diff --git a/lib/rdf/dataset.ex b/lib/rdf/dataset.ex
index 9e3ab47..81b8dfc 100644
--- a/lib/rdf/dataset.ex
+++ b/lib/rdf/dataset.ex
@@ -48,10 +48,12 @@ defmodule RDF.Dataset do
 
   ## Examples
 
-      RDF.Dataset.new({EX.S, EX.p, EX.O})
-
       RDF.Dataset.new(name: EX.GraphName)
 
+      RDF.Dataset.new(init: {EX.S, EX.p, EX.O})
+
+      RDF.Dataset.new({EX.S, EX.p, EX.O})
+
   """
   @spec new(input | keyword) :: t
   def new(data_or_options)
@@ -59,7 +61,8 @@ defmodule RDF.Dataset do
   def new(data_or_options)
       when is_list(data_or_options) and length(data_or_options) != 0 do
     if Keyword.keyword?(data_or_options) do
-      new([], data_or_options)
+      {data, options} = Keyword.pop(data_or_options, :init)
+      new(data, options)
     else
       new(data_or_options, [])
     end
@@ -75,6 +78,9 @@ defmodule RDF.Dataset do
   Available options:
 
   - `name`: the name of the dataset to be created
+  - `init`: some data with which the dataset should be initialized; the data can be
+    provided in any form accepted by `add/3` and above that also with a function returning
+    the initialization data in any of these forms
 
   """
   @spec new(input, keyword) :: t
@@ -87,9 +93,13 @@ defmodule RDF.Dataset do
   def new(data, options) do
     %__MODULE__{}
     |> new(options)
-    |> add(data)
+    |> init(data)
   end
 
+  defp init(dataset, nil), do: dataset
+  defp init(dataset, fun) when is_function(fun), do: add(dataset, fun.())
+  defp init(dataset, data), do: add(dataset, data)
+
   @doc """
   Returns the dataset name IRI of `dataset`.
   """
diff --git a/lib/rdf/diff.ex b/lib/rdf/diff.ex
index f98eb6f..9676fbd 100644
--- a/lib/rdf/diff.ex
+++ b/lib/rdf/diff.ex
@@ -35,7 +35,7 @@ defmodule RDF.Diff do
   defp coerce_graph(%Description{} = description),
     do: if(Enum.empty?(description), do: Graph.new(), else: Graph.new(description))
 
-  defp coerce_graph(data), do: Graph.new(data)
+  defp coerce_graph(data), do: Graph.new(init: data)
 
   @doc """
   Computes a diff between two `RDF.Graph`s or `RDF.Description`s.
diff --git a/lib/rdf/graph.ex b/lib/rdf/graph.ex
index 566ac86..8748684 100644
--- a/lib/rdf/graph.ex
+++ b/lib/rdf/graph.ex
@@ -56,17 +56,19 @@ defmodule RDF.Graph do
   @doc """
   Creates an `RDF.Graph`.
 
-  If a keyword list is given an empty graph is created.
+  If a keyword list with options is given an empty graph is created.
   Otherwise an unnamed graph initialized with the given data is created.
 
   See `new/2` for available arguments and the different ways to provide data.
 
   ## Examples
 
-      RDF.Graph.new({EX.S, EX.p, EX.O})
-
       RDF.Graph.new(name: EX.GraphName)
 
+      RDF.Graph.new(init: {EX.S, EX.p, EX.O})
+
+      RDF.Graph.new({EX.S, EX.p, EX.O})
+
   """
   @spec new(input | keyword) :: t
   def new(data_or_options)
@@ -74,7 +76,8 @@ defmodule RDF.Graph do
   def new(data_or_options)
       when is_list(data_or_options) and length(data_or_options) != 0 do
     if Keyword.keyword?(data_or_options) do
-      new([], data_or_options)
+      {data, options} = Keyword.pop(data_or_options, :init)
+      new(data, options)
     else
       new(data_or_options, [])
     end
@@ -99,6 +102,9 @@ defmodule RDF.Graph do
     and will be used for example when serializing in a format with prefix support
   - `base_iri`: a base IRI which should be stored alongside the graph
     and will be used for example when serializing in a format with base IRI support
+  - `init`: some data with which the graph should be initialized; the data can be
+    provided in any form accepted by `add/3` and above that also with a function returning
+    the initialization data in any of these forms
 
   ## Examples
 
@@ -121,11 +127,15 @@ defmodule RDF.Graph do
   end
 
   def new(data, options) do
-    %__MODULE__{}
+    new()
     |> new(options)
-    |> add(data)
+    |> init(data)
   end
 
