It is generally assumed that carnivore social organizations evolved directionally from a solitary ancestor into progressively more advanced forms of group living. Although alternative explanations exist, this evolutionary hypothesis has never been tested. Here, I used literature data and maximum likelihood reconstruction on a complete carnivore phylogeny to test this hypothesis against two others: one assuming directional evolution from a non-solitary ancestor, and one assuming parallel evolutions from a socially flexible ancestor, that is, an ancestor with abilities to live in a variety of social organizations. The phylogenetic reconstructions did not support any of the three hypotheses of social evolution at the root of Carnivora. At the family level, however, there was support for a non-solitary and socially flexible ancestor to Canidae, a socially flexible or solitary ancestor to Mustelidae, a solitary or socially flexible ancestor to Mephitidae, a solitary or group living ancestor to Phocidae, a group living ancestor to Otariidae and a solitary ancestor to Ursidae, Felidae, Herpestidae and Viverridae. There was equivocal support for the ancestral state of Procyonidae and Hyaenidae. It is unclear whether the common occurrence of a solitary ancestry at the family level was caused by a solitary ancestor at the root of Carnivora or by multiple transitions into a solitary state. The failure to support a solitary ancestor to Carnivora calls for caution when using this hypothesis in an evolutionary framework, and I suggest continued investigations of the pathways of the evolution of carnivore social organizations.