diff --git a/src/backtrack.rs b/src/backtrack.rs
index 2eaeb72e5..6100c1730 100644
--- a/src/backtrack.rs
+++ b/src/backtrack.rs
@@ -115,8 +115,8 @@ impl<'a, 'm, 'r, 's, I: Input> Bounded<'a, 'm, 'r, 's, I> {
         // Then we reset all existing allocated space to 0.
         // Finally, we request more space if we need it.
         //
-        // This is all a little circuitous, but doing this unsafely
-        // doesn't seem to have a measurable impact on performance.
+        // This is all a little circuitous, but doing this using unchecked
+        // operations doesn't seem to have a measurable impact on performance.
         // (Probably because backtracking is limited to such small
         // inputs/regexes in the first place.)
         let visited_len =
diff --git a/src/dfa.rs b/src/dfa.rs
index 3bb4a5205..9ac0c2c39 100644
--- a/src/dfa.rs
+++ b/src/dfa.rs
@@ -848,7 +848,7 @@ impl<'a> Fsm<'a> {
     /// next_si transitions to the next state, where the transition input
     /// corresponds to text[i].
     ///
-    /// This elides bounds checks, and is therefore unsafe.
+    /// This elides bounds checks, and is therefore not safe.
     #[cfg_attr(feature = "perf-inline", inline(always))]
     unsafe fn next_si(&self, si: StatePtr, text: &[u8], i: usize) -> StatePtr {
         // What is the argument for safety here?
@@ -1688,7 +1688,7 @@ impl Transitions {
         self.num_byte_classes * mem::size_of::<StatePtr>()
     }
 
-    /// Like `next`, but uses unchecked access and is therefore unsafe.
+    /// Like `next`, but uses unchecked access and is therefore not safe.
     unsafe fn next_unchecked(&self, si: StatePtr, cls: usize) -> StatePtr {
         debug_assert!((si as usize) < self.table.len());
         debug_assert!(cls < self.num_byte_classes);
diff --git a/src/expand.rs b/src/expand.rs
index fd2ab03ac..70dbf91f4 100644
--- a/src/expand.rs
+++ b/src/expand.rs
@@ -144,7 +144,8 @@ fn find_cap_ref(replacement: &[u8]) -> Option<CaptureRef> {
     }
     // We just verified that the range 0..cap_end is valid ASCII, so it must
     // therefore be valid UTF-8. If we really cared, we could avoid this UTF-8
-    // check with either unsafe or by parsing the number straight from &[u8].
+    // check via an unchecked conversion or by parsing the number straight from
+    // &[u8].
     let cap =
         str::from_utf8(&rep[i..cap_end]).expect("valid UTF-8 capture name");
     Some(CaptureRef {