diff --git a/interface/src/avatar/AvatarActionHold.cpp b/interface/src/avatar/AvatarActionHold.cpp
index a1826076fa..5fb9c9a0ee 100644
--- a/interface/src/avatar/AvatarActionHold.cpp
+++ b/interface/src/avatar/AvatarActionHold.cpp
@@ -569,7 +569,7 @@ void AvatarActionHold::lateAvatarUpdate(const AnimPose& prePhysicsRoomPose, cons
     }
 
     btTransform worldTrans = rigidBody->getWorldTransform();
-    AnimPose worldBodyPose(1.0f, bulletToGLM(worldTrans.getRotation()), bulletToGLM(worldTrans.getOrigin()));
+    AnimPose worldBodyPose(glm::vec3(1), bulletToGLM(worldTrans.getRotation()), bulletToGLM(worldTrans.getOrigin()));
 
     // transform the body transform into sensor space with the prePhysics sensor-to-world matrix.
     // then transform it back into world uisng the postAvatarUpdate sensor-to-world matrix.
diff --git a/interface/src/avatar/MyAvatar.cpp b/interface/src/avatar/MyAvatar.cpp
index 6fef47da8e..63c80ed27a 100755
--- a/interface/src/avatar/MyAvatar.cpp
+++ b/interface/src/avatar/MyAvatar.cpp
@@ -2980,7 +2980,7 @@ void MyAvatar::postUpdate(float deltaTime, const render::ScenePointer& scene) {
         auto animSkeleton = _skeletonModel->getRig().getAnimSkeleton();
 
         // the rig is in the skeletonModel frame
-        AnimPose xform(1.0f, _skeletonModel->getRotation(), _skeletonModel->getTranslation());
+        AnimPose xform(glm::vec3(1), _skeletonModel->getRotation(), _skeletonModel->getTranslation());
 
         if (_enableDebugDrawDefaultPose && animSkeleton) {
             glm::vec4 gray(0.2f, 0.2f, 0.2f, 0.2f);
@@ -3025,7 +3025,7 @@ void MyAvatar::postUpdate(float deltaTime, const render::ScenePointer& scene) {
     updateHoldActions(_prePhysicsRoomPose, postUpdateRoomPose);
 
     if (_enableDebugDrawDetailedCollision) {
-        AnimPose rigToWorldPose(1.0f, getWorldOrientation() * Quaternions::Y_180, getWorldPosition());
+        AnimPose rigToWorldPose(glm::vec3(1.0f), getWorldOrientation() * Quaternions::Y_180, getWorldPosition());
         const int NUM_DEBUG_COLORS = 8;
         const glm::vec4 DEBUG_COLORS[NUM_DEBUG_COLORS] = {
             glm::vec4(1.0f, 1.0f, 1.0f, 1.0f),
@@ -4821,7 +4821,7 @@ void MyAvatar::FollowHelper::prePhysicsUpdate(MyAvatar& myAvatar, const glm::mat
     swingTwistDecomposition(hipsinWorldSpace, avatarUpWorld, resultingSwingInWorld, resultingTwistInWorld);
 
     // remove scale present from sensorToWorldMatrix
-    followWorldPose.scale() = 1.0f;
+    followWorldPose.scale() = glm::vec3(1.0f);
 
     if (isActive(Rotation)) {
             //use the hmd reading for the hips follow
diff --git a/interface/src/avatar/MySkeletonModel.cpp b/interface/src/avatar/MySkeletonModel.cpp
index 253cc891ee..26d69841d0 100755
--- a/interface/src/avatar/MySkeletonModel.cpp
+++ b/interface/src/avatar/MySkeletonModel.cpp
@@ -41,7 +41,7 @@ static AnimPose computeHipsInSensorFrame(MyAvatar* myAvatar, bool isFlying) {
     if (myAvatar->isJointPinned(hipsIndex)) {
         Transform avatarTransform = myAvatar->getTransform();
         AnimPose result = AnimPose(worldToSensorMat * avatarTransform.getMatrix() * Matrices::Y_180);
-        result.scale() = 1.0f;
+        result.scale() = glm::vec3(1.0f, 1.0f, 1.0f);
         return result;
     }
 
@@ -108,7 +108,7 @@ void MySkeletonModel::updateRig(float deltaTime, glm::mat4 parentTransform) {
 
     Rig::ControllerParameters params;
 
