I am looking at the performance of three zoom lenses for Pentax cameras.
They are somewhat different animals, if not for the fact that they come for cheap:

• Cosina "Super Cosina" MC 80-200mm f4.5-5.6, manual focus (out of production)
• (Pentax) Takumar* 70-200 f4-5.6, autofocus (out of production)
• Tamron 70-300 f4-5.6 Di MACRO (a product you can buy today)

* according to the best about-Pentax online sources, at the time the "Takumar" lenses were similar or equivalent to the Pentax-F, where the "Takumar" brand was used to mark cheaper versions that lack the multi-coating of the better siblings. You must specify "at the time" because the Takumar brand has been previously used for other purposes, also for first-class lenses

PREAMBLE

I admit that I don't have the time to perform this sort of tests properly, so some of the results may be wrong for a number of reasons.
In spite of the fact that I took some precautions and all of the shots have been taken with Shake Reduction ON, a few may suffer from a bit of shaking. Let me remind everyone that Shake-Reduction is sensor-based on the Pentax K50 (as in all Pentax DSLR, for what I know), so it applies to all lenses, no matter their age.
The targets have been shot twice for each test; for the manual-focus lens (the Cosina), I focused once with Live View and once with my eyes only; for the other two lenses I once relied on autofocus and once on Live View.
As for the borders, everytime I have selected just the best one out of the eight available (4 per each of the two shots); same applies for the two available center spots.
The numbers and scales on the targets are meaningless because I didn't use them properly, so I'm just basing my judgement on what I see; all samples are at full resolution.
Sorry, you cannot judge fall-off from these shots because the targets were not evenly illuminated. I know, I've been lazy.

I'd like two share some more few considerations.

It may be obvious, but I don't want to take for granted that only skilled photographers may read this page, so...: remember that I'm working with APS-C here, hence the "real borders" of the lens input are cut out. If you are going to attach any of these lenses to a full-frame camera, you don't know what quality to expect at borders through this evaluation.
Most lenses are not perfectly aligned and borders are not avenly at focus. This is, sadly, a very common issue, well known to those who seriously test photographic equipment: there is nothing odd in the fact that one border, e.g. the one on the top-left, appears much better than the corner on the (e.g.) bottom-right.
Also, tests on targets at a given, relatively short distance may not tell the whole story. It is a very good idea to shoot at infinity and see what comes out; some lenses may behave better or worse in comparison to what target-based tests say.
That is why I am also adding some outdoor shots, and I will likely based my personal choice on the images taken outdoor instead of doing otherwise.