Previous electron microscopic studies of synaptic terminal distributions in the lateral geniculate nucleus have been flawed by potential sampling biases favoring larger synapses. We have thus re-investigated this in the geniculate A-laminae of the cat with an algorithm to correct this sampling bias. We used serial reconstructions with the electron microscope to determine the size of each terminal and synaptic type. We observed that RL (retinal) terminals are largest, F (local, GABAergic, inhibitory) terminals are intermediate in size, and RS (cortical and brainstem) terminals are smallest. We also found that synapses from RL terminals are largest, and thus most oversampled, and we used synaptic size data to correct for sampling errors. Doing so, we found that the relative synaptic percentages overall are 11.7% for RL terminals, 27.5% for F, and 60.8% for RS. Furthermore, we distinguished between relay cells and interneurons with post-embedding immunocytochemistry for GABA (relay cells are GABA negative and interneurons are GABA positive). Onto relay cells, RL terminals contributed 7.1%, F terminals contributed 30.9%, and RS terminals contributed 62.0%. Onto interneurons, RL terminals contributed 48.7%, F terminals contributed 24.4%, and RS terminals contributed 26.9%. We also found that RL terminals included many more separate synaptic contact zones (9.1 +/- 1.6) than did F terminals (2.6 +/- 0.2) or RS terminals (1.02 +/- 0.02). We used these data plus the calculation of overall percentages of each synaptic type to compute the relative percentage of each terminal type in the neuropil: RL terminals represent 1.8%, F terminals represent 14.5%, and RS terminals represent 83.7%. We argue that this relative synaptic paucity is typical for driver inputs (from retina), whereas modulator inputs (all others) require many more synapses to achieve their function.