Uniformity, specificity and variability of corticocortical connectivity

In many studies of the mammalian brain, subjective assessments of connectiv ity patterns and connection strengths have been used to subdivide the corte x into separate but linked areas and to make deductions about the flow of i nformation through the cortical network. Here we describe the results of ap plying statistical analyses to quantitative corticocortical connection data , and the conclusions that can be drawn from such quantitative approaches. Injections of the tracer WGA-HRP were made into different visual areas eith er side of the middle suprasylvian sulcus (MSS) in 11 adult cats. Retrograd ely labelled cells produced by these injections were counted in selected co ronal sections taken at regularly spaced intervals (1 mm) through the entir e visual cortex, and their cumulative sums and relative proportions in each of 16 recognized visual cortical areas were computed. The surface dimensio ns of these areas were measured in each cat, from contour lines made on enl arged drawings of the same sections. A total of 116 149 labelled neurons we re assigned to all visual cortical areas in the 11 cats, with 5212 others e xcluded because of their uncertain location. The distribution of relative c onnection strengths, that is, the percentage of labelled cells per cortical area, was evaluated using non-parametric cluster analyses and Monte Carlo simulation, and relationships between connection strength and area size wer e examined by linear regression. The absolute size of each visual cortical area was uniform across individua l cats, whereas the strengths of connections between the same area pairs we re extremely variable for injections in different animals. The overall dist ribution of labelling strengths for corticocortical connections was continu ous and monotonic, rather than inherently clustered, with the highest frequ encies presented by the absent (zero density) and the very-low-density conn ections. These two categories could not, on analytical grounds, be separate d from each other. Thus it seems that any subjective description of cortico cortical connectivity strengths by ordinal classes (such as 'absent', 'weak ', 'moderate' or 'strong') imposes a categorization on the data, rather tha n recognizes a structure inherent in the data themselves. Despite the great variability of connections, similarities in the distribut ion profiles for the relative strengths of labelled cells in all areas coul d be used to identify clusters of different injection sites in the MSS. Thi s supported the conclusion that there are four connectionally distinct subd ivisions of this cortex, corresponding to areas 21a, PMLS and AMLS (in the medial bank) and to area PLLS (in the lateral bank). Even for tracer deposi ts in the same cortical subdivision, however, the strength of connections p rojecting to the site from other cortical areas varied greatly across injec tion in different individual animals. We further demonstrated that, on aver age, the strength of connections originating from any given cortical area w as positively and linearly correlated with the size of its surface dimensio ns. When analysed by specific injection site location, however, this relati onship was shown to hold for the individual connections to the medial bank MSS areas, but not for connections leading to the lateral bank area. The da ta suggest that connectivity of the cat's visual cortex possesses a number of uniform global features, which are locally organized in such a way as to give each cortical area unique characteristics.