diff --git a/Datasheet.md b/Datasheet.md
index 18c9f0b..db53430 100644
--- a/Datasheet.md
+++ b/Datasheet.md
@@ -19,7 +19,7 @@
 > Q. Do tables express heterogeneous data, or must data be homogenized?
 
 
-> Q. Do tables capture missing data and, if so, how?
+> Q. Do tables capture missing data and, if so, how? Do missing values affect the output constraints of any operations, for example `groupBy`?
 
 
 > Q. Are mutable tables supported? Are there any limitations?
diff --git a/README.md b/README.md
index b36989e..7083ada 100644
--- a/README.md
+++ b/README.md
@@ -2,7 +2,7 @@
 
 The Brown Benchmark for Table Types
 
-Version 1.1
+Version 1.2
 
 ## Context
 
diff --git a/TableAPI.md b/TableAPI.md
index 4012d47..b28bb85 100644
--- a/TableAPI.md
+++ b/TableAPI.md
@@ -1032,7 +1032,7 @@ Sorts the rows of a `Table` in ascending order by using a sequence of specified
 - `header(t2)` is equal to `["value", "count"]`
 - `schema(t2)["value"]` is equal to `schema(t1)[c]`
 - `schema(t2)["count"]` is equal to `Number`
-- `nrows(t2)` is equal to `length(removeDuplicates(getColumn(t1, c)))`
+- `nrows(t2)` is equal to `length(removeDuplicates(getColumn(t1, c)))` Note that if there are missing values in the input, this constraint requires one row for missing values in the output.
 
 #### Description
 
@@ -1150,6 +1150,8 @@ Partitions rows into groups and summarize each group with the functions in `agg`
 - `schema(t2)` is equal to `schema(r3)`
 - `nrows(t2)` is equal to `length(removeDuplicates(ks))`, where `ks` is the results of applying `key` to each row of `t1`. `ks` can be defined with `select` and `getColumn`.
 
+Note that these constraints assume a first class representation for missing values.
+
 #### Description
 
 Groups the rows of a table according to a specified key selector function and creates a result value from each group and its key. The rows of each group are projected by using a specified function.