We should resize the balls based not on the languages they directly influenced, but on the transitive closure on the relation.<p>I was rather shocked to see ISWIM having such a small node, given it influenced basically the whole statically typed functional branch. Miranda got a correspondingly undeserved treatment.<p>Ideally, for the influence network, the size of the ball should correspond only to the influences that where innovations in the considered language. That may be too much to compute, though.