Recognition and interpretation of dogs emotional and motivational states from visual behavioural signs are important for public safety and dog welfare. This study used an online survey to explore the ability of members of the public (n = 4,133) to recognise the underlying emotional or motivational states of dogs in silent videos (n=30). Participants scored each video for nine pre-determined emotional and motivational states on a scale from 0 to 15 and rated the relative difficulty of scoring each video. Participants could also select \"I am uncertain\" for individual states which translated to missing values. Public scores were compared with those of eleven dog behaviour experts. The states \"nervous/anxious\", \"stressed\", \"relaxed\", \"comfortable\", \"playful\", \"interested/curious\", \"excited\", and \"frustrated\" showed high inter-expert agreement and were used in further analysis. \"Boredom\" was removed due to low inter-expert agreement. Principal components and cluster analyses on both datasets were used to collapse categories into two dimensions, identify groupings and compare overall perception. Results indicate similarity in perception of underlying states between public and experts. Correlation between expert difficulty rating, and both inter-expert agreement and public accuracy, indicates that experts effectively assessed the relative difficulty of determining underlying state. Members of the public perceived playful, excited, and curious dogs as easier to interpret than anxious and stressed dogs; however, this was not reflected in how accurately they scored videos (i.e., how different a participants scores were from the expert scores) and instead was reflected by how likely a participant was to score a video in full, rather than selecting that they were \"uncertain\" in response to any of the listed states. Findings of this study inform human behaviour change interventions to improve public interpretation of dog emotional and motivational states.