diff --git a/libraries/animation/src/AnimBlendLinear.cpp b/libraries/animation/src/AnimBlendLinear.cpp
index 4e8c42dad3..c066dc92eb 100644
--- a/libraries/animation/src/AnimBlendLinear.cpp
+++ b/libraries/animation/src/AnimBlendLinear.cpp
@@ -37,6 +37,19 @@ const AnimPoseVec& AnimBlendLinear::evaluate(const AnimVariantMap& animVars, con
     } else if (_children.size() == 1) {
         _poses = _children[0]->evaluate(animVars, context, dt, triggersOut);
         context.setDebugAlpha(_children[0]->getID(), parentDebugAlpha, _children[0]->getType());
+    } else if (_children.size() == 2 && _blendType != AnimBlendType_Normal) {
+        // special case for additive blending
+        float alpha = glm::clamp(_alpha, 0.0f, 1.0f);
+        const size_t prevPoseIndex = 0;
+        const size_t nextPoseIndex = 1;
+        evaluateAndBlendChildren(animVars, context, triggersOut, alpha, prevPoseIndex, nextPoseIndex, dt);
+
+        // for animation stack debugging
+        float weight2 = alpha;
+        float weight1 = 1.0f - weight2;
+        context.setDebugAlpha(_children[prevPoseIndex]->getID(), weight1 * parentDebugAlpha, _children[prevPoseIndex]->getType());
+        context.setDebugAlpha(_children[nextPoseIndex]->getID(), weight2 * parentDebugAlpha, _children[nextPoseIndex]->getType());
+
     } else {
         float clampedAlpha = glm::clamp(_alpha, 0.0f, (float)(_children.size() - 1));
         size_t prevPoseIndex = glm::floor(clampedAlpha);
diff --git a/libraries/animation/src/AnimUtil.cpp b/libraries/animation/src/AnimUtil.cpp
index 621e714f97..e5f7cf4182 100644
--- a/libraries/animation/src/AnimUtil.cpp
+++ b/libraries/animation/src/AnimUtil.cpp
@@ -28,19 +28,22 @@ void blend(size_t numPoses, const AnimPose* a, const AnimPose* b, float alpha, A
 
 // additive blend
 void blendAdd(size_t numPoses, const AnimPose* a, const AnimPose* b, float alpha, AnimPose* result) {
+
+    const glm::quat identity = glm::quat();
     for (size_t i = 0; i < numPoses; i++) {
         const AnimPose& aPose = a[i];
         const AnimPose& bPose = b[i];
 
         result[i].scale() = lerp(aPose.scale(), bPose.scale(), alpha);
 
-        // adjust sign of bPose.rot() if necessary
-        glm::quat bTemp = bPose.rot();
-        float dot = glm::dot(aPose.rot(), bTemp);
-        if (dot < 0.0f) {
-            bTemp = -bTemp;
+        // ensure that delta has the same "polarity" as the identity quat.
+        // we don't need to do a full dot product, just sign of w is sufficient.
+        glm::quat delta = bPose.rot();
+        if (delta.w < 0.0f) {
+            delta = -delta;
         }
-        result[i].rot() = glm::normalize((alpha * bTemp) * aPose.rot());
+        delta = glm::lerp(identity, delta, alpha);
+        result[i].rot() = glm::normalize(delta * aPose.rot());
 
         result[i].trans() = aPose.trans() + (alpha * bPose.trans());
     }