diff --git a/doc/devel/color_changes.rst b/doc/devel/color_changes.rst
new file mode 100644
index 000000000000..9703171d9513
--- /dev/null
+++ b/doc/devel/color_changes.rst
@@ -0,0 +1,133 @@
+.. _color_changes:
+
+*********************
+Default Color changes
+*********************
+
+As discussed at length elsewhere [insert links], ``jet`` is an
+empirically bad color map and should not be the default color map.
+Due to the position that changing the appearance of the plot breaks
+backward compatibility, this change has been put off for far longer
+than it should have been.  In addition to changing the default color
+map we plan to take the chance to change the default color-cycle on
+plots and to adopt a different color map for filled plots (``imshow``,
+``pcolor``, ``contourf``, etc) and for scatter like plots.
+
+
+Default Heat Map Colormap
+-------------------------
+
+The choice of a new color map is fertile ground to bike-shedding ("No,
+it should be _this_ color") so we have a proposed set criteria (via
+Nathaniel Smith) to evaluate proposed color maps.
+
+- it should be a sequential colormap, because diverging colormaps are
+  really misleading unless you know where the "center" of the data is,
+  and for a default colormap we generally won't.
+
+- it should be perceptually uniform, i.e., human subjective judgments
+  of how far apart nearby colors are should correspond as linearly as
+  possible to the difference between the numerical values they
+  represent, at least locally.
+
+- it should have a perceptually uniform luminance ramp, i.e. if you
+  convert to greyscale it should still be uniform. This is useful both
+  in practical terms (greyscale printers are still a thing!) and
+  because luminance is a very strong and natural cue to magnitude.
+
+- it should also have some kind of variation in hue, because hue
+  variation is a really helpful additional cue to perception, having
+  two cues is better than one, and there's no reason not to do it.
+
+- the hue variation should be chosen to produce reasonable results
+  even for viewers with the more common types of
+  colorblindness. (Which rules out things like red-to-green.)
+
+- For bonus points, it would be nice to choose a hue ramp that still
+  works if you throw away the luminance variation, because then we
+  could use the version with varying luminance for 2d plots, and the
+  version with just hue variation for 3d plots. (In 3d plots you
+  really want to reserve the luminance channel for lighting/shading,
+  because your brain is *really* good at extracting 3d shape from
+  luminance variation. If the 3d surface itself has massively varying
+  luminance then this screws up the ability to see shape.)
+
+- Not infringe any existing IP
+
+Example script
+++++++++++++++
+
+Proposed Colormaps
+++++++++++++++++++
+
+Default Scatter Colormap
+------------------------
+
+For heat-map like applications it can be desirable to cover as much of
+the luminence scale as possible, however when color mapping markers,
+having markers too close to white can be a problem.  For that reason
+we propose using a different (but maybe related) color map to the
+heat map for marker-based.  The design parameters are the same as
+above, only with a more limited luminence variation.
+
+
+Example script
+++++++++++++++
+::
+   import numpy as np
+   import matplotlib.pyplot as plt
+
+   np.random.seed(1234)
+
+   fig, (ax1, ax2) = plt.subplots(1, 2)
+
+   N = 50
+   x = np.random.rand(N)
+   y = np.random.rand(N)
+   colors = np.random.rand(N)
+   area = np.pi * (15 * np.random.rand(N))**2  # 0 to 15 point radiuses
+
+   ax1.scatter(x, y, s=area, c=colors, alpha=0.5)
+
+
+   X,Y = np.meshgrid(np.arange(0, 2*np.pi, .2),
+                     np.arange(0, 2*np.pi, .2))
+   U = np.cos(X)
+   V = np.sin(Y)
+   Q = ax2.quiver(X, Y, U, V, units='width')
+   qd = np.random.rand(np.prod(X.shape))
+   Q.set_array(qd)
+
+Proposed Colormaps
+++++++++++++++++++
+
+Color Cycle / Qualitative color map
+-----------------------------------
+
+When plotting lines it is frequently desirable to plot multiple lines
+or artists which need to be distinguishable, but there is no inherent
+ordering.
+
+
+Example script
+++++++++++++++
+::
+   import numpy as np
+   import matplotlib.pyplot as plt
+
+   fig, (ax1, ax2) = plt.subplots(1, 2)
+
+   x = np.linspace(0, 1, 10)
+
+   for j in range(10):
+       ax1.plot(x, x * j)
+
+
+   th = np.linspace(0, 2*np.pi, 1024)
+   for j in np.linspace(0, np.pi, 10):
+       ax2.plot(th, np.sin(th + j))
+
+   ax2.set_xlim(0, 2*np.pi)
+
+Proposed Color cycle
+++++++++++++++++++++
diff --git a/doc/devel/index.rst b/doc/devel/index.rst
index 5b3c01eebd5a..9b1872dd154d 100644
--- a/doc/devel/index.rst
+++ b/doc/devel/index.rst
@@ -21,3 +21,4 @@ The Matplotlib Developers' Guide
    release_guide.rst
    transformations.rst
    add_new_projection.rst
+   color_changes