diff --git a/README.md b/README.md
index f235aa5..da0887e 100644
--- a/README.md
+++ b/README.md
@@ -68,10 +68,12 @@ Selectors can be combined; e.g. `".class:not([attribute]) element.class"`
 
 ###Limitations
 - Attribute values in selectors currently cannot contain any spaces, since space is interpreted as a delimiter between the `ancestor` and `descendant`, `parent` and `>`, or `>` and `child` parts of the selector
-- Likewise, for the `parent > child` relation, the spaces before and after the `>` are mandatory
+- Consequently, for the `parent > child` relation, the spaces before and after the `>` are mandatory
+- Attribute values in selectors currently also cannot contain any of `#`, `.`, `[`, `]`, `:`, `(`, or `)`
 - `<!` elements are not parsed, including doctype, comments, and CDATA
 - Textnodes are not seperate entries in the tree, so the content of `<p>line1<br />line2</p>` is plainly `"line1<br />line2"`
-- All start and end tags should be explicitly specified in the text to be parsed; omitted tags (as [permitted](http://www.w3.org/TR/html5/syntax.html#optional-tags) by the the HTML spec) are NOT implied. Only the [void](http://www.w3.org/TR/html5/syntax.html#void-elements) elements naturally don't need an end tag
+- All start and end tags should be explicitly specified in the text to be parsed; omitted tags (as [permitted](http://www.w3.org/TR/html5/syntax.html#optional-tags) by the the HTML spec) are NOT implied. Only the [void](http://www.w3.org/TR/html5/syntax.html#void-elements) elements naturally don't need (and mustn't have) an end tag
+- The HTML text is not validated in any way; tag and attribute names and the nesting of different tags is completely arbitrary. The only HTML-specific part of the parser is that it knows which tags are void elements
 
