diff --git a/lib/rdf/model/literal.rb b/lib/rdf/model/literal.rb
index 28e06f04..302b0c6c 100644
--- a/lib/rdf/model/literal.rb
+++ b/lib/rdf/model/literal.rb
@@ -194,7 +194,7 @@ def initialize(value, language: nil, datatype: nil, direction: nil, lexical: nil
       @string   = @string.encode(Encoding::UTF_8).freeze if instance_variable_defined?(:@string)
       @object   = @string if instance_variable_defined?(:@string) && @object.is_a?(String)
       @language = language.to_s.downcase.to_sym if language
-      @direction = direction.to_s.downcase.to_sym if direction
+      @direction = direction.to_s.to_sym if direction
       @datatype = RDF::URI(datatype).freeze if datatype
       @datatype ||= self.class.const_get(:DATATYPE) if self.class.const_defined?(:DATATYPE)
       @datatype ||= if instance_variable_defined?(:@language) && @language &&
diff --git a/lib/rdf/ntriples/reader.rb b/lib/rdf/ntriples/reader.rb
index 832f8942..4fbb21a1 100644
--- a/lib/rdf/ntriples/reader.rb
+++ b/lib/rdf/ntriples/reader.rb
@@ -51,14 +51,14 @@ class Reader < RDF::Reader
     # @see http://www.w3.org/TR/turtle/
     ##
     # Unicode regular expressions.
-    U_CHARS1         = Regexp.compile(<<-EOS.gsub(/\s+/, ''))
-                         [\\u00C0-\\u00D6]|[\\u00D8-\\u00F6]|[\\u00F8-\\u02FF]|
-                         [\\u0370-\\u037D]|[\\u037F-\\u1FFF]|[\\u200C-\\u200D]|
-                         [\\u2070-\\u218F]|[\\u2C00-\\u2FEF]|[\\u3001-\\uD7FF]|
-                         [\\uF900-\\uFDCF]|[\\uFDF0-\\uFFFD]|[\\u{10000}-\\u{EFFFF}]
-                       EOS
-    U_CHARS2         = Regexp.compile("\\u00B7|[\\u0300-\\u036F]|[\\u203F-\\u2040]").freeze
-    IRI_RANGE        = Regexp.compile("[[^<>\"{}\|\^`\\\\]&&[^\\x00-\\x20]]").freeze
+    U_CHARS1             = Regexp.compile(<<-EOS.gsub(/\s+/, ''))
+                             [\\u00C0-\\u00D6]|[\\u00D8-\\u00F6]|[\\u00F8-\\u02FF]|
+                             [\\u0370-\\u037D]|[\\u037F-\\u1FFF]|[\\u200C-\\u200D]|
+                             [\\u2070-\\u218F]|[\\u2C00-\\u2FEF]|[\\u3001-\\uD7FF]|
+                             [\\uF900-\\uFDCF]|[\\uFDF0-\\uFFFD]|[\\u{10000}-\\u{EFFFF}]
+                           EOS
+    U_CHARS2             = Regexp.compile("\\u00B7|[\\u0300-\\u036F]|[\\u203F-\\u2040]").freeze
+    IRI_RANGE            = Regexp.compile("[[^<>\"{}\|\^`\\\\]&&[^\\x00-\\x20]]").freeze
 
     PN_CHARS_BASE        = /[A-Z]|[a-z]|#{U_CHARS1}/.freeze
     PN_CHARS_U           = /_|#{PN_CHARS_BASE}/.freeze
diff --git a/spec/ntriples_spec.rb b/spec/ntriples_spec.rb
index f2828aff..0d42ff09 100644
--- a/spec/ntriples_spec.rb
+++ b/spec/ntriples_spec.rb
@@ -322,6 +322,7 @@
         {
           "language" => '<http://subj> <http://pred>  "Hello"@en .',
           "direction" => '<http://subj> <http://pred>  "Hello"@en--ltr .',
+          "direction2" => '<http://subj> <http://pred>  "Hello"@en--rtl .',
         }.each_pair do |name, triple|
           specify "test #{name}" do
             stmt = reader.new(triple, rdfstar: true).first
@@ -518,6 +519,10 @@
         %q(<http://example/s> <http://example/p> "string"@--ltr .),
         %r(Expected end of statement)
       ],
+      "xx bad dir 3" => [
+        %q(<http://example/s> <http://example/p> "string"@en--LTR .),
+        %r(Invalid Literal)
+      ],
       "nt-syntax-bad-string-05" => [
         %q(<http://example/s> <http://example/p> """abc""" .),
         %r(Expected end of statement \(found: .* \."\))