tions that existed in the past， paleoclimatologists invoke

　　four principal criteria. First， the material-rocks， lakes，

　　vegetation， etc.-on which the method relies must be

　　（5） widespread enough to provide plenty of information，

　　since analysis of material that is rarely encountered will

　　not permit correlation with other regions or with other

　　periods of geological history. Second， in the process of

　　formation， the material must have received an environ-

　　（10） mental signal that reflects a change in climate and that

　　can be deciphered by modern physical or chemical

　　means. Third， at least some of the material must have

　　retained the signal unaffected by subsequent changes in

　　the environment. Fourth， it must be possible to deter-

　　（15） mine the time at which the inferred climatic conditions

　　held. This last criterion is more easily met in dating

　　marine sediments， because dating of only a small

　　number of layers in a marine sequence allows the age of

　　other layers to be estimated fairly reliably by extrapola-

　　（20） tion and interpolation. By contrast， because sedimenta-

　　tion is much less continuous in continental regions， esti-

　　mating the age of a continental bed from the known

　　ages of beds above and below is more risky.

　　One very old method used in the investigation of past

　　（25） climatic conditions involves the measurement of water

　　levels in ancient lakes. In temperate regions， there are

　　enough lakes for correlations between them to give us a

　　reliable picture. In arid and semiarid regions， on the

　　other hand， the small number of lakes and the great

　　（30） distances between them reduce the possibilities for corre-

　　lation. Moreover， since lake levels are controlled by rates

　　of evaporation as well as by precipitation， the interpreta-

　　tion of such levels is ambiguous. For instance， the fact

　　that lake levels in the semiarid southwestern United

　　（35） States appear to have been higher during the last ice age

　　than they are now was at one time attributed to

　　increased precipitation. On the basis of snow-line eleva-

　　tions， however， it has been concluded that the climate

　　then was not necessarily wetter than it is now， but rather