Top-down processing envisages a far more proactive mind. Your top-down mind is seen as actively formulating hypotheses all the time as to what might be going on. These hypotheses are, of course, founded upon experience, recent or remote. With its hypotheses in hand, the mind sets off into the world searching for related evidence. Your mind forms the relatively few questions necessary to interrogate the evidence to decide whether or not what it suspects is taking place is what really is taking place. Having actively sought the answers to its questions and found (often) its hypotheses confirmed, your mind may then begin to make decisions as to how to take things usefully or entertainingly forward and start seeking the answers to the new questions which will arise as a result. Because your mind, in this mode, is looking for relevant evidence, it discards, or ignores most of the incalculable amounts of material with which it is continually bombarded. It is selective in pursuit of its hypotheses. (If it had no hypotheses it could not sort information by relevance – information can only be relevant to something.)

Your mind is probably a curious, selective and proactive hypothesis-generating device continually searching for that minimum of data which will enable it to confirm or refute its hypotheses. Your mind probably interrogates the world unceasingly; however, it probably does not examine all of everything presented to it, but ‘skims’ incoming data intelligently.

We can easily see how top-down processing enables us to make good decisions quickly, based on much less, or less perfect, information than if we had to rely on pure, data-driven, bottom-up processing. An example: my son is upstairs in his room playing the ‘Sergeant Pepper’ album. There are two doors between us, muffling the sound of the music such that I can barely hear it. It is, though, an album I know backwards. I feel I can hear the words and music clearly. This is nonsense, though. The truth is that I know it so well that I need only the tiniest and most imperfect hint of relevant musical information to make me think I can hear it all in detail. My mind suggests, from its extensive library of Beatles’ music, what I am almost able to hear; this and the tiny amount of real evidence I actually have is enough for me to experience it almost fully. If my son were to play an album I know less intimately I would experience nothing beyond very distant sounds.

A silly oversimplification as another example:

When