Spatial summation of responses in striate neurons in cats under N2O/O2 anaesthesia was examined quantitatively both along the line of the optimal stimulus orientation (length summation) using moving light bars and single light and dark edge stimuli, and at right angles to the optimal orientation (width summation) using stationary flashing bars. Activity profiles and length-response curves were prepared from simple, complex and hypercomplex I and II cells. An activity profile indicates the responsiveness of a cell at locations along the length of its receptive field. The activity profiles from all cell types were usually well fitted by Gaussian functions. Length summation occurs both in end-free (simple and complex) and, to a lesser extent, in end-stopped (hypercomplex I and II) cells over a wide range of stimulus contrasts (0.13 to 0.95). The linearity of length summation was tested either by comparing the recorded length-response curves with the curves predicted from the linear integration of the activity profiles or by comparing the response to the activation of two regions of the receptive field with the sum of the responses to each region activated separately. Although length summation was usually non-linear (either greater than or less than direct proportionality) it was more nearly linear in complex than it was in simple and hypercomplex I cells. Mechanisms responsible for non-linear length summation were studied, including a threshold for discharge, response saturation and summation of end-zone inhibition. Complex cells show little width summation for bars wider than 0.3 degrees. In simple and hypercomplex I cells there was also relatively little width summation either in an ON or an OFF discharge region at contrasts above about 0.4 but at lower contrasts width summation may be apporximately linear. Spatial summation of responses does not appear to be a useful characteristic for distinguishing one striate cell type from another.