diff --git a/Box2D/Collision/b2BroadPhase.cpp b/Box2D/Collision/b2BroadPhase.cpp
index 0af21cd..52c8979 100755
--- a/Box2D/Collision/b2BroadPhase.cpp
+++ b/Box2D/Collision/b2BroadPhase.cpp
@@ -37,9 +37,9 @@ b2BroadPhase::~b2BroadPhase()
 	b2Free(m_pairBuffer);
 }
 
-int32 b2BroadPhase::CreateProxy(const b2AABB& aabb, void* userData)
+int32 b2BroadPhase::CreateProxy(const b2AABB& aabb, void* userData, uint16 categoryBits)
 {
-	int32 proxyId = m_tree.CreateProxy(aabb, userData);
+	int32 proxyId = m_tree.CreateProxy(aabb, userData, categoryBits);
 	++m_proxyCount;
 	BufferMove(proxyId);
 	return proxyId;
diff --git a/Box2D/Collision/b2BroadPhase.h b/Box2D/Collision/b2BroadPhase.h
index 54ee9c4..65cad89 100755
--- a/Box2D/Collision/b2BroadPhase.h
+++ b/Box2D/Collision/b2BroadPhase.h
@@ -46,7 +46,7 @@ class b2BroadPhase
 
 	/// Create a proxy with an initial AABB. Pairs are not reported until
 	/// UpdatePairs is called.
-	int32 CreateProxy(const b2AABB& aabb, void* userData);
+	int32 CreateProxy(const b2AABB& aabb, void* userData, uint16 categoryBits);
 
 	/// Destroy a proxy. It is up to the client to remove any pairs.
 	void DestroyProxy(int32 proxyId);
@@ -87,7 +87,7 @@ class b2BroadPhase
 	/// @param input the ray-cast input data. The ray extends from p1 to p1 + maxFraction * (p2 - p1).
 	/// @param callback a callback class that is called for each proxy that is hit by the ray.
 	template <typename T>
-	void RayCast(T* callback, const b2RayCastInput& input) const;
+	void RayCast(T* callback, const b2RayCastInput& input, uint16 maskBits) const;
 
 	/// Get the height of the embedded tree.
 	int32 GetTreeHeight() const;
@@ -220,9 +220,9 @@ inline void b2BroadPhase::Query(T* callback, const b2AABB& aabb) const
 }
 
 template <typename T>
-inline void b2BroadPhase::RayCast(T* callback, const b2RayCastInput& input) const
+inline void b2BroadPhase::RayCast(T* callback, const b2RayCastInput& input, uint16 maskBits) const
 {
-	m_tree.RayCast(callback, input);
+	m_tree.RayCast(callback, input, maskBits);
 }
 
 inline void b2BroadPhase::ShiftOrigin(const b2Vec2& newOrigin)
diff --git a/Box2D/Collision/b2DynamicTree.cpp b/Box2D/Collision/b2DynamicTree.cpp
index 5e9a8fa..800d963 100755
--- a/Box2D/Collision/b2DynamicTree.cpp
+++ b/Box2D/Collision/b2DynamicTree.cpp
@@ -85,6 +85,7 @@ int32 b2DynamicTree::AllocateNode()
 	m_nodes[nodeId].child1 = b2_nullNode;
 	m_nodes[nodeId].child2 = b2_nullNode;
 	m_nodes[nodeId].height = 0;
+	m_nodes[nodeId].categoryBits = _I16_MAX;
 	m_nodes[nodeId].userData = NULL;
 	m_nodes[nodeId].moved = false;
 	++m_nodeCount;
@@ -105,7 +106,7 @@ void b2DynamicTree::FreeNode(int32 nodeId)
 // Create a proxy in the tree as a leaf node. We return the index
 // of the node instead of a pointer so that we can grow
 // the node pool.
-int32 b2DynamicTree::CreateProxy(const b2AABB& aabb, void* userData)
+int32 b2DynamicTree::CreateProxy(const b2AABB& aabb, void* userData, uint16 categoryBits)
 {
 	int32 proxyId = AllocateNode();
 
@@ -115,8 +116,8 @@ int32 b2DynamicTree::CreateProxy(const b2AABB& aabb, void* userData)
 	m_nodes[proxyId].aabb.upperBound = aabb.upperBound + r;
 	m_nodes[proxyId].userData = userData;
 	m_nodes[proxyId].height = 0;
+	m_nodes[proxyId].categoryBits = categoryBits;
 	m_nodes[proxyId].moved = true;
-
 	InsertLeaf(proxyId);
 
 	return proxyId;
@@ -286,6 +287,7 @@ void b2DynamicTree::InsertLeaf(int32 leaf)
 	m_nodes[newParent].userData = NULL;
 	m_nodes[newParent].aabb.Combine(leafAABB, m_nodes[sibling].aabb);
 	m_nodes[newParent].height = m_nodes[sibling].height + 1;
+	m_nodes[newParent].categoryBits = m_nodes[leaf].categoryBits | m_nodes[sibling].categoryBits;
 
 	if (oldParent != b2_nullNode)
 	{
@@ -328,7 +330,7 @@ void b2DynamicTree::InsertLeaf(int32 leaf)
 
 		m_nodes[index].height = 1 + b2Max(m_nodes[child1].height, m_nodes[child2].height);
 		m_nodes[index].aabb.Combine(m_nodes[child1].aabb, m_nodes[child2].aabb);
-
+		m_nodes[newParent].categoryBits = m_nodes[child1].categoryBits | m_nodes[child2].categoryBits;
 		index = m_nodes[index].parent;
 	}
 