 ##Examples
 See `./doc/samples.lua`
diff --git a/tst/init.lua b/tst/init.lua
index 1349809..e3b164e 100644
--- a/tst/init.lua
+++ b/tst/init.lua
@@ -27,9 +27,6 @@ function test_void()
 		else
 			assert_equal("br", n.name, "name")
 			assert_equal("", n:getcontent(), "content")
-			for _ in pairs(n.attributes) do
-				fail("should not have attributes")
-			end
 		end
 	end
 end
@@ -37,18 +34,12 @@ end
 function test_attr()
 	local tree = htmlparser.parse([[
 		<n a1 a2= a3='' a4=""
-			a5='a"5"' a6="a'6'" a7='a 7' a8='a=8'
-			a9='en-gb' a10='enen'
-			a11='one two three'
-		></n>
-		<m a9="en-us" a10></m>
-		<l a9="enen" a11="three four five"></l>
+			a5='a"5"' a6="a'6'" a7='#.[] :()' a8='|*+-=?$^%&/'
+			a9=a9
+		a10></n>
 	]])
-	assert_equal(3, #tree.nodes, "top level")
-	local n
-	for _,v in ipairs(tree.nodes) do
-		if v.name == "n" then n = v break end
-	end
+	assert_equal(1, #tree.nodes, "top level")
+	local n = tree.nodes[1]
 	assert(tree("[a1]")[n], "a1")
 	assert(tree("[a2]")[n], "a2")
 	assert(tree("[a3]")[n], "a3")
@@ -57,27 +48,91 @@ function test_attr()
 	assert(tree("[a6]")[n], "a6")
 	assert(tree("[a7]")[n], "a7")
 	assert(tree("[a8]")[n], "a8")
+	assert(tree("[a9]")[n], "a9")
+	assert(tree("[a10]")[n], "a10")
+end
+
+function test_attr_equal()
+	local tree = htmlparser.parse([[
+		<n a1 a2= a3='' a4=""
+			a5='a"5"' a6="a'6'" a7='#.[] :()' a8='|*+-=?$^%&/'
+			a9=a9
+		a10></n>
+	]])
+	assert_equal(1, #tree.nodes, "top level")
+	local n = tree.nodes[1]
 	assert(tree("[a1='']")[n], "a1=''")
 	assert(tree("[a2='']")[n], "a2=''")
 	assert(tree("[a3='']")[n], "a3=''")
 	assert(tree("[a4='']")[n], "a4=''")
 	assert(tree("[a5='a\"5\"']")[n], "a5='a\"5\"'")
 	assert(tree("[a6=\"a'6'\"]")[n], "a6=\"a'6'\"")
-	assert(tree("[a8='a=8']")[n], "a8='a=8'")
-	assert_equal(1, tree("[a10=]"):len(), "a10=")
-	assert_equal(1, tree("[a10='']"):len(), "a10=''")
-	assert_equal(2, tree("[a10!='enen']"):len(), "a10!='enen'")
-	assert_equal(2, tree("[a10!='']"):len(), "a10!=''")
-	assert_equal(3, tree("[a0!='']"):len(), "a0!=''")
-	assert_equal(0, tree("[a0='']"):len(), "a0=''")
-	assert_equal(2, tree("[a9|='en']"):len(), "a9|='en'")
-	assert_equal(3, tree("[a9^='en']"):len(), "a9^='en'")
-	assert_equal(1, tree("[a9$='en']"):len(), "a9$='en'")
-	assert_equal(1, tree("[a11~='two']"):len(), "a1~='two'")
-	assert_equal(2, tree("[a11~='three']"):len(), "a1~='three'")
-	assert_equal(1, tree("[a11~='four']"):len(), "a1~='four'")
-	assert_equal(1, tree("[a7*='7']"):len(), "a7*='7'")
-	assert_equal(1, tree("[a11*='f']"):len(), "a11*='f'")
+	-- not these characters
+	-- (because these have a special meaning as id, class, or attribute selector, hierarchy separator, or filter command)
+	-- they can occur in the HTML, but not in a selector string
+	-- assert(tree("[a7='#.[] :()']")[n], "a7='#.[] :()'")
+	assert(tree("[a8='|*+-=?$^%&/']")[n], "a8='|*+-=?$^%&/'")
+	assert(tree("[a9='a9']")[n], "a9='a9'")
+	assert(tree("[a10='']")[n], "a10=''")
+	assert(tree("[a10=]")[n], "a10=")
+end
+
+function test_attr_notequal()
+	local tree = htmlparser.parse([[
+		<n a1="a1"></n>
+		<n a1="a2"></n>
+		<n a1></n>
+		<n></n>
+	]])
+	assert_equal(4, #tree.nodes, "top level")
+	assert_equal(3, tree("[a1!='a1']"):len(), "a1!='a1'")
+	assert_equal(4, tree("[a1!='b1']"):len(), "a1!='b1'")
+	assert_equal(3, tree("[a1!='']"):len(), "a1!=''")
+	assert_equal(3, tree("[a1!=]"):len(), "a1!=")
+end
+
+function test_attr_prefix_start_end()
+	local tree = htmlparser.parse([[
+		<n a1="en-gb"></n>
+		<n a1="en-us"></n>
+		<n a1="en"></n>
+		<n a1="enen"></n>
+		<n></n>
+	]])
+	assert_equal(5, #tree.nodes, "top level")
+	assert_equal(3, tree("[a1|='en']"):len(), "a1|='en'")
+	assert_equal(4, tree("[a1^='en']"):len(), "a1^='en'")
+	assert_equal(2, tree("[a1$='en']"):len(), "a1$='en'")
+end
+
+function test_attr_word()
+	local tree = htmlparser.parse([[
+		<n a1="one two three"></n>
+		<n a1="three four five"></n>
+		<n a1></n>
+		<n></n>
+	]])
+	assert_equal(4, #tree.nodes, "top level")
+	assert_equal(1, tree("[a1~='two']"):len(), "a1~='two'")
+	assert_equal(2, tree("[a1~='three']"):len(), "a1~='three'")
+	assert_equal(1, tree("[a1~='four']"):len(), "a1~='four'")
+end
+
+function test_attr_contains()
+	local tree = htmlparser.parse([[
+		<n a1="one"></n>
+		<n a1="one two three"></n>
+		<n a1="three four five"></n>
+		<n a1=""></n>
+		<n a1></n>
+		<n></n>
+	]])
+	assert_equal(6, #tree.nodes, "top level")
+	assert_equal(2, tree("[a1*='one']"):len(), "a1*='one'")
+	assert_equal(2, tree("[a1*='t']"):len(), "a1*='t'")
+	assert_equal(1, tree("[a1*='f']"):len(), "a1*='f'")
+	assert_equal(5, tree("[a1*='']"):len(), "a1*=''")
+	assert_equal(5, tree("[a1*=]"):len(), "a1*=")
 end
 
 function test_descendants()
@@ -104,8 +159,7 @@ function test_descendants()
 			<child>not</child>
 		</arbitrary>
 	]])
-	local sel = tree("parent child")
-	assert_equal(8, sel:len(), 'parent child')
+	assert_equal(8, tree("parent child"):len(), 'parent child')
 end
 
 function test_children()
@@ -132,6 +186,5 @@ function test_children()
 			<child>not</child>
 		</arbitrary>
 	]])
-	local sel = tree("parent > child")
-	assert_equal(4, sel:len(), 'parent > child')
+	assert_equal(4, tree("parent > child"):len(), 'parent > child')
 end
\ No newline at end of file