-    AnimPose avatarToRigPose(1.0f, Quaternions::Y_180, glm::vec3(0.0f));
+    AnimPose avatarToRigPose(glm::vec3(1.0f), Quaternions::Y_180, glm::vec3(0.0f));
 
     glm::mat4 rigToAvatarMatrix = Matrices::Y_180;
     glm::mat4 avatarToWorldMatrix = createMatFromQuatAndPos(myAvatar->getWorldOrientation(), myAvatar->getWorldPosition());
@@ -127,7 +127,7 @@ void MySkeletonModel::updateRig(float deltaTime, glm::mat4 parentTransform) {
         // preMult 180 is necessary to convert from avatar to rig coordinates.
         // postMult 180 is necessary to convert head from -z forward to z forward.
         glm::quat headRot = Quaternions::Y_180 * head->getFinalOrientationInLocalFrame() * Quaternions::Y_180;
-        params.primaryControllerPoses[Rig::PrimaryControllerType_Head] = AnimPose(1.0f, headRot, glm::vec3(0.0f));
+        params.primaryControllerPoses[Rig::PrimaryControllerType_Head] = AnimPose(glm::vec3(1.0f), headRot, glm::vec3(0.0f));
         params.primaryControllerFlags[Rig::PrimaryControllerType_Head] = 0;
     }
 
diff --git a/libraries/animation/src/AnimClip.cpp b/libraries/animation/src/AnimClip.cpp
index 71b876ff8c..1adc04ee1b 100644
--- a/libraries/animation/src/AnimClip.cpp
+++ b/libraries/animation/src/AnimClip.cpp
@@ -140,10 +140,10 @@ void AnimClip::copyFromNetworkAnim() {
                 postRot = animSkeleton.getPostRotationPose(animJoint);
 
                 // cancel out scale
-                preRot.scale() = 1.0f;
-                postRot.scale() = 1.0f;
+                preRot.scale() = glm::vec3(1.0f);
+                postRot.scale() = glm::vec3(1.0f);
 
-                AnimPose rot(1.0f, hfmAnimRot, glm::vec3());
+                AnimPose rot(glm::vec3(1.0f), hfmAnimRot, glm::vec3());
 
                 // adjust translation offsets, so large translation animatons on the reference skeleton
                 // will be adjusted when played on a skeleton with short limbs.
@@ -155,7 +155,7 @@ void AnimClip::copyFromNetworkAnim() {
                     boneLengthScale = glm::length(relDefaultPose.trans()) / glm::length(hfmZeroTrans);
                 }
 
-                AnimPose trans = AnimPose(1.0f, glm::quat(), relDefaultPose.trans() + boneLengthScale * (hfmAnimTrans - hfmZeroTrans));
+                AnimPose trans = AnimPose(glm::vec3(1.0f), glm::quat(), relDefaultPose.trans() + boneLengthScale * (hfmAnimTrans - hfmZeroTrans));
 
                 _anim[frame][skeletonJoint] = trans * preRot * rot * postRot;
             }
diff --git a/libraries/animation/src/AnimInverseKinematics.cpp b/libraries/animation/src/AnimInverseKinematics.cpp
index 7af9e81889..a1809f3438 100644
--- a/libraries/animation/src/AnimInverseKinematics.cpp
+++ b/libraries/animation/src/AnimInverseKinematics.cpp
@@ -552,7 +552,7 @@ void AnimInverseKinematics::solveTargetWithCCD(const AnimContext& context, const
                 AnimPose accum = absolutePoses[_hipsIndex];
                 AnimPose baseParentPose = absolutePoses[_hipsIndex];
                 for (int i = (int)chainDepth - 1; i >= 0; i--) {
-                    accum = accum * AnimPose(1.0f, jointChainInfoOut.jointInfoVec[i].rot, jointChainInfoOut.jointInfoVec[i].trans);
+                    accum = accum * AnimPose(glm::vec3(1.0f), jointChainInfoOut.jointInfoVec[i].rot, jointChainInfoOut.jointInfoVec[i].trans);
                     postAbsPoses[i] = accum;
                     if (jointChainInfoOut.jointInfoVec[i].jointIndex == topJointIndex) {
                         topChainIndex = i;
@@ -734,7 +734,7 @@ void AnimInverseKinematics::computeAndCacheSplineJointInfosForIKTarget(const Ani
         glm::mat3 m(u, v, glm::cross(u, v));
         glm::quat rot = glm::normalize(glm::quat_cast(m));
 
-        AnimPose pose(1.0f, rot, spline(t));
+        AnimPose pose(glm::vec3(1.0f), rot, spline(t));
         AnimPose offsetPose = pose.inverse() * defaultPose;
 
         SplineJointInfo splineJointInfo = { index, ratio, offsetPose };
@@ -767,7 +767,7 @@ void AnimInverseKinematics::solveTargetWithSpline(const AnimContext& context, co
     const int baseIndex = _hipsIndex;
 
