diff --git a/babel/numbers.py b/babel/numbers.py
index b684d8647..541e9ed85 100644
--- a/babel/numbers.py
+++ b/babel/numbers.py
@@ -525,16 +525,16 @@ def format_scientific(
 
     The format pattern can also be specified explicitly:
 
-    >>> format_scientific(1234567, u'##0E00', locale='en_US')
-    u'1E06'
+    >>> format_scientific(1234567, u'##0.##E00', locale='en_US')
+    u'1.23E06'
 
     By default the locale is allowed to truncate and round a high-precision
     number by forcing its format pattern onto the decimal part. You can bypass
     this behavior with the `decimal_quantization` parameter:
 
-    >>> format_scientific(1234.9876, locale='en_US')
-    u'1E3'
-    >>> format_scientific(1234.9876, locale='en_US', decimal_quantization=False)
+    >>> format_scientific(1234.9876, u'#.##E0', locale='en_US')
+    u'1.23E3'
+    >>> format_scientific(1234.9876, u'#.##E0', locale='en_US', decimal_quantization=False)
     u'1.2349876E3'
 
     :param number: the number to format
@@ -796,8 +796,12 @@ def apply(
             frac_prec = (get_currency_precision(currency), ) * 2
 
         # Bump decimal precision to the natural precision of the number if it
-        # exceeds the one we're about to use.
-        if not decimal_quantization:
+        # exceeds the one we're about to use. This adaptative precision is only
+        # triggered if the decimal quantization is disabled or if a scientific
+        # notation pattern has a missing mandatory fractional part (as in the
+        # default '#E0' pattern). This special case has been extensively
+        # disccused at https://github.com/python-babel/babel/pull/494#issuecomment-307649969 .
+        if not decimal_quantization or (self.exp_prec and frac_prec == (0, 0)):
             frac_prec = (frac_prec[0], max([frac_prec[1], get_decimal_precision(value)]))
 
         # Render scientific notation.
diff --git a/tests/test_numbers.py b/tests/test_numbers.py
index 03164ebbb..03a4c9b08 100644
--- a/tests/test_numbers.py
+++ b/tests/test_numbers.py
@@ -467,14 +467,25 @@ def test_format_percent_quantization():
 def test_format_scientific():
     assert numbers.format_scientific(10000, locale='en_US') == u'1E4'
     assert numbers.format_scientific(4234567, u'#.#E0', locale='en_US') == u'4.2E6'
-    assert numbers.format_scientific(4234567, u'0E0000', locale='en_US') == u'4E0006'
-    assert numbers.format_scientific(4234567, u'##0E00', locale='en_US') == u'4E06'
-    assert numbers.format_scientific(4234567, u'##00E00', locale='en_US') == u'42E05'
-    assert numbers.format_scientific(4234567, u'0,000E00', locale='en_US') == u'4,235E03'
+    assert numbers.format_scientific(4234567, u'0E0000', locale='en_US') == u'4.234567E0006'
+    assert numbers.format_scientific(4234567, u'##0E00', locale='en_US') == u'4.234567E06'
+    assert numbers.format_scientific(4234567, u'##00E00', locale='en_US') == u'42.34567E05'
+    assert numbers.format_scientific(4234567, u'0,000E00', locale='en_US') == u'4,234.567E03'
     assert numbers.format_scientific(4234567, u'##0.#####E00', locale='en_US') == u'4.23457E06'
+    assert numbers.format_scientific(4234567, u'##0.##E00', locale='en_US') == u'4.23E06'
     assert numbers.format_scientific(42, u'00000.000000E0000', locale='en_US') == u'42000.000000E-0003'
 
 
+def test_default_scientific_format():
+    """ Check the scientific format method auto-correct the rendering pattern
+    in case of a missing fractional part.
+    """
+    assert numbers.format_scientific(12345, locale='en_US') == u'1.2345E4'
+    assert numbers.format_scientific(12345.678, locale='en_US') == u'1.2345678E4'
+    assert numbers.format_scientific(12345, u'#E0', locale='en_US') == u'1.2345E4'
+    assert numbers.format_scientific(12345.678, u'#E0', locale='en_US') == u'1.2345678E4'
+
+
 @pytest.mark.parametrize('input_value, expected_value', [
     ('10000', '1E4'),
     ('1', '1E0'),