diff --git a/CHANGELOG.md b/CHANGELOG.md
index b1741ff32..de7c11d61 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -15,6 +15,8 @@ This section is for changes merged to the `major` branch that are not also merge
 - Successive commas in brace expansions are handled in less surprising manner (`{,,,}` expands to four empty strings rather than an empty string, a comma and an empty string again). (#3002, #4632).
 - `%` is no longer used for process and job expansion. `$fish_pid` and `$last_pid` have taken the place of `%self` and `%last` respectively. (#4230, #1202)
 - The new `math` builtin (see below) does not support logical expressions; `test` should be used instead (#4777).
+- The `?` wildcard has been removed (#4520).
+- The `^` caret redirection for stderr has been removed (#4394). To redirect stderr, `2>/some/path` may be used, or `2>|` as a pipe.
 
 ## Notable fixes and improvements
 - `wait` builtin is added for waiting on processes (#4498).
@@ -51,7 +53,6 @@ This section is for changes merged to the `major` branch that are not also merge
 - The machine hostname, where available, is now exposed as `$hostname` which is now a reserved variable. This drops the dependency on the `hostname` executable (#4422).
 - `functions --handlers` can be used to show event handlers (#4694).
 - Variables set in `if` and `while` conditions are available outside the block (#4820).
-- The `?` wildcard has been removed (#4520).
 
 ## Other significant changes
 - Command substitution output is now limited to 10 MB by default (#3822).
diff --git a/src/common.cpp b/src/common.cpp
index a1612a06e..f5fcf65c3 100644
--- a/src/common.cpp
+++ b/src/common.cpp
@@ -1007,7 +1007,6 @@ static void escape_string_script(const wchar_t *orig_in, size_t in_len, wcstring
                 case L'$':
                 case L' ':
                 case L'#':
-                case L'^':
                 case L'<':
                 case L'>':
                 case L'(':
diff --git a/src/fish_tests.cpp b/src/fish_tests.cpp
index fb080f632..d748e6e27 100644
--- a/src/fish_tests.cpp
+++ b/src/fish_tests.cpp
@@ -534,7 +534,7 @@ static void test_tokenizer() {
 
     const wchar_t *str =
         L"string <redirection  2>&1 'nested \"quoted\" '(string containing subshells "
-        L"){and,brackets}$as[$well (as variable arrays)] not_a_redirect^ ^ ^^is_a_redirect "
+        L"){and,brackets}$as[$well (as variable arrays)] not_a_redirect^ 2> 2>^is_a_redirect "
         L"&&& ||| "
         L"&& || & |"
         L"Compress_Newlines\n  \n\t\n   \nInto_Just_One";
@@ -608,8 +608,6 @@ static void test_tokenizer() {
     // Test redirection_type_for_string.
     if (redirection_type_for_string(L"<") != redirection_type_t::input)
         err(L"redirection_type_for_string failed on line %ld", (long)__LINE__);
-    if (redirection_type_for_string(L"^") != redirection_type_t::overwrite)
-        err(L"redirection_type_for_string failed on line %ld", (long)__LINE__);
     if (redirection_type_for_string(L">") != redirection_type_t::overwrite)
         err(L"redirection_type_for_string failed on line %ld", (long)__LINE__);
     if (redirection_type_for_string(L"2>") != redirection_type_t::overwrite)
diff --git a/src/tokenizer.cpp b/src/tokenizer.cpp
index afd0903da..75d4ce71f 100644
--- a/src/tokenizer.cpp
+++ b/src/tokenizer.cpp
@@ -70,10 +70,8 @@ bool tokenizer_t::next(struct tok_t *result) {
     return true;
 }
 
-/// Tests if this character can be a part of a string. The redirect ^ is allowed unless it's the
-/// first character. Hash (#) starts a comment if it's the first character in a token; otherwise it
-/// is considered a string character. See issue #953.
-static bool tok_is_string_character(wchar_t c, bool is_first) {
+/// Tests if this character can be a part of a string.
+static bool tok_is_string_character(wchar_t c) {
     switch (c) {
         case L'\0':
         case L' ':
@@ -84,15 +82,9 @@ static bool tok_is_string_character(wchar_t c, bool is_first) {
         case L'\r':
         case L'<':
         case L'>':
-        case L'&': {
-            // Unconditional separators.
+        case L'&':
             return false;
-        }
-        case L'^': {
-            // Conditional separator.
-            return !is_first;
-        }
-        default: { return true; }
+        default: return true;
     }
 }
 