     // build spline from tip to base
-    AnimPose tipPose = AnimPose(1.0f, target.getRotation(), target.getTranslation());
+    AnimPose tipPose = AnimPose(glm::vec3(1.0f), target.getRotation(), target.getTranslation());
     AnimPose basePose = absolutePoses[baseIndex];
     CubicHermiteSplineFunctorWithArcLength spline;
     if (target.getIndex() == _headIndex) {
@@ -815,7 +815,7 @@ void AnimInverseKinematics::solveTargetWithSpline(const AnimContext& context, co
             glm::mat3 m(u, v, glm::cross(u, v));
             glm::quat rot = glm::normalize(glm::quat_cast(m));
 
-            AnimPose desiredAbsPose = AnimPose(1.0f, rot, trans) * splineJointInfo.offsetPose;
+            AnimPose desiredAbsPose = AnimPose(glm::vec3(1.0f), rot, trans) * splineJointInfo.offsetPose;
 
             // apply flex coefficent
             AnimPose flexedAbsPose;
@@ -965,7 +965,7 @@ const AnimPoseVec& AnimInverseKinematics::overlay(const AnimVariantMap& animVars
                     }
 
                     _relativePoses[_hipsIndex] = parentAbsPose.inverse() * absPose;
-                    _relativePoses[_hipsIndex].scale() = 1.0f;
+                    _relativePoses[_hipsIndex].scale() = glm::vec3(1.0f);
                 }
 
                 // if there is an active jointChainInfo for the hips store the post shifted hips into it.
@@ -1753,7 +1753,7 @@ void AnimInverseKinematics::setSecondaryTargets(const AnimContext& context) {
     AnimPose rigToGeometryPose = AnimPose(glm::inverse(context.getGeometryToRigMatrix()));
     for (auto& iter : _secondaryTargetsInRigFrame) {
         AnimPose absPose = rigToGeometryPose * iter.second;
-        absPose.scale() = 1.0f;
+        absPose.scale() = glm::vec3(1.0f);
 
         AnimPose parentAbsPose;
         int parentIndex = _skeleton->getParentIndex(iter.first);
@@ -1825,7 +1825,7 @@ void AnimInverseKinematics::debugDrawSpineSplines(const AnimContext& context, co
         const int baseIndex = _hipsIndex;
 
         // build spline
-        AnimPose tipPose = AnimPose(1.0f, target.getRotation(), target.getTranslation());
+        AnimPose tipPose = AnimPose(glm::vec3(1.0f), target.getRotation(), target.getTranslation());
         AnimPose basePose = _skeleton->getAbsolutePose(baseIndex, _relativePoses);
 
         CubicHermiteSplineFunctorWithArcLength spline;
diff --git a/libraries/animation/src/AnimManipulator.cpp b/libraries/animation/src/AnimManipulator.cpp
index c75c9865bb..1146cbb19a 100644
--- a/libraries/animation/src/AnimManipulator.cpp
+++ b/libraries/animation/src/AnimManipulator.cpp
@@ -172,5 +172,5 @@ AnimPose AnimManipulator::computeRelativePoseFromJointVar(const AnimVariantMap&
             break;
     }
 
-    return AnimPose(1.0f, relRot, relTrans);
+    return AnimPose(glm::vec3(1), relRot, relTrans);
 }
diff --git a/libraries/animation/src/AnimPoleVectorConstraint.cpp b/libraries/animation/src/AnimPoleVectorConstraint.cpp
index 96e4e67261..f017fe2348 100644
--- a/libraries/animation/src/AnimPoleVectorConstraint.cpp
+++ b/libraries/animation/src/AnimPoleVectorConstraint.cpp
@@ -95,7 +95,7 @@ const AnimPoseVec& AnimPoleVectorConstraint::evaluate(const AnimVariantMap& anim
     AnimPose tipPose = ikChain.getAbsolutePoseFromJointIndex(_tipJointIndex);
 
     // Look up refVector from animVars, make sure to convert into geom space.
-    glm::vec3 refVector = midPose.xformVector(_referenceVector);
+    glm::vec3 refVector = midPose.xformVectorFast(_referenceVector);
     float refVectorLength = glm::length(refVector);
 
     glm::vec3 axis = basePose.trans() - tipPose.trans();
diff --git a/libraries/animation/src/AnimPose.cpp b/libraries/animation/src/AnimPose.cpp
index 7b80e96ed5..d77514e691 100644
--- a/libraries/animation/src/AnimPose.cpp
+++ b/libraries/animation/src/AnimPose.cpp
@@ -14,14 +14,15 @@
 #include <glm/gtc/matrix_transform.hpp>
 #include "AnimUtil.h"
 
