diff --git a/CHANGES b/CHANGES
index f44cc15e..5648234f 100644
--- a/CHANGES
+++ b/CHANGES
@@ -1,3 +1,5 @@
+ * delay() can return milliseconds to delay sending next request.
+
 wrk 4.0.0
 
  * The wrk global variable is the only global defined by default.
diff --git a/SCRIPTING b/SCRIPTING
index 7e9c3f07..c15a03e4 100644
--- a/SCRIPTING
+++ b/SCRIPTING
@@ -38,6 +38,7 @@ Overview
 
     global setup    -- called during thread setup
     global init     -- called when the thread is starting
+    global delay    -- called to get the request delay
     global request  -- called to generate the HTTP request
     global response -- called with HTTP response data
     global done     -- called with results of run
@@ -64,6 +65,7 @@ Setup
 Running
 
   function init(args)
+  function delay()
   function request()
   function response(status, headers, body)
 
@@ -73,6 +75,9 @@ Running
   The init() function receives any extra command line arguments for the
   script which must be separated from wrk arguments with "--".
 
+  delay() returns the number of milliseconds to delay sending the next
+  request.
+
   request() returns a string containing the HTTP request. Building a new
   request each time is expensive, when testing a high performance server
   one solution is to pre-generate all requests in init() and do a quick
diff --git a/scripts/delay.lua b/scripts/delay.lua
new file mode 100644
index 00000000..a9897cca
--- /dev/null
+++ b/scripts/delay.lua
@@ -0,0 +1,6 @@
+-- example script that demonstrates adding a random
+-- 10-50ms delay before each request
+
+function delay()
+   return math.random(10, 50)
+end
diff --git a/src/script.c b/src/script.c
index a8a4d571..8c90564c 100644
--- a/src/script.c
+++ b/src/script.c
@@ -141,6 +141,14 @@ void script_init(lua_State *L, thread *t, int argc, char **argv) {
     lua_pop(t->L, 1);
 }
 
+uint64_t script_delay(lua_State *L) {
+    lua_getglobal(L, "delay");
+    lua_call(L, 0, 1);
+    uint64_t delay = lua_tonumber(L, -1);
+    lua_pop(L, 1);
+    return delay;
+}
+
 void script_request(lua_State *L, char **buf, size_t *len) {
     int pop = 1;
     lua_getglobal(L, "request");
@@ -189,6 +197,10 @@ bool script_want_response(lua_State *L) {
     return script_is_function(L, "response");
 }
 
+bool script_has_delay(lua_State *L) {
+    return script_is_function(L, "delay");
+}
+
 bool script_has_done(lua_State *L) {
     return script_is_function(L, "done");
 }
diff --git a/src/script.h b/src/script.h
index f6861d4d..1bf820da 100644
--- a/src/script.h
+++ b/src/script.h
@@ -16,12 +16,14 @@ void script_setup(lua_State *, thread *);
 void script_done(lua_State *, stats *, stats *);
 
 void script_init(lua_State *, thread *, int, char **);
+uint64_t script_delay(lua_State *);
 void script_request(lua_State *, char **, size_t *);
 void script_response(lua_State *, int, buffer *, buffer *);
 size_t script_verify_request(lua_State *L);
 
 bool script_is_static(lua_State *);
 bool script_want_response(lua_State *L);
+bool script_has_delay(lua_State *L);
 bool script_has_done(lua_State *L);
 void script_summary(lua_State *, uint64_t, uint64_t, uint64_t);
 void script_errors(lua_State *, errors *);
diff --git a/src/wrk.c b/src/wrk.c
index cd99fa97..b0208b06 100644
--- a/src/wrk.c
+++ b/src/wrk.c
@@ -5,13 +5,14 @@
 #include "main.h"
 
 static struct config {
-    uint64_t threads;
     uint64_t connections;
     uint64_t duration;
+    uint64_t threads;
     uint64_t timeout;
     uint64_t pipeline;
-    bool     latency;
+    bool     delay;
     bool     dynamic;
+    bool     latency;
     char    *script;
     SSL_CTX *ctx;
 } cfg;
@@ -107,7 +108,8 @@ int main(int argc, char **argv) {
 
         if (i == 0) {
             cfg.pipeline = script_verify_request(t->L);
-            cfg.dynamic = !script_is_static(t->L);
+            cfg.dynamic  = !script_is_static(t->L);
+            cfg.delay    = script_has_delay(t->L);
             if (script_want_response(t->L)) {
                 parser_settings.on_header_field = header_field;
                 parser_settings.on_header_value = header_value;
@@ -212,6 +214,7 @@ void *thread_main(void *arg) {
         c->ssl     = cfg.ctx ? SSL_new(cfg.ctx) : NULL;
         c->request = request;
         c->length  = length;
+        c->delayed = cfg.delay;
         connect_socket(thread, c);
     }
 
@@ -282,6 +285,13 @@ static int record_rate(aeEventLoop *loop, long long id, void *data) {
     return RECORD_INTERVAL_MS;
 }
 
+static int delay_request(aeEventLoop *loop, long long id, void *data) {
+    connection *c = data;
+    c->delayed = false;
+    aeCreateFileEvent(loop, c->fd, AE_WRITABLE, socket_writeable, c);
+    return AE_NOMORE;
+}
+
 static int header_field(http_parser *parser, const char *at, size_t len) {
     connection *c = parser->data;
     if (c->state == VALUE) {
@@ -331,6 +341,7 @@ static int response_complete(http_parser *parser) {
         if (!stats_record(statistics.latency, now - c->start)) {
             thread->errors.timeout++;
         }
+        c->delayed = cfg.delay;
         aeCreateFileEvent(thread->loop, c->fd, AE_WRITABLE, socket_writeable, c);
     }
 
@@ -371,6 +382,13 @@ static void socket_writeable(aeEventLoop *loop, int fd, void *data, int mask) {
     connection *c = data;
     thread *thread = c->thread;
 
+    if (c->delayed) {
+        uint64_t delay = script_delay(thread->L);
+        aeDeleteFileEvent(loop, fd, AE_WRITABLE);
+        aeCreateTimeEvent(loop, delay, delay_request, c, NULL);
+        return;
+    }
+
     if (!c->written) {
         if (cfg.dynamic) {
             script_request(thread->L, &c->request, &c->length);
diff --git a/src/wrk.h b/src/wrk.h
index af64ee64..e5036761 100644
--- a/src/wrk.h
+++ b/src/wrk.h
@@ -51,6 +51,7 @@ typedef struct connection {
     } state;
     int fd;
     SSL *ssl;
+    bool delayed;
     uint64_t start;
     char *request;
     size_t length;