Attention to the ideas of context-free grammars and absurdly specialized vocabularies - prompted by the acceptance some years ago of a computer-generated paper as having been legitimately authored, and recalling the infamous Sokol hoax - reminds me of my earliest introduction to the notion, as well as to that of self-reference.
Published in 1901, Gelett Burgess' Nonsense Book included an item called "A Permutative System." The paragraph I reproduce here is made up of sentences designed to be equally meaningless, but equally suggestive of meaning, in any order. Further, uncross your eyes from your first reading, and you'll see that the apparent subject of the nonsense is itself.
The book this is copied from is one of my very few tangible links to my childhood, having graced my shelf since I was about nine years old:
It may be doubted that any system of thought arranged upon the lines herewith proposed can be a success. The fact of its accomplishment, alone, important as it must be, is no proof of method. For instance, the correct relation between any two facts is one that must be investigated along the lines of thought best correlated to these facts. And in spite of what, at first sight, be called irrelevancy, there is this to be observed, no matter what bearing the above may have to the subject in hand, that the relation of one part to any other may or may not be true. And here must be noted the importance of the demand that such types of thought do exist. This is, no doubt, a quality of subjects, rather than of relativity between modes of expression. So, too, are questions affecting the expression of coherent symbols of equal importance with the methods by which these symbols are expressed. But, at the same time, there must of necessity be a certain divergence in form between the types of questions to be discussed.