diff --git a/Objects/floatobject.c b/Objects/floatobject.c
index a42be71..b571ca8 100644
--- a/Objects/floatobject.c
+++ b/Objects/floatobject.c
@@ -289,7 +289,7 @@
  * may lose info from fractional bits.  Converting the integer to a double
  * also has two failure modes:  (1) a long int may trigger overflow (too
  * large to fit in the dynamic range of a C double); (2) even a C long may have
- * more bits than fit in a C double (e.g., on a a 64-bit box long may have
+ * more bits than fit in a C double (e.g., on a 64-bit box long may have
  * 63 bits of precision, but a C double probably has only 53), and then
  * we can falsely claim equality when low-order integer bits are lost by
  * coercion to double.  So this part is painful too.