In many experiments which compare a large number of different treatments or varieties, it is necessary to decide which sub-sets of treatments or varieties do not differ significantly from each other. Multiple comparison procedures provide a commonly used, but much criticized, way of tackling this problem. As an alternative, techniques known collectively as cluster analysis can be used. An advantage of using certain types of cluster analysis is that their results can be displayed as a dendrogram or tree diagram which gives a very useful picture of the overall relationship between treatments or varieties. These procedures can also be extended readily to multivariate observations from each plot.