diff --git a/include/gcache/ghost_cache.h b/include/gcache/ghost_cache.h
index 9a0d627..23dc4c4 100644
--- a/include/gcache/ghost_cache.h
+++ b/include/gcache/ghost_cache.h
@@ -32,7 +32,7 @@ class GhostKvCache;
  * Templated type Meta must have a field size_idx; in almost all cases, this
  * field does not need to specified; it is only useful if there is some
  * additional per-page metadata to be carried.
- * To support cache size of differnet scales, we support templates numerical
+ * To support cache size of different scales, we support templates numerical
  * type for cache size.
  */
 template <typename Hash = ghash, typename Meta = GhostMeta<uint32_t>,
@@ -245,7 +245,7 @@ GhostCache<Hash, Meta, SizeType, HashType>::access_impl(SizeType block_id,
    * 1) every node at boundary with size_idx < X should increase its size_idx
    *    and the boundary pointer shall move to its next.
    * 2) X's size_idx should be set to 0.
-   * 3) if X itself is a boundary, set that boundary to X's next (sucessor).
+   * 3) if X itself is a boundary, set that boundary to X's next (successor).
    */
   SizeType size_idx;
   if (s) {  // No new insertion
diff --git a/include/gcache/ghost_kv_cache.h b/include/gcache/ghost_kv_cache.h
index b1cefd0..f188946 100644
--- a/include/gcache/ghost_kv_cache.h
+++ b/include/gcache/ghost_kv_cache.h
@@ -131,7 +131,7 @@ class SampledGhostKvCache {
     }
     return curve;
     // should be implicitly moved by compiler
-    // avoid explict move for Return Value Optimization (RVO)
+    // avoid explicit move for Return Value Optimization (RVO)
   }
 };
 }  // namespace gcache
diff --git a/include/gcache/lru_cache.h b/include/gcache/lru_cache.h
index 66f4c2e..4e0af63 100644
--- a/include/gcache/lru_cache.h
+++ b/include/gcache/lru_cache.h
@@ -94,7 +94,7 @@ class LRUCache {
   void pin(Handle_t handle);
 
   /**
-   * The normal opeartions (`insert`/`lookup`/`release`) will only cause a node
+   * The normal operations (`insert`/`lookup`/`release`) will only cause a node
    * to flow among the lru list, the in-use list, and the free list.
    * Only `erase` will force a node to jump out of circulation, and only
    * `install` may add a node. When a node is erased, its value is trashed
@@ -196,7 +196,7 @@ class LRUCache {
   // This list is only maintained for memory efficiency purposes.
   Node_t erased_;
 
-  // Pool for additionaly allocated handles.
+  // Pool for additionally allocated handles.
   std::vector<Node_t*> extra_pool_;
 
   template <typename H, typename M, typename ST, typename HT>
@@ -242,7 +242,7 @@ inline LRUCache<Key_t, Value_t, Hash>::~LRUCache() {
     delete[] pool_;
     delete table_;
   }
-  /* `extrac_pool_` is always owned by this instance. */
+  /* `extra_pool_` is always owned by this instance. */
   for (auto e : extra_pool_) delete e;
 }
 
diff --git a/tests/test_ghost.cpp b/tests/test_ghost.cpp
index a9301f6..700e2c6 100644
--- a/tests/test_ghost.cpp
+++ b/tests/test_ghost.cpp
@@ -180,7 +180,7 @@ void bench1() {
   auto ts3 = rdtsc();
 
   for (uint32_t i = 0; i < bench_size / 8; ++i) {
-    // manually flatten the loop, becaues we cannot turn on compiler
+    // manually flatten the loop, because we cannot turn on compiler
     // optimization which will remove the whole loop since it produces nothing.
     gcache::ghash{}(i * 8);
     gcache::ghash{}(i * 8 + 1);
@@ -391,7 +391,7 @@ int main() {
   test3();   // test checkpoint and recover
   bench1();  // ghost cache w/o sampling
   bench2();  // ghost cache w/ sampling
-  bench3();  // hit rate comparsion
+  bench3();  // hit rate comparison
   bench4();  // large bench: may exceed CPU cache size
   bench5();  // real random access
   return 0;