diff --git a/NEWS.md b/NEWS.md
index d0c1cf00..7a71f99e 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -13,6 +13,9 @@
   
 * `str_replace()` and `str_replace_all()` now behave correctly when a
   replacement string contains `$`s, `\\\\1`, etc. (#83, @gagolews).
+  
+* `boundary()` has a different default argument which works for splitting on
+   sentence boundaries (#58, @lmullen).
 
 # stringr 1.0.0
 
diff --git a/R/modifiers.r b/R/modifiers.r
index 5e2ee492..ce7648a4 100644
--- a/R/modifiers.r
+++ b/R/modifiers.r
@@ -115,12 +115,18 @@ regex <- function(pattern, ignore_case = FALSE, multiline = FALSE,
 
 #' @param type Boundary type to detect.
 #' @param skip_word_none Ignore "words" that don't contain any characters
-#'   or numbers - i.e. punctuation.
+#'   or numbers - i.e. punctuation. Default \code{NA} will skip such "words"
+#'   only when splitting on \code{word} boundaries.
 #' @export
 #' @rdname modifiers
 boundary <- function(type = c("character", "line_break", "sentence", "word"),
-                    skip_word_none = TRUE, ...) {
+                    skip_word_none = NA, ...) {
   type <- match.arg(type)
+
+  if (identical(skip_word_none, NA)) {
+    skip_word_none <- type == "word"
+  }
+
   options <- stri_opts_brkiter(
     type = type,
     skip_word_none = skip_word_none,
diff --git a/man/modifiers.Rd b/man/modifiers.Rd
index 1b27dd1f..732165fa 100644
--- a/man/modifiers.Rd
+++ b/man/modifiers.Rd
@@ -16,7 +16,7 @@ regex(pattern, ignore_case = FALSE, multiline = FALSE, comments = FALSE,
   dotall = FALSE, ...)
 
 boundary(type = c("character", "line_break", "sentence", "word"),
-  skip_word_none = TRUE, ...)
+  skip_word_none = NA, ...)
 }
 \arguments{
 \item{pattern}{Pattern to modify behaviour.}
@@ -43,7 +43,8 @@ default, only match the start and end of the input.}
 \item{type}{Boundary type to detect.}
 
 \item{skip_word_none}{Ignore "words" that don't contain any characters
-or numbers - i.e. punctuation.}
+or numbers - i.e. punctuation. Default \code{NA} will skip such "words"
+only when splitting on \code{word} boundaries.}
 }
 \description{
 \describe{
diff --git a/tests/testthat/test-split.r b/tests/testthat/test-split.r
index 942bf9e0..33c070ea 100644
--- a/tests/testthat/test-split.r
+++ b/tests/testthat/test-split.r
@@ -67,3 +67,12 @@ test_that("n sets exact number of splits in str_split_fixed", {
     equals(c("Subject", "Roger: his drinking problems")))
 
 })
+
+test_that("str_split can split sentences correctly", {
+ test <- "This is a sentence. Is this a sentence? Why, yes it is."
+ expect_that(length(str_split(test, boundary("sentence"))[[1]]),
+             equals(3))
+ expect_that(str_split(test, boundary("sentence")),
+             equals(list(c("This is a sentence. ", "Is this a sentence? ",
+                           "Why, yes it is."))))
+})