OBJECTIVES:To describe preliminary use of a forced-choice preferential looking task for the clinical assessment of vision in dogs. MATERIALS AND METHODS:The vision of 18 pet dogs was investigated in two separate studies using a forced-choice preferential looking task: multiple observers watched eye, head and body movements on video recordings to identify cues suggesting when a dog had seen the feature of interest. Human observer reliability was determined using eight dogs and computer-generated stimuli. Visual acuity was assessed using computer-generated grating stimuli: in real-time, an observer watched each dog's eye movement patterns and behaviour to decide whether each grating was seen. Stimuli were presented in a step-wise manner and were controlled by the observer. Acuity was estimated as the highest spatial frequency the dog was determined to have seen. RESULTS:Median estimated visual acuity was better at 1 m compared to that at 3 m. Average test time was longer at a 3-m distance than at 1 m. Inter- and intra-observer reliability was better from 1 m than from 3 m. CLINICAL SIGNIFICANCE:Preliminary use of a forced-choice preferential looking task for measurement of visual acuity in dogs has potential use as a clinical tool for the assessment of vision in dogs.