Dogs are known to be skilled at using human social signals such as pointing at a target, gaze, visual direction of attention, and facial emotional cues. Two nonmutually exclusive hypotheses have been proposed to explain these abilities: the domestication hypothesis and the ``Two-Stage'' hypothesis. One way to test the Two-Stage hypothesis is to compare subpopulations of dogs with different histories with humans. For example, the abilities of pet dogs, who live in human homes and have developed strong affiliative bonds with humans, can be compared with those of shelter dogs, who live,in social isolation and are deprived of extended contact with humans. Here, we review the extant literature on studies comparing these 2 subpopulations using identical protocols. Pet dogs perform better than shelter dogs at following human pointing and at estimating humans' attentional state or direction of visual attention. Shelter dogs seem to be more socially driven to gaze and interact with humans compared to pet dogs. Shelter dogs' impoverished contact with humans is the best candidate explanation for these results. We summarize survey results highlighting the importance of life experience and learning in determining dogs' abilities to use human social cues and argue that shelter dogs may have learned not to respond to human cues that are not useful to them or have lost some previously acquired skills due to a lack of exposure to humans. Finally, we encourage further research that adds to both our theoretical and practical understanding of the impaired abilities of shelter dogs to use human social cues, and its link with the effect of life experiences. (C) 2016 Elsevier Inc. All rights reserved.