diff --git a/readme.md b/readme.md
index 9768385..bf4c935 100644
--- a/readme.md
+++ b/readme.md
@@ -14,10 +14,13 @@ https://www.mathworks.com/matlabcentral/fileexchange/34087-cbrewer-colorbrewer-s
 The file is included here for convenience; copyright remains with the original author.
 
 The following are two examples comparing the default outputs of `scatter(.)` and `scatter_nice(.)`.
-In the first example, the linear colormap used by `scatter` is completely useless due to a skewed distribution of data points. `scatter_nice` fixes this by using an automatically balanced nonlinear colormap.
+In the first example, the linear colormap used by `scatter` is completely useless due to a skewed distribution of data points.
+`scatter_nice` fixes this by using an automatically balanced nonlinear colormap.
+Moreover, outliers near the beginning of the input vector are completely invisible in the first plot since they are hidden by later datapoints.
+Using `scatter_nice`, they are easily visible because of a) the nonlinear colormap and b) *randomized-order plotting*.
 In the second example, simply plotting two groups after another (with `hold on`) using `scatter`, the second group will completely hide the first.
 `scatter_nice` offers a simple interface for plotting multiple groups and plots them in *randomized order*.
-In both cases, `scatter_nice` adaptively chooses a point size that is (to my eyes, at least) better suited to the number of datapoints, and selects a fitting and well-crafted colormap from `cbrewer`.
+In both examples, `scatter_nice` adaptively chooses a point size that is (to my eyes, at least) better suited to the number of datapoints, and selects a fitting and well-crafted colormap from `cbrewer`.
 
 ![An example plot.](demo.png)