diff --git a/chirho/indexed/internals.py b/chirho/indexed/internals.py
index b3038148..d684b8ad 100644
--- a/chirho/indexed/internals.py
+++ b/chirho/indexed/internals.py
@@ -1,6 +1,5 @@
-import functools
 import numbers
-from typing import Dict, Hashable, List, Optional, TypeVar, Union
+from typing import Dict, Hashable, Optional, TypeVar, Union
 
 import pyro
 import pyro.infer.reparam
@@ -38,21 +37,6 @@ def _gather_number(
     )
 
 
-@functools.singledispatch
-def index_select_from_array_like(arr, dim: int, indices: List[int]):
-    raise NotImplementedError(f"index_select not implemented for {type(arr)}")
-
-
-@index_select_from_array_like.register
-def _index_select_from_array_like_tensor(
-    arr: torch.Tensor, dim: int, indices: List[int]
-) -> torch.Tensor:
-    indices_tensor = torch.tensor(indices, device=arr.device, dtype=torch.long)
-    return arr.index_select(dim=dim, index=indices_tensor)
-
-
-# TODO _gather_tensor now works for any array like with an index_select_from_array_like implementation.
-#  Can we dispatch for array like objects generally?
 @gather.register
 def _gather_tensor(
     value: torch.Tensor,
@@ -75,8 +59,9 @@ def _gather_tensor(
         dim = name_to_dim[name] - event_dim
         if len(result.shape) < -dim or result.shape[dim] == 1:
             continue
-        result = index_select_from_array_like(
-            result, name_to_dim[name] - event_dim, list(sorted(indices))
+        result = result.index_select(
+            name_to_dim[name] - event_dim,
+            torch.tensor(list(sorted(indices)), device=value.device, dtype=torch.long),
         )
     return result