How many things can we mentally juggle?
In many cases the answer is “Miller’s law”1: seven plus or minus two. This number comes from experiments on the number of “chunks” of information a person can keep in memory. Some experiments suggest that human working memory should be even smaller (perhaps only one to four).
In other cases “Dunbar’s number”2 (about 150) fits better. It’s the number of stable human relationships a person can maintain. This number is extrapolated from our neocortex size, together with observations of other primates. There is some experimental evidence for Dunbar’s number (for example, from analysis of Twitter messages), and other researchers suggest the number may be twice as high.
Maybe in the kalahari, those numbers are enough. In the information age though, even Dunbar’s number seems small.
Perhaps working memory plays a role in selective attention. Is it a coincidence that this video has eight moving objects in addition to the “invisible gorilla”?
So why is any of this important? A well designed user interface or business process should not tax our abilities too much. That doesn’t necessarily mean a GUI, document, process or database can’t involve more than a few hundred items. It does mean that working with that that data shouldn’t require keeping too much in your head at one time.
For a long time I was puzzled by the lack of popularity of the “naked object”3,4 paradigm. The idea is so elegant, the benefits so great. Too bad users don’t like this kind of interface. Maybe it’s because navigation through the domain objects requires some sort of mental map of the relationships between different kinds of information. The vertices in that map could easily overwhelm Miller’s law, and it’s not rare for them to exceed Dunbar’s number as well.
To fix the problem of user acceptance, without discarding the advantages of naked objects, we developed tools for rapidly building applications that display information from databases in the form of documents, so that unskilled users can see, update, and otherwise manipulate richly connected data with no navigation other than scrolling.
Griesser, Art (1997). “A generic editor”. OOPSLA ‘97 Addendum to the 1997 ACM SIGPLAN conference on Object-oriented programming, systems, languages, and applications 50-55. doi:10.1145/274567.274577 ↩