In a CWB corpus, every token has positional attributes. While s-attributes
cover a range of tokens, every single token in the token stream of a corpus
will have a set of positional attributes (such as part-of-speech, or lemma).
The available p-attributes are returned by the p_attributes-method.