Microphone arrays are still designed for sound recordings devoted to multichannel applications. Some of them are based on empirical principles inherited from the stereophonic recording techniques and extended to 5.1 restitution setups. These systems are mainly non-coincident microphone arrays for which one single microphone is generally assigned to one of the five channels. Other techniques (e.g. Ambisonics) rely on the exact representation on the sound field. Although these systems are aiming at reproducing the original sound scene in the most exact way, some timbre problems have been sometimes reported. This paper is aimed at comparing four microphone arrays (two of each kind) from global preference and four verbal attributes known to underlie the spatial audio quality. The four microphone arrays were assessed by expert and naive listeners.