There are 2 reasons you want a large sample size: (1) to have enough subjects to expect a reasonably representative random sample (typically ~30+ for social science [1]) and (2) to have sufficient statistical power. There's nothing inherently wrong with n=40 and n=40 from an unbiased sample is better than n=400 from a biased sample.
Physicists are generally looking for very very small effects, hence very high n (higher n = higher power = more sensitive to treatment effects). This doesn't mean lower samples sizes are insufficient for other areas of research.
Anyway, the real issue is over-reliance on convenience sampling.
Physicists are generally looking for very very small effects, hence very high n (higher n = higher power = more sensitive to treatment effects). This doesn't mean lower samples sizes are insufficient for other areas of research.
Anyway, the real issue is over-reliance on convenience sampling.
[1] http://sph.bu.edu/otlt/mph-modules/bs/bs704_probability/BS70...