Actually, the game keeps track of this for you. Any dwarf born or who immigrates to the site will have a persistent prefs page, even after they die.
This means that any non-experiment fortress can be datamined for that information, if the overseer(s) in question feel compelled to do so. (you can see how unconstrained breeding swirls traits around perfectly well with that info)
This experiment does not use a control, because it is more of a raw data gathering experiment, and keeping the control isolated is too difficult. That is to say, the intent here is to measure the effect of culling, and to track rates of culling over several generations for phases 1 and 2 of the project.
The reason that such information is useful, is, among other things, that it provides a mathematical baseline for predicting the timetables of eugenics projects, allowing a cost-benefit analysis. It also will answer basic questions about dwarven inheritence, such as the previous question about Darwinian-vs-Lamarkian inheritence. If you can pen up your proletariat beating on wild giant armadillos, and get magically, and profoundly superior baseline dwarves from the practice, that alone is worth something to know about.
This experiment is simply to get raw data, for the most part. Not explicitly to test a hypothesis, though there is in an intrinsic one, in that culling *should* result in multi-generation improvements over time. However, again, I dont really need a control to show if this is the case or not. Here's why:
Let's say that inherited traits are PURELY random, with no inheritance whatsoever. The rate of culling will be a linear line. It will never deviate, despite having been done for generations. The rate of negative traits spontaneously appearing will likewise be flat, etc. (sure, it will be a little jaggy due to the RNG being involved, but the overall confidence zone will occupy a flat horizontal bar on the plot.) 50 generations should silence most noise, and 100 generations should silence nearly all noise. (this project will take a *LONG* time. to expedite the process, I have set baby:0 and child:1. I may need to set child:0.)
However, should there be an inheritance model of any sort, the culling represents a disruption in the equilibrium, and will manifest as a log-like curve over time. The exact fitting of that curve will give valuable insights into the impact of that inheritance model on resulting offspring. Those insights pave the way for more structured experiments. These kinds of experiments and studies are known as "Basic Research." From the data collected, better questions can be asked. (Such as experiments that involve controls: Does "beating up poor giant armadillos" positively influence the inheritence rate curve? etc.)
So, to collect useful data, all I have to really do is religiously cull offspring, and studiously label and track all births and unions, then process the data at the end of phase 2.
Like all science, this kind of experiment is *INTENDED* to be repeated by other people looking to verify and refine the data, and to be repeated as a baseline for other empirical studies.
Again, since one can just dump all the pref pages of all the dwarves that have EVER been members of their fortresses, and many members have multigenerational fortresses already, all I have to do is ask politely, and I can get a guaranteed contamination-free reference to plot curves against without adding further complications to an already highly ambitious project.