diff --git a/dataframely/_base_schema.py b/dataframely/_base_schema.py
index 7ae64e3d..b875e6bf 100644
--- a/dataframely/_base_schema.py
+++ b/dataframely/_base_schema.py
@@ -21,12 +21,12 @@
 else:
     from typing_extensions import Self
 
-
 _COLUMN_ATTR = "__dataframely_columns__"
 _RULE_ATTR = "__dataframely_rules__"
 
 ORIGINAL_COLUMN_PREFIX = "__DATAFRAMELY_ORIGINAL__"
 
+
 # --------------------------------------- UTILS -------------------------------------- #
 
 
@@ -84,6 +84,25 @@ class Metadata:
     rules: dict[str, RuleFactory] = field(default_factory=dict)
 
     def update(self, other: Self) -> None:
+        """Merge another Metadata instance into this one.
+
+        Overlapping keys are allowed if and only if they refer to the *same* underlying
+        object. This accommodates multiple-inheritance / diamond patterns where the same
+        base schema is visited more than once.
+        """
+        # Detect conflicting column definitions: same name, different Column instance
+        duplicated_column_names = self.columns.keys() & other.columns.keys()
+        conflicting_columns = {
+            name
+            for name in duplicated_column_names
+            if self.columns[name] is not other.columns[name]
+        }
+        if conflicting_columns:
+            raise ImplementationError(
+                f"Columns {conflicting_columns} are duplicated with conflicting definitions."
+            )
+
+        # All clear
         self.columns.update(other.columns)
         self.rules.update(other.rules)
 
@@ -203,6 +222,8 @@ def _get_metadata(source: dict[str, Any]) -> Metadata:
             k: v for k, v in source.items() if not k.startswith("__")
         }.items():
             if isinstance(value, Column):
+                if (col_name := value.alias or attr) in result.columns:
+                    raise ImplementationError(f"Column {col_name!r} is duplicated.")
                 result.columns[value.alias or attr] = value
             if isinstance(value, RuleFactory):
                 # We must ensure that custom rules do not clash with internal rules.
diff --git a/tests/columns/test_alias.py b/tests/columns/test_alias.py
index 0dd364a7..7276f049 100644
--- a/tests/columns/test_alias.py
+++ b/tests/columns/test_alias.py
@@ -2,8 +2,10 @@
 # SPDX-License-Identifier: BSD-3-Clause
 
 import polars as pl
+import pytest
 
 import dataframely as dy
+from dataframely.exc import ImplementationError
 
 
 class AliasSchema(dy.Schema):
@@ -36,3 +38,21 @@ def test_alias_unset() -> None:
     no_alias_col = dy.Int32()
     assert no_alias_col.alias is None
     assert no_alias_col.name == ""
+
+
+def test_duplicate_alias_same_schema() -> None:
+    with pytest.raises(ImplementationError, match="'a' is duplicated"):
+
+        class MySchema(dy.Schema):
+            a = dy.Int64(alias="a")
+            b = dy.String(alias="a")
+
+
+def test_duplicate_alias_inherited_schema() -> None:
+    class MySchema(dy.Schema):
+        a = dy.Int64(alias="a")
+
+    with pytest.raises(ImplementationError, match="'a'.*duplicated"):
+
+        class MySchema2(MySchema):
+            b = dy.Int64(alias="a")