Context : HD 50138 is a Herbig B [ e ] star with a circumstellar disc detected at infrared and millimeter wavelength .
Its brightness makes it a good candidate for near-infrared interferometry observations .

Aims : We aim to resolve , spatially and spectrally , the continuum and hydrogen emission lines in the 2.12-2.47 micron region , to shed light on the immediate circumstellar environment of the star .

Methods : Very Large Telescope-Interferometer/ AMBER ( VLTI/ AMBER ) K-band observations provide spectra , visibilities , differential phases , and closure phases along three long baselines for the continuum , and H i emission in Br \gamma and five high-n Pfund lines .
By computing the pure-line visibilities , we derive the angular size of the different line-emitting regions .
A simple local thermodynamic equilibrium ( LTE ) model was created to constrain the physical conditions of H i emitting region .

Results : The continuum region can not be reproduced by a geometrical two-dimensional ( 2D ) elongated Gaussian fitting model .
We estimate the size of the region to be 1 au .
We find the detected hydrogen lines ( Br \gamma and Pfund lines ) come from a more compact region of size 0.4 au .
The Br \gamma line exhibits an S-shaped differential phase , indicative of rotation .
The continuum and Br \gamma line closure phase show offsets of \sim -25 \pm 5 ° and 20 \pm 10° respectively .
This is evidence of an asymmetry in their origin , but with opposing directions .
We find that we can not converge on constraints for the H i physical parameters without a more detailed model .

Conclusions : Our analysis reveals that HD 50138 hosts a complex circumstellar environment .
Its continuum emission can not be reproduced by a simple disc brightness distribution .
Similarly , several components must be evoked to reproduce the interferometric observables within the Br \gamma line .
Combining the spectroscopic and interferometric data of the Br \gamma and Pfund lines favours an origin in a wind region with a large opening angle .
Finally , although we can not exclude the possibility that HD 50138 is a young star our results point to an evolved source .