Identifiers, on the other hand, are syntax objects (elements
that make up Scheme expressions).  In plain R5RS (and R6RS
sans syntax-case), the programmer has no way of manipulating
the program's identifiers.  You can access the value of a
lexical variable, or you can assign to it, but you have no
way of getting your hand on it in the same way you can hold
symbols, pairs, and other data structures.  You can think of
identifiers as a tuple containing the name of the identifier
(usually represented by a symbol, but that's not strictly
necessary) and its lexical information (marks/substitutions
pair in some implementations).

Operations on identifiers include:
(datum->syntax id datum) which converts the datum to a
   syntax object containing the same lexical information
   as the identifier
(literal-identifier=? id1 id1) which returns true if
   the two identifiers have the same name
(free-identifier=? id1 id2) which returns true if the
   two identifiers refer to the exact same binding (the
   binding may be imported or lexical).
(bound-identifier=? id1 id2) which returns true if
   binding one of the identifiers would also bind the
   other (usually used to perform duplicate identifier
   check for binding forms like let and letrec).

So, symbols and identifiers are not the same thing at
all. They support different operations and each serves
a different purpose.