It is instructive to consider the effects of corrupting the database in each case. Suppose we wrote each clause on a punched card, and shuffled the cards, then this would only matter to (A).
If we were to delete a clause from each set, this would not matter much to (A), (C) or (D), but it would break (B) into two chains, whose relative position would then be unknown. This shows that the effect of errors on information depends very much on how the information is represented. This idea is often used in psychological experiments. Given some kind of error (e.g. a brain lesion), you can try to deduce from the behavioural changes it causes, how information is stored in the brain.
Now let's consider databases - not the Prolog database, but commercial conmputer databases. Let's suppose that we have a file of employees and their salaries: for example,
Chavasse 34000 Linton 12500
Hannington 20500 Matthews 9842
Latner 13000 Slinger 30000
If we want to search, as we might in a real database, for all
employees with a salary of less than 12000, then as the file stands we
would
have to examine every entry. However, if we were to sort it in order of
salary, we could stop as soon as we found one that was 12000 or over.
Alternatively, we could build some kind of index giving the position of
the first person earning 9000, 10000, 11000, and so on.
You can see how the time taken to find the information depends on the way in which we order and index the files. This is true not only of files, but of any data structure, and the study of such matters is a large and important part of computing.
Conversely, given some system, one can try to discover the internal representation by seeing how its response time varies with certain inputs - Supplement 1 had an example of this, under inference. As with errors, psychologists use response-time experiments to try and infer things about how our brains represent information.