diff --git a/icu4c/source/i18n/nfsubs.cpp b/icu4c/source/i18n/nfsubs.cpp
index 52d7478cfcc..3e17f7b1e1f 100644
--- a/icu4c/source/i18n/nfsubs.cpp
+++ b/icu4c/source/i18n/nfsubs.cpp
@@ -681,7 +681,8 @@ ModulusSubstitution::doParse(const UnicodeString& text,
         ruleToUse->doParse(text, parsePosition, FALSE, upperBound, result);
 
         if (parsePosition.getIndex() != 0) {
-            double tempResult = result.getDouble();
+            UErrorCode status = U_ZERO_ERROR;
+            double tempResult = result.getDouble(status);
             tempResult = composeRuleValue(tempResult, baseValue);
             result.setDouble(tempResult);
         }
@@ -881,7 +882,7 @@ FractionalPartSubstitution::doParse(const UnicodeString& text,
                 }
                 if (fmt) {
                     fmt->parse(workText, temp, workPos);
-                    digit = temp.getLong();
+                    digit = temp.getLong(status);
                 }
             }
 
@@ -950,7 +951,7 @@ NumeratorSubstitution::doSubstitution(double number, UnicodeString& toInsertInto
         int64_t nf =longNF;
         int32_t len = toInsertInto.length();
         while ((nf *= 10) < denominator) {
-            toInsertInto.insert(pos + getPos(), (UChar)0x0020);
+            toInsertInto.insert(pos + getPos(), gSpace);
             ruleSet->format((int64_t)0, toInsertInto, pos + getPos());
         }
         pos += toInsertInto.length() - len;
@@ -997,12 +998,10 @@ NumeratorSubstitution::doParse(const UnicodeString& text,
     if (withZeros) {
         ParsePosition workPos(1);
         Formattable temp;
-        int32_t digit;
 
         while (workText.length() > 0 && workPos.getIndex() != 0) {
             workPos.setIndex(0);
             getRuleSet()->parse(workText, workPos, 1, temp); // parse zero or nothing at all
-            digit = temp.getLong(status);
             if (workPos.getIndex() == 0) {
                 // we failed, either there were no more zeros, or the number was formatted with digits
                 // either way, we're done
@@ -1012,14 +1011,14 @@ NumeratorSubstitution::doParse(const UnicodeString& text,
             ++zeroCount;
             parsePosition.setIndex(parsePosition.getIndex() + workPos.getIndex());
             workText.remove(0, workPos.getIndex());
-            while (workText.length() > 0 && workText.charAt(0) == 0x0020) {
+            while (workText.length() > 0 && workText.charAt(0) == gSpace) {
                 workText.remove(0, 1);
                 parsePosition.setIndex(parsePosition.getIndex() + 1);
             }
         }
 
         workText = text;
-        workText.remove(0, (int32_t)parsePosition.getIndex()); // arrgh!
+        workText.remove(0, (int32_t)parsePosition.getIndex());
         parsePosition.setIndex(0);
     }
 
@@ -1031,7 +1030,7 @@ NumeratorSubstitution::doParse(const UnicodeString& text,
         // force this to not bother trying all the base values?
 
         // compute the 'effective' base and prescale the value down
-        int64_t n = result.getLong();
+        int64_t n = result.getLong(status); // force conversion!
         int64_t d = 1;
         int32_t pow = 0;
         while (d <= n) {
@@ -1064,6 +1063,8 @@ NumeratorSubstitution::getDynamicClassID() const {
     return getStaticClassID();
 }
 
+const UChar NumeratorSubstitution::LTLT[] = { 0x003c, 0x003c };
+        
 //===================================================================
 // NullSubstitution
 //===================================================================
diff --git a/icu4c/source/i18n/nfsubs.h b/icu4c/source/i18n/nfsubs.h
index 8b4c2509d01..91bde26e957 100644
--- a/icu4c/source/i18n/nfsubs.h
+++ b/icu4c/source/i18n/nfsubs.h
@@ -481,8 +481,7 @@ public:
         : NFSubstitution(_pos, _ruleSet, formatter, fixdesc(description), status), denominator(_denominator) 
     {
         ldenominator = util64_fromDouble(denominator);
-        UChar LTLT[] = { 0x003c, 0x003c, 0 };
-        withZeros = description.endsWith(LTLT);
+        withZeros = description.endsWith(LTLT, 2);
     }
     
