C. ESTABLE 327 



Dendro-dendritic and dcndro-somatic synapses can be described under 

 many types: (i) with special modelling of the expansions ni contact 

 (glomeruli, Fig. 2; chalices, Fig. 3); (2) with dendrites not configuring a 

 special articular contact splicing or connection device (Figs. 5, 6); (3) with 

 an equivalent interchange of dendrites between the neurones united by the 

 synapse (Fig. 4) ; (4) with an unequal participation ot dendritic endings in 

 the dendro-dcndrito-somatic reciprocal synapses (Fig. 13); (5) with the 

 dendritic participation of one neurone and the perikaryon and the dendrite 

 of another neurone, that is, dendro-somato-dendritic synapse in one way 

 and dendro-dendritic in the opposite way; (6) with the dendritic participa- 

 tion of a neurone and the perikaryon of another or other neurones (dcn- 

 dro-somatic synapse) ; (7) of parallel dendritic contacts ; (8) of cross-dendritic 

 contacts; (9) complex synapses, dcndro and axo-dcndritic at the same 

 time. 



Functionally there are two kinds of direct interneuronal relations: (a) one 

 way, in which pre- and post-synaptic pars of different nature constitute 

 polarized and irreversible synapses; (b) reciprocal, in which pre- and post- 

 synaptic pars of similar nature constitute reversible synapses. The former 

 are exclusively axo-somatic and axo-dendritic; in such cases, the neurone 

 whose axon models the pre-synaptic pars controls the other. Prevalence of 

 reciprocity may be suggested for the dendro-dendritic and dendro-somatic 

 synapses and the somato-somatic and the axo-axonic contacts. Such useful 

 reciprocity might be for isofunctional neurones of a same centre, which 

 lack recurrent collaterals, to correlate themselves directly by means of 

 dendro-dendritic, dendro-somatic and somato-somatic contacts (Figs. 2, 

 12). Thus the small neurones improperly called cerebellar grains, are 

 reciprocally connected only by means of particularly modelled dendro- 

 dendritic branches and close somatic contacts, but never interconnected 

 through axonic collaterals. On the contrary, Purkinje neurones are 

 reciprocally comiected by means of double polarized pathways (axo- 

 somatic and axo-dendritic synapses at the end of recurrent collaterals) 

 never by means of dendritic interchange. The rule we have formulated 

 does not exclude the co-existence, in certain centres, of the two kinds of 

 neuronal connections of reciprocal action. 



Cajal did not commit himself when considering the existence of dendro- 

 dendritic synapses: dealing with the neuronal connections of the sym- 

 pathetic ganglia (Cajal, 1 891) he wisely and cautiously wrote: 'It is difficult 

 to reject the possibility of the passage of a nervous impulse, either between 

 non-medullated contiguous nerve fibres or between juxtapositional 

 protoplasmic branches. To exclude this possibility would mean giving the 



