With a data base consisting of hundreds of largely independent findings, it'd take one hell of a systematic bias to not get at least something right. Granted, they're obviously abstracting here, but we should avoid underplaying this achievement as much as overestimating it.