Canine perinatal mortality is known to be relatively high. However, the literature on perinatal mortality in dogs is still sparse and often refers to a single or only a few breeds. The aim of this large-scale observational study was to describe the perinatal mortality in purebred dogs of various breeds at both puppy and litter level. In addition, the influence of breed, breed size, litter size, age of the bitch, litter number and season for whelping on the risk of perinatal mortality at litter level was studied and the mean litter size at eight days and eight wks after birth was calculated. A retrospective cohort study was performed by studying 10,810 litters of 224 breeds registered in the Norwegian Kennel Club in 2006 and 2007. Perinatal mortality was defined as the sum of stillborn puppies and puppies that died during the first wk after birth (early neonatal mortality) and was present in 24.6% of the litters. Eight percent of the puppies died before eight days after birth, with 4.3% as stillbirth and 3.7% as early neonatal mortality. For most breeds the perinatal mortality was low, but for some breeds a higher perinatal mortality was found. The mean litter size at eight days and eight wks after birth was 4.97 (±0.02) and 4.92 (±0.02) puppies, respectively. Of all puppies born, only 1% died during the period from eight days to eight wks after birth. Random effects logistic regression analysis indicated that increasing litter size and age of the bitch were associated with an increased risk of stillbirth, early neonatal mortality and total perinatal mortality at the litter level (P < 0.001). The random breed effect was significant for all outcomes. Litter number also had a significant effect on stillbirth, early neonatal mortality and total perinatal mortality at the litter level, with the highest risk of perinatal mortality found in the first litter (P < 0.001). Further, the risk of early neonatal mortality was doubled in litters with stillborn puppies. No significant effect of whelping season on perinatal mortality at litter level was found. An interaction existed between the age of the bitch and litter number and the risk of stillbirth was three times as high (odds ratio = 3.00) in litters from bitches having their first litter after the age of six y. Breed was a more important determinant of perinatal mortality in litters than breed size. However, more than 90% of the variation in perinatal mortality was found at the individual litter level and efforts to minimize puppy mortality should be targeted at the management of the individual litter rather than at the breed level.