diff --git a/scripts/create_test_subset.py b/scripts/create_test_subset.py
index 250abf284..6605c2347 100755
--- a/scripts/create_test_subset.py
+++ b/scripts/create_test_subset.py
@@ -166,34 +166,65 @@
     """
 )
 
+COHORT_QUERY = gql(
+    """
+    query CohortQuery($project: String!) {
+        project(name: $project) {
+            cohorts {
+                id
+                sequencingGroups {
+                    sample {
+                        id
+                    }
+                }
+            }
+        }
+    }
+    """
+)
+
 
 def main(
     project: str,
     samples_n: int,
     families_n: int,
+    cohort_samples_n: int,
     additional_families: set[str],
     additional_samples: set[str],
+    cohorts: set[str],
     skip_ped: bool = True,
 ):
     """
     Script creates a test subset for a given project.
-    A new project with a prefix -test is created, and for any files in sample/meta,
+    A new project with a suffix -test is created, and for any files in sample/meta,
     sequence/meta, or analysis/output a copy in the -test namespace is created.
     """
-
-    if not any([additional_families, additional_samples, samples_n, families_n]):
+    if not any(
+        [additional_families, additional_samples, samples_n, families_n, cohorts]
+    ):
         raise ValueError('Come on, what exactly are you asking for?')
 
+    if cohorts and not cohort_samples_n:
+        raise ValueError(
+            'You must specify the number of samples to transfer from the cohort.'
+        )
+
     # for reproducibility
     logger.info('Setting random seed to 42')
     random.seed(42)
 
-    # 1. Find and SG IDs to be moved by Family ID -test.
+    # 1. Find SG IDs to be moved by Family ID to -test.
     if families_n or additional_families:
         additional_samples.update(
             get_sids_for_families(project, families_n, additional_families)
         )
 
+    # 1.5 Find SG IDs to be moved by Cohort ID to -test.
+    if cohorts:
+        additional_samples.update(
+            get_sids_for_cohorts(project, cohorts, cohort_samples_n)
+        )
+
     # 2. Get all sample IDs and their SG IDs in project.
     logger.info(f'Querying all samples in {project}')
     sid_output = query(SG_ID_QUERY, variables={'project': project})
@@ -206,7 +237,7 @@ def main(
     )
 
     # 4. Query all the samples from the selected sgs
-    logger.info(f'Transfering {len(additional_samples)} samples. Querying metadata.')
+    logger.info(f'Transferring {len(additional_samples)} samples. Querying metadata.')
     original_project_subset_data = query(
         QUERY_ALL_DATA, {'project': project, 'sids': list(additional_samples)}
     )
@@ -627,6 +658,30 @@ def get_sids_for_families(
     return included_sids
 
 
+def get_sids_for_cohorts(
+    project: str, cohorts: set[str], cohort_samples_n: int
+) -> set[str]:
+    """Returns cohort_samples_n specific samples for given cohort IDs."""
+
+    cohort_sid_output = query(COHORT_QUERY, {'project': project})
+
+    all_cohort_groups = cohort_sid_output.get('project', {}).get('cohorts', [])
+
+    all_cohorts_sample_ids_subset: set[str] = set()
+    for cohort in all_cohort_groups:
+        sids_for_cohort: list[str] = []
+        if cohort.get('id') in cohorts:
+            seq_groups = cohort.get('sequencingGroups', [])
+            for seq_group in seq_groups:
+                sample = seq_group.get('sample')
+                sids_for_cohort.append(sample['id'])
+        all_cohorts_sample_ids_subset.update(
+            random.sample(sids_for_cohort, cohort_samples_n)
+        )
+
+    return all_cohorts_sample_ids_subset
+
+
 def transfer_families(
     initial_project: str, target_project: str, internal_participant_ids: list[int]
 ) -> list[int]:
@@ -852,8 +907,16 @@ def file_exists(path: str) -> bool:
     parser.add_argument(
         '--project', required=True, help='The sample-metadata project ($DATASET)'
     )
-    parser.add_argument('-n', type=int, help='# Random Samples to copy', default=0)
+    parser.add_argument(
+        '-n', type=int, help='# Random Samples to copy', default=DEFAULT_SAMPLES_N
+    )
     parser.add_argument('-f', type=int, help='# Random families to copy', default=0)
+    parser.add_argument(
+        '-nsamples-cohort',
+        type=int,
+        help='# Random samples to copy from each cohort',
+        default=0,
+    )
     # Flag to be used when there isn't available pedigree/family information.
     parser.add_argument(
         '--skip-ped',
@@ -874,6 +937,13 @@ def file_exists(path: str) -> bool:
         type=str,
         default={},
     )
+    parser.add_argument(
+        '--cohorts',
+        nargs='+',
+        help='Cohorts to take random samples from.',
+        type=str,
+        default={},
+    )
     parser.add_argument(
         '--noninteractive', action='store_true', help='Skip interactive confirmation'
     )
@@ -886,7 +956,9 @@ def file_exists(path: str) -> bool:
         project=args.project,
         samples_n=args.n,
         families_n=args.f,
+        cohort_samples_n=args.nsamples_cohort,
         additional_samples=set(args.samples),
         additional_families=set(args.families),
         skip_ped=args.skip_ped,
+        cohorts=set(args.cohorts),
     )