diff --git a/Lib/_pyrepl/commands.py b/Lib/_pyrepl/commands.py
index 10127e58897a58..25a58e792ef272 100644
--- a/Lib/_pyrepl/commands.py
+++ b/Lib/_pyrepl/commands.py
@@ -169,7 +169,7 @@ class unix_word_rubout(KillCommand):
     def do(self) -> None:
         r = self.reader
         for i in range(r.get_arg()):
-            self.kill_range(r.bow(), r.pos)
+            self.kill_range(r.bow_ws(), r.pos)
 
 
 class kill_word(KillCommand):
diff --git a/Lib/_pyrepl/reader.py b/Lib/_pyrepl/reader.py
index 9ab92f64d1ef63..4834fbaec1055b 100644
--- a/Lib/_pyrepl/reader.py
+++ b/Lib/_pyrepl/reader.py
@@ -417,6 +417,24 @@ def bow(self, p: int | None = None) -> int:
             p -= 1
         return p + 1
 
+    def bow_ws(self, p: int | None = None) -> int:
+        """Return the 0-based index of the whitespace-delimited word break
+        preceding p most immediately.
+
+        p defaults to self.pos; only whitespace is considered a word
+        boundary, matching the behavior of unix-word-rubout in bash/readline.
+        See https://github.com/python/cpython/issues/146044
+        """
+        if p is None:
+            p = self.pos
+        b = self.buffer
+        p -= 1
+        while p >= 0 and b[p] in " \n\t":
+            p -= 1
+        while p >= 0 and b[p] not in " \n\t":
+            p -= 1
+        return p + 1
+
     def eow(self, p: int | None = None) -> int:
         """Return the 0-based index of the word break following p most
         immediately.
diff --git a/Lib/test/test_pyrepl/test_reader.py b/Lib/test/test_pyrepl/test_reader.py
index b1b6ae16a1e592..de7558eb6e39c1 100644
--- a/Lib/test/test_pyrepl/test_reader.py
+++ b/Lib/test/test_pyrepl/test_reader.py
@@ -358,6 +358,34 @@ def test_setpos_from_xy_for_non_printing_char(self):
         reader.setpos_from_xy(8, 0)
         self.assertEqual(reader.pos, 7)
 
+    def test_bow_ws_stops_at_whitespace(self):
+        # See https://github.com/python/cpython/issues/146044
+        reader = prepare_reader(prepare_console([]))
+        reader.buffer = list("foo.bar baz")
+        reader.pos = len(reader.buffer)
+        self.assertEqual(reader.bow_ws(), 8)
+
+    def test_bow_ws_includes_punctuation_in_word(self):
+        reader = prepare_reader(prepare_console([]))
+        reader.buffer = list("foo.bar(baz) qux")
+        reader.pos = 12
+        self.assertEqual(reader.bow_ws(), 0)
+
+    def test_bow_vs_bow_ws(self):
+        reader = prepare_reader(prepare_console([]))
+        reader.buffer = list("foo.bar")
+        reader.pos = len(reader.buffer)
+        # bow() stops at '.' so we return the index of 'b' in "bar"
+        self.assertEqual(reader.bow(), 4)
+        # bow_ws() treats entire "foo.bar" as one word
+        self.assertEqual(reader.bow_ws(), 0)
+
+    def test_bow_ws_with_tabs(self):
+        reader = prepare_reader(prepare_console([]))
+        reader.buffer = list("foo\tbar")
+        reader.pos = len(reader.buffer)
+        self.assertEqual(reader.bow_ws(), 4)
+
 @force_colorized_test_class
 class TestReaderInColor(ScreenEqualMixin, TestCase):
     def test_syntax_highlighting_basic(self):
diff --git a/Misc/NEWS.d/next/Library/2026-03-20-00-00-01.gh-issue-146044.WsCtrl.rst b/Misc/NEWS.d/next/Library/2026-03-20-00-00-01.gh-issue-146044.WsCtrl.rst
new file mode 100644
index 00000000000000..d71ad1828fd5d8
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2026-03-20-00-00-01.gh-issue-146044.WsCtrl.rst
@@ -0,0 +1,3 @@
+Fix ``unix-word-rubout`` (Ctrl-W) in the REPL to use whitespace-only word
+boundaries, matching behavior of the basic REPL. Previously it used
+syntax-table boundaries which treated punctuation as word separators.