diff --git a/layeredcache.go b/layeredcache.go
index bb57a3c..1027d1b 100644
--- a/layeredcache.go
+++ b/layeredcache.go
@@ -192,7 +192,7 @@ func (c *LayeredCache) worker() {
 			}
 		case item := <-c.deletables:
 			if item.element == nil {
-				item.promotions = -2
+				atomic.StoreInt32(&item.promotions, -2)
 			} else {
 				c.size -= item.size
 				if c.onDelete != nil {
@@ -206,13 +206,13 @@ func (c *LayeredCache) worker() {
 
 func (c *LayeredCache) doPromote(item *Item) bool {
 	// deleted before it ever got promoted
-	if item.promotions == -2 {
+	if atomic.LoadInt32(&item.promotions) == -2 {
 		return false
 	}
 	if item.element != nil { //not a new item
 		if item.shouldPromote(c.getsPerPromote) {
 			c.list.MoveToFront(item.element)
-			item.promotions = 0
+			atomic.StoreInt32(&item.promotions, 0)
 		}
 		return false
 	}
diff --git a/readme.md b/readme.md
index 8a7efa2..bb399db 100644
--- a/readme.md
+++ b/readme.md
@@ -122,7 +122,7 @@ user := item.Value()   //will be nil if "user:4" didn't exist in the cache
 item.Release()  //can be called even if item.Value() returned nil
 ```
 
-In practive, `Release` wouldn't be called until later, at some other place in your code.
+In practice, `Release` wouldn't be called until later, at some other place in your code.
 
 There's a couple reason to use the tracking mode if other parts of your code also hold references to objects. First, if you're already going to hold a reference to these objects, there's really no reason not to have them in the cache - the memory is used up anyways.