diff --git a/exercises/exercise-03.md b/exercises/exercise-03.md
new file mode 100644
index 0000000000000000000000000000000000000000..7d7ca3033b5f78d9d35bdcf410cbab70ed7163ce
--- /dev/null
+++ b/exercises/exercise-03.md
@@ -0,0 +1,80 @@
+# Exercise Session 3
+
+This week we will play with genericity and OO concepts.
+
+A binary search tree is a binary tree such that, for a node, all elements in the left sub-tree are smaller than the element at the node, and all elements in the right sub-tree are greater than the element at the node. Therefore, binary search trees do not contain duplicate elements.
+
+Because we want to build a generic tree structure, we also need the notion of a comparator, or a less-than-or-equal operator (denoted `leq`) for two generic elements which satisfies the following properties:
+
+- Transitivity:   `leq(a, b) && leq(b, c) => leq(a, c)`
+- Reflexivity:    `leq(a, a)` is `true`.
+- Anti-symmetry:  `leq(a, b) && leq(b, a) => a == b`
+- Totality:       either `leq(a, b)` or `leq(b, a)` is `true` (or both)
+
+Note that the above defines a total order.
+
+Here is the structure we will be using for implementing these trees:
+
+```scala
+trait Tree[T] { ... }
+case class EmptyTree[T](leq: (T, T) => Boolean) extends Tree[T] { ... }
+case class Node[T](left: Tree[T], elem: T, right: Tree[T], leq: (T, T) => Boolean) extends Tree[T] { ... }
+```
+
+For consistency, all subtrees must contain the same leq parameter.
+Creating an empty binary tree for integers can be then done as follows:
+
+```scala
+val intLeq: (Int, Int) => Boolean = (x, y) => x <= y
+val emptyIntTree: Tree[Int] = new EmptyTree(intLeq)
+```
+
+## Question 1
+
+Given only `leq` for comparison, how can you test for equality? How about strictly-less-than?
+
+## Question 2
+
+Define the size method on `Tree[T]`, which returns its size, i.e. the number of Nodes in the tree.
+
+```scala
+trait Tree[T] {
+  def size: Int
+  ...
+}
+```
+
+Implement it in two ways:
+
+1. within `Tree[T]`, using pattern matching,
+2. in the subclasses of `Tree[T]`.
+
+## Question 3
+
+Define the `add` method, that adds an element to a `Tree[T]`, and returns the resulting tree:
+
+```scala
+def add(t: T): Tree[T] = ???
+```
+
+Remember that trees do not have duplicate values. If t is already in the tree, the result should be unchanged.
+
+## Question 4
+
+Define the function `toList`, which returns the sorted list representation for a tree. For example, `emptyIntTree.add(2).add(1).add(3).toList` should return `List(1, 2, 3)`
+
+```scala
+def toList: List[T] = ???
+```
+
+You can use the `Nil` operator for creating an empty list, and the `::` operator for adding a new element to the head of a list: `1 :: List(2, 3) == List(1, 2, 3)`. You are naturally free to define any auxiliary functions as necessary.
+
+## Question 5
+
+Define the function `sortedList`, which takes an unsorted list where no two elements are equal, and returns a new list that contains all the elements of the previous list (and only those), in increasing order.
+
+```scala
+def sortedList[T](leq: (T, T) => Boolean, ls: List[T]): List[T] = ???
+```
+
+_Hint_: you might need to define some auxiliary functions.