If you look at the Phase Response Curve you made you see that, as you follow the curve through the 24-hour cycle, you first encounter a dead zone during the subjective day (VT0 - CT 12) during which light pulses exert no or little effect on the phase of the clock. The line, then, turns down (negative slope) into the delay portion of the curve until it reaches a maximal delay in the early night. It reverses its direction then and goes up (positive slope) until it reaches maximal phase-advances in the late night. Finally, it falls back down again (negative slope) until it jons the X-axis again.
Let's say that, in your experiment, you have used light pulses that were 6 hours long (duration), 100 lux strong (intensity), and containing the full spectrum of visible light (quality). The PRC will tell you how your animals would entrain to a murky, cloudy mid-winter day in high latitudes.
If you used 14-hour pulses of white light of 2000 lux, you would have built a PRC describing entrainment to a nice clear summer day in North Carolina.
These two PRCs would look somewhat similar to each other. All Phase Response Curves are qualitatively the same. There are only quantitative differences.
Here is an example of two human PRCs generated with two intensities of light:
Keeping duration and intensity constant, but systematically varying quality - using for instance, blue, green, yellow, orange and red light pulses - will result in a series of PRCs of similar shape, yet the sizes of phase-shifts would differ. This series of PRCs can tell us about the spectral sensitivity of the photoreceptive pigment involved in the transduction of light information from the environment to the clock. For instance, strongest responses to orange-red portion of the spectrum suggest rhodopsin or a similar pigment (e.g,. melanopsin). On the other hand, peak response to the blue light suggests a pigment like cryptochrome.
Systematic varying of either duration or intensity of light would also result in construction of a class of PRCs. As intensity (or duration) increases, the sizes of phase-shifts (both advances and delays) also increase (and vice-versa also holds: I have seen nice, low-amplitude PRCs to light-pulses measured in miliseconds). At the same time, you will notice a couple of other things: first, the dead zone is getting progressively narrower, and second, as you look at your raw data, you will notice that fewer and fewer days of transients are needed for the rhythm to achieve the new steady-state after the perturbation by light.
As you keep increasing intensity (or duration) of the light pulse and producing new PRCs, there will be a point - a treshold - at which there will be no more dead zone and no more transients. At this point, each light-pulse elicits an immediate large phase-shift, some as large as 12-hour shifts. It becomes impossible to differentiate between advances and delays (transients used to be a guide - but they are gone now), so the convention is to plot all the data as phase-delays.
This treshold, explained by some elegant yet complex math by Arthur Winfree in a series of books and papers, denotes a switch from a Slow-Resetting PRC (Type I PRC) and Fast-Resetting PRC (Type 0 PRC). The point (of intensity or duration) at which Type I turns into Type 0 PRC is species-specific. Mammals, especially rodents (burrowing nocturnal non-migratory animals) tend to have very small shifts and it requires enormous amounts of light energy to make the switch from Type I to Type 0 PRC. On the other hand, relatively weak stimuli result in Type 0 resetting in a number of plants, fungi and prostists, as well as in some migratory animals, e.g., Japanese quail. Thus the shape and size of PRC can tell us something about the phylogeny and ecology of the organism we are studying.
Phase-Response Curves to other stimuli (e.g., dark pulses on a constant light background) tend to have a different shape. Here is a paper that contains some PRCs to various chemicals, including melatonin, serotonin and glutamate. Notice how the curve for glutamate closely matches the curve for light, suggesting that this neurotransmitter may be involved in transmission of light information from the retina to the clock.
Notice also how few data-points were neccessary for the completion of the glutamate PRC. Once the PRC to light has been generated for a particular species, further (more expensive and involved) experiments can be done by applying stimuli (e.g., chemicals) at only three time-points: the phase of greatest advance, the phase of greatest delay, and the dead zone (as control). In experiments performed in this manner, it has been reported that some neurotransmitters and neuromodulators are involved only in phase-delays and others only in phase-advances. This shows how formal analysis aids in the study of underlying physiological mechanisms.
Next: how to use a PRC to study entrainment.
Does an organism have to be free running before you can demonstrate a phase response?
No. There are eight methods for deriving a PRC. The usual one is to have organism freerunning in constant conditions (usually constant darkness) and apply light pulses. But other methods involve entrainment, followed by step-wise phase-delays or advances of the entraining cycle (requires that the organism shows transients), or entrainment followed by a pulse followed by a freerun. I have published a paper that demonstrates that one of the methods - entrainment to non-24h cycles - provides the exact same PRC as the standard method, thus validating this alternative method.
I described my paper and the entire issue of alternative methods here.