@@ -117,7 +109,6 @@ tok_t tokenizer_t::read_string() {
     std::vector<char> expecting;
     int slice_offset = 0;
     const wchar_t *const buff_start = this->buff;
-    bool is_first = true;
 
     while (true) {
         wchar_t c = *this->buff;
@@ -219,7 +210,7 @@ tok_t tokenizer_t::read_string() {
                 break;
             }
         }
-        else if (mode == tok_mode::regular_text && !tok_is_string_character(c, is_first)) {
+        else if (mode == tok_mode::regular_text && !tok_is_string_character(c)) {
             break;
         }
 
@@ -233,7 +224,6 @@ tok_t tokenizer_t::read_string() {
 #endif
 
         this->buff++;
-        is_first = false;
     }
 
     if ((!this->accept_unfinished) && (mode != tok_mode::regular_text)) {
@@ -292,7 +282,7 @@ static maybe_t<parsed_redir_or_pipe_t> read_redirection_or_fd_pipe(const wchar_t
     size_t idx = 0;
 
     // Determine the fd. This may be specified as a prefix like '2>...' or it may be implicit like
-    // '>' or '^'. Try parsing out a number; if we did not get any digits then infer it from the
+    // '>'. Try parsing out a number; if we did not get any digits then infer it from the
     // first character. Watch out for overflow.
     long long big_fd = 0;
     for (; iswdigit(buff[idx]); idx++) {
@@ -313,10 +303,6 @@ static maybe_t<parsed_redir_or_pipe_t> read_redirection_or_fd_pipe(const wchar_t
                 result.fd = STDIN_FILENO;
                 break;
             }
-            case L'^': {
-                result.fd = STDERR_FILENO;
-                break;
-            }
             default: {
                 errored = true;
                 break;
@@ -325,12 +311,11 @@ static maybe_t<parsed_redir_or_pipe_t> read_redirection_or_fd_pipe(const wchar_t
     }
 
     // Either way we should have ended on the redirection character itself like '>'.
-    // Don't allow an fd with a caret redirection - see #1873
     wchar_t redirect_char = buff[idx++];  // note increment of idx
-    if (redirect_char == L'>' || (redirect_char == L'^' && idx == 1)) {
+    if (redirect_char == L'>') {
         result.redirection_mode = redirection_type_t::overwrite;
         if (buff[idx] == redirect_char) {
-            // Doubled up like ^^ or >>. That means append.
+            // Doubled up like >>. That means append.
             result.redirection_mode = redirection_type_t::append;
             idx++;
         }
@@ -505,8 +490,7 @@ maybe_t<tok_t> tokenizer_t::tok_next() {
             break;
         }
         case L'>':
-        case L'<':
-        case L'^': {
+        case L'<': {
             // There's some duplication with the code in the default case below. The key difference
             // here is that we must never parse these as a string; a failed redirection is an error!
             auto redir_or_pipe = read_redirection_or_fd_pipe(this->buff);
@@ -615,9 +599,7 @@ bool move_word_state_machine_t::consume_char_punctuation(wchar_t c) {
 }
 
 bool move_word_state_machine_t::is_path_component_character(wchar_t c) {
-    // Always treat separators as first. All this does is ensure that we treat ^ as a string
-    // character instead of as stderr redirection, which I hypothesize is usually what is desired.
-    return tok_is_string_character(c, true) && !wcschr(L"/={,}'\"", c);
+    return tok_is_string_character(c) && !wcschr(L"/={,}'\"", c);
 }
 
 bool move_word_state_machine_t::consume_char_path_components(wchar_t c) {
diff --git a/tests/gen_output.fish b/tests/gen_output.fish
index 7bf7f3ec3..788e1a421 100755
--- a/tests/gen_output.fish
+++ b/tests/gen_output.fish
@@ -8,10 +8,10 @@
 #
 
 for i in $argv
-	set template_out (basename $i .in).out
-	set template_err (basename $i .in).err
-	set template_status (basename $i .in).status
+    set template_out (basename $i .in).out
+    set template_err (basename $i .in).err
+    set template_status (basename $i .in).status
 