-const AnimPose AnimPose::identity = AnimPose(1.0f, glm::quat(), glm::vec3(0.0f));
+const AnimPose AnimPose::identity = AnimPose(glm::vec3(1.0f),
+                                             glm::quat(),
+                                             glm::vec3(0.0f));
 
 AnimPose::AnimPose(const glm::mat4& mat) {
     static const float EPSILON = 0.0001f;
-    glm::vec3 scale = extractScale(mat);
+    _scale = extractScale(mat);
     // quat_cast doesn't work so well with scaled matrices, so cancel it out.
-    glm::mat4 tmp = glm::scale(mat, 1.0f / scale);
-    _scale = extractUniformScale(scale);
+    glm::mat4 tmp = glm::scale(mat, 1.0f / _scale);
     _rot = glm::quat_cast(tmp);
     float lengthSquared = glm::length2(_rot);
     if (glm::abs(lengthSquared - 1.0f) > EPSILON) {
@@ -39,15 +40,25 @@ glm::vec3 AnimPose::xformPoint(const glm::vec3& rhs) const {
     return *this * rhs;
 }
 
+// really slow, but accurate for transforms with non-uniform scale
 glm::vec3 AnimPose::xformVector(const glm::vec3& rhs) const {
+    glm::vec3 xAxis = _rot * glm::vec3(_scale.x, 0.0f, 0.0f);
+    glm::vec3 yAxis = _rot * glm::vec3(0.0f, _scale.y, 0.0f);
+    glm::vec3 zAxis = _rot * glm::vec3(0.0f, 0.0f, _scale.z);
+    glm::mat3 mat(xAxis, yAxis, zAxis);
+    glm::mat3 transInvMat = glm::inverse(glm::transpose(mat));
+    return transInvMat * rhs;
+}
+
+// faster, but does not handle non-uniform scale correctly.
+glm::vec3 AnimPose::xformVectorFast(const glm::vec3& rhs) const {
     return _rot * (_scale * rhs);
 }
 
 AnimPose AnimPose::operator*(const AnimPose& rhs) const {
-    float scale = _scale * rhs._scale;
-    glm::quat rot = _rot * rhs._rot;
-    glm::vec3 trans = _trans + (_rot * (_scale * rhs._trans));
-    return AnimPose(scale, rot, trans);
+    glm::mat4 result;
+    glm_mat4u_mul(*this, rhs, result);
+    return AnimPose(result);
 }
 
 AnimPose AnimPose::inverse() const {
@@ -60,10 +71,11 @@ AnimPose AnimPose::mirror() const {
 }
 
 AnimPose::operator glm::mat4() const {
-    glm::vec3 xAxis = _rot * glm::vec3(_scale, 0.0f, 0.0f);
-    glm::vec3 yAxis = _rot * glm::vec3(0.0f, _scale, 0.0f);
-    glm::vec3 zAxis = _rot * glm::vec3(0.0f, 0.0f, _scale);
-    return glm::mat4(glm::vec4(xAxis, 0.0f), glm::vec4(yAxis, 0.0f), glm::vec4(zAxis, 0.0f), glm::vec4(_trans, 1.0f));
+    glm::vec3 xAxis = _rot * glm::vec3(_scale.x, 0.0f, 0.0f);
+    glm::vec3 yAxis = _rot * glm::vec3(0.0f, _scale.y, 0.0f);
+    glm::vec3 zAxis = _rot * glm::vec3(0.0f, 0.0f, _scale.z);
+    return glm::mat4(glm::vec4(xAxis, 0.0f), glm::vec4(yAxis, 0.0f),
+        glm::vec4(zAxis, 0.0f), glm::vec4(_trans, 1.0f));
 }
 
 void AnimPose::blend(const AnimPose& srcPose, float alpha) {
diff --git a/libraries/animation/src/AnimPose.h b/libraries/animation/src/AnimPose.h
index 4d6dee1987..fb693dc31d 100644
--- a/libraries/animation/src/AnimPose.h
+++ b/libraries/animation/src/AnimPose.h
@@ -21,13 +21,16 @@ class AnimPose {
 public:
     AnimPose() {}
     explicit AnimPose(const glm::mat4& mat);
-    explicit AnimPose(const glm::quat& rotIn) : _rot(rotIn), _trans(0.0f), _scale(1.0f) {}
-    AnimPose(const glm::quat& rotIn, const glm::vec3& transIn) : _rot(rotIn), _trans(transIn), _scale(1.0f) {}
-    AnimPose(float scaleIn, const glm::quat& rotIn, const glm::vec3& transIn) : _rot(rotIn), _trans(transIn), _scale(scaleIn) {}
+
+    explicit AnimPose(const glm::quat& rotIn) : _scale(1.0f), _rot(rotIn), _trans(0.0f) {}
+    AnimPose(const glm::quat& rotIn, const glm::vec3& transIn) : _scale(1.0f), _rot(rotIn), _trans(transIn) {}
+    AnimPose(const glm::vec3& scaleIn, const glm::quat& rotIn, const glm::vec3& transIn) : _scale(scaleIn), _rot(rotIn), _trans(transIn) {}
+
     static const AnimPose identity;
 
