AKA summated rating.
A Likert scale is a polling format in which a statement is presented, and the respondent is asked to indicate how much they agree with it. It was developed by psychologist Rensis Likert, who used it for his work in investigating corporate management styles.
Likert Scales are usually 5-point scales, but they can have any number of points; three and seven are also popular numbers. Even-numbered scales are rarer, because it is not uncommon for a respondent to be uncertain or to not have an opinion; including a 'neutral' midpoint allows for this to be reflected in your data. You will rarely see a Likert scale with more than seven points -- any more than that is more likely to be confusing than helpful.
Example:
------------------------------------------------------------
Neither
Agree or
Disagree
Disagree Agree
Strongly | | | | | | Completely
| 1 | 2 | 3 | 4 | 5 |
-------------------------------------------------
I Like Cheese | | | | | |
-------------------------------------------------
Milk: I got it | | | | | |
-------------------------------------------------
Ice Cream Rules | | | | | |
------------------------------------------------------------
This type of scale is useful because it is easy to code the data, the format is uncomplicated and familiar to most respondents, and it can be easily modified to almost any subject.
------------------------------------------------------------
Respondents Name: Bob D.
Neither
Agree or
Disagree
Disagree Agree
Strongly | | | | | | Completely
| 1 | 2 | 3 | 4 | 5 |
-------------------------------------------------
I Like Cheese | | | | | x |
-------------------------------------------------
Milk: I got it | | | x | | |
-------------------------------------------------
Ice Cream Rules | | | | x | |
------------------------------------------------------------
Bob has a dairy index of 5+3+4 = 12. To put this in context, you'll want to note that this is 12 out of a possible 15, where nine is neutral; Bob's dairy index is .80, or 80%. The mathematically minded among us will notice that this index number is only a general indicator at best*. There are some misleading factors; an index of .60 or lower is a neutral or negative overall response, and a neutral answer (which would include 'no opinion' and 'don't know') is categorized the same as an actual opinion. It should also be noted that T-tests are not strictly valid when comparing data between Likert scale questions, because the range of answers is discrete, not continuous. If the results of a question resemble a normal 'mound-shaped' distribution, then a T-test may provide useful perspective, but be careful. A chi-square is a better bet.
You can do things to make the results more useful. You might use the mode instead of mean. You might display the results as a bar chart or a table instead of as a number. You might drop all 'neutral' answers (you could even leave out the center option in the first place, using a four point scale).
When you see results from a Likert scale, remember that not everyone cares if data is represented accurately. If you see an index-type number, like the one I included for Bob, be suspicious. But then, you should be suspicious whenever you see statistics...
* If you really want to use an index number and have it mean something, you'd do better to rename the points -2, -1, 0, 1, and 2, add them up, and give the resulting number. In this case Bob would score 3, out of a -6 to 6 scale. But then, at this point you're probably better off just making a chart to represent it visually. | 
