diff --git a/lib/docx/elements/text.rb b/lib/docx/elements/text.rb
index fb89401..dcac793 100644
--- a/lib/docx/elements/text.rb
+++ b/lib/docx/elements/text.rb
@@ -13,6 +13,8 @@ def content
 
       def content=(args)
         @node.content = args
+        # Without xml:space="preserve", Word drops leading and trailing spaces.
+        @node['xml:space'] = 'preserve' if args.to_s.match?(/\A\s|\s\z/)
       end
 
       def initialize(node)
diff --git a/spec/docx/elements/text_spec.rb b/spec/docx/elements/text_spec.rb
index c15e979..3f9fbc2 100644
--- a/spec/docx/elements/text_spec.rb
+++ b/spec/docx/elements/text_spec.rb
@@ -23,5 +23,33 @@
 
       expect(node.content).to eq('world')
     end
+
+    # Regression tests for #187: without xml:space="preserve", Word drops
+    # leading and trailing spaces of a w:t.
+    it 'marks the node xml:space="preserve" when the content has a trailing space' do
+      text.content = 'Jane '
+
+      expect(node.to_xml).to eq('<w:t xml:space="preserve">Jane </w:t>')
+    end
+
+    it 'marks the node xml:space="preserve" when the content has a leading space' do
+      text.content = ' end'
+
+      expect(node.to_xml).to eq('<w:t xml:space="preserve"> end</w:t>')
+    end
+
+    it 'does not add xml:space when the content has no surrounding whitespace' do
+      text.content = 'Jane'
+
+      expect(node.to_xml).to eq('<w:t>Jane</w:t>')
+    end
+
+    it 'keeps an existing xml:space="preserve"' do
+      node = w_node('<w:t xml:space="preserve"> a </w:t>')
+
+      described_class.new(node).content = 'b'
+
+      expect(node.to_xml).to eq('<w:t xml:space="preserve">b</w:t>')
+    end
   end
 end