It's incredible what you can discover by reading the antiscientific sources. For instance, did you know that before October 26, 1978, nobody knew anything at all about what the Arctic sea ice cover was like? Amazing. I certainly didn't know that. I don't think the Navy realized that in the 1950s when it issued an atlas of Arctic ice cover normals. (I have a copy.) The folks fishing and shipping across high latitudes would also be surprised to discover that they'd missed the chance to use the northwest passage, or to fish in the Beaufort Sea in late summer, or .... Just incredible, in the literal sense of 'not credible'.
The reality is simple -- we do indeed have observations of the Arctic ice pack running back before what we normally call the satellite period. This includes from a different satellite (ESMR) in the early 1970s, Naval operations, shipping and fishing observations, flight observations, and so on. As an antiscientific sort, though, one ignores the possibility that there could be more data available and draws the conclusion without bothering to look.
If, instead, you go look, say at the National Snow and Ice Data Center (a name in the news fairly routinely since summer 2007 when the major Arctic decline started) for what data are available on sea ice, you can rapidly find this listing of sea ice data that they have: http://nsidc.org/data/seaice/all.html . Looking through the table for 'temporal coverage', we see the AARI maps, 1953-1990, Arctic ocean ship tracks (irregularly back to 1872), Arctic ocean back to 1901 (see qualifiers as well), US-Russian atlas back to 1950, 'Morphometric characteristics of ice and snow ...' (irregularly back to 1928), the ESMR data (1973-1976). And, for some reason not listed there but I knew of it to look for it, information in the Nordic seas back to 1750: http://www.nsidc.org/data/g02169.html
So if we have data back to 1950 or so on most or all of the Arctic, and with more qualifiers back to 1901, why do Cryosphere Today (link on right) and NSIDC (ditto) show curves only from October 1978? The answer is, because they're trying to be scientifically honest. With the start of SMMR data in October 1978, we have a nearly continuous series of data which was all collected in nearly the same way (polar orbiting microwave satellite, only slight differences in details of the instrument). Before that period, the available data and their processing were distinctly different. This, the careful scientist knows, could introduce artefacts -- false trends or steps -- in a time series that spanned different sorts of sources. Since we expected the changes in the Arctic ice pack to be slight, these artefacts could overwhelm the slight -- but real -- changes that researchers were trying to detect.
On the other hand, we have the extraordinary low covers of 2007 and 2008. These are so drastically far from what has happened before that the minor differences in extent that would be due to changes in details of map construction and data sources are entirely overwhelmed.
So yes, we do know that the minima of 2007 and 2008 are far from what's previously been observed. And our knowledge extend much farther back than the 30 years of the satellite record. You can safely weed out sources that try to imply that the 2007 and 2008 minimum isn't unusual.
On arguing by analogy
1 hour ago