-	fish <$i >$template_out ^$template_err
-	echo $status >$template_status
+    fish <$i >$template_out 2>$template_err
+    echo $status >$template_status
 end
diff --git a/tests/interactive.fish b/tests/interactive.fish
index 41552f615..ad10cbf54 100644
--- a/tests/interactive.fish
+++ b/tests/interactive.fish
@@ -32,8 +32,9 @@ else
     set files_to_test *.expect
 end
 
-source test_util.fish (status -f) $argv; or exit
-cat interactive.config >> $XDG_CONFIG_HOME/fish/config.fish
+source test_util.fish (status -f) $argv
+or exit
+cat interactive.config >>$XDG_CONFIG_HOME/fish/config.fish
 
 say -o cyan "Testing interactive functionality"
 if not type -q expect
@@ -46,7 +47,7 @@ function test_file
     echo -n "Testing file $file ... "
     begin
         set -lx TERM dumb
-        expect -n -c 'source interactive.expect.rc' -f $file >$file.tmp.out ^$file.tmp.err
+        expect -n -c 'source interactive.expect.rc' -f $file >$file.tmp.out 2>$file.tmp.err
     end
     set -l exit_status $status
     set -l res ok
diff --git a/tests/jobs.in b/tests/jobs.in
index 9b2ba6c22..326f20e59 100644
--- a/tests/jobs.in
+++ b/tests/jobs.in
@@ -3,7 +3,7 @@ sleep 5 &
 sleep 5 &
 jobs -c
 jobs -q; echo $status
-bg -23 1 ^/dev/null
+bg -23 1 2>/dev/null
 or echo bg: invalid option -23 >&2
 fg 3
 bg 3
diff --git a/tests/printf.in b/tests/printf.in
index 366b34921..4ee7e95cd 100644
--- a/tests/printf.in
+++ b/tests/printf.in
@@ -28,7 +28,7 @@ printf 'a\cb'
 echo
 
 # Bogus printf specifier, should produce no stdout
-printf "%5" 10 ^/dev/null
+printf "%5" 10 2>/dev/null
 
 # Octal escapes produce literal bytes, not characters
 # \376 is 0xFE
diff --git a/tests/read.in b/tests/read.in
index fcd2e071d..7230e4cfc 100644
--- a/tests/read.in
+++ b/tests/read.in
@@ -141,7 +141,7 @@ set line abcdefghijklmnopqrstuvwxyz
 # Ensure the `read` command terminates if asked to read too much data. The var
 # should be empty. We throw away any data we read if it exceeds the limit on
 # what we consider reasonable.
-yes $line | dd bs=1024 count=(math "1 + $fish_read_limit / 1024") ^/dev/null | read --null x
+yes $line | dd bs=1024 count=(math "1 + $fish_read_limit / 1024") 2>/dev/null | read --null x
 if test $status -ne 122
     echo reading too much data did not terminate with failure status
 end
@@ -188,7 +188,7 @@ end
 
 # Same as previous test but limit the amount of data fed to `read` rather than
 # using the `--nchars` flag.
-yes $line | dd bs=1024 count=(math "$fish_read_limit / 1024") ^/dev/null | read --null x
+yes $line | dd bs=1024 count=(math "$fish_read_limit / 1024") 2>/dev/null | read --null x
 if test $status -ne 0
     echo the read of the max amount of data failed unexpectedly
 end
diff --git a/tests/test.fish b/tests/test.fish
index c42e4a08e..0a0dd5605 100644
--- a/tests/test.fish
+++ b/tests/test.fish
@@ -35,7 +35,7 @@ function test_file
 
     echo -n "Testing file $file ... "
 
-    ../test/root/bin/fish <$file >$base.tmp.out ^$base.tmp.err
+    ../test/root/bin/fish <$file >$base.tmp.out 2>$base.tmp.err
     set -l exit_status $status
     set -l res ok
 
diff --git a/tests/test6.in b/tests/test6.in
index 0ea772bb9..b64d2048b 100644
--- a/tests/test6.in
+++ b/tests/test6.in
@@ -88,7 +88,7 @@ if begin; rm -rf test6.tmp.dir; and mkdir test6.tmp.dir; end
     end
     popd
     if begin
-            set -l PATH $PWD/test6.tmp.dir $PATH ^/dev/null
+            set -l PATH $PWD/test6.tmp.dir $PATH 2>/dev/null
             complete -C$dir | grep "^$dir/.*Directory" >/dev/null
         end
         echo "incorrect implicit cd from PATH"