In two previous studies, we had demonstrated the influence of eye posi
tion on neuronal discharges in the middle temporal area, medial superi
or temporal area, lateral intraparietal area and area 7A of the awake
monkey (Bremmer et al., 1997a,b), Eye position effects also have been
found in visual cortical areas V3A and V6 and even in the premotor cor
tex and the supplementary eye field. These effects are generally discu
ssed in light of a coordinate transformation of visual signals into a
non-retinocentric frame of reference. Neural network studies dealing w
ith the eye position effect succeeded in constructing such non-retinoc
entric representations by using model neurones whose response characte
ristics resembled those of 'real' neurones. However, to our knowledge,
response properties of real neurones never acted as input into these
neural networks. In the present study, we thus investigated whether, t
heoretically, eye position could be estimated from the population disc
harge of the (previously) recorded neurones and, if so, we intended to
develop an encoding algorithm for the position of the eyes in the orb
it. The optimal linear estimator proved the capability of the ensemble
activity for determining correctly eye position. We then developed th
e so-called subpopulation encoding of eye position. This algorithm is
based on the partition of the ensemble of neurones into two pairs of s
ubpopulations. Eye position is represented by the differences of activ
ity levels within each pair of subpopulations. Considering this result
, encoding of the location of an object relative to the head could eas
ily be accomplished by combining eye position information with the int
rinsic knowledge about the retinal location of a visual stimulus. Take
n together, these results show that throughout the monkey's visual cor
tical system information is available which can be used in a fairly si
mple manner in order to generate a non-retinocentric representation of
visual information.