Books 2010

Books 2009

« ...has faded away | Main | Walking round London 2 »

Friday, 13 July 2007


Clustering correctness (re coment by Jno): that was my point. In mutivariate analysis there are a number of techniques for clustering. In different applications different approaches have different utility, I was making a more general point than cladistics, which is a technique whch is indeed appropriate for understanding evolution. However other techniques may be more useful if indeed you wish to select on observable morphology (as most gardeners do). I think we are essentially in agreement.


DP wrote:
> Now there is no "correct" method of clustering

I think the point is that there *is* a correct method of clustering for the purposes of evolutionary biology, viz. cladistics, but the tools and information to cluster organisms in this way are only recently available. The good old Gilbert White approach to natural history, of classifying organisms on the basis of their observable features, results in groups like fish (and possibly zebras) which are not natural groups; that is they do not map to clades.

Thanks for the comment and the Oyster Band track! Dancing Cat

Fascinating response from Dark Puss - and a very pleasing one from Rhys. There's another, entirely frivolous, side to this - the fun you have on safari telling the earnest guide that there's no such thing as a zebra!!

A number of interesting issues are raised by this post. Disregarding evolution (I.e. ignoring any underlying mechanism) I suppose we could regard classifcation of animals/plants/galaxies/music etc. as an exercise in clustering in a multidimensional parameter space. Now there is no "correct" method of clustering, and indeed there are a whole variety of techniques available to the analyst. The question often boils down to "is it useful"? That is if there is a traning set that we can develop the technique with will the technique have a good efficency at separating out the test data with reasonable accuracy. Imperfect sampling, particularly when there is some desirable characteristic, can often generate articicial splits in otherwise conrinuous distributions. For example short dark-skinned, black-haired women look very different from pale-skinned, tall red-headed women, but they are not distinct in terms of the underlying genetic information. Because of our exceptional human ability to recognise patterns, we are very good at making such distinctions and the problem then comes when we try to associate morphology with the evolutionary sequence.

Returning the the Zebra theme, the now extinct Quagga was one of the first (possibly the first) animal to be subject to ancient DNA analysis. The results suggest strongly that the Quagga evolved from the Plains Zebra somewhere between about 150k and 250k years ago, and illustrates how rapidly the coat colour could evolve in an isolated population, possibly under external climatic influences. Another example, from horticulture, the whole classifaction of the genus Rhododendron was revised significantly by Cullen and Chamberlain (Notes from the Royal Botanic Garden Edinburgh 39:1 and 39:2 1980) in a move away from the horticulturally-based approach of Balfour to a more herbarium (i.e. anatomical/cytological/phytochemical) based approach. This caused some uproar in hoticultural circles at the time when beloved distinct species were "shown" to be just variants selected by the great plant hunters who were often particularly on the look out for horticulturally intersting variations. Indeed many were funded by nurseries and private collectors.

To conclude, is there a zebra? I think yes in the sense that there is a usefully distinctive group of animals looked at from an external perspective. Are they uniquely distinct, in the same sense that boron is not beryllium, no they are not? DNA based classifcations help us greatly to unravel evolutionary trajectories and relationships, but they won't tell you which "species" to grow in your flower beds if you seek variety.

Dark Puss

This is all very interesring and thought provoking for a Saturday morning blog walk.

The comments to this entry are closed.


  • Nothing is of greater consolation to the author of a novel than the disovery of readings he had not conceived but which are then prompted by his readers. (Umberto Eco, Reflections on The Name of the Rose)
  • ... relatively few persons in London ... can afford the luxury of one or more servants. No fewer than 3,700,000 have no servants at all, and of the half million that have servants 227,000 have only one. (The Times, 6 June 1895)
  • Standing among savage scenery, the hotel offers stupendous revelations. There is a French widow in every bedroom, affording delightful prospects. (Tyrolean inn brochure, according to Gerard Hoffnung)
  • (A doctor is at an elderly relative's deathbed) "The old sawbones, eh?" he bellowed ... "Just in the nick, perhaps. Haul the old girl back by the short hairs, if you ask me. Devilish smart at his work ... Always take a fence with more confidence when I know he's out with us."
  • Too often, when a man of Monty Godkin's mental powers is plunged in thought, nothing happens at all. The machinery just whirs for a while, and that is the end of it. (P G Wodehouse, Heavy Weather)
  • ...the breed that take their pleasures as Saint Laurence took his grid (Kipling, The Five nations)

Photo Albums

Blog powered by Typepad