diff --git a/samples/c_cxx/imagenet/cached_interpolation.h b/samples/c_cxx/imagenet/cached_interpolation.h
index 43f0a9223f..f8bd1ed0fb 100644
--- a/samples/c_cxx/imagenet/cached_interpolation.h
+++ b/samples/c_cxx/imagenet/cached_interpolation.h
@@ -20,7 +20,7 @@ limitations under the License.
 struct CachedInterpolation {
   int64_t lower;  // Lower source index used in the interpolation
   int64_t upper;  // Upper source index used in the interpolation
-  // 1-D linear iterpolation scale (see:
+  // 1-D linear interpolation scale (see:
   // https://en.wikipedia.org/wiki/Bilinear_interpolation)
   float lerp;
 };
diff --git a/winml/test/concurrency/ConcurrencyTests.h b/winml/test/concurrency/ConcurrencyTests.h
index 045b96295e..46fb2813dc 100644
--- a/winml/test/concurrency/ConcurrencyTests.h
+++ b/winml/test/concurrency/ConcurrencyTests.h
@@ -37,13 +37,13 @@ WINML_TEST_CLASS_END()
 static constexpr uint32_t TABBY_CAT_INDEX = 281;
 static constexpr uint32_t TENCH_INDEX = 0;
 // concurrency bugs are often race conditions and hard to catch deterministically.
-//
-// there are several approachs to find them, from consistent testing to random
+
+// there are several approaches to find them, from consistent testing to random
 // style stress/fuzz testing
 //
 // the testing strategy for *this* test is:
 // - use a consistent and reasonable number of threads (10 vs. 1000)
-// - run them for a consistent and long enough period of time (60 seconds) .
+// - run them for a consistent and long enough period of time (60 seconds).
 // - the smaller number of threads is also to make sure memory pressure is not an issue
 //   on pre checkin CI test machines
 static constexpr uint32_t NUM_THREADS = 10;
diff --git a/winml/test/image/imageTestHelper.cpp b/winml/test/image/imageTestHelper.cpp
index d6578541db..52aeddffd8 100644
--- a/winml/test/image/imageTestHelper.cpp
+++ b/winml/test/image/imageTestHelper.cpp
@@ -70,7 +70,7 @@ namespace ImageTestHelper {
         // else if()
         // TODO: for Gray8
         else {
-            std::cerr << "Unsupportted pixelFormat";
+            std::cerr << "Unsupported pixelFormat";
         }
         return tf;
     }
@@ -91,7 +91,7 @@ namespace ImageTestHelper {
         float* pCPUTensor;
         uint32_t uCapacity;
 
-        // CPU tensorization
+        // CPU tensor initialization
         TensorFloat tf = TensorFloat::Create(shape);
         com_ptr<ITensorNative> itn = tf.as<ITensorNative>();
         itn->GetBuffer(reinterpret_cast<BYTE**>(&pCPUTensor), &uCapacity);
diff --git a/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp b/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp
index 31e2b53a8c..418bfe865c 100644
--- a/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp
+++ b/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp
@@ -222,7 +222,7 @@ static void Scenario5AsyncEval() {
 }
 
 //! Scenario6: use BindInputWithProperties - BitmapBounds, BitmapPixelFormat
-// apparently this scenario is cut for rs5. - not cut, just rewprked. move props
+// apparently this scenario is cut for rs5. - not cut, just reworked. move props
 // to the image value when that is checked in.
 static void Scenario6BindWithProperties() {
   // load a model
@@ -609,7 +609,7 @@ void SubmitEval(LearningModel model, SwapChainEntry* sessionBindings, int swapch
   // return without waiting for the submit to finish, setup the completion handler
 }
 
-//Scenario14:Load single model, run it mutliple times on a single gpu device using a fast swapchain pattern
+//Scenario14:Load single model, run it multiple times on a single gpu device using a fast swapchain pattern
 static void Scenario14RunModelSwapchain() {
   const int swapchainentrycount = 3;
   SwapChainEntry sessionBindings[swapchainentrycount];