During the last years many authors suggested that a proportion of the long-term warming trends of the last hundred years in many global, hemispheric and even regional mean temperature records could be partially due to their contamination by urban growth effects. The urban heat island effect has been documented for many cities and the method for quantifying its intensity is to compare the urban temperature record with a neighboring rural station. It has been suggested that on the annual basis, the urban bias effect is a non-linear function of population. In this paper, we present a review of the results we found in order to demonstrate that population growth is not the only aspect that must be taken into account in correcting temperature series by urbanization effects. Linear correlation coefficients between yearly mean urban to rural temperature differences (URTD) and rural temperature (RT) are negative for most of the U.S., Argentine and Australian cities analyzed suggesting that the urban heat island effect depends among other parameters on the temperature itself. This negative correlation is not a statistical artifact, but the consequence that the interannual temperature variability in an urban environment is generally lower than in its rural surrounding. Therefore, it can be expected a reduction of heat island effect during warming periods. Linear trends in URTD and RT were also studied showing that urban stations are prone to have lower trends in absolute value than rural ones. So, regional datasets including records from urban stations, in addition to urban growth bias may have a second type of urban bias associated to the urban growth bias and of opposite sign. As an exercise, it was estimated the second urban bias trend for the contiguous U.S. for the period 1901-84. It was found that this bias is almost of the same order as the urban growth effect and of opposite sign. Therefore, it is suggested that the second bias in urban trends could be enhancing the urban growth bias during cooling periods and compensating it during warming ones. If these results could be extended to hemispheric or global scales, it could be expected the contamination of temperature trends by urbanization growth during warming periods will be offset by the diminishing of the urban heat island intensity.