AB STRACT: Kan ga roo care (KC) has been the inter ven tion for preterm infants in numer ous pub lished stud i es. However, most well designed stud ies to date have used a one-group repeated mea sure design. This meth odol ogy is not as defin i tive as an exper i men tal design. Because of the absence of a com pa ra b le con trol group, change between pre test and posttest may be due to any other envi ron men tal vari ables or nor mal vari a tion of sub jects (Kirk, 1995). This ran dom ized con trolled trial (RCT) was done to test t he hypothe ses that KC infants would have higher mean tym panic tem per a tures, less weight loss, more opti mal behav ioral states, and lower acu ity (length of stay). Thirty-four eli gi ble mother-infant dyads were randomly assigned to the KC or the con trol group by com put er ized minimization on the day fol low i ng birth. Strat i fi ca tion vari ables included infant gen der, birth weight, deliv ery method, and par ity. KC infants com pared to con trol infants had higher mean tym panic tem per a ture (37.3 °C vs. 37.0 °C), more quiet sleep (62% vs. 22%), and less cry ing (2% vs. 6%) all at p = .000. No sig nif i cant dif fer ence was found for weight loss and acu ity (length of stay). These find ings can be used for evi dence-based nursing prac tice in Tai wan. With the knowl edge attained from this RCT, nurses can edu cate and moti vate moth ers to keep their sta ble preterm infants warm by skin-to-skin con tact inside their cloth ing, thereby encour ag ing self-regulatory feed ing.Key Words: preterm in fant, kan ga roo care, ran dom ized con trolled trial.