diff --git a/manuscript/05-agnostic-decomposition.Rmd b/manuscript/05-agnostic-decomposition.Rmd
index d1ad13e6..dec4d2f9 100644
--- a/manuscript/05-agnostic-decomposition.Rmd
+++ b/manuscript/05-agnostic-decomposition.Rmd
@@ -26,7 +26,7 @@ $$y = \hat{f}(x_1, x_2) = 2 + e^{x_1} - x_2 + x_1 \cdot x_2$$
 Think of the function as a machine learning model.
 We can visualize the function with a 3D plot or a heatmap with contour lines:
 
-```{r, fig.cap = "Prediction surface of a function with two features $X_1$ and $X_2$."}
+```{r decomp-example, fig.cap = "Prediction surface of a function with two features $X_1$ and $X_2$."}
 x1 = seq(from = -1, to = 1, length.out = 30)
 x2 = seq(from = -1, to = 1, length.out = 30)
 f = function(x1, x2){
@@ -64,7 +64,7 @@ The intercept is $\hat{f}_0\sim3.18$.
 Since the other components are functions, we can visualize them:
 
 
-```{r, fig.cap = "Decomposition of a function."}
+```{r decomp-example-continued, fig.cap = "Decomposition of a function."}
 pred.fun = function(model = NULL, newdata){
   f(newdata$x1, newdata$x2)
 }