     virtual UBool operator==(const NFSubstitution& rhs) const;
@@ -503,8 +502,9 @@ public:
     virtual UChar tokenChar() const { return (UChar)0x003c; } // '<'
 private:
     static const char fgClassID;
+    static const UChar LTLT[2];
     static UnicodeString fixdesc(const UnicodeString& desc) {
-      if (desc.endsWith("<<")) {
+      if (desc.endsWith(LTLT, 2)) {
         UnicodeString result(desc, 0, desc.length()-1);
         return result;
       }
diff --git a/icu4c/source/test/intltest/itrbnf.cpp b/icu4c/source/test/intltest/itrbnf.cpp
index c4d047498c6..248f4ad3c92 100644
--- a/icu4c/source/test/intltest/itrbnf.cpp
+++ b/icu4c/source/test/intltest/itrbnf.cpp
@@ -77,21 +77,54 @@ void IntlTestRBNF::runIndexedTest(int32_t index, UBool exec, const char* &name,
 
 void IntlTestRBNF::TestHebrewFraction() {
     // this is the expected output for 123.45, with no '<' in it.
-    UChar text[] = { 
-        0x05de, 0x05d0, 0x05d4,0x0020, 
+    UChar text1[] = { 
+        0x05de, 0x05d0, 0x05d4, 0x0020, 
         0x05e2, 0x05e9, 0x05e8, 0x05d9, 0x05dd, 0x0020,
         0x05d5, 0x05e9, 0x05dc, 0x05d5, 0x05e9, 0x0020, 
         0x05e0, 0x05e7, 0x05d5, 0x05d3, 0x05d4, 0x0020,
         0x05d0, 0x05e8, 0x05d1, 0x05e2, 0x05d9, 0x05dd, 0x0020,
         0x05d5, 0x05d7, 0x05de, 0x05e9, 0x0000,
     };
+    UChar text2[] = { 
+        0x05DE, 0x05D0, 0x05D4, 0x0020, 
+        0x05E2, 0x05E9, 0x05E8, 0x05D9, 0x05DD, 0x0020, 
+        0x05D5, 0x05E9, 0x05DC, 0x05D5, 0x05E9, 0x0020, 
+        0x05E0, 0x05E7, 0x05D5, 0x05D3, 0x05D4, 0x0020, 
+        0x05D0, 0x05E4, 0x05E1, 0x0020, 
+        0x05D0, 0x05E4, 0x05E1, 0x0020, 
+        0x05D0, 0x05E8, 0x05D1, 0x05E2, 0x05D9, 0x05DD, 0x0020, 
+        0x05D5, 0x05D7, 0x05DE, 0x05E9, 0x0000,
+    };
     UErrorCode status = U_ZERO_ERROR;
-    RuleBasedNumberFormat* formatter = new RuleBasedNumberFormat(URBNF_SPELLOUT, "he_IW", status);
-    UnicodeString expected(text);
+    RuleBasedNumberFormat* formatter = new RuleBasedNumberFormat(URBNF_SPELLOUT, "he_IL", status);
     UnicodeString result;
-    formatter->format(123.450000, result);
-    if (result != expected) {
-        errln((UnicodeString)"expected '" + TestUtility::hex(expected) + "'\nbut got: '" + TestUtility::hex(result) + "'");
+    Formattable parseResult;
+    ParsePosition pp(0);
+    {
+        UnicodeString expected(text1);
+        formatter->format(123.45, result);
+        if (result != expected) {
+            errln((UnicodeString)"expected '" + TestUtility::hex(expected) + "'\nbut got: '" + TestUtility::hex(result) + "'");
+        } else {
+            formatter->parse(result, parseResult, pp);
+            if (parseResult.getDouble() != 123.45) {
+                errln("expected 123.45 but got: %g", parseResult.getDouble());
+            }
+        }
+    }
+    {
+        UnicodeString expected(text2);
+        result.remove();
+        formatter->format(123.0045, result);
+        if (result != expected) {
+            errln((UnicodeString)"expected '" + TestUtility::hex(expected) + "'\nbut got: '" + TestUtility::hex(result) + "'");
+        } else {
+            pp.setIndex(0);
+            formatter->parse(result, parseResult, pp);
+            if (parseResult.getDouble() != 123.0045) {
+                errln("expected 123.0045 but got: %g", parseResult.getDouble());
+            }
+        }
     }
     delete formatter;
 }