diff --git a/CanlabCore/@fmri_data/model_brain_pathway.m b/CanlabCore/@fmri_data/model_brain_pathway.m
index 1c569a45..07190407 100644
--- a/CanlabCore/@fmri_data/model_brain_pathway.m
+++ b/CanlabCore/@fmri_data/model_brain_pathway.m
@@ -323,8 +323,8 @@
     [~,~,xs_pathway_three, ys_pathway_three] = plsregress(source_one_train',target_two_train',ndim);
     [~,~,xs_pathway_four, ys_pathway_four] = plsregress(source_two_train',target_two_train',ndim);
     
-    Z_pathway_one = pinv(source_one_train') * ys_pathway_one;  % Z = pattern across X voxels (source), predicting latent Y target
-    V_pathway_one = pinv(target_one_train') * xs_pathway_one;  % V = pattern across Y voxels (target), predicting latent X source
+    Z_pathway_one = pinv(source_one_train') * ys_pathway_one;  % Z = pattern across Y voxels (target), predicting latent Y target
+    V_pathway_one = pinv(target_one_train') * xs_pathway_one;  % V = pattern across Z voxels (source), predicting latent X source
     
     Z_pathway_two = pinv(source_two_train') * ys_pathway_two;
     V_pathway_two = pinv(target_one_train') * xs_pathway_two;
@@ -361,25 +361,28 @@
     
     % optimized models
     % Revision 6/4/2021, Phil Kragel and Tor Wager - documentation
-    
+
     % YS: linear combinations of target voxels chosen to covary with X (source)
-    YS_Test_target_one_pathway_one = Ytest_target_one * V_pathway_one;  % Fitted (predicted) latent Y, based on observed Y. V = pattern across Y (target), predicting latent X source
+    % YS = T_hat: Predicted latent timeseries of Source (X) covarying with Target (Y)
+    % Linear combination of target test data and spatial patterns of X (V)
+    YS_Test_target_one_pathway_one = Ytest_target_one * V_pathway_one;
     YS_Test_target_one_pathway_two = Ytest_target_one * V_pathway_two;
     YS_Test_target_two_pathway_three = Ytest_target_two * V_pathway_three;
     YS_Test_target_two_pathway_four = Ytest_target_two * V_pathway_four;
-    
-    % XS: linear combinations of source voxels chosen to covary with Y (target)
-    XS_Test_source_one_pathway_one = Xtest_source_one * Z_pathway_one; % Fitted (predicted) latent X, based on observed Y; Z = pattern across X (source), predicting latent Y target  
+
+    % XS = U_hat: Predicted latent timeseries of Target (Y) covarying with Source (X)
+    % Linear combination of source test data and spatial patterns of Y (Z)
+    XS_Test_source_one_pathway_one = Xtest_source_one * Z_pathway_one; % Fitted (predicted) latent X, based on observed Y; Z = pattern across X (source), predicting latent Y target
     XS_Test_source_two_pathway_two = Xtest_source_two * Z_pathway_two;
     XS_Test_source_one_pathway_three = Xtest_source_one * Z_pathway_three;
     XS_Test_source_two_pathway_four = Xtest_source_two * Z_pathway_four;
-    
-    stats.latent_timeseries_source(indices==k, 1) = XS_Test_source_one_pathway_one(:, 1);   % (cross-validated) time series of the latent var for X, pathway 1 
+
+    stats.latent_timeseries_source(indices==k, 1) = XS_Test_source_one_pathway_one(:, 1);
     stats.latent_timeseries_source(indices==k, 2) = XS_Test_source_two_pathway_two(:, 1);
     stats.latent_timeseries_source(indices==k, 3) = XS_Test_source_one_pathway_three(:, 1);
     stats.latent_timeseries_source(indices==k, 4) = XS_Test_source_two_pathway_four(:, 1);   
 
-    stats.latent_timeseries_target(indices==k, 1) = YS_Test_target_one_pathway_one(:, 1);   % (cross-validated) time series of the latent var for Y, pathway 1
+    stats.latent_timeseries_target(indices==k, 1) = YS_Test_target_one_pathway_one(:, 1);
     stats.latent_timeseries_target(indices==k, 2) = YS_Test_target_one_pathway_two(:, 1);    
     stats.latent_timeseries_target(indices==k, 3) = YS_Test_target_two_pathway_three(:, 1);
     stats.latent_timeseries_target(indices==k, 4) = YS_Test_target_two_pathway_four(:, 1);