The historical temperature record shows the fluctuations of the temperature of the atmosphere and the oceans throughout history, and in particular since 1850.
The Historical Period: from 1850
The phrase historical temperature record is generally used for the time over which reasonably reliable near-surface temperature records exist from actual observations from thermometers. This is generally considered to start in about 1850 - earlier records exist, but coverage and instrument standardisation are less.
In the present day most meteorological observations are taken for use in weather forecasts. Centres such as ECMWF show instantaneous map of their coverage (http://www.ecmwf.int/products/forecasts/d/overview/monitoring/coverage/dcover!ssmi!00!pop!od!oper!w_coverage!latest!obs/); or the Hadley Centre show the coverage for the average of the year 2000 (http://www.meto.gov.uk/research/hadleycentre/CR_data/Annual/HadCRUTanm_2000.gif). Coverage for earlier in the 20th and 19th centuries would be significantly less.
Most of the warming occurred during two periods: 1910 to 1945 and 1976 to 2000. See  (http://www.grida.no/climate/ipcc_tar/wg1/figspm-1.htm) for a picture of the temperature record. Attribution of the temperature change to natural or anthropogenic factors is an important question: see global warming and attribution of recent climate change.
The data for the record come from thermometer measurements from land stations and ships  (http://www.met-office.gov.uk/research/hadleycentre/pubs/talks/sld011.html), which independently show much the same warming since 1860. The data from these stations show an average surface temperature increase of 0.4 to 0.8 Celsius degrees during the 20th century. The Intergovernmental Panel on Climate Change (IPCC) concludes:
- "The best estimate of global surface temperature change is a 0.6°C increase since the late 19th century with a 95% confidence interval of 0.4 to 0.8°C"  (http://www.grida.no/climate/ipcc_tar/wg1/049.htm).
The US National Academy of Science, both in its 2002 report to President George W. Bush, and in its latest publications, has strongly endorsed evidence of an average global temperature increase in the 20th century.
There are important concerns about the historical temperature record, which essentially divide into the fraction of the globe covered; and the effect of changing thermometer designs and observing practice and effects of changing land-use around the observing stations.
There exists a parallel record of marine observations from surface ships. These too suffer from changing practices (such as the switch from collecting water in canvas buckets to measuring the temperature from engine intakes) but they are at least immune to the urban heat island effect. The land and marine records can be compared.  (http://www.grida.no/climate/ipcc_tar/wg1/fig2-6.htm)
The need for (or otherwise), and magnitude of, corrections to the land station data to account for the urban heat island effect is discussed on that page.
Tropospheric temperature (the satellite and balloon temperature records)
Satellites have been measuring the temperature of the troposphere since 1979. Ballon measurements begin to show an approximation of global coverage in the 1950's. Further discussion can be found on the Satellite temperature measurements page.
Secondary evidence for temperature changes can be obtained by observing things that are predicted to be affected by temperature changes, such as variations in the snow cover and ice extent  (http://www.grida.no/climate/ipcc_tar/wg1/062.htm), sea level rise, precipitation  (http://www.grida.no/climate/ipcc_tar/wg1/077.htm), cloud cover  (http://www.grida.no/climate/ipcc_tar/wg1/082.htm), El Niño and extreme weather events  (http://www.grida.no/climate/ipcc_tar/wg1/089.htm). For example, satellite data shows a 10% decrease of snow cover since the late 1960s  (http://www.grida.no/climate/ipcc_tar/wg1/061.htm#fig213), and the Northern Hemisphere spring and summer sea-ice extent has decreased by about 10% to 15% since the 1950s and there has been a widespread retreat of mountain glaciers in non-polar regions throughout the 20th century. (http://www.grida.no/climate/ipcc_tar/wg1/064.htm).
The global temperature changes are not uniform over the globe, nor would they be expected to be, whether the changes were naturally or human forced. Certain places, such as the north shore of Alaska, showing dramatic rises temperature far above the average for the globe as a whole  (http://www.nytimes.com/2003/09/13/opinion/13KRIS.html?th). The Antarctic peninsula has warmed by 2.5 °C in the past 5 decades in certain places  (http://www.antarctica.ac.uk/met/gjma/) meanwhile East Antarctic has not significantly warmed  (http://huey.colorado.edu/LTER/news/NaturePaperReport_011302.html).
Proxies: tree rings, ice cores: the last 1000 years
Longer records exist from proxies: quantities such as tree-ring widths, coral growth or isotope variations in ice cores. From these, proxy temperature reconstruction of the last 1000 years have been made for the northern hemisphere.  (http://www.grida.no/climate/ipcc_tar/wg1/figspm-1.htm)  (http://www.grida.no/climate/ipcc_tar/wg1/069.htm#fig220) However, coverage of these proxies is sparse: even the best proxy records contain far fewer observations than the worst periods of the observational record. Also, problems exist in connecting the proxies (e.g. tree ring width) to the variable of interest (e.g. temperature).
The quantitative temperature reconstructions mentioned above show a general slow decline in temperature over the last 1000 years, followed by a faster temperature rise during the 20th century. This appears to contradict our knowledge of temperature fluctuation in this period such as the Medieval Warm Period, which is generally believed to have allowed the Vikings to colonise Greenland; and the Little Ice Age that followed, during which these colonies were abandoned. The explanation given for this apparent discrepancy is that the quantative reconstructions are for hemispheric temperature, whereas the Little Ice Age may have been more localised.
See-also: Temperature record of the past 1000 years
Indirect historical proxies
As well as natural, numerical proxies (tree-ring widths, for example) there exist records from the human historical period which can be used to infer climate variations, often in a less directly numerical way: reports of Frost Fairs on the Thames; records of good and bad harvests; dates of spring blossom or lambing; extraordinary falls of rain and snow, and unusual floods or droughts. These too can be used to infer historical temperatures, but generally in a more qualitative manner than the natural proxies discussed above.
The long term ice core record: the last 400,000 years
Even longer term records exist for few sites: one ice core (from Vostok, Antarctica) stretches back 420,000 years  (http://cdiac.esd.ornl.gov/trends/temp/vostok/graphics/tempplot5.gif); details from  (http://cdiac.esd.ornl.gov/trends/temp/vostok/jouz_tem.htm); many others reach more than 100,000 years. The Vostok core covers four glacial/interglacial cycles. Two cores GRIP (http://www.ngdc.noaa.gov/paleo/icecore/greenland/summit/document/gripinfo.htm), GISP (http://earth.agu.org/revgeophys/mayews01/mayews01.html) from Greenland stretch back as far as the previous interglacial. Whilst the large-scale signals from the cores are clear, there are problems interpreting the detail, and connecting the isotopic variation to the temperature signal.
- Global average temperature for the last 150 years (http://www.grida.no/climate/ipcc_tar/wg1/figspm-1.htm) and discussion (http://www.grida.no/climate/ipcc_tar/wg1/056.htm#fig27) of trends
- Preliminary data from the last 2000 years (ftp://holocene.evsc.virginia.edu/pub/mann/mannjones03.pdf)
Various other records exist and are even longer (principally ocean sediment cores) but they await a competent describer...