Context. Contrary to theoretical expectation, surprisingly low concentrations of molecular oxygen, O 2 , have been found in the interstellar medium. Telluric absorption makes ground based O 2 observations essentially impossible and observations had to be done from space. Millimetre-wave telescopes on space platforms were necessarily small, which resulted in large, several arcminutes wide, beam patterns. Observations of the (N J = 1 1 −1 0 ) ground state transition of O 2 with the Odin satellite resulted in a 5σ detection toward the dense core ρ Oph A. At the frequency of the line, 119 GHz, the Odin telescope has a beam width of 10 , larger than the size of the dense core.Aims. The precise nature of the emitting source and its exact location and extent are therefore unknown. The current investigation is intended to remedy this. Methods. Although the Earth's atmosphere is entirely opaque to low-lying O 2 transitions, it allows ground based observations of the much rarer 16 O 18 O in favourable conditions and at much higher angular resolution with larger telescopes. In addition, ρ Oph A exhibits both multiple radial velocity systems and considerable velocity gradients. Extensive mapping of the region in the proxy C 18 O (J = 3−2) line can be expected to help identify the O 2 source on the basis of its line shape and Doppler velocity. Line opacities were determined from observations of optically thin 13 C 18 O (J = 3−2). During several observing periods, two C 18 O intensity maxima in ρ Oph A were searched for O 18 O in the (2 1 −0 1 ) line at 234 GHz with the 12 m APEX telescope. These positions are associated also with peaks in the mm-continuum emission from dust.