Scandanavian characters display O.K in HTML text but not in form fields, why?

[An...@DCSTRANS.COM]

We have an application that uses XSLT to build HTML screens. When
Scandanavian characters are output as HTML text
they  display  correctly  in  the browser. However, when they are defaulted
into form fields their &# value is displayed in the browser.
I  have  debugged  the  application and the data is correct before the XSLT
transformation takes place, but is corrupted in the
results of the transformation.

But  it's only where the data appears in form fields.  The output-method we
are using is 'xml' - why would it make any difference which result tags the
data is inside?
Should  we be using the 'html' output method?  (We tried this years ago and
it didn't work for us, so we've stuck to 'xml' by default ever since.)

To add a further complication this does not occur in all environments where
our application is installed.
Are there any general pointers regarding defaulting special characters into
form fields? Does Xalan use any system/envionment
settings to resolve character encoding?