drivers and only 500 female drivers，and if SO then only 10 Percent．of the men had accidents as opposed to 20 per cent of the women．Or perhaps the community is not a typical sample of the American population；or perhaps the police records included only some accidents，not all；or perhaps these figures cover a wide range of accidents，from mild bumps to head-on collisions．
Perhaps a great many things．As you see，statistics must be handled carefully．If it is too much to expect all writers to be trained in statistical method，it is not too much to ask them to SUbject any statistical data they use to the common-sense criteria of accuracy，relevancy，and completeness．So tested，statistics are good evidence．
Most of the faults of inductive reasoning follow from igno—ring these criteria，or，even worse，from ignoring the spirit be—hind them．Induction begins with facts，and it stays with facts until it has established their truth．It does not select or distort facts to fit a preconceived notion．It is easy，for example，to blame j uvenile delinquency on comic books by ignoring the corn—plexity of forces that create j uvenile crime．Such an argument may strike US for a moment，but only for a moment．It is neither true nor honest．And in argumentation，as in murder,truth willout．
We have stressed here the problems of reasoning well，for that is the essence of argumentation．Yet to be fully effective an argument must be not only well reasoned but well expressed．Its organization must be clear．The writer should make plain at the very beginning what he is arguing for or what he is contending against，and his paragraphs should march in perfect order from premise to inference or from evidence to conclusion．
展开
