diff --git a/doc/stages/filters.merge.rst b/doc/stages/filters.merge.rst
index b751af867e..d1a382ec31 100644
--- a/doc/stages/filters.merge.rst
+++ b/doc/stages/filters.merge.rst
@@ -5,21 +5,23 @@ filters.merge
 
 The merge filter combines input from multiple sources into a single output.
 In most cases, this happens automatically on output and use of the merge
-filter is unnecessary.
-However, there may be special cases where merging points prior to a particular
-filter or writer is necessary or desirable.  The merge filter will log a
-warning if its input point sets are based on different spatial references.
-No checks are made to ensure that points from various sources being merged
-have similar dimensions or are generally compatible.  Notably, dimensions
-are not initialized when points merged from various sources do not have
-dimensions in common.
+filter is unnecessary.  However, there may be special cases where
+merging points prior to a particular filter or writer is necessary
+or desirable.
+
+The merge filter will log a warning if its input point sets are based on
+different spatial references.  No checks are made to ensure that points
+from various sources being merged have similar dimensions or are generally
+compatible.  Notably, dimensions are not initialized when points merged
+from various sources do not have dimensions in common.
 
 Example 1
 ---------
 
-This pipeline will create an output file "output.las" that contcatantes
-the points from "file1", "file2" and file3".  Note that the explicit
-use of the merge filter is unnecessary in this case.
+This pipeline will create an output file "output.las" that contcatenates
+the points from "file1", "file2" and "file3".  Note that the explicit
+use of the merge filter is unnecessary in this case (removing the merge
+filter will yield the same result).
 
 .. code-block:: json
 
@@ -38,10 +40,42 @@ use of the merge filter is unnecessary in this case.
 Example 2
 ---------
 
+Here are a pair of unlikely pipelines that show one way in which a merge filter
+might be used.  The first pipeline simply reads the input files "utm1.las",
+"utm2.las" and "utm3.las".  Since the points from each input set are
+carried separately through the pipeline, three files are created as output,
+"out1.las", "out2.las" and "out3.las".  "out1.las" contains the points
+in "utm1.las".  "out2.las" contains the points in "utm2.las" and "out3.las"
+contains the points in "utm3.las".
+
 .. code-block:: json
 
     {
       "pipeline": [
-        
+        "utm1.las"
+        "utm2.las",
+        "utm3.las",
+        "out#.las"
       ]
     }
+
+Here is the same pipeline with a merge filter added.  The merge filter will
+combine the points in its input: "utm1.las" and "utm2.las".  Then the result
+of the merge filter is passed to the writer along with "utm3.las".  This
+results in two output files: "out1.las" contains the points from "utm1.las"
+and "utm2.las", while "out2.las" contains the points from "utm3.las".
+
+.. code-block:: json
+
+    {
+      "pipeline": [
+        "utm1.las"
+        "utm2.las",
+        {
+            "type" : "filters.merge"
+        },
+        "utm3.las",
+        "out#.las"
+      ]
+    }
+