diff --git a/IBDReduce/ibdreduce_v3.py b/IBDReduce/ibdreduce_v3.py
index b637bb8..9c4af6e 100755
--- a/IBDReduce/ibdreduce_v3.py
+++ b/IBDReduce/ibdreduce_v3.py
@@ -83,6 +83,30 @@ def parse_line(line: str) -> Tuple[int, int, np.ndarray, np.ndarray]:
     return chrom, pos, obs_parts, vals
 
 
+def write_permutation_values(
+    data: np.ndarray, idx: int, offset: int, vals: np.ndarray, nperm: int
+) -> None:
+    """Insert observed and permutation statistics into the shared matrix.
+
+    Args:
+        data: The matrix tracking observed and permutation statistics for all
+            breakpoints.
+        idx: Row index of the breakpoint being updated.
+        offset: Zero-based permutation run offset.
+        vals: Array containing the observed value followed by permutation
+            statistics for the current run.
+        nperm: Number of permutations produced per run.
+    """
+
+    if offset == 0:
+        data[idx, 0] = vals[0]
+        data[idx, 1 : (nperm + 1)] = vals[1:]
+    else:
+        start = 1 + offset * nperm
+        end = 1 + (offset + 1) * nperm
+        data[idx, start:end] = vals[1:]
+
+
 def compute_fdr_adjusted_pvalues(empp: np.ndarray, permutation_pvalues: np.ndarray) -> np.ndarray:
     """Calculate FDR-adjusted p-values from permutation p-values.
 
@@ -267,11 +291,7 @@ def main():
                                 )
                             ibdfrac[idx] = (float(dis) - float(predis)) / total
                             predis = dis
-                            data[idx, offset:(args.nperm + 1)] = vals
-                        else:
-                            data[
-                                idx, (offset * args.nperm):((offset + 1) * args.nperm)
-                            ] = vals[1:]
+                        write_permutation_values(data, idx, offset, vals, args.nperm)
                     except ValueError as e:
                         print(f"Error at {i} {j} {idx} {offset} {fpath}", file=sys.stderr)
                         raise e
diff --git a/tests/data/sample_chr1_run1.txt b/tests/data/sample_chr1_run1.txt
new file mode 100644
index 0000000..43a246e
--- /dev/null
+++ b/tests/data/sample_chr1_run1.txt
@@ -0,0 +1,3 @@
+1 1000 0.1 0.2 0.3 0.9 0.8 0.7
+1 2000 0.2 0.1 0.3 0.6 0.5 0.4
+1 3000 0.1 0.1 0.1 0.3 0.2 0.1
diff --git a/tests/test_permutation_columns.py b/tests/test_permutation_columns.py
new file mode 100644
index 0000000..db19c59
--- /dev/null
+++ b/tests/test_permutation_columns.py
@@ -0,0 +1,42 @@
+import pathlib
+import sys
+
+import numpy as np
+
+
+PROJECT_ROOT = pathlib.Path(__file__).resolve().parents[1]
+sys.path.append(str(PROJECT_ROOT / "IBDReduce"))
+
+from ibdreduce_v3 import parse_line, write_permutation_values
+
+
+def test_permutation_columns_do_not_overlap():
+    data_dir = PROJECT_ROOT / "tests" / "data"
+    run_paths = [data_dir / f"sample_chr1_run{j}.txt" for j in range(2)]
+
+    first_line = run_paths[0].read_text(encoding="utf-8").splitlines()[0]
+    _, _, _, first_vals = parse_line(first_line)
+    nperm = len(first_vals) - 1
+
+    with run_paths[0].open("r", encoding="utf-8") as handle:
+        breakpoints = sum(1 for _ in handle)
+
+    data = np.zeros((breakpoints, nperm * len(run_paths) + 1), dtype=float)
+
+    for offset, path in enumerate(run_paths):
+        idx = 0
+        with path.open("r", encoding="utf-8") as handle:
+            for line in handle:
+                _, _, _, vals = parse_line(line)
+                write_permutation_values(data, idx, offset, vals, nperm)
+                idx += 1
+
+    for offset, path in enumerate(run_paths):
+        start = 1 + offset * nperm
+        end = 1 + (offset + 1) * nperm
+        idx = 0
+        with path.open("r", encoding="utf-8") as handle:
+            for line in handle:
+                _, _, _, vals = parse_line(line)
+                np.testing.assert_allclose(data[idx, start:end], vals[1:])
+                idx += 1