@@ -0,0 +1,4 @@
+Polar errorbar caps are rotated
+-------------------------------
+When plotting errorbars on a polar plot, the caps are now rotated so they are
+perpendicular to the errorbars.
@@ -3522,10 +3522,11 @@ def _upcast_err(err):
         eb_cap_style['color'] = ecolor
 
         barcols = []
-        caplines = []
+        caplines = {'x': [], 'y': []}
 
         # Vectorized fancy-indexer.
-        def apply_mask(arrays, mask): return [array[mask] for array in arrays]
+        def apply_mask(arrays, mask):
+            return [array[mask] for array in arrays]
 
         # dep: dependent dataset, indep: independent dataset
         for (dep_axis, dep, err, lolims, uplims, indep, lines_func,
@@ -3571,7 +3572,7 @@ def apply_mask(arrays, mask): return [array[mask] for array in arrays]
                     line = mlines.Line2D(indep_masked, indep_masked,
                                          marker=marker, **eb_cap_style)
                     line.set(**{f"{dep_axis}data": lh_masked})
-                    caplines.append(line)
+                    caplines[dep_axis].append(line)
             for idx, (lims, hl) in enumerate([(lolims, high), (uplims, low)]):
                 if not lims.any():
                     continue
@@ -3585,15 +3586,36 @@ def apply_mask(arrays, mask): return [array[mask] for array in arrays]
                 line = mlines.Line2D(x_masked, y_masked,
                                      marker=hlmarker, **eb_cap_style)
                 line.set(**{f"{dep_axis}data": hl_masked})
-                caplines.append(line)
+                caplines[dep_axis].append(line)
                 if capsize > 0:
-                    caplines.append(mlines.Line2D(
+                    caplines[dep_axis].append(mlines.Line2D(
                         x_masked, y_masked, marker=marker, **eb_cap_style))
-
-        for l in caplines:
-            self.add_line(l)
+        if self.name == 'polar':
+            for axis in caplines:
+                if not caplines[axis]:
+                    continue
+                lo, hi = caplines[axis]
+                for (lo_theta, lo_r,
+                     hi_theta, hi_r) in zip(
+                            lo.get_xdata(), lo.get_ydata(),
+                            hi.get_xdata(), hi.get_ydata()):
+                    # Rotate caps to be perpendicular to the error bars
+                    rotation = (lo_theta + hi_theta)/2
+                    if axis == 'x':
+                        rotation += np.pi / 2
+                    ms = mmarkers.MarkerStyle(marker=marker)
+                    ms._transform = mtransforms.Affine2D().rotate(rotation)
+                    self.add_line(mlines.Line2D([lo_theta], [lo_r],
+                                  marker=ms, **eb_cap_style))
+                    self.add_line(mlines.Line2D([hi_theta], [hi_r],
+                                  marker=ms, **eb_cap_style))
+        else:
+            for axis in caplines:
+                for l in caplines[axis]:
+                    self.add_line(l)
 
         self._request_autoscale_view()
+        caplines = caplines['x'] + caplines['y']
         errorbar_container = ErrorbarContainer(
             (data_line, tuple(caplines), tuple(barcols)),
             has_xerr=(xerr is not None), has_yerr=(yerr is not None),
 
@@ -3600,6 +3600,16 @@ def test_errorbar():
     ax.set_title("Simplest errorbars, 0.2 in x, 0.4 in y")
 
 
+@image_comparison(['errorbar_polar_caps'], extensions=['png'])
+def test_errorbar_polar_caps():
+    fig = plt.figure()
+    ax = plt.subplot(111, projection='polar')
+    theta = np.arange(0, 2*np.pi, np.pi / 4)
+    r = theta / np.pi / 2 + 0.5
+    ax.errorbar(theta, r, xerr=0.15, yerr=0.1)
+    ax.set_rlim(0, 2)
+
+
 def test_errorbar_colorcycle():
 
     f, ax = plt.subplots()