diff --git a/icu4c/source/common/symtable.h b/icu4c/source/common/symtable.h
index 85d8dc104a5..d26baf17660 100644
--- a/icu4c/source/common/symtable.h
+++ b/icu4c/source/common/symtable.h
@@ -1,6 +1,6 @@
 /*
 **********************************************************************
-*   Copyright (c) 2000, International Business Machines
+*   Copyright (c) 2000-2003, International Business Machines
 *   Corporation and others.  All Rights Reserved.
 **********************************************************************
 *   Date        Name        Description
@@ -21,15 +21,27 @@ class UnicodeSet;
 class UnicodeString;
 
 /**
- * An interface that maps strings to objects.  This interface defines
- * both lookup protocol and parsing.  This allows different components
- * to share a symbol table and to handle name parsing uniformly.  It
- * is expected that client parse code look for the SYMBOL_REF
- * character and, when seen, attempt to parse the characters after it
- * using parseReference().
+ * An interface that defines both lookup protocol and parsing of
+ * symbolic names.
  *
- * <p>Currently, RuleBasedTransliterator and UnicodeSet use this
- * interface to share variable definitions.
+ * <p>A symbol table maintains two kinds of mappings.  The first is
+ * between symbolic names and their values.  For example, if the
+ * variable with the name "start" is set to the value "alpha"
+ * (perhaps, though not necessarily, through an expression such as
+ * "$start=alpha"), then the call lookup("start") will return the
+ * char[] array ['a', 'l', 'p', 'h', 'a'].
+ *
+ * <p>The second kind of mapping is between character values and
+ * UnicodeMatcher objects.  This is used by RuleBasedTransliterator,
+ * which uses characters in the private use area to represent objects
+ * such as UnicodeSets.  If U+E015 is mapped to the UnicodeSet [a-z],
+ * then lookupMatcher(0xE015) will return the UnicodeSet [a-z].
+ *
+ * <p>Finally, a symbol table defines parsing behavior for symbolic
+ * names.  All symbolic names start with the SYMBOL_REF character.
+ * When a parser encounters this character, it calls parseReference()
+ * with the position immediately following the SYMBOL_REF.  The symbol
+ * table parses the name, if there is one, and returns it.
  */
 class U_COMMON_API SymbolTable /* not : public UObject because this is an interface/mixin class */ {
 public:
@@ -47,26 +59,37 @@ public:
     /**
      * Lookup the characters associated with this string and return it.
      * Return <tt>NULL</tt> if no such name exists.  The resultant
-     * string may have length zero.
+     * array may have length zero.
+     * @param s the symbolic name to lookup
+     * @return a char array containing the name's value, or NULL if
+     * there is no mapping for s.
      */
     virtual const UnicodeString* lookup(const UnicodeString& s) const = 0;
 
     /**
      * Lookup the UnicodeMatcher associated with the given character, and
-     * return it.  Return <tt>null</tt> if not found.
+     * return it.  Return <tt>NULL</tt> if not found.
+     * @param ch a 32-bit code point from 0 to 0x10FFFF inclusive.
+     * @return the UnicodeMatcher object represented by the given
+     * character, or NULL if there is no mapping for ch.
      */
     virtual const UnicodeFunctor* lookupMatcher(UChar32 ch) const = 0;
 
     /**
      * Parse a symbol reference name from the given string, starting
      * at the given position.  If no valid symbol reference name is
-     * found, return an empty string.
+     * found, return NULL and leave pos unchanged.  That is, if the
+     * character at pos cannot start a name, or if pos is at or after
+     * text.length(), then return an empty string.  This indicates an
+     * isolated SYMBOL_REF character.
      * @param text the text to parse for the name
      * @param pos on entry, the index of the first character to parse.
      * This is the character following the SYMBOL_REF character.  On
-     * exit, the index after the last parsed character.
+     * exit, the index after the last parsed character.  If the parse
+     * failed, pos is unchanged on exit.
      * @param limit the index after the last character to be parsed.
-     * @return the parsed name or an empty string.
+     * @return the parsed name, or an empty string if there is no
+     * valid symbolic name at the given position.
      */
     virtual UnicodeString parseReference(const UnicodeString& text,
                                          ParsePosition& pos, int32_t limit) const = 0;