     glm::vec3 xformPoint(const glm::vec3& rhs) const;
     glm::vec3 xformVector(const glm::vec3& rhs) const;  // really slow, but accurate for transforms with non-uniform scale
+    glm::vec3 xformVectorFast(const glm::vec3& rhs) const;  // faster, but does not handle non-uniform scale correctly.
 
     glm::vec3 operator*(const glm::vec3& rhs) const; // same as xformPoint
     AnimPose operator*(const AnimPose& rhs) const;
@@ -36,8 +39,8 @@ public:
     AnimPose mirror() const;
     operator glm::mat4() const;
 
-    float scale() const { return _scale; }
-    float& scale() { return _scale; }
+    const glm::vec3& scale() const { return _scale; }
+    glm::vec3& scale() { return _scale; }
 
     const glm::quat& rot() const { return _rot; }
     glm::quat& rot() { return _rot; }
@@ -49,13 +52,13 @@ public:
 
 private:
     friend QDebug operator<<(QDebug debug, const AnimPose& pose);
+    glm::vec3 _scale { 1.0f };
     glm::quat _rot;
     glm::vec3 _trans;
-    float _scale { 1.0f };  // uniform scale only.
 };
 
 inline QDebug operator<<(QDebug debug, const AnimPose& pose) {
-    debug << "AnimPose, trans = (" << pose.trans().x << pose.trans().y << pose.trans().z << "), rot = (" << pose.rot().x << pose.rot().y << pose.rot().z << pose.rot().w << "), scale =" << pose.scale();
+    debug << "AnimPose, trans = (" << pose.trans().x << pose.trans().y << pose.trans().z << "), rot = (" << pose.rot().x << pose.rot().y << pose.rot().z << pose.rot().w << "), scale = (" << pose.scale().x << pose.scale().y << pose.scale().z << ")";
     return debug;
 }
 
diff --git a/libraries/animation/src/Rig.cpp b/libraries/animation/src/Rig.cpp
index d09de36a14..a9c57a4a15 100644
--- a/libraries/animation/src/Rig.cpp
+++ b/libraries/animation/src/Rig.cpp
@@ -1326,7 +1326,7 @@ static bool findPointKDopDisplacement(const glm::vec3& point, const AnimPose& sh
         }
         if (slabCount == (DOP14_COUNT / 2) && minDisplacementLen != FLT_MAX) {
             // we are within the k-dop so push the point along the minimum displacement found
-            displacementOut = shapePose.xformVector(minDisplacement);
+            displacementOut = shapePose.xformVectorFast(minDisplacement);
             return true;
         } else {
             // point is outside of kdop
@@ -1335,7 +1335,7 @@ static bool findPointKDopDisplacement(const glm::vec3& point, const AnimPose& sh
     } else {
         // point is directly on top of shapeInfo.avgPoint.
         // push the point out along the x axis.
-        displacementOut = shapePose.xformVector(shapeInfo.points[0]);
+        displacementOut = shapePose.xformVectorFast(shapeInfo.points[0]);
         return true;
     }
 }
diff --git a/libraries/avatars-renderer/src/avatars-renderer/Avatar.cpp b/libraries/avatars-renderer/src/avatars-renderer/Avatar.cpp
index daef0e411a..1d62aed5b4 100644
--- a/libraries/avatars-renderer/src/avatars-renderer/Avatar.cpp
+++ b/libraries/avatars-renderer/src/avatars-renderer/Avatar.cpp
@@ -1977,12 +1977,12 @@ float Avatar::getUnscaledEyeHeightFromSkeleton() const {
         auto& rig = _skeletonModel->getRig();
 
         // Normally the model offset transform will contain the avatar scale factor, we explicitly remove it here.
-        AnimPose modelOffsetWithoutAvatarScale(1.0f, rig.getModelOffsetPose().rot(), rig.getModelOffsetPose().trans());
+        AnimPose modelOffsetWithoutAvatarScale(glm::vec3(1.0f), rig.getModelOffsetPose().rot(), rig.getModelOffsetPose().trans());
         AnimPose geomToRigWithoutAvatarScale = modelOffsetWithoutAvatarScale * rig.getGeometryOffsetPose();
 
