The assertion that there is an intrinsic excess of binaries with mass ratios q 1 -the twin hypothesis -is investigated. A strong version of this hypothesis (H s ), due to Lucy & Ricco (1979, AJ, 84, 401) and Tokovinin (2000, A&A, 360, 997), refers to a narrow peak in the distribution function ψ(q) for q > ∼ 0.95. A weak version (H w ), due to Halbwachs et al. (2003, A&A, 397, 159), refers to a broad peak for q > ∼ 0.8. Current data on SB2's is analysed and H s is found to be statistically significant for a sample restricted to orbits of high precision. But claims that H s is significant for binaries with special characteristics are not confirmed since the sample sizes are well below the minimum required for a reliable test. With regard to H w , additional observational evidence is not presented, but evidence to the contrary in the form of Hogeveen's (1992, Ap&SS, 196, 299) model of biased sampling with ψ ∝ q −2 is criticized. Specifically, his success in thus fitting catalogued data depends on implausible assumptions about the research methodologies of binary-star spectroscopists.