diff --git a/reformat/AWGS.cpp b/reformat/AWGS.cpp
index 136ec41..8556fa5 100644
--- a/reformat/AWGS.cpp
+++ b/reformat/AWGS.cpp
@@ -388,7 +388,7 @@ int ReformatAWGS_WP::PrintParagraph(const uint8_t* ptr, long maxLen)
             RTFTab();
             break;
         default:
-            RTFPrintExtChar(ConvertMacRomanToUTF16(uch));
+            RTFPrintUTF16Char(ConvertMacRomanToUTF16(uch));
             break;
         }
     }
diff --git a/reformat/ReformatBase.h b/reformat/ReformatBase.h
index 524c23b..484d4c0 100644
--- a/reformat/ReformatBase.h
+++ b/reformat/ReformatBase.h
@@ -342,22 +342,25 @@ protected:
     // (only use this if we're in RTF mode)
     inline void RTFPrintChar(uint8_t ch) {
         char pch = PrintableChar(ch);
-        RTFPrintExtChar(pch);
+        RTFPrintUTF16Char(pch);
     }
-    // output an RTF-escaped char, allowing high ASCII
+    // output an RTF-escaped char, allowing UTF-16 Unicode values
     // (only use this if we're in RTF mode)
-    inline void RTFPrintExtChar(uint16_t ch) {
-        if (ch == '\\')
+    inline void RTFPrintUTF16Char(uint16_t ch) {
+        if (ch == '\\') {
             fExpBuf.Printf("\\\\");
-        else if (ch == '{')
+        } else if (ch == '{') {
             fExpBuf.Printf("\\{");
-        else if (ch == '}')
+        } else if (ch == '}') {
             fExpBuf.Printf("\\}");
-        else if (ch < 256) {
-            // TODO: should be \'xx for 0x80-ff?
+        } else if (ch >= 0x20 && ch < 0x80) {
+            // don't use Unicode escapes for these, or the output will be
+            // unreadable by mere humans
             fExpBuf.Printf("%c", ch);
         } else {
-            fExpBuf.Printf("\\u%d?", ch);
+            // must print as a *signed* 16-bit decimal value, though it
+            // looks like most parsers work either way
+            fExpBuf.Printf("\\u%d?", (int16_t)ch);
         }
     }
     // output a char, doubling up double quotes (for .CSV)
@@ -378,7 +381,8 @@ protected:
     }
 
     // Convert a Mac OS Roman character value (from a IIgs document) to
-    // its UTF-16 Unicode equivalent.
+    // its UTF-16 Unicode equivalent.  This also includes a conversion
+    // for the control characters.
     uint16_t ConvertMacRomanToUTF16(uint8_t ch) {
         return kUTF16Conv[ch];
     }
diff --git a/reformat/Teach.cpp b/reformat/Teach.cpp
index 18a3cb2..21f657e 100644
--- a/reformat/Teach.cpp
+++ b/reformat/Teach.cpp
@@ -206,7 +206,7 @@ int ReformatTeach::Process(const ReformatHolder* pHolder,
             } else if (uch == '\t') {
                 RTFTab();
             } else {
-                RTFPrintExtChar(ConvertMacRomanToUTF16(uch));
+                RTFPrintUTF16Char(ConvertMacRomanToUTF16(uch));
             }
             dataBuf++;
             dataLen--;