         // This factor can be used to scale distances in the geometry frame into the unscaled rig frame.
         // Typically it will be the unit conversion from cm to m.
-        float scaleFactor = geomToRigWithoutAvatarScale.scale();
+        float scaleFactor = geomToRigWithoutAvatarScale.scale().x;  // in practice this always a uniform scale factor.
 
         int headTopJoint = rig.indexOfJoint("HeadTop_End");
         int headJoint = rig.indexOfJoint("Head");
diff --git a/libraries/render-utils/src/AnimDebugDraw.cpp b/libraries/render-utils/src/AnimDebugDraw.cpp
index 8944ae7996..bf528ee5f0 100644
--- a/libraries/render-utils/src/AnimDebugDraw.cpp
+++ b/libraries/render-utils/src/AnimDebugDraw.cpp
@@ -374,7 +374,7 @@ void AnimDebugDraw::update() {
             glm::vec4 color = std::get<3>(iter.second);
 
             for (int i = 0; i < skeleton->getNumJoints(); i++) {
-                const float radius = BONE_RADIUS / (absPoses[i].scale() * rootPose.scale());
+                const float radius = BONE_RADIUS / (absPoses[i].scale().x * rootPose.scale().x);
 
                 // draw bone
                 addBone(rootPose, absPoses[i], radius, color, v);
@@ -394,16 +394,16 @@ void AnimDebugDraw::update() {
             glm::vec3 pos = std::get<1>(iter.second);
             glm::vec4 color = std::get<2>(iter.second);
             const float radius = POSE_RADIUS;
-            addBone(AnimPose::identity, AnimPose(1.0f, rot, pos), radius, color, v);
+            addBone(AnimPose::identity, AnimPose(glm::vec3(1), rot, pos), radius, color, v);
         }
 
-        AnimPose myAvatarPose(1.0f, DebugDraw::getInstance().getMyAvatarRot(), DebugDraw::getInstance().getMyAvatarPos());
+        AnimPose myAvatarPose(glm::vec3(1), DebugDraw::getInstance().getMyAvatarRot(), DebugDraw::getInstance().getMyAvatarPos());
         for (auto& iter : myAvatarMarkerMap) {
             glm::quat rot = std::get<0>(iter.second);
             glm::vec3 pos = std::get<1>(iter.second);
             glm::vec4 color = std::get<2>(iter.second);
             const float radius = POSE_RADIUS;
-            addBone(myAvatarPose, AnimPose(1.0f, rot, pos), radius, color, v);
+            addBone(myAvatarPose, AnimPose(glm::vec3(1), rot, pos), radius, color, v);
         }
 
         // draw rays from shared DebugDraw singleton
diff --git a/libraries/render-utils/src/CauterizedModel.cpp b/libraries/render-utils/src/CauterizedModel.cpp
index b70925201a..81a81c5602 100644
--- a/libraries/render-utils/src/CauterizedModel.cpp
+++ b/libraries/render-utils/src/CauterizedModel.cpp
@@ -122,7 +122,7 @@ void CauterizedModel::updateClusterMatrices() {
 
             if (_useDualQuaternionSkinning) {
                 auto jointPose = _rig.getJointPose(cluster.jointIndex);
-                Transform jointTransform(jointPose.rot(), glm::vec3(jointPose.scale()), jointPose.trans());
+                Transform jointTransform(jointPose.rot(), jointPose.scale(), jointPose.trans());
                 Transform clusterTransform;
                 Transform::mult(clusterTransform, jointTransform, _rig.getAnimSkeleton()->getClusterBindMatricesOriginalValues(meshIndex, clusterIndex).inverseBindTransform);
                 state.clusterDualQuaternions[j] = Model::TransformDualQuaternion(clusterTransform);
@@ -138,7 +138,7 @@ void CauterizedModel::updateClusterMatrices() {
     if (!_cauterizeBoneSet.empty()) {
 
         AnimPose cauterizePose = _rig.getJointPose(_rig.indexOfJoint("Neck"));
-        cauterizePose.scale() = 0.0001f;
+        cauterizePose.scale() = glm::vec3(0.0001f, 0.0001f, 0.0001f);
 
         static const glm::mat4 zeroScale(
             glm::vec4(0.0001f, 0.0f, 0.0f, 0.0f),
@@ -161,7 +161,7 @@ void CauterizedModel::updateClusterMatrices() {
                         // not cauterized so just copy the value from the non-cauterized version.
                         state.clusterDualQuaternions[j] = _meshStates[i].clusterDualQuaternions[j];
                     } else {
-                        Transform jointTransform(cauterizePose.rot(), glm::vec3(cauterizePose.scale()), cauterizePose.trans());
+                        Transform jointTransform(cauterizePose.rot(), cauterizePose.scale(), cauterizePose.trans());
                         Transform clusterTransform;
                         Transform::mult(clusterTransform, jointTransform, _rig.getAnimSkeleton()->getClusterBindMatricesOriginalValues(meshIndex, clusterIndex).inverseBindTransform);
                         state.clusterDualQuaternions[j] = Model::TransformDualQuaternion(clusterTransform);
diff --git a/libraries/render-utils/src/Model.cpp b/libraries/render-utils/src/Model.cpp
index 260e25009e..0206bd6963 100644
--- a/libraries/render-utils/src/Model.cpp
+++ b/libraries/render-utils/src/Model.cpp
@@ -1364,7 +1364,7 @@ void Model::updateClusterMatrices() {
 
             if (_useDualQuaternionSkinning) {
                 auto jointPose = _rig.getJointPose(cluster.jointIndex);
-                Transform jointTransform(jointPose.rot(), glm::vec3(jointPose.scale()), jointPose.trans());
+                Transform jointTransform(jointPose.rot(), jointPose.scale(), jointPose.trans());
                 Transform clusterTransform;
                 Transform::mult(clusterTransform, jointTransform, _rig.getAnimSkeleton()->getClusterBindMatricesOriginalValues(meshIndex, clusterIndex).inverseBindTransform);
                 state.clusterDualQuaternions[j] = Model::TransformDualQuaternion(clusterTransform);
diff --git a/libraries/render-utils/src/Model.h b/libraries/render-utils/src/Model.h
index c55178a21a..aadfca78ba 100644
--- a/libraries/render-utils/src/Model.h
+++ b/libraries/render-utils/src/Model.h
@@ -298,9 +298,9 @@ public:
         TransformDualQuaternion() {}
         TransformDualQuaternion(const glm::mat4& m) {
             AnimPose p(m);
-            _scale.x = p.scale();
-            _scale.y = p.scale();
-            _scale.z = p.scale();
+            _scale.x = p.scale().x;
+            _scale.y = p.scale().y;
+            _scale.z = p.scale().z;
             _scale.w = 0.0f;
             _dq = DualQuaternion(p.rot(), p.trans());
         }
diff --git a/tests/animation/src/AnimTests.cpp b/tests/animation/src/AnimTests.cpp
index 2a49846b6b..0cd9571e22 100644
--- a/tests/animation/src/AnimTests.cpp
+++ b/tests/animation/src/AnimTests.cpp
@@ -22,11 +22,9 @@
 #include <ResourceRequestObserver.h>
 #include <StatTracker.h>
 #include <test-utils/QTestExtensions.h>
-#include <test-utils/GLMTestUtils.h>
 
 QTEST_MAIN(AnimTests)
 
-
 const float TEST_EPSILON = 0.001f;
 
 void AnimTests::initTestCase() {
@@ -374,10 +372,16 @@ void AnimTests::testAnimPose() {
     const glm::quat ROT_Y_180 = glm::angleAxis(PI, glm::vec3(0.0f, 1.0, 0.0f));
     const glm::quat ROT_Z_30 = glm::angleAxis(PI / 6.0f, glm::vec3(1.0f, 0.0f, 0.0f));
 
-    std::vector<float> scaleVec = {
-        1.0f,
-        2.0f,
-        0.5f
+    std::vector<glm::vec3> scaleVec = {
+        glm::vec3(1),
+        glm::vec3(2.0f, 1.0f, 1.0f),
+        glm::vec3(1.0f, 0.5f, 1.0f),
+        glm::vec3(1.0f, 1.0f, 1.5f),
+        glm::vec3(2.0f, 0.5f, 1.5f),
+        glm::vec3(-2.0f, 0.5f, 1.5f),
+        glm::vec3(2.0f, -0.5f, 1.5f),
+        glm::vec3(2.0f, 0.5f, -1.5f),
+        glm::vec3(-2.0f, -0.5f, -1.5f),
     };
 
     std::vector<glm::quat> rotVec = {
@@ -407,7 +411,7 @@ void AnimTests::testAnimPose() {
             for (auto& trans : transVec) {
 
                 // build a matrix the old fashioned way.
-                glm::mat4 scaleMat = glm::scale(glm::mat4(), glm::vec3(scale));
+                glm::mat4 scaleMat = glm::scale(glm::mat4(), scale);
                 glm::mat4 rotTransMat = createMatFromQuatAndPos(rot, trans);
                 glm::mat4 rawMat = rotTransMat * scaleMat;
 
@@ -425,7 +429,7 @@ void AnimTests::testAnimPose() {
             for (auto& trans : transVec) {
 
