Paul has argued that values should have emphasis placed on them because their contents are generally more important than their label (go read his post for his reasoning, then come back here to read the rest of this). I, however, feel there’s a good reason that labels should be emphasized as they usually are:
If you have a series of label/value pairs, the content of the value is going to vary HUGELY across datasets. However, labels are known quantities. Because of this, our eyes can quickly pick up the word “name”, and then shift to read the value much quicker than to scan a list of unknown values and then shift to find out what the value is for (ie, what the label is). By emphasizing the values with both size/strength/positioning relative to the value consistently, it makes it that much easier for our eyes to pick out what the label is, and thus find the value.And of course, because labels remain consistent, it becomes easier for our eyes to pick out and recognize those symbols.
Even scanning your post, it takes longer in your second example for me to find that “Paul Jarvis” is your name, rathe than in the first example to find that your “name” is Paul Jarvis. Now expand that list of label/value pairs to say, 10 or 15, or even worse, 50. Suddenly, finding individual data is much harder, so the ability to scan quickly becomes even more important.