@@ -380,7 +382,7 @@ void b2DynamicTree::RemoveLeaf(int32 leaf)
 
 			m_nodes[index].aabb.Combine(m_nodes[child1].aabb, m_nodes[child2].aabb);
 			m_nodes[index].height = 1 + b2Max(m_nodes[child1].height, m_nodes[child2].height);
-
+			m_nodes[index].categoryBits = m_nodes[child1].categoryBits | m_nodes[child2].categoryBits;
 			index = m_nodes[index].parent;
 		}
 	}
@@ -460,6 +462,8 @@ int32 b2DynamicTree::Balance(int32 iA)
 
 			A->height = 1 + b2Max(B->height, G->height);
 			C->height = 1 + b2Max(A->height, F->height);
+			A->categoryBits = B->categoryBits | G->categoryBits;
+			C->categoryBits = A->categoryBits | F->categoryBits;
 		}
 		else
 		{
@@ -471,6 +475,9 @@ int32 b2DynamicTree::Balance(int32 iA)
 
 			A->height = 1 + b2Max(B->height, F->height);
 			C->height = 1 + b2Max(A->height, G->height);
+
+			A->categoryBits = B->categoryBits | F->categoryBits;
+			C->categoryBits = A->categoryBits | G->categoryBits;
 		}
 
 		return iC;
@@ -520,6 +527,9 @@ int32 b2DynamicTree::Balance(int32 iA)
 
 			A->height = 1 + b2Max(C->height, E->height);
 			B->height = 1 + b2Max(A->height, D->height);
+
+			A->categoryBits = C->categoryBits | E->categoryBits;
+			B->categoryBits = A->categoryBits | D->categoryBits;
 		}
 		else
 		{
@@ -531,6 +541,9 @@ int32 b2DynamicTree::Balance(int32 iA)
 
 			A->height = 1 + b2Max(C->height, D->height);
 			B->height = 1 + b2Max(A->height, E->height);
+
+			A->categoryBits = C->categoryBits | D->categoryBits;
+			B->categoryBits = A->categoryBits | E->categoryBits;
 		}
 
 		return iB;
@@ -576,6 +589,10 @@ float32 b2DynamicTree::GetAreaRatio() const
 	return totalArea / rootArea;
 }
 
+void b2DynamicTree::SetCategoryBits(uint16 categoryBits)
+{
+}
+
 // Compute the height of a sub-tree.
 int32 b2DynamicTree::ComputeHeight(int32 nodeId) const
 {
@@ -672,6 +689,9 @@ void b2DynamicTree::ValidateMetrics(int32 index) const
 	b2Assert(aabb.lowerBound == node->aabb.lowerBound);
 	b2Assert(aabb.upperBound == node->aabb.upperBound);
 
+	uint16 categoryBits = m_nodes[child1].categoryBits | m_nodes[child2].categoryBits;
+	b2Assert(node->categoryBits == categoryBits);
+
 	ValidateMetrics(child1);
 	ValidateMetrics(child2);
 }
@@ -777,6 +797,7 @@ void b2DynamicTree::RebuildBottomUp()
 		parent->child2 = index2;
 		parent->height = 1 + b2Max(child1->height, child2->height);
 		parent->aabb.Combine(child1->aabb, child2->aabb);
+		parent->categoryBits = child1->categoryBits | child2->categoryBits;
 		parent->parent = b2_nullNode;
 
 		child1->parent = parentIndex;
diff --git a/Box2D/Collision/b2DynamicTree.h b/Box2D/Collision/b2DynamicTree.h
index 28a56e8..ff95190 100755
--- a/Box2D/Collision/b2DynamicTree.h
+++ b/Box2D/Collision/b2DynamicTree.h
@@ -23,7 +23,6 @@
 #include <Box2D/Common/b2GrowableStack.h>
 
 #define b2_nullNode (-1)
-
 /// A node in the dynamic tree. The client does not interact with this directly.
 struct b2TreeNode
 {
@@ -37,6 +36,9 @@ struct b2TreeNode
 
 	void* userData;
 
+	// Category bits for collision filtering
+	uint16 categoryBits;
+
 	union
 	{
 		int32 parent;
@@ -70,7 +72,7 @@ class b2DynamicTree
 	~b2DynamicTree();
 
 	/// Create a proxy. Provide a tight fitting AABB and a userData pointer.
-	int32 CreateProxy(const b2AABB& aabb, void* userData);
+	int32 CreateProxy(const b2AABB& aabb, void* userData, uint16 categoryBits);
 
