diff --git a/doc/changelog.qmd b/doc/changelog.qmd
index 4da468bc6..f44b4bc9d 100644
--- a/doc/changelog.qmd
+++ b/doc/changelog.qmd
@@ -63,6 +63,9 @@ title: Changelog
 
 - Fixed using `facet_grid` with a column named `key`. ({{< issue 734 >}})
 
+- Fixed using legend when using an identity scale and reordering the
+  breaks. ({{< issue 735 >}})
+
 ### Enhancements
 
 - All `__all__` variables are explicitly assigned to help static typecheckers
diff --git a/plotnine/guides/guide_legend.py b/plotnine/guides/guide_legend.py
index 086bea1fd..9e5ec80c3 100644
--- a/plotnine/guides/guide_legend.py
+++ b/plotnine/guides/guide_legend.py
@@ -69,7 +69,7 @@ def train(self, scale, aesthetic=None):
         scale name is one of the aesthetics: `x`, `y`, `color`,
         `fill`, `size`, `shape`, `alpha`, `stroke`.
 
-        Returns this guide if trainning is successful and None
+        Returns this guide if training is successful and None
         if it fails
         """
         if aesthetic is None:
diff --git a/plotnine/scales/scale_discrete.py b/plotnine/scales/scale_discrete.py
index a334444b5..123ececb3 100644
--- a/plotnine/scales/scale_discrete.py
+++ b/plotnine/scales/scale_discrete.py
@@ -237,21 +237,20 @@ def get_breaks(
         The form is suitable for use by the guides e.g.
             ['fair', 'good', 'very good', 'premium', 'ideal']
         """
-        if limits is None:
-            limits = self.limits
-
         if self.is_empty():
             return []
 
-        if self.breaks is True:
-            breaks = list(limits)
-        elif self.breaks in (False, None):
+        if limits is None:
+            limits = self.limits
+
+        if self.breaks in (None, False):
             breaks = []
+        elif self.breaks is True:
+            breaks = list(limits)
         elif callable(self.breaks):
             breaks = self.breaks(limits)
         else:
-            _wanted_breaks = set(self.breaks)
-            breaks = [l for l in limits if l in _wanted_breaks]
+            breaks = list(self.breaks)
 
         return breaks
 
@@ -264,10 +263,8 @@ def get_bounded_breaks(
         if limits is None:
             limits = self.limits
 
-        lookup = set(limits)
-        breaks = self.get_breaks()
-        strict_breaks = [b for b in breaks if b in lookup]
-        return strict_breaks
+        lookup_limits = set(limits)
+        return [b for b in self.get_breaks() if b in lookup_limits]
 
     def get_labels(
         self, breaks: Optional[ScaleDiscreteBreaks] = None
@@ -291,13 +288,15 @@ def get_labels(
             return self.labels(breaks)
         # if a dict is used to rename some labels
         elif isinstance(self.labels, dict):
-            labels = [
+            return [
                 str(self.labels[b]) if b in self.labels else str(b)
                 for b in breaks
             ]
-            return labels
         else:
-            return self.labels
+            # Return the labels in the order that they match with
+            # the breaks.
+            label_lookup = dict(zip(self.get_breaks(), self.labels))
+            return [label_lookup[b] for b in breaks]
 
     def transform_df(self, df: pd.DataFrame) -> pd.DataFrame:
         """
diff --git a/tests/baseline_images/test_scale_internals/test_legend_ordering_with_identity_scale.png b/tests/baseline_images/test_scale_internals/test_legend_ordering_with_identity_scale.png
new file mode 100644
index 000000000..421709c82
Binary files /dev/null and b/tests/baseline_images/test_scale_internals/test_legend_ordering_with_identity_scale.png differ
diff --git a/tests/test_scale_internals.py b/tests/test_scale_internals.py
index 4a32fe818..05d1ea7c2 100644
--- a/tests/test_scale_internals.py
+++ b/tests/test_scale_internals.py
@@ -6,6 +6,7 @@
 import pandas as pd
 import pytest
 
+import plotnine as p9
 from plotnine import (
     aes,
     annotate,
@@ -710,6 +711,26 @@ def test_legend_ordering_added_scales():
     assert p == "legend_ordering_added_scales"
 
 
+def test_legend_ordering_with_identity_scale():
+    data = pd.DataFrame(
+        {
+            "x": [1, 2, 3, 4],
+            "y": [1, 2, 3, 4],
+            "color": ["blue", "blue", "red", "red"],
+        }
+    )
+
+    p = (
+        ggplot(data, aes("x", "y", color="color"))
+        + geom_point()
+        + p9.scale_color_identity(
+            breaks=["red", "blue"], labels=["Red", "Blue"], guide="legend"
+        )
+    )
+
+    assert p == "test_legend_ordering_with_identity_scale"
+
+
 def test_breaks_and_labels_outside_of_limits():
     data = pd.DataFrame({"x": range(5, 11), "y": range(5, 11)})
     p = (