+  defp init(graph, nil), do: graph
+  defp init(graph, fun) when is_function(fun), do: add(graph, fun.())
+  defp init(graph, data), do: add(graph, data)
+
   @doc """
   Removes all triples from `graph`.
 
diff --git a/test/unit/dataset_test.exs b/test/unit/dataset_test.exs
index 0a56eec..55dd0e7 100644
--- a/test/unit/dataset_test.exs
+++ b/test/unit/dataset_test.exs
@@ -150,6 +150,53 @@ defmodule RDF.DatasetTest do
                {EX.Subject, EX.predicate(), EX.Object, EX.GraphName}
              )
     end
+
+    @tag skip: "This case is currently not supported, since it's indistinguishable from Keywords"
+    test "creating a dataset with a list of subject-predications pairs" do
+      ds =
+        Dataset.new([
+          {EX.S1,
+           [
+             {EX.P1, EX.O1},
+             %{EX.P2 => [EX.O2]}
+           ]}
+        ])
+
+      assert dataset_includes_statement?(ds, {EX.S1, EX.P1, EX.O1})
+      assert dataset_includes_statement?(ds, {EX.S1, EX.P2, EX.O2})
+    end
+
+    test "with init data" do
+      ds =
+        Dataset.new(
+          init: [
+            {EX.S1,
+             [
+               {EX.P1, EX.O1},
+               %{EX.P2 => [EX.O2]}
+             ]}
+          ]
+        )
+
+      assert unnamed_dataset?(ds)
+      assert dataset_includes_statement?(ds, {EX.S1, EX.P1, EX.O1})
+      assert dataset_includes_statement?(ds, {EX.S1, EX.P2, EX.O2})
+
+      ds =
+        Dataset.new(
+          name: EX.Dataset,
+          init: {EX.S, EX.p(), EX.O}
+        )
+
+      assert named_dataset?(ds, RDF.iri(EX.Dataset))
+      assert dataset_includes_statement?(ds, {EX.S, EX.p(), EX.O})
+    end
+
+    test "with an initializer function" do
+      ds = Dataset.new(init: fn -> {EX.S, EX.p(), EX.O} end)
+      assert unnamed_dataset?(ds)
+      assert dataset_includes_statement?(ds, {EX.S, EX.p(), EX.O})
+    end
   end
 
   test "name/1" do
diff --git a/test/unit/graph_test.exs b/test/unit/graph_test.exs
index 59afe2b..dcba210 100644
--- a/test/unit/graph_test.exs
+++ b/test/unit/graph_test.exs
@@ -156,6 +156,53 @@ defmodule RDF.GraphTest do
       g = Graph.new(Graph.new(prefixes: %{ex: XSD, rdfs: RDFS}), prefixes: prefix_map)
       assert g.prefixes == PrefixMap.new(ex: EX, rdfs: RDFS)
     end
+
+    @tag skip: "This case is currently not supported, since it's indistinguishable from Keywords"
+    test "creating a graph with a list of subject-predications pairs" do
+      g =
+        Graph.new([
+          {EX.S1,
+           [
+             {EX.P1, EX.O1},
+             %{EX.P2 => [EX.O2]}
+           ]}
+        ])
+
+      assert graph_includes_statement?(g, {EX.S1, EX.P1, EX.O1})
+      assert graph_includes_statement?(g, {EX.S1, EX.P2, EX.O2})
+    end
+
+    test "with init data" do
+      g =
+        Graph.new(
+          init: [
+            {EX.S1,
+             [
+               {EX.P1, EX.O1},
+               %{EX.P2 => [EX.O2]}
+             ]}
+          ]
+        )
+
+      assert unnamed_graph?(g)
+      assert graph_includes_statement?(g, {EX.S1, EX.P1, EX.O1})
+      assert graph_includes_statement?(g, {EX.S1, EX.P2, EX.O2})
+
+      g =
+        Graph.new(
+          name: EX.Graph,
+          init: {EX.S, EX.p(), EX.O}
+        )
+
+      assert named_graph?(g, RDF.iri(EX.Graph))
+      assert graph_includes_statement?(g, {EX.S, EX.p(), EX.O})
+    end
+
+    test "with an initializer function" do
+      g = Graph.new(init: fn -> {EX.S, EX.p(), EX.O} end)
+      assert unnamed_graph?(g)
+      assert graph_includes_statement?(g, {EX.S, EX.p(), EX.O})
+    end
   end
 
   test "clear/1" do