The 'urban heat island' (UHI) effect refers to the fact that many temperature measurement sites are located in cities. These locations are generally warmer than the surrounding countryside, due to many factors, including heating, air conditioning, and absorption of solar radiation by man-made materials.  This means that using temperature readings from urban locations is  misleading (temperatures are too high). Furthermore, as cities expand, the UHI effect gets greater, leading to a false increase in temperatures.

The AR4 SPM claims that these effects are small (page 5): "Urban heat island effects are real but local, and have a negligible influence (less than 0.006°C per decade over land and zero over the oceans)."

The statement about UHI being 'local' makes no sense at all - since these urban readings are used to obtain averages over a wide area, the effects are wide-reaching.

The main text of AR4 WG1 (chapter 3, page 244) makes no attempt to justify the figure of 0.006° per decade, but merely repeats this figure from its previous report

Papers on the UHI effect include:

E Kalnay and M Cai, Impact of urbanization and land-use change on climate, Nature 423, 528-531 (2003).

They find an effect of 0.027° per decade, including both UHI and land-use change effects, much greater then the figure quoted by the IPCC.

For a detailed article on the UHI effect, with many more references, see  this article

Meteorologist Anthony Watts has examined over one-third of the 1,221 weather stations making up the U.S. Historical Climatology Network and published the results at his surfacestations website. Of those examined, more than half fall short of federal guidelines for optimum placement. Some examples include weather stations placed near sewage treatment plants, parking lots, and near cars, buildings and air-conditioners - all artificial heat sources which cause spurious higher temperature readings, providing physical confirmation of a root cause for a significant UHI effect on the record. Here is an example of a poorly positioned site (Maryville, CA). Note the parking area and air conditioning units just 10 feet from the temperature sensor.

The result is that the temperature record for Marysville shows a spurious 0.13 degree per decade increase, compared with a nearby rural site. This site is still being used as part of the NASA GISS temperature analysis. It is claimed that 'adjustments' are used to correct for these effects, by using nearby rural sites. But this makes no sense at all. The data from this site is clearly useless, and should not be used at all.  Furthermore, in many cases there are no nearby rural sites used by GISS that can be used to do a comparison (for example, in the South-East of Australia, all the sites used by GISS aare at airports in cities!)

A 2006 paper in Geophys. Res. Lett.  on land use/land cover change effects on temperature trends at rural U.S. weather stations showed that there was no significant temperature trend prior to the period of greatest land use changes; “in contrast, after the period of greatest LULC change was observed, 95% of the stations that exhibited significant trends (minimum, maximum, or mean temperature) displayed warming trends”.

A 2002 NASA study said that cities could be as much as 5 degrees warmer than their surroundings, and that this UHI  could lead to significant impacts such as increased rainfall nearby.

Update (2008):

A paper by Chinese scientists, Ren, G., Zhou, Y., Chu, Z., Zhou, J., Zhang, A., Guo, J. and Liu, X. (2008) Urbanization effects on observed surface air temperature trends in north China. Journal of Climate 21: 1333-1348, has now shown conclusively that the IPCC's claim that the UHI is negligible is false. Ren et al did the obvious thing to check for the UHI effect: compare trends from rural stations with those from urban ones. They found that the trend (degrees C per decade) was 0.18 for rural stations, 0.25 in small cities and 0.34 in large cities, with an overall trend of 0.29. They estimate that the UHI effect is responsible for a trend of about 0.11 degrees per decade in China; that is almost 20 times as great as the IPCC's claim.