A general survey is made of the feasibility of using trends in long-period temperature records as representations of climatic change. A breakdown of all probable significant sources of influence on such records, together with quantitative estimates of instrumental, observational, and local environmental factors, points up the generally critical influence of non-climatic factors, which for the most part induce apparent secular temperature rises. Three independent studies of city influence are presented. In the first, recent overlapping observations between the New Haven city and airport stations are used to estimate the local city influence which in turn is used to revise the secular station trend. In the second, evidence of negligible city influence but of real climatic change at Blue Hill Observatory since 1890 is discussed. In the third, a statistical study involving 77 stations in the United States, whose temperature records were observationally homogeneous between 1900 and 1940, bears out the prevalence of important city influence in this country. Except in the period of rapid climatic temperature change occurring since about 1890, observed temperature records, with few individual exceptions, are concluded to be very misleading as direct measures of macroclimatic change over periods longer than a few decades. With their use in climatic studies, particularly those extending back of 1900, isolation of the effects of widespread urban development and frequent thermometer relocation is imperative. At average stations in the United States, urban development has contributed local temperature rises at the rate of more than 1F in a century. The influence of very large cities has not been in proportion.