In the field of statistics, a sampling frame is the set of people or items that are to be sampled from. In other words, the sampling frame is the list of things that you plan to actually look at. This list may be different than the list of things that you would ideally like to sample. (The list of things that you want to sample is the sampling universe, AKA the sampling population).
In some cases the sampling frame and the sampling universe will be the same thing. For example, the United States Census wants to know about American Citizens. They do this by sampling, in theory, every single person in the US. The sampling universe is "every US citizen": the sampling frame is "every US citizen".
But other, less well-funded, researchers must settle for less thorough sampling frames; for example, I might try to get a representative sample of Americans by calling every hundredth name in each phonebook for the 100 biggest US cities. This would give me a very large and rather diverse sample. It would not give me exactly the same results as if I were able to sample everyone; my results would be biased towards people who live in big cities and own a home phone and list their numbers in the phone book. It is important to specify your sampling frame both so that you can see what pitfalls you might fall into, and so that others, looking at your data, can point out all the pitfalls that you fell into.
Knowing a researcher's sampling frame can highlight problems (and strengths) of a researcher's data. If a pollster reports that 60% of Americans support the legalization of marijuana, it would be useful to know if he took his sample on the streets of Berkeley, CA, or in Salt Lake City, UT -- or if he intentionally left these cities out in hopes of getting the views of the 'average American' rather than the extremes.
Unfortunately, very few news reports will give you the background information on the data they are reporting. You are lucky to get regular statements of the targeted sampling universe (we are supposed to assume that the universe = America); for a full statement of the sampling frame you will usually have to go back to the original published report or paper.