Captions, like other display copy, are generally styled in contrast to the running text for aesthetic and practical reasons: The appearance of display type offers a change of pace from the uniform appearance of running text and makes it easier for readers to distinguish it from the default content. The font size should be the same as the running text, or smaller.
Captions come in several forms, and they can vary to some extent within a print publication or on a website according to function. At their simplest, captions can consist merely of the name of the person pictured, or can identify, in a few words, an object or a location. A longer phrase, in the form of an incomplete sentence and without punctuation, might describe whatever is pictured. Or the caption might consist of one or more complete sentences that explain the contents of the photograph or illustration. Occasionally, a brief article is itself formatted as a caption. (Be cautious about not overwhelming readers with information, however. Caption copy can be redundant to a passage in the running text but should not go into excessive detail.)
It’s best to minimize the different forms a caption can take in one publication. When captioning a headshot photograph of a person, for example, be consistent about whether you simply label the photo with the person’s name or whether you provide additional information. Because such photos tend to be small, printing the subject’s name alone is recommended. But a caption for a larger candid photo, as opposed to a posed headshot, for example, should provide some context.
Choose between using incomplete sentences or complete sentences when providing more detail; it’s better form to be consistent with one approach or the other.
When writing captions for images in textbooks, magazines or training materials I always try to include as much information as possible so visual learners will find a way into the information detailed within the body of the content. It’s also useful to include key terms in the caption and bold them in the body copy. Easier for visual learns to scan the pages and find the information when they want or need it.