diff --git a/lib/Date/Language/Czech.pm b/lib/Date/Language/Czech.pm
index 4273162..6c3868d 100644
--- a/lib/Date/Language/Czech.pm
+++ b/lib/Date/Language/Czech.pm
@@ -56,7 +56,7 @@ sub strftime {
       my $ref = shift;
       my @a = @_;
       $a[0] =~ s/(%[do]\.?\s?)%B/$1%Q/;
-      $ref->SUPER::time2str(@a);
+      $ref->SUPER::strftime(@a);
       }
 
 1;
diff --git a/t/czech-strftime.t b/t/czech-strftime.t
new file mode 100644
index 0000000..23d7d02
--- /dev/null
+++ b/t/czech-strftime.t
@@ -0,0 +1,35 @@
+use strict;
+use warnings;
+use Test::More;
+use Date::Language;
+
+# Regression test: Czech strftime() was calling SUPER::time2str()
+# instead of SUPER::strftime(), producing garbage dates because
+# time2str() interprets its second argument as epoch seconds,
+# not as a localtime() array reference.
+
+my $cz = Date::Language->new('Czech');
+
+# Tue Sep  7 13:02:42 1999 GMT
+my @tm = (42, 2, 13, 7, 8, 99, 2, 249, 0);
+
+# strftime takes (format, time_ref) — same as time2str but with arrayref
+my $result = $cz->strftime('%d. %B %Y', \@tm);
+
+# Must match time2str output for the same date
+my $epoch = 936709362; # same timestamp
+my $expected = $cz->time2str('%d. %B %Y', $epoch, 'GMT');
+
+is($result, $expected, 'Czech strftime produces same result as time2str');
+
+# Verify the date components are sane
+like($result, qr/^ ?7\. .+ 1999$/, 'Czech strftime: day 7, year 1999');
+
+# Also test the %o format (Czech day ordinal)
+is($cz->strftime('%o', \@tm), '7.', 'Czech strftime: %o produces day with dot');
+
+# Test that time2str still works (both methods should apply the B→Q substitution)
+like($cz->time2str('%d. %B %Y', $epoch, 'GMT'), qr/1999/, 'Czech time2str still works');
+like($cz->strftime('%d. %B %Y', \@tm), qr/1999/, 'Czech strftime returns correct year');
+
+done_testing;