diff --git a/curve25519-dalek/CHANGELOG.md b/curve25519-dalek/CHANGELOG.md
index 1715e3b..190958a 100644
--- a/curve25519-dalek/CHANGELOG.md
+++ b/curve25519-dalek/CHANGELOG.md
@@ -123,7 +123,7 @@ major series.
 
 ### 2.1.2
 
-* Multiple documenation typo fixes.
+* Multiple documentation typo fixes.
 * Fix `alloc` feature working with stable rust.
 
 ### 2.1.1
diff --git a/curve25519-dalek/docs/ifma-notes.md b/curve25519-dalek/docs/ifma-notes.md
index c6fd3b3..faf8928 100644
--- a/curve25519-dalek/docs/ifma-notes.md
+++ b/curve25519-dalek/docs/ifma-notes.md
@@ -351,7 +351,7 @@ This computation requires 25 `vpmadd52luq` and 25 `vpmadd52huq`
 operations.  For 256-bit vectors, IFMA operations execute on an
 i3-8121U with latency 4 cycles, throughput 0.5 cycles, so executing 50
 instructions requires 25 cycles' worth of throughput.  Accumulating
-terms with coefficient \\(1\\) and \\(2\\) seperately means that the
+terms with coefficient \\(1\\) and \\(2\\) separately means that the
 longest dependency chain has length 5, so the critical path has length
 20 cycles and the bottleneck is throughput.
 
diff --git a/ed25519-dalek/tests/validation_criteria.rs b/ed25519-dalek/tests/validation_criteria.rs
index fc5b8a5..7c45a96 100644
--- a/ed25519-dalek/tests/validation_criteria.rs
+++ b/ed25519-dalek/tests/validation_criteria.rs
@@ -22,7 +22,7 @@ const VERIFY_ALLOWED_EDGECASES: &[Flag] = &[
 const VERIFY_STRICT_ALLOWED_EDGECASES: &[Flag] =
     &[Flag::LowOrderComponentA, Flag::LowOrderComponentR];
 
-/// Each variant describes a specfiic edge case that can occur in an Ed25519 signature. Refer to
+/// Each variant describes a specific edge case that can occur in an Ed25519 signature. Refer to
 /// the test vector [README][] for more info.
 ///
 /// [README]: https://github.com/C2SP/CCTV/blob/5ea85644bd035c555900a2f707f7e4c31ea65ced/ed25519vectors/README.md