I'm not sure how, in his formulation, words are anything other than "hollow icons", all stroke and no fill.<p>The author also seems not to have thought out his argument completely; in one breath, he speaks of parsing words as shapes rather than reading individual letters, and in the next he argues that there being only 26 letters in the English alphabet makes English easier to parse than the theoretically infinite set of hollow icons. If letters don't matter to English parsing, why does it matter how many there are? If they <i>do</i> matter, then how is an English word significantly different from a hollow icon?