The tw i n method has been cr i ti ci sed for i ts al l eged non-gener al i sability. When population parameters of intellectual abilities are estimated from a twin sample, critics point to the twinsi ngl eton di ffer ences i n i ntr auter i ne and fami l y envi r onments. These di ffer ences ar e suggested to l ead to subopti mal cogni ti ve devel opment i n tw i ns. A l though pr evi ous studi es have r epor ted tw i nsi ngl eton di ffer ences i n i ntel l i gence, these studi es had tw o major dr aw back s: they tested young tw i ns, and tw i ns w er e compar ed w i th (geneti cal l y) unr el ated si ngl etons. To test accur atel y w hether tw i n-si ngl eton di ffer ences i n i ntel l i gence exi st, a gr oup of adul t tw i ns and thei r non-tw i n siblings were administered the Dutch WAIS-III. The group was large enough to detect twinsingleton differences of magnitudes reported in earlier investigations. The data were analysed using maximum likelihood model fitting. No evidence of differences between adult twins and their non-tw i n si bl i ngs on cogni ti ve per for mance w as found. I t i s concl uded that tw i n studi es pr ovi de reliable estimates of heritabilities of intellectual abilities which can be generalised to the singleton popul ati on. Twin Research (2000) 3, 83-87.