This happens on trunk, p3yk branch and 2.5.

Trying:
    doctest.testfile('test_doctest4.txt', encoding='utf-8')
Expecting:
    (0, 4)
test test_doctest crashed -- <type 'exceptions.UnicodeEncodeError'>: 'ascii' codec can't encode characters in position 301-302: ordinal not in range(128)
Traceback (most recent call last):
  File "./Lib/test/regrtest.py", line 557, in runtest_inner
    indirect_test()
  File "/home/neal/python/dev/r25/Lib/test/test_doctest.py", line 2402, in test_main
    test_support.run_doctest(test_doctest, verbosity=True)
  File "/home/neal/python/dev/r25/Lib/test/test_support.py", line 467, in run_doctest
    f, t = doctest.testmod(module, verbose=verbosity)
  File "/home/neal/python/dev/r25/Lib/doctest.py", line 1799, in testmod
    runner.run(test)
  File "/home/neal/python/dev/r25/Lib/doctest.py", line 1345, in run
    return self.__run(test, compileflags, out)
  File "/home/neal/python/dev/r25/Lib/doctest.py", line 1261, in __run
    self.report_failure(out, test, example, got)
  File "/home/neal/python/dev/r25/Lib/doctest.py", line 1125, in report_failure
    self._checker.output_difference(example, got, self.optionflags))
UnicodeEncodeError: 'ascii' codec can't encode characters in position 301-302: ordinal not in range(128)

The basic issue here is that running in verbose mode echoes back the
expected values from the file, so the results from non-ascii doctest
files must be encoded before printing.

It looks to me like the DocTestRunner class must grow an '_encoding'
attribute to keep track of how to encode any verbose output. I've
attached a patch that does this.

Happy post bug day!
Mike

Here's a patch for test_doctest.py that checks the problem has been fixed.

This test still fail on Win7 with Py2.6.3rc1 in verbose mode, it works
fine in normal mode.
I attached a file with the traceback.