diff --git a/lib/text/simple_text_displayer.js b/lib/text/simple_text_displayer.js
index 63604c8fdc..e2343caa3f 100644
--- a/lib/text/simple_text_displayer.js
+++ b/lib/text/simple_text_displayer.js
@@ -121,16 +121,28 @@ shaka.text.SimpleTextDisplayer = class {
     // We don't want to modify the array or objects passed in, since we don't
     // technically own them.  So we build a new array and replace certain items
     // in it if they need to be flattened.
-    const flattenedCues = cues.map((cue) => {
-      if (cue.nestedCues.length) {
-        const flatCue = cue.clone();
-        flatCue.nestedCues = [];
-        flatCue.payload = flattenPayload(cue);
-        return flatCue;
-      } else {
-        return cue;
+    // We also don't want to flatten the text payloads starting at a container
+    // element; otherwise, for containers encapsulating multiple caption lines,
+    // the lines would merge into a single cue. This is undesirable when a
+    // subset of the captions are outside of the append time window. To fix
+    // this, we only call flattenPayload() starting at elements marked as
+    // isContainer = false.
+    const getCuesToFlatten = (cues, result) => {
+      for (const cue of cues) {
+        if (cue.isContainer) {
+          // Recurse to find the actual text payload cues.
+          getCuesToFlatten(cue.nestedCues, result);
+        } else {
+          // Flatten the payload.
+          const flatCue = cue.clone();
+          flatCue.nestedCues = [];
+          flatCue.payload = flattenPayload(cue);
+          result.push(flatCue);
+        }
       }
-    });
+      return result;
+    };
+    const flattenedCues = getCuesToFlatten(cues, []);
 
     // Convert cues.
     const textTrackCues = [];
diff --git a/test/text/simple_text_displayer_unit.js b/test/text/simple_text_displayer_unit.js
index 88c4dc86b2..206387f6e4 100644
--- a/test/text/simple_text_displayer_unit.js
+++ b/test/text/simple_text_displayer_unit.js
@@ -127,6 +127,33 @@ describe('SimpleTextDisplayer', () => {
           [shakaCue]);
     });
 
+    it('flattens nested cue payloads correctly', () => {
+      const level0ContainerCue = new shaka.text.Cue(10, 30, '');
+      level0ContainerCue.isContainer = true;
+
+      const level1NonContainerCueA = new shaka.text.Cue(10, 20, '');
+      const level1NonContainerCueB = new shaka.text.Cue(20, 30, '');
+
+      // Add a trailing whitespace character to get a space-delimited expected
+      // result.
+      const cueANestedCue0 = new shaka.text.Cue(10, 20, 'Cue A Test0 ');
+      const cueANestedCue1 = new shaka.text.Cue(10, 20, 'Cue A Test1');
+      const cueBNestedCue0 = new shaka.text.Cue(20, 30, 'Cue B Test0 ');
+      const cueBNestedCue1 = new shaka.text.Cue(20, 30, 'Cue B Test1');
+
+      level1NonContainerCueA.nestedCues = [cueANestedCue0, cueANestedCue1];
+      level1NonContainerCueB.nestedCues = [cueBNestedCue0, cueBNestedCue1];
+      level0ContainerCue.nestedCues =
+          [level1NonContainerCueA, level1NonContainerCueB];
+
+      verifyHelper(
+          [
+            {startTime: 10, endTime: 20, text: 'Cue A Test0 Cue A Test1'},
+            {startTime: 20, endTime: 30, text: 'Cue B Test0 Cue B Test1'},
+          ],
+          [level0ContainerCue]);
+    });
+
     // Regression test for b/159050711
     it('maintains the styles of the parent cue', () => {
       const shakaCue = new shaka.text.Cue(10, 20, '');