diff --git a/R/data.table.R b/R/data.table.R
index 1f5af08be..f14e993d5 100644
--- a/R/data.table.R
+++ b/R/data.table.R
@@ -394,6 +394,18 @@ chmatch2 <- function(x, table, nomatch=NA_integer_) {
         isub = substitute(i)
         # Fixes 4994: a case where quoted expression with a "!", ex: expr = quote(!dt1); dt[eval(expr)] requires 
         # the "eval" to be checked before `as.name("!")`. Therefore interchanged.
+        restore.N = remove.N = FALSE
+        if (exists(".N", envir=parent.frame(), inherits=FALSE)) {
+           old.N = get(".N", envir=parent.frame(), inherits=FALSE)
+           if (!bindingIsLocked(".N", parent.frame())) {
+               assign(".N", nrow(x), envir=parent.frame(), inherits=FALSE)
+               restore.N = TRUE
+           } # binding locked when .SD[.N] but that's ok as that's the .N we want anyway
+           # TO DO: change isub at C level s/.N/nrow(x); changing a symbol to a constant should be ok
+        } else {
+           assign(".N", nrow(x), envir=parent.frame(), inherits=FALSE)
+           remove.N = TRUE
+        }
         if (is.call(isub) && isub[[1L]]=="eval") {  # TO DO: or ..()
             isub = eval(.massagei(isub[[2L]]), parent.frame(), parent.frame())
             if (is.expression(isub)) isub=isub[[1L]]
@@ -416,7 +428,9 @@ chmatch2 <- function(x, table, nomatch=NA_integer_) {
             i = eval(isub, order_env, parent.frame())             # for optimisation of 'order' to 'forder'
             # that forder returns integer(0) is taken care of internally within forder
         } else if (!is.name(isub)) i = eval(.massagei(isub), x, parent.frame())
-        else i = eval(isub, parent.frame(), parent.frame())
+          else i = eval(isub, parent.frame(), parent.frame())
+        if (restore.N) assign(".N", old.N, envir=parent.frame())
+        if (remove.N) rm(list=".N", envir=parent.frame())
         if (is.matrix(i)) stop("i is invalid type (matrix). Perhaps in future a 2 column matrix could return a list of elements of DT (in the spirit of A[B] in FAQ 2.14). Please let datatable-help know if you'd like this, or add your comments to FR #1611.")
         if (is.logical(i)) {
             if (notjoin) {
diff --git a/README.md b/README.md
index 592e09ec8..2bfa1d824 100644
--- a/README.md
+++ b/README.md
@@ -149,6 +149,8 @@ We moved from R-Forge to GitHub on 9 June 2014, including history.
   
   25. `setorder()` and `setorderv()` gain `na.last = TRUE/FALSE`. Closes [#706](https://github.com/Rdatatable/data.table/issues/706).
 
+  26. `.N` is now available in `i`, [FR#724](https://github.com/Rdatatable/data.table/issues/724). Thanks to newbie indirectly [here](http://stackoverflow.com/a/24649115/403310) and Farrel directly [here](http://stackoverflow.com/questions/24685421/how-do-you-extract-a-few-random-rows-from-a-data-table-on-the-fly).
+
 #### BUG FIXES
 
   1. `fread()`:
diff --git a/inst/tests/tests.Rraw b/inst/tests/tests.Rraw
index 7292021eb..4f00b551f 100644
--- a/inst/tests/tests.Rraw
+++ b/inst/tests/tests.Rraw
@@ -4872,6 +4872,13 @@ DT <- data.table(id=rep(1:2, c(3,2)), k=c(letters[1:3], letters[1:2]), v=1:5)
 test(1347.1, dcast.data.table(DT, id ~ k, fun.aggregate=last, value.var="v"), error="Aggregating function provided to argument 'fun.aggregate' should always return a length 1 vector")
 test(1347.2, dcast.data.table(DT, id ~ k, fun.aggregate=last, value.var="v", fill=NA_integer_), data.table(id=1:2, a=c(1L, 4L), b=c(2L,5L), c=c(3L,NA_integer_), key="id"))
 
+# .N now available in i
+DT = data.table(a=1:3,b=1:6)
+test(1348, DT[.N], DT[6])
+test(1349, DT[.N-1:3], DT[5:3])
+test(1350, DT[.N+1], DT[NA])
+
+
 ##########################