Indexed punctuation
Key_word
Underscores are treated as a letter of the alphabet, which is why you can [url=http://www.google.com/search?q=_]search for an underscore directly[/url]. Use underscores in content if your visitors include an underscore when searching (e.g. if you had a programming site).
Key&word
Ampersands or 'and symbols' have fairly unique handling. They're both [url=http://www.google.com/search?q=%26]indexed[/url] and also treated as the equivalent of word "and". If there are no spaces separating the symbol and the adjacent letters, the search results are an approximate equivalent of combining results for ["key and word"] and ["key & word"] (note the phrase matching). Use ampersands in copy as is natural for your target audience.
Explicit search operators
Many punctuation characters are explicit search operators, with a documented effect on results. Search operators are not indexed (or at least, they can't be searched for) and so are usually treated as word separators when found within website copy:
Key¦word
An (unbroken) pipe character is the equivalent of boolean OR: a search for [key OR word]. It can be a handy shortcut when conducting complex queries.
Key"word
A double quote triggers an exact or phrase search for the proceeding words (whether you include a closing double quote or not). So in this instance, it's the equivalent of a search for [key word] since a single word can't be a phrase. ["key word] is the same as searching for ["key word"].
Key*word
An asterisk is a wildcard search for zero or more words: [key ... word]. Putting numbers on both sides will trigger the calculator. Occasionally, Google delivers (strange!) results if you search for an asterisk directly.
Key~word
A tilde triggers Google's related word operator - in this instance, a search for both 'key' and 'word', as well as other words related to 'word' - like 'Microsoft', 'dictionary' and others.
Search operator oddities
Key-word
A hyphen (as is probably consistent with language use) returns a mix of results for the words both used separately, and joined together - somewhere between [key word] and [keyword]. It's the preferred word separator within website URLs, since other punctuation characters that are treated as a word-separator have specific functions within a URL.
Others
A few punctuation characters have a strange impact on results - returning far fewer results than for either separated or concatenated words. They are neither known search operators, or indexed characters. These are . / \ @ = :
Usualy they and all other punctuation characters are treated as simply a space or word separator.
