Dogs (Canis familiaris) are a highly social species and within a shelter environment pair-housing is recommended to prevent the stress associated with social isolation. Separation of individuals which may have formed bonds in this environment is a usual occurrence, as a result of rehoming or euthanasia. To investigate the impact of separation, the behaviour, cognitive bias, faecal S-IgA and cortisol levels were examined in 12 adult pair-housed dogs, maintained in a private animal shelter. Prior to separation, dogs engaged in more affiliative than agonistic behaviour with conspecifics (means of 3 and 0.1% of time respectively). Following separation, increased activity was observed in the form of more running and grooming (P = 0.02), circling (P = 0.006), figure of 8 movement (P = 0.01), posture changes (P = 0.003) and stretching (P = 0.005), and less play behaviour was observed (P = 0.01). Secretory IgA increased (P = 0.02) after separation (mean = 443.7 ± 182.5 ng/mL; before separation mean = 370.1 ± 108.2 ng/mL). Cortisol concentrations were not affected by separation (P = 0.26, mean before separation = 792 ng/g; mean after separation = 874 ng/g). There was no indication from cognitive bias testing that the dogs’ emotional valency was affected, as latencies to reach ambiguous cues before and after separation did not differ significantly (P = 0.33). These results demonstrate that separation of a dog from a conspecific negatively affected behaviour and stimulated the immune system, changes which could be indicative of stress.