diff --git a/Tests/test_image_resize.py b/Tests/test_image_resize.py
index f37fd4b04e7..9dc24f47b68 100644
--- a/Tests/test_image_resize.py
+++ b/Tests/test_image_resize.py
@@ -282,6 +282,18 @@ def test_reducing_gap_8(
 
         assert_image_equal(ref, im)
 
+    @pytest.mark.parametrize("mode", ("RGBA", "LA"))
+    def test_reducing_gap_alpha(self, mode: str) -> None:
+        with Image.open("Tests/images/transparent.png") as im:
+            im_converted = im.convert(mode)
+
+        # Premultiply the alpha manually
+        premultiplied = im_converted.convert({"RGBA": "RGBa", "LA": "La"}[mode])
+        expected = premultiplied.resize((16, 16), reducing_gap=2).convert(mode)
+
+        # Test that resizing the image applies the alpha in the same way
+        assert_image_equal(im_converted.resize((16, 16), reducing_gap=2), expected)
+
     @pytest.mark.parametrize(
         "box, epsilon",
         (((0, 0, 512, 512), 5.5), ((0.9, 1.7, 128, 128), 9.5)),
diff --git a/docs/releasenotes/13.0.0.rst b/docs/releasenotes/13.0.0.rst
index 0763dccf8fb..a9fcc0dc510 100644
--- a/docs/releasenotes/13.0.0.rst
+++ b/docs/releasenotes/13.0.0.rst
@@ -183,6 +183,14 @@ installed on the system with the expected name (e.g. ``libharfbuzz.so.0``).
 All wheels are now linked against the HarfBuzz that Pillow ships,
 which may slightly change how text is rendered in scripts that rely on the auto-hinter.
 
+Resizing images with alpha
+^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+The ``reducing_gap`` argument to :py:meth:`~PIL.Image.Image.resize` is now
+applied to RGBA and LA images. Previously, these modes ignored the argument,
+including when called by :py:meth:`~PIL.Image.Image.thumbnail`, which uses
+``reducing_gap=2.0`` by default.
+
 Preserved transparency when padding or expanding images in ImageOps
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
diff --git a/src/PIL/Image.py b/src/PIL/Image.py
index a55cdccd710..6dec284d086 100644
--- a/src/PIL/Image.py
+++ b/src/PIL/Image.py
@@ -2420,7 +2420,7 @@ def resize(
 
         if self.mode in ["LA", "RGBA"] and resample != Resampling.NEAREST:
             im = self.convert({"LA": "La", "RGBA": "RGBa"}[self.mode])
-            im = im.resize(size, resample, box)
+            im = im.resize(size, resample, box, reducing_gap)
             return im.convert(self.mode)
 
         self.load()