One aspect of welfare that is measured in protocols used for the assessment of farm animal welfare is the human-animal relationship. The aim of the present study was to develop a 'human approach test' to measure the human-animal relationship in shelter cats and to investigate aspects of feasibility, reliability and validity of this test in a surveillance setting. Cat behaviour towards humans was assessed using a standardized approach in their home pens. Cats were categorized as 'no contact possible' when they withdrew, 'froze' or showed signs of aggression. 'Contact possible' was assigned to cats that stayed near and investigated the experimenter. In total, 725 cats in 30 shelters were tested and it was possible to initiate contact with 59 +/- 21% (Mean +/- S.D.) of the cats per shelter. To evaluate test-retest reliability of the approach test at shelter level, ten shelters were visited twice (mean number of days between visits: 58). Spearman rank correlation of the proportion of cats per shelter showing contact behaviour resulted in r(s) = 0.68, P= 0.032. No significant difference was found between visit 1 and visit 2 (P= 0.799). Between-experimenter repeatability was tested at pen level in 8 shelters on a total of 221 cats in 36 housing units. The ratings of the first rater and the three second raters resulted in rs = 0.99, p < 0.001. We attempted to assess the validity of the approach test by investigating relationships with attitudes of staff and frequency of daily care reported by the shelter manager. To assess attitudes to cats, 127 members of shelter staff completed a questionnaire. Overall, shelter staff had a quite positive attitude to cats (e.g. positive characteristics of cats: 5.6 +/- 0.9 on a scale from one to seven). Negative quality of interactions was seldom reported (coercive handling: 1.8 +/- 1). Relationships between cat behaviour and the attitude of several subsamples of shelter staff (based on measures of time spent working with cats) were analysed at shelter level. Of 60 correlations, only two in different subsamples were significant. However, more frequent provision of fresh water (r(s) = 0.38, p = 0.044, N = 28) and more frequent cleaning of food bowls (r(s) = 0.38, p = 0.036, N =30) were related to an increased proportion of cats showing contact behaviour. Overall, the test was feasible and showed very good to moderate reliability, but cat behaviour did not reflect caretaker attitudes in the shelters we visited. To develop valid animal-based measures of the human-shelter cat relationship further studies are needed. (C) 2016 Elsevier B.V. All rights reserved.