diff --git a/dlib/graph_cuts/find_max_factor_graph_potts_abstract.h b/dlib/graph_cuts/find_max_factor_graph_potts_abstract.h
index 668adf3e3..5943734db 100644
--- a/dlib/graph_cuts/find_max_factor_graph_potts_abstract.h
+++ b/dlib/graph_cuts/find_max_factor_graph_potts_abstract.h
@@ -174,13 +174,14 @@ namespace dlib
             - for all valid i and j:
                 - prob.factor_value_disagreement(i,j) == prob.factor_value_disagreement(j,i)
         ensures
-            - computes the model score for the given potts_problem.  To define this
-              precisely:
+            - computes the model score for the given potts_problem.  We define this
+              precisely below:
                 - let L(i) == the boolean label of the ith variable in prob.  Or in other 
                   words, L(i) == (prob.get_label(i) != 0).
-                - let F == the sum of all values of prob.factor_value(i) where whenever L(i) == true.
+                - let F == the sum of all values of prob.factor_value(i) for only i values
+                  where L(i) == true.
                 - Let D == the sum of all values of prob.factor_value_disagreement(i,j) 
-                  whenever the following conditions are true about i and j:
+                  for only i and j values which meet the following conditions:
                     - i and j are neighbors in the graph defined by prob, that is,
                       it is valid to call prob.factor_value_disagreement(i,j).
                     - L(i) != L(j)