Tw i n studi es of chi l d temper ament usi ng objecti ve measur es consi stentl y suggest moder ate heritability for most dimensions. However, parent rating measures produce unusual patter ns of r esul ts. I ntr acl ass cor r el ati ons for i denti cal (M Z) tw i ns ar e typi cal l y hi gh, w her eas fr ater nal (DZ) tw i n i ntr acl ass cor r el ati ons ar e much l ow er than w oul d be pr edi cted fr om an addi ti ve geneti c model . The 'too l ow ' DZ cor r el ati ons can be expl ai ned by par ent-r ati ng bi ases that ei ther exagger ate the di ffer ences betw een DZ tw i ns (contr ast effects) or that i nfl ate the si mi l ar i ty of M Z tw i ns (assi mi l ati on effects), or by the pr esence of non-addi ti ve geneti c var i ance. To eval uate the thr ee possi bl e expl anati ons, w e used model -fi tti ng pr ocedur es appl i ed to par ent-r ati ng data aver aged acr oss 14, 20, 24, and 36 months of age i n a sampl e of 196 tw i n pai r s par ti ci pati ng i n the M acA r thur Longi tudi nal Tw i n Study. The data w er e best descr i bed by a model that i ncl uded contr ast effects. I mpl i cati ons for non-tw i n r esear ch ar e di scussed. Twin Research (2000) 3, 224-233. Keyw or ds: tw i ns, temperament, geneti cs, rater bi as, model fi tti ng I ntr oducti onResearch empl oyi ng the tw i n desi gn provi des the strongest evi dence of geneti c i nfl uences on i ndi vi dual di fferences i n temperament duri ng i nfancy and earl y chi l dhood. How ever, tw i n studi es assessi ng temperament vi a parent-rati ng measures, the most common method empl oyed to assess temperament, frequentl y produce an unusual pattern of resul ts. Wi th such measures, i denti cal (monozygoti c, M Z) co-tw i n resembl ance for temperament di mensi ons, as i ndexed by i ntracl ass correl ati ons, i s typi cal l y moderate; w hereas fraternal (di zygoti c, DZ) co-tw i n resembl ance i s very l ow, often near zero or even sl i ghtl y negati ve. [1][2][3] This puzzling outcome is particul arl y evi dent i n, but not excl usi ve to, rati ng measures that requi re parents to make gl obal judgements of thei r chi l d's behavi our. 4 Because M Z tw i ns are more si mi l ar than DZ tw i ns for parent-report measures of temperament, these resul ts provi de evi dence of geneti c i nfl uence. Nonethel ess, the l ow DZ tw i n resembl ance i s puzzl i ng because the si mpl e (additi ve) geneti c model predi cts that DZ tw i n si mi l ari ty shoul d be at l east half that of the M Z tw i ns due to the fact that DZ tw i ns are, on average, 50% si mi l ar geneti cal l y, w hereas M Z tw i ns are geneti cal l y i dentical . The pattern of very l ow DZ correl ati ons i s si gni fi cant because i t i mpl i es that DZ tw i ns are percei ved as hardl y any more si mi l ar than tw o randoml y pai red chi l dren.
Possible explanations for low DZ resemblanceThe probl em of 'too l ow ' DZ correl ati ons, as i t has come to be descri bed i n the i nfant and chi l d temperament l i terature, has been expl ai ned by assi mi l ati on effects, contrast effec...