                 // build a matrix the old fashioned way.
-                glm::mat4 scaleMat = glm::scale(glm::mat4(), glm::vec3(scale));
+                glm::mat4 scaleMat = glm::scale(glm::mat4(), scale);
                 glm::mat4 rotTransMat = createMatFromQuatAndPos(rot, trans);
                 glm::mat4 rawMat = rotTransMat * scaleMat;
 
@@ -441,91 +445,6 @@ void AnimTests::testAnimPose() {
     }
 }
 
-void AnimTests::testAnimPoseMultiply() {
-    const float PI = (float)M_PI;
-    const glm::quat ROT_X_90 = glm::angleAxis(PI / 2.0f, glm::vec3(1.0f, 0.0f, 0.0f));
-    const glm::quat ROT_Y_180 = glm::angleAxis(PI, glm::vec3(0.0f, 1.0, 0.0f));
-    const glm::quat ROT_Z_30 = glm::angleAxis(PI / 6.0f, glm::vec3(1.0f, 0.0f, 0.0f));
-
-    std::vector<float> scaleVec = {
-        1.0f,
-        2.0f,
-        0.5f,
-    };
-
-    std::vector<glm::quat> rotVec = {
-        glm::quat(),
-        ROT_X_90,
-        ROT_Y_180,
-        ROT_Z_30,
-        ROT_X_90 * ROT_Y_180 * ROT_Z_30,
-        -ROT_Y_180
-    };
-
-    std::vector<glm::vec3> transVec = {
-        glm::vec3(),
-        glm::vec3(10.0f, 0.0f, 0.0f),
-        glm::vec3(0.0f, 5.0f, 0.0f),
-        glm::vec3(0.0f, 0.0f, 7.5f),
-        glm::vec3(10.0f, 5.0f, 7.5f),
-        glm::vec3(-10.0f, 5.0f, 7.5f),
-        glm::vec3(10.0f, -5.0f, 7.5f),
-        glm::vec3(10.0f, 5.0f, -7.5f)
-    };
-
-    const float TEST_EPSILON = 0.001f;
-
-    std::vector<glm::mat4> matrixVec;
-    std::vector<AnimPose> poseVec;
-
-    for (auto& scale : scaleVec) {
-        for (auto& rot : rotVec) {
-            for (auto& trans : transVec) {
-
-                // build a matrix the old fashioned way.
-                glm::mat4 scaleMat = glm::scale(glm::mat4(), glm::vec3(scale));
-                glm::mat4 rotTransMat = createMatFromQuatAndPos(rot, trans);
-                glm::mat4 rawMat = rotTransMat * scaleMat;
-
-                matrixVec.push_back(rawMat);
-
-                // use an anim pose to build a matrix by parts.
-                AnimPose pose(scale, rot, trans);
-                poseVec.push_back(pose);
-            }
-        }
-    }
-
-    for (int i = 0; i < matrixVec.size(); i++) {
-        for (int j = 0; j < matrixVec.size(); j++) {
-
-            // multiply the matrices together
-            glm::mat4 matrix = matrixVec[i] * matrixVec[j];
-
-            // convert to matrix (note this will remove sheer from the matrix)
-            AnimPose resultA(matrix);
-
-            // multiply the poses together directly
-            AnimPose resultB = poseVec[i] * poseVec[j];
-
-            /*
-            qDebug() << "matrixVec[" << i << "] =" << matrixVec[i];
-            qDebug() << "matrixVec[" << j << "] =" << matrixVec[j];
-            qDebug() << "matrixResult =" << resultA;
-
-            qDebug() << "poseVec[" << i << "] =" << poseVec[i];
-            qDebug() << "poseVec[" << j << "] =" << poseVec[j];
-            qDebug() << "poseResult =" << resultB;
-            */
-
-            // compare results.
-            QCOMPARE_WITH_ABS_ERROR(resultA.scale(), resultB.scale(), TEST_EPSILON);
-            QCOMPARE_WITH_ABS_ERROR(resultA.rot(), resultB.rot(), TEST_EPSILON);
-            QCOMPARE_WITH_ABS_ERROR(resultA.trans(), resultB.trans(), TEST_EPSILON);
-        }
-    }
-}
-
 void AnimTests::testExpressionTokenizer() {
     QString str = "(10 +  x) >= 20.1 && (y != !z)";
     AnimExpression e("x");
diff --git a/tests/animation/src/AnimTests.h b/tests/animation/src/AnimTests.h
index 637611e8c4..439793f21d 100644
--- a/tests/animation/src/AnimTests.h
+++ b/tests/animation/src/AnimTests.h
@@ -27,7 +27,6 @@ private slots:
     void testVariant();
     void testAccumulateTime();
     void testAnimPose();
-    void testAnimPoseMultiply();
     void testExpressionTokenizer();
     void testExpressionParser();
     void testExpressionEvaluator();