diff --git a/src/builtin_math.cpp b/src/builtin_math.cpp
index 41622d602..52ef83bae 100644
--- a/src/builtin_math.cpp
+++ b/src/builtin_math.cpp
@@ -141,6 +141,13 @@ static wcstring math_describe_error(te_error_t& error) {
 
 /// Return a formatted version of the value \p v respecting the given \p opts.
 static wcstring format_double(double v, const math_cmd_opts_t &opts) {
+    // As a special-case, a scale of 0 means to truncate to an integer
+    // instead of rounding.
+    if (opts.scale == 0) {
+        v = std::trunc(v);
+        return format_string(L"%.*f", opts.scale, v);
+    }
+
     wcstring ret = format_string(L"%.*f", opts.scale, v);
     // If we contain a decimal separator, trim trailing zeros after it, and then the separator
     // itself if there's nothing after it. Detect a decimal separator as a non-digit.
diff --git a/tests/math.out b/tests/math.out
index e1913225e..4cafd7a30 100644
--- a/tests/math.out
+++ b/tests/math.out
@@ -3,7 +3,7 @@
 # Validate basic expressions
 1.5
 1.666667
-2
+1
 1
 1.667
 4
@@ -27,7 +27,7 @@
 100000000000000
 -1000000000000000
 0
-2
+1
 500000000000000
 
 ####################