diff --git a/doc.go b/doc.go
index 23c125a..c8c9add 100644
--- a/doc.go
+++ b/doc.go
@@ -83,9 +83,9 @@
 // The idea behind a query path is to allow quick access to any element, or set
 // of elements within TOML document, with a single expression.
 //
-//   result := tree.Query("$.foo.bar.baz")  // result is 'nil' if the path is not present
+//   result, err := tree.Query("$.foo.bar.baz")
 //
-// This is equivalent to:
+// This is roughly equivalent to:
 //
 //   next := tree.Get("foo")
 //   if next != nil {
@@ -96,6 +96,11 @@
 //   }
 //   result := next
 //
+// err is nil if any parsing exception occurs.
+//
+// If no node in the tree matches the query, result will simply contain an empty list of
+// items.
+//
 // As illustrated above, the query path is much more efficient, especially since
 // the structure of the TOML file can vary.  Rather than making assumptions about
 // a document's structure, a query allows the programmer to make structured
diff --git a/query_test.go b/query_test.go
index 81818ba..9bf7bad 100644
--- a/query_test.go
+++ b/query_test.go
@@ -57,3 +57,14 @@
 	}
 	assertArrayContainsInAnyOrder(t, values, "pelletier", "mypassword")
 }
+
+func TestQueryPathNotPresent(t *testing.T) {
+	config, _ := Load(`a = "hello"`)
+	results, err := config.Query("$.foo.bar")
+	if err != nil {
+		t.Fatalf("err should be nil. got %s instead", err)
+	}
+	if len(results.items) != 0 {
+		t.Fatalf("no items should be matched. %d matched instead", len(results.items))
+	}
+}