> If there is no capitalized word

what is a capitalised "word" here, is it the first text node or do you
mean that you have to tokenise the content and find any capitilised word
in the content?

eg: <par class="Komm_H3" xml:lang="de">EINKOMMENSTEUER<linebreak/>
				<inline class="*Paragraph">B' 3

--> EINKOMMENSTEUER


		<par class="Komm_H3" xml:lang="de">
				<inline class="*Paragraph">B' 4 Abs 4 Z 


--> nothing, get previous one

			<par class="Komm_H3" xml:lang="de">BAO, DOPPELBESTEUERUNG</par>


--> ??

if you just want to get the nearest par class="Komm_H3" that has a
non-white text node then


seelct="(.|preceding-sibling::par[@class='Komm_H3'])[text()[1][normalize-space()]][last()]

David

________________________________________________________________________
This e-mail has been scanned for all viruses by Star. The
service is powered by MessageLabs. For more information on a proactive
anti-virus service working around the clock, around the globe, visit:
http://www.star.net.uk
________________________________________________________________________