diff --git a/tsconvert/newick.py b/tsconvert/newick.py
index 9a933b9..1d54827 100644
--- a/tsconvert/newick.py
+++ b/tsconvert/newick.py
@@ -43,11 +43,15 @@ def from_ms(string):
     if tree_lines == "":
         raise ValueError(
             "Malformed input: no lines starting with [."
-            " Make sure you run ms with the -T and -r flags.")
+            " Make sure you run ms with the -T and -r flags."
+        )
 
     trees = dendropy.TreeList.get(
-        data=tree_lines, schema="newick", extract_comment_metadata=True,
-        rooting="force-rooted")
+        data=tree_lines,
+        schema="newick",
+        extract_comment_metadata=True,
+        rooting="force-rooted",
+    )
 
     if len(trees) == 0:
         raise ValueError("No valid trees in ms file")
@@ -55,22 +59,28 @@ def from_ms(string):
     spans = []
     for i, tree in enumerate(trees):
         try:
-            spans.append(float(tree.comments[0]))  # the initial [X] is the first comment
+            spans.append(
+                float(tree.comments[0])
+            )  # the initial [X] is the first comment
         except (ValueError, IndexError):
             raise ValueError(
-                "Problem reading integer # of positions spanned in tree {}".format(i) +
-                " (in ms format this preceeds the tree, in square braces)")
+                "Problem reading integer # of positions spanned in tree {}".format(i)
+                + " (in ms format this preceeds the tree, in square braces)"
+            )
         if len(trees.taxon_namespace) != sum([1 for l in tree.leaf_node_iter()]):
             raise ValueError(
                 "Tree {} does not have all {} expected tips".format(
-                    i, len(trees.taxon_namespace)))
+                    i, len(trees.taxon_namespace)
+                )
+            )
         # below we might want to set is_force_max_age=True to allow ancient tips
         # and work around branch length precision errors in ms
         tree.calc_node_ages()
         node_ages = [n.age for n in tree.ageorder_node_iter(include_leaves=False)]
         if len(set(node_ages)) != len(node_ages):
             raise ValueError(
-                "Tree {}: cannot have two internal nodes with the same time".format(i))
+                "Tree {}: cannot have two internal nodes with the same time".format(i)
+            )
 
     # NB: here we could check that the sequence_length == nsites, where nsites is given
     # in the ms_line, as the second number following the -r switch
@@ -102,6 +112,73 @@ def from_ms(string):
     return tables.tree_sequence()
 
 
+def from_beast(string, precision=6):
+    """
+    Reads in a BEAST nexus description.
+    """
+    trees = dendropy.TreeList.get(data=string, schema="nexus")
+
+    print("Got", len(trees))
+    positions = []
+    for i, tree in enumerate(trees):
+        label_toks = tree.label.split()
+        assert label_toks[0] == "STATE"
+        positions.append(int(label_toks[1]))
+        # TODO likely some parameter tweaks would be needed to make this work reliably
+        tree.calc_node_ages()
+        # FIXME assuming for now what all samples are contemporary.
+        node_ages = [
+            round(n.age, precision)
+            for n in tree.ageorder_node_iter(include_leaves=False)
+        ]
+        if len(set(node_ages)) != len(node_ages):
+            raise ValueError(
+                "Tree {}: cannot have two internal nodes with the same time".format(i)
+            )
+
+    assert len(positions) > 1
+    assert positions[0] == 0
+    diff = positions[-1] - positions[-2]
+    positions.append(positions[-1] + diff)
+
+    tables = tskit.TableCollection(positions[-1])
+    sample_id_map = {}
+    # Get the samples from the first tree.
+    for node in trees[0].leaf_node_iter():
+        node_id = tables.nodes.add_row(flags=tskit.NODE_IS_SAMPLE, time=node.age)
+        sample_id_map[node.taxon.label] = node_id
+
+    age_id_map = {}
+    left = 0
+    for right, tree in zip(positions[1:], trees):
+        for node in tree.ageorder_node_iter(include_leaves=False):
+            age = round(node.age, precision)
+            # print("node", age, node.annotations)
+            # print(node.description())
+            children = list(node.child_nodes())
+            if node.age not in age_id_map:
+                age_id_map[age] = tables.nodes.add_row(flags=0, time=age)
+            parent_id = age_id_map[age]
+            for child in children:
+                if child.is_leaf():
+                    child_id = sample_id_map[child.taxon.label]
+                else:
+                    child_age = round(child.age, precision)
+                    child_id = age_id_map[child_age]
+                tables.edges.add_row(left, right, parent_id, child_id)
+        left = right
+
+    # import numpy as np
+    # print(tables.nodes)
+    # t = tables.nodes.time
+    # np.sort(t)
+    # print(t)
+    tables.sort()
+    # Simplify will squash together any edges, removing redundancy.
+    tables.simplify()
+    return tables.tree_sequence()
+
+
 def to_ms(ts):
     """
     Returns an ms-formatted version of the specified tree sequence.