The cosmological effects of a cosmological constant are not difficult to understand [42, 74, 21]. In the early universe, the density of energy and matter is far more important than the term on the r.h.s. of the Friedmann equation. But the average matter density decreases as the universe expands, and at a rather low redshift (z ~ 0.2 for m = 0.3, = 0.7) the term finally becomes dominant. Around this redshift, the term almost balances the attraction of the matter, and the scale factor a (1 + z)-1 increases very slowly, although it ultimately starts increasing exponentially as the universe starts inflating under the influence of the increasingly dominant term. The existence of a period during which expansion slows while the clock runs explains why t0 can be greater than for = 0, but this also shows that there is an increased likelihood of finding galaxies in the redshift interval when the expansion slowed, and a correspondingly increased opportunity for lensing by these galaxies of quasars (which mostly lie at higher redshift z 2).
The observed frequency of such optical lensed quasars is about what would be expected in a standard = 1, = 0 cosmology, so this data sets fairly stringent upper limits: 0.70 at 90% C.L. [81, 69], with more recent data giving even tighter constraints: < 0.66 at 95% confidence if m + = 1 . This limit could perhaps be weakened if there were (a) significant extinction by dust in the E/S0 galaxies responsible for the lensing or (b) rapid evolution of these galaxies, but there is much evidence that these galaxies have little dust and have evolved only passively for z 1 [120, 78, 112]. An alternative analysis  of some of the same optical lensing data gives a value = 0.64-0.26+0.15. My group  (cf. ) showed that edge-on disk galaxies can lens quasars very effectively, and discussed a case in which optical extinction is significant. But the radio observations discussed by , which give a 2 limit < 0.73, are not affected by extinction, so those are the ones quoted in the Table above. Recently a reanalysis  of lensing using new models of the evolution of elliptical galaxies gave = 0.7+0.1-0.2, but Kochanek et al.  (see especially Fig. 4) show that the available evidence disfavors such models.
A model-dependent constraint appeared to come from simulations of CDM  and OpenCDM  COBE-normalized models with h = 0.7, m = 0.3, and either = 0.7 or, for the open case, = 0. These models have too much power on small scales to be consistent with observations, unless there is strong scale-dependent antibiasing of galaxies with respect to dark matter. However, recent high-resolution simulations  find that merging and destruction of galaxies in dense environments lead to exactly the sort of scale-dependent antibiasing needed for agreement with observations for the CDM model. Similar results have been found using simulations plus semi-analytic methods  (but cf. ).
Another constraint on from simulations is a claim  that the number of long arcs in clusters is in accord with observations for an open CDM model with m = 0.3 but an order of magnitude too low in a CDM model with the same m. This apparently occurs because clusters with dense cores form too late in such models. This is potentially a powerful constraint, and needs to be checked and understood. It is now known that including cluster galaxies does not alter these results [83, 44].