. Could I go out the first mile at 6:18 and watch the wheels fall off at mile two? Yep. Did that the previous weekend. Not pretty in the slightest.
Many citizen-athletes swear by - and as I have said on many occasions, often trust too much without knowing the technology - the consumer-grade GPS receiver. The last two receivers I've used (Garmin) have a "virtual pacer" function; the user sets a desired pace and can receive feedback whether they are ahead or behind, both in terms of time and distance. I first tried it last year, with satisfactory results, when Suzanne and I ran together (for most of the) Rock n' Roll Virginia Beach Half Marathon.
The virtual pacer is a more-gradual form of feedback when compared to the near-real-time "stopwatch and pace" output most runners tend to use. The GPS receiver will pick up the signals every two-to-three seconds, and even the most consistent of runners will receive inaccurate information - radio frequency interference, buildings, cloud cover and the U.S. government - a ballpark figure of how fast you moved since the last receipt. You might see a 6:18 at one point, and a 7:15 at another.
This is a case where all I wanted to know was how close I was to 7:00 pace, and the Garmin pretty much did its job. Once I got to the third mile split and saw I was five seconds faster than my desired time I knew all I had to do was push the last 187 yards to the finish. Personal best? Not at all...I have a solid minute per mile to go before I even approach that...and it might never happen. But it was the first time I've run three consistent one-mile splits for a 5,000 meter race on the roads.
Data is good when you race. Make certain you're receiving the right kind of data.