db.test.insert(    {       "_id" : ObjectId("5ba4d6ca79f29bad7b3fa913"),       "firstArray": [          {          },          {             "secondArray": ["a6", "a7", "a8", "a9", "a10"]          },          {             "secondArray": ["a11", "a12", "a13", "a14", "a15"]          },       ]     } )   Then run the following aggregation: db.test.aggregate([    {       "$match": {          "_id": ObjectId("5ba4d6ca79f29bad7b3fa913") {{      }}}    },    {       "$project": {          "array": {             "$arrayElemAt": ["$firstArray.secondArray", 0]           } {{      }}}     } ]) Returns: {{{}}    "array":  ["a6", "a7", "a8", "a9", "a10"] } This is confusing. I would expect it to return: {{{}}    "array": null } or whatever mongo uses to indicate the aggregate returns nothing.   The reason I expect it to return nothing, is because the first element of firstArray }}doesn't contain a property {{secondArray , so the index of 0 should first be selecting the first element of firstArray (i.e.,  , ) and then return  secondArray which doesn't exit.   What is actually happening is mongo is getting all the array elements that contain  secondArray  and then applying the index to that array. Hence why  }}{{"$arrayElemAt": ["$firstArray.secondArray", 2]  returns nothing, and not the last element in firstArray.   Am I missing something? The documentation doesn't make it clear that this is the expected behaviour. Either the documentation should be updated to make it clear what the index supplied to  $arrayElemAt actually applies to or this bug should be fixed. Many thanks!