diff --git a/docs/bgm-format.md b/docs/bgm-format.md
index 8970580..d0d2519 100644
--- a/docs/bgm-format.md
+++ b/docs/bgm-format.md
@@ -37,9 +37,11 @@ bruce,[]
 
 ### Inline vs file
 
-`$variants` can be a file/URL path *or* an inline CSV string. If the value
-contains a newline it is inline CSV; otherwise it is a path. In YAML a block
-scalar (`|`) is the natural way to write inline CSV; in JSON you'd use `\n`.
+`$variants` can be a single source or an array of sources. Each source is a
+file/URL path if its first line ends in `.csv`, otherwise it is inline CSV.
+This keeps the two forms self-documenting and applies the same rule to single
+values and array elements alike. In YAML a block scalar (`|`) is the natural
+way to write inline CSV; in JSON you'd use `\n`.
 
 ```yaml
 $variants: |
@@ -47,9 +49,21 @@ $variants: |
   string,string,[number;number;number;number]
   fish,Fish,[0;0;5;2]
   grain,Grain,[1;0;5;2]
-  wood,Wood,[2;0;5;2]
 ```
 
+An array of sources concatenates their rows. This lets one part definition
+pull from several CSVs with different schemas — e.g. a deck where the regular
+cards share a face sheet but the jokers have their own:
+
+```yaml
+$variants:
+  - ./cards.csv
+  - ./jokers.csv
+```
+
+Each source is parsed with its own schema, and its rows extend the original
+object independently.
+
 ### CSV conventions
 
 CSV is parsed with `typed-csv`:
diff --git a/packages/bgm/src/variants.test.ts b/packages/bgm/src/variants.test.ts
index 8954f47..acfe213 100644
--- a/packages/bgm/src/variants.test.ts
+++ b/packages/bgm/src/variants.test.ts
@@ -48,7 +48,7 @@ describe('parseCsvData', () => {
 });
 
 describe('expandVariants', () => {
-  it('parses inline CSV when the value contains a newline', () => {
+  it('parses inline CSV when the first line does not end in .csv', () => {
     const rows = expandVariants('a,b\nstring,number\nx,1', 'pkg/def.yaml', new Map(), 'src');
     expect(rows).toEqual([{ a: 'x', b: 1 }]);
   });
@@ -61,15 +61,30 @@ describe('expandVariants', () => {
     expect(rows).toEqual([{ seat: 0 }, { seat: 1 }]);
   });
 
+  it('concatenates rows from an array of csv paths', () => {
+    const defs = new Map<string, DefFile[]>([
+      ['pkg/parts/a.csv', [defFile('pkg/parts/a.csv', 'id\nstring\nred')]],
+      ['pkg/parts/b.csv', [defFile('pkg/parts/b.csv', 'id\nstring\nblack')]],
+    ]);
+    const rows = expandVariants(['./a.csv', './b.csv'], 'pkg/parts/board.yaml', defs, 'src');
+    expect(rows).toEqual([{ id: 'red' }, { id: 'black' }]);
+  });
+
   it('throws when the referenced csv is missing', () => {
     expect(() => expandVariants('./nope.csv', 'pkg/def.yaml', new Map(), 'src')).toThrow(
       /CSV not found/,
     );
   });
 
-  it('throws when $variants is not a string', () => {
+  it('throws when $variants is not a string or string array', () => {
     expect(() => expandVariants(42, 'pkg/def.yaml', new Map(), 'src')).toThrow(
-      /must be a path or inline CSV/,
+      /must be a path or inline CSV string, or an array of them/,
+    );
+  });
+
+  it('throws on an empty $variants array', () => {
+    expect(() => expandVariants([], 'pkg/def.yaml', new Map(), 'src')).toThrow(
+      /must not be empty/,
     );
   });
 });
\ No newline at end of file
diff --git a/packages/bgm/src/variants.ts b/packages/bgm/src/variants.ts
index 466e471..ded4495 100644
--- a/packages/bgm/src/variants.ts
+++ b/packages/bgm/src/variants.ts
@@ -9,8 +9,8 @@
  * - Rows are validated against a schema derived from the type row.
  * - A cell for an array/tuple type uses `;` as the element separator
  *   (`[0;0;5;2]`), because `,` is the CSV delimiter.
- * - `$variants` can be a file/URL path *or* an inline CSV string: a value
- *   containing a newline is inline CSV, otherwise it is a path.
+ * - `$variants` can be a single source or an array of them. A source is a
+ *   file/URL path if its first line ends in `.csv`, otherwise inline CSV.
  *
  * Parsing is delegated to `typed-csv`'s `parseCsv`, which implements exactly
  * this header/schema/data layout and validates each row against a schema
@@ -74,7 +74,8 @@ export function parseCsvByName(
 /**
  * Expand a `$variants` value into rows.
  *
- * @param value the `$variants` value: a path or inline CSV
+ * @param value the `$variants` value: a path or inline CSV string, or an
+ *   array of them
  * @param baseName the path-style name of the referencing def file; a path
  *   value resolves relative to its directory
  * @param defs the virtual def map, for resolving the path
@@ -86,14 +87,41 @@ export function expandVariants(
   defs: Map<string, DefFile[]>,
   source: string,
 ): Record<string, unknown>[] {
-  if (typeof value !== 'string') {
-    throw new BgmError('`$variants` must be a path or inline CSV string', source);
+  const sources = Array.isArray(value) ? value : [value];
+  if (sources.length === 0) {
+    throw new BgmError('`$variants` array must not be empty', source);
   }
 
-  if (value.includes('\n')) {
-    return parseCsvData(value, source).rows;
+  const rows: Record<string, unknown>[] = [];
+  for (const item of sources) {
+    if (typeof item !== 'string') {
+      throw new BgmError(
+        '`$variants` must be a path or inline CSV string, or an array of them',
+        source,
+      );
+    }
+    rows.push(...expandVariantsOne(item, baseName, defs, source));
   }
+  return rows;
+}
 
-  const name = path.posix.join(path.posix.dirname(baseName), value);
-  return parseCsvByName(name, defs, source).rows;
+/**
+ * Expand a single `$variants` source: a path or inline CSV.
+ *
+ * A source is a path when its first line ends in `.csv`; otherwise it is
+ * inline CSV. This keeps the two forms self-documenting and applies the same
+ * rule to single values and array elements alike.
+ */
+function expandVariantsOne(
+  value: string,
+  baseName: string,
+  defs: Map<string, DefFile[]>,
+  source: string,
+): Record<string, unknown>[] {
+  const firstLine = value.split('\n', 1)[0] ?? value;
+  if (/[.]csv$/i.test(firstLine)) {
+    const name = path.posix.join(path.posix.dirname(baseName), value);
+    return parseCsvByName(name, defs, source).rows;
+  }
+  return parseCsvData(value, source).rows;
 }
\ No newline at end of file