An eigenanalysis-based ordination method, also known as reciprocal averaging. See Correspondence Analysis.
* Correspondence Analysis has been discovered independently by different scientists.
* Reciprocal Averaging means that sample scores are calculated as a weighted average (or centroid) of species scores, and species scores are calculated as a weighted average (or centroid) of sample scores, and iterations continue until there is no change. However, other algorithms are possible.
* Correspondence Analysis simultaneously ordinates species and samples. There are as many axes as there are species or samples, whichever is less.
* The number of axes worth interpreting is a matter of taste, but the size of eigenvalues can be a guide.
* Correspondence Analysis maximizes the correlation between species scores and sample scores. The eigenvalue is equal to the correlation coefficient. The eigenvectors are either species scores or sample scores.
* An eigenvalue of 1.0 implies that one sample (or group of samples) shares no species with all other samples.
* One can put new points in a Correspondence Analysis without affecting the ordination.
* As with all the other eigenanalysis techniques, it is possible to define "passive samples" or "passive species".
* Correspondence Analysis has a problem called the arch effect. This effect is caused by nonlinearity of species response curves.
* The arch is not as serious as the horseshoe effect of PCA, because the ends of the gradient are not convoluted.
* Another related problem of Correspondence Analysis is that the ends of the gradient are compressed.
* Detrended Correspondence Analysis was designed to correct for the arch effect and gradient compression, as described above.

This definition is in context to Ordination. See more contextual defintions for Correspondence Analysis.