How rich is a text’s vocabulary? Divide the number of DISTINCT words (types) by the total words (tokens). A short passage scores near 1; but the longer the text, the more words repeat, so the ratio always FALLS — which is why you can’t compare it across lengths without care. Slide the text length and watch the ratio slide down.
Type-Token Ratio = |types| / |tokens|: unique word-forms over running words. It measures lexical diversity, but is length-dependent — as tokens accumulate, repetition rises and TTR monotonically declines toward the text’s underlying vocabulary rate (which is why standardized variants like MTLD or a fixed-window MATTR exist). A fail-loud self-check throws unless a longer sample of the same source has a TTR no higher than a shorter one. ◆ real computational stylistics, node-verified.
The sample stream is illustrative; the ratio and its monotonic fall with length are exact. Raw TTR should NOT be compared across texts of different lengths — that length-sensitivity is the point this sphere makes.