Publié en 1984

now lets take a pair of words like        « generous » and « thrifty » say we could probably find an axis        that ran through them unfortunately we could find many more        than one axis but lets take an axis an axis is a good        word it suggests so much a kind of space through with        it runs a kind of semantic globe domain ? hyperspace ?        anyway lets call it an axis        […]        but what could that mean that        a word could lie closer to the same axis than another it could        mean that we will have to find only pure opposites or antonyms        lying at ends of feature axes and that all the words in the        system will have to be plotted by the intersection of various axes        their spatial coordinates in some kind of hyperspace so that        we know just how far off the axis of « closed » and « open »        « generous » and « thrifty » may really be and how close they lie        to an axis of « big » and « small » for example or « soft » and        « hard »        how many features will we need to map any lexeme        how many features are there is there a feature axis that        can be constructed by drawing a line between any two words        that can be regarded as opposites seen from some point of        view will we have to connect every word with every        other word in practice in principle dollars and doughnuts        if not will        there be a finite set of such contrasts ?        […]        now the reason i chose to talk about tuning        i was proposing a way of looking at how we understand        things how we come to understand things come to an        understanding with each other about things through        language has something to do with a notion process        i would like to call tuning

clichés are the        commonplaces by which we begin a discourse        there is no way to        put down cliché in the beginning because if there was        nothing in common in the utterances with which we addressed        each other thered be no way that we could understand what we        were saying in some sense there is no adequate theory of how        we understand what we say to each other one of the most        depressing things about the present attempts at knowledge        is the array of formal machines we have for explaining how we        know what we say and how poorly they explain it