diff --git a/Initialization.md b/Initialization.md
new file mode 100644
index 0000000..81756b3
--- /dev/null
+++ b/Initialization.md
@@ -0,0 +1,56 @@
+`set` phrases, data sections, and type and function return default values all
+support value initialization. There are two types of initialization in ARF:
+simple and complex.
+
+Simple initialization is just a single argument that is sufficient for setting
+the value of a simple thing such as an Int. However, objects and arrays require
+more than that. This is where complex initialization comes in.
+
+An indented block under a...
+
+- Data section
+- Type section
+- Enum section member
+- Objt section member
+- Set phrase
+- Function ouput declaration
+
+... is to be treated as a complex initialization. There are two types of complex
+initialization: object and array. Array initialization is simple, it is just a
+list of values of the same type. The following array initializations are
+syntactically and semantically valid:
+
+```
+set x:{Int 4}
+	3849 92348 93 -21
+	
+set y:{Int 5}
+	23 2
+	39 695 1
+
+set z:{Int 5}
+	23
+	2
+	39
+	695
+	1
+```
+
+Object initializations allow setting the values for an entire object, including
+its constituent parts:
+
+```
+set x:SomeObject
+	.someMember 5
+	.memberWithMembers
+		.bird0
+		.bird1
+	.arrayMember
+		9853 943850 2093 399
+		394 93
+```
+
+The parser is able to differentiate between the two because in an object
+initialization block, every member is preceded by a dot. Object initialization
+members may nest - they may have array initializations in them, or even other
+object initializations.
\ No newline at end of file