I'd first of all call $node->normalize, to make sure the text nodes aren't weirdly split apart. Then call $node->firstChild (both documented in perldoc XML::LibXML::Node). You might also compare $node->nodeType == XML_TEXT_NODE (see perldoc XML::LibXML "EXPORT TAGS" section for all the node types) to make sure it's a text node. I believe the "node name" of text nodes is '#text', so you could even try that.