Functional magnetic resonance imaging (fMRI) of the spinal cord has been the subject of intense research for the last ten years. An important motivation for this technique is its ability to detect non-invasively neuronal activity in the spinal cord related to sensorimotor functions in various conditions, such as after spinal cord lesions. Although promising results of spinal cord fMRI have arisen from previous studies, the poor reproducibility of BOLD activations and their characteristics remain a major drawback. In the present study we investigated the reproducibility of BOLD fMRI in the spinal cord of cats (N=9) by repeating the same stimulation protocol over a long period (approximately 2 h). Cats were anaesthetized with ketamine, and spinal cord activity was induced by electrical stimulation of cutaneous nerves of the hind limbs. As a result, task-related signals were detected in most cats with relatively good spatial specificity. However, BOLD response significantly varied within and between cats. This variability was notably attributed to the moderate intensity of the stimulus producing a low amplitude haemodynamic response, variation in end-tidal CO(2) during the session, low signal-to-noise ratio (SNR) in spinal fMRI time series and animal-specific vascular anatomy. Original contributions of the present study are: (i) first spinal fMRI experiment in ketamine-anaesthetized animals, (ii) extensive study of intra- and inter-subject variability of activation, (iii) characterisation of static and temporal SNR in the spinal cord and (iv) investigation on the impact of CO(2) end-tidal level on the amplitude of BOLD response.