diff --git a/curve25519-dalek/CHANGELOG.md b/curve25519-dalek/CHANGELOG.md
index 619aa32..5052a47 100644
--- a/curve25519-dalek/CHANGELOG.md
+++ b/curve25519-dalek/CHANGELOG.md
@@ -216,7 +216,7 @@ besides the `rand_core` version bump, there are no other user-visible changes.
   for very large input sizes.
 * Add a more extensive randomized multiscalar multiplication consistency check
   to the test suite to prevent regressions.
-* Ensure that that multiscalar and NAF computations work correctly on extremal
+* Ensure that multiscalar and NAF computations work correctly on extremal
   `Scalar` values constructed via `from_bits`.
 
 ### 1.2.0
diff --git a/curve25519-dalek/README.md b/curve25519-dalek/README.md
index e851636..eebf300 100644
--- a/curve25519-dalek/README.md
+++ b/curve25519-dalek/README.md
@@ -99,7 +99,7 @@ Curve arithmetic is implemented and used by one of the following backends:
 
 At runtime, `curve25519-dalek` selects an arithmetic backend from the set of backends it was compiled to support. For Intel x86-64 targets, unless otherwise specified, it will build itself with `simd` support, and default to `serial` at runtime if the appropriate CPU features aren't detected. See [SIMD backend] for more details.
 
-In the future, `simd` backend may be extended to cover more instruction sets. This change will be non-breaking as this is considered as implementation detail.
+In the future, `simd` backend may be extended to cover more instruction sets. This change will be non-breaking as this is considered an implementation detail.
 
 ## Manual Backend Override
 
diff --git a/curve25519-dalek/docs/ifma-notes.md b/curve25519-dalek/docs/ifma-notes.md
index faf8928..9bd84d5 100644
--- a/curve25519-dalek/docs/ifma-notes.md
+++ b/curve25519-dalek/docs/ifma-notes.md
@@ -136,7 +136,7 @@ i3-8121U) executes 512-bit IFMA instructions at half rate compared to
 using 512-bit IFMA instructions, and this implementation uses 256-bit
 vectors.
 
-To extend this to 512-bit vectors, it's only only necessary to achieve
+To extend this to 512-bit vectors, it's only necessary to achieve
 2-way parallelism, and it's possible (with a small amount of overhead)
 to create a hybrid strategy that operates entirely within 128-bit
 lanes.  This means that cross-lane operations can use the faster
diff --git a/curve25519-dalek/src/backend/serial/curve_models/mod.rs b/curve25519-dalek/src/backend/serial/curve_models/mod.rs
index 1343d37..f524889 100644
--- a/curve25519-dalek/src/backend/serial/curve_models/mod.rs
+++ b/curve25519-dalek/src/backend/serial/curve_models/mod.rs
@@ -69,7 +69,7 @@
 //! and passed to \\( \mathbb P\^2 \\) by setting \\( x = W\_1 / W\_3
 //! \\), \\(y = W\_2 / W\_3 \\).
 //! Up to variable naming, this is the projective representation
-//! introduced in in [_Twisted Edwards
+//! introduced in [_Twisted Edwards
 //! Curves_][bernstein-birkner-joye-lange-peters-2008] by Bernstein,
 //! Birkner, Joye, Lange, and Peters.  In `curve25519-dalek`, it is
 //! represented by the `ProjectivePoint` struct.
diff --git a/curve25519-dalek/src/backend/serial/scalar_mul/pippenger.rs b/curve25519-dalek/src/backend/serial/scalar_mul/pippenger.rs
index f60d9b9..a5469b8 100644
--- a/curve25519-dalek/src/backend/serial/scalar_mul/pippenger.rs
+++ b/curve25519-dalek/src/backend/serial/scalar_mul/pippenger.rs
@@ -31,7 +31,7 @@ use crate::traits::VartimeMultiscalarMul;
 ///    Bucket 0 is not needed as it would contain points multiplied by 0.
 /// 2. Convert scalars to a radix-`2^w` representation with signed digits in `[-2^w/2, 2^w/2]`.
 ///    Note: only the last digit may equal `2^w/2`.
-/// 3. Starting with the last window, for each point `i=[0..n)` add it to a a bucket indexed by
+/// 3. Starting with the last window, for each point `i=[0..n)` add it to a bucket indexed by
 ///    the point's scalar's value in the window.
 /// 4. Once all points in a window are sorted into buckets, add buckets by multiplying each
 ///    by their index. Efficient way of doing it is to start with the last bucket and compute two sums:
diff --git a/curve25519-dalek/src/scalar.rs b/curve25519-dalek/src/scalar.rs
index 6afd74e..e400028 100644
--- a/curve25519-dalek/src/scalar.rs
+++ b/curve25519-dalek/src/scalar.rs
@@ -2043,7 +2043,7 @@ pub(crate) mod test {
 
     // Check that a * b == a.reduce() * a.reduce() for ANY scalars a,b, even ones that violate
     // invariant #1, i.e., a,b > 2^255. Old versions of ed25519-dalek did multiplication where a
-    // was reduced and b was clamped and unreduced. This checks that that was always well-defined.
+    // was reduced and b was clamped and unreduced. This checks that was always well-defined.
     #[test]
     fn test_mul_reduction_invariance() {
         let mut rng = rand::thread_rng();