330 BRAIN MECHANISMS AND LEARNING 



layer; [d) very slender transversal filaments rising from Miiller's fibrocells. 

 It is in this extremely dense plexus of heterogeneous nature that Dr Villegas 

 (in her unpublished observations) has discovered the existence, in fish, of a 

 sort of neuropilc similar to that appearing in other centres (Villegas, 

 personal communication) ; we have also observed it in the retina of 

 mammals. 



Dendro-dendritic, dendro-somatic and dendro-somato-dendritic syn- 

 apses are most abundant in the retina and appear clearly in the horizontal 

 neurones and in the amacrine cells, for instance. The synaptic spectrum 

 of the horizontal neurones shows a prevalence of the dendro-dendritic, 

 dendro-somatic and dendro-somato-dendritic synapses (Figs. 5, 6). 

 However, not all horizontal neurones have the same synaptic spectrum. 



The influence ot some areas or foci over other areas or foci of the same 

 retina has remained unexplained. However, we may take into account that 

 there exist transversal contacts ((7) of the filiform pedal expansions of cones 

 and rods; (/)) of the dendritic tufts of bipolars; of the dendrites of gan- 

 glionic neurones, whose axons very rarely give oft collaterals. 



Despite these connections, the intrinsic retinal associations are mainly 

 due to dendritic connections of two kinds of neurones, the horizontal and 

 the amacrine cells (Fig. ja). 



Cajal discovered the retinopetal fibres which form synapses with the 

 body of the amacrine cells (Fig. 7/)). 'Two more categories of retinopetal 

 fibres arc to be considered; those establishing direct synapses with the 

 dendrites of ganglionic cells; others, more dispersed, with their collaterals 

 spreading out, covering the field of the dendrites of these neurones and the 

 amacrine cells (this last category was confirmed by Poliak (Poliak, 1941)). 



According to Cajal, the centrifugal optic fibres are undivicied in all their 

 intra-retinal course, and end in a little bunch of short collaterals on the 

 soma of an amacrine cell. The abot'c-iiiciitioiicd retinopetal fibres would lack 

 functional nieanino (a) // the conduction in the dendrites of the amacrine cells were 

 not 'antidromic', (b) // the dendro-dendritic contacts were not functional, or if the 

 excitatory state did not propagate through it — in other words, if it did not con- 

 stitute a real synapse. 



We have already mentioned three kinds of retinopetal fibres. There are 

 fibres too, which, following the same course, end in the capillaries of the 

 retina which may be taken for visual fibres if their vascular ending is not 

 noticeci. 



As concerns the neuro-muscular junctions, let us mention a type of 

 synapse in which the ending of the axon or its collaterals go into the 