 	/// Destroy a proxy. This asserts if the id is invalid.
 	void DestroyProxy(int32 proxyId);
@@ -104,7 +106,7 @@ class b2DynamicTree
 	/// @param input the ray-cast input data. The ray extends from p1 to p1 + maxFraction * (p2 - p1).
 	/// @param callback a callback class that is called for each proxy that is hit by the ray.
 	template <typename T>
-	void RayCast(T* callback, const b2RayCastInput& input) const;
+	void RayCast(T* callback, const b2RayCastInput& input, uint16 maskBits) const;
 
 	/// Validate this tree. For testing.
 	void Validate() const;
@@ -120,6 +122,8 @@ class b2DynamicTree
 	/// Get the ratio of the sum of the node areas to the root area.
 	float32 GetAreaRatio() const;
 
+	void SetCategoryBits(uint16 categoryBits);
+
 	/// Build an optimal tree. Very expensive. For testing.
 	void RebuildBottomUp();
 
@@ -218,7 +222,7 @@ inline void b2DynamicTree::Query(T* callback, const b2AABB& aabb) const
 }
 
 template <typename T>
-inline void b2DynamicTree::RayCast(T* callback, const b2RayCastInput& input) const
+inline void b2DynamicTree::RayCast(T* callback, const b2RayCastInput& input, uint16 maskBits) const
 {
 	b2Vec2 p1 = input.p1;
 	b2Vec2 p2 = input.p2;
@@ -256,7 +260,7 @@ inline void b2DynamicTree::RayCast(T* callback, const b2RayCastInput& input) con
 
 		const b2TreeNode* node = m_nodes + nodeId;
 
-		if (b2TestOverlap(node->aabb, segmentAABB) == false)
+		if ((node->categoryBits & maskBits) == 0 || (b2TestOverlap(node->aabb, segmentAABB) == false))
 		{
 			continue;
 		}
diff --git a/Box2D/Dynamics/b2Fixture.cpp b/Box2D/Dynamics/b2Fixture.cpp
index ff75b10..1e08025 100755
--- a/Box2D/Dynamics/b2Fixture.cpp
+++ b/Box2D/Dynamics/b2Fixture.cpp
@@ -130,7 +130,7 @@ void b2Fixture::CreateProxies(b2BroadPhase* broadPhase, const b2Transform& xf)
 	{
 		b2FixtureProxy* proxy = m_proxies + i;
 		m_shape->ComputeAABB(&proxy->aabb, xf, i);
-		proxy->proxyId = broadPhase->CreateProxy(proxy->aabb, proxy);
+		proxy->proxyId = broadPhase->CreateProxy(proxy->aabb, proxy, m_filter.categoryBits);
 		proxy->fixture = this;
 		proxy->childIndex = i;
 	}
diff --git a/Box2D/Dynamics/b2World.cpp b/Box2D/Dynamics/b2World.cpp
index d3a0518..f0bf8e7 100755
--- a/Box2D/Dynamics/b2World.cpp
+++ b/Box2D/Dynamics/b2World.cpp
@@ -1119,7 +1119,7 @@ struct b2WorldRayCastWrapper
 	b2RayCastCallback* callback;
 };
 
-void b2World::RayCast(b2RayCastCallback* callback, const b2Vec2& point1, const b2Vec2& point2) const
+void b2World::RayCast(b2RayCastCallback* callback, const b2Vec2& point1, const b2Vec2& point2, uint16 maskBits) const
 {
 	b2WorldRayCastWrapper wrapper;
 	wrapper.broadPhase = &m_contactManager.m_broadPhase;
@@ -1128,7 +1128,7 @@ void b2World::RayCast(b2RayCastCallback* callback, const b2Vec2& point1, const b
 	input.maxFraction = 1.0f;
 	input.p1 = point1;
 	input.p2 = point2;
-	m_contactManager.m_broadPhase.RayCast(&wrapper, input);
+	m_contactManager.m_broadPhase.RayCast(&wrapper, input, maskBits);
 	for (b2ParticleSystem* p = m_particleSystemList; p; p = p->GetNext())
 	{
 		if (callback->ShouldQueryParticleSystem(p))
diff --git a/Box2D/Dynamics/b2World.h b/Box2D/Dynamics/b2World.h
index ce0b8c1..af68817 100755
--- a/Box2D/Dynamics/b2World.h
+++ b/Box2D/Dynamics/b2World.h
@@ -166,7 +166,7 @@ class b2World
 	/// @param callback a user implemented callback class.
 	/// @param point1 the ray starting point
 	/// @param point2 the ray ending point
-	void RayCast(b2RayCastCallback* callback, const b2Vec2& point1, const b2Vec2& point2) const;
+	void RayCast(b2RayCastCallback* callback, const b2Vec2& point1, const b2Vec2& point2, uint16 maskBits) const;
 
 	/// Get the world body list. With the returned body, use b2Body::GetNext to get
 	/// the next body in the world list. A NULL body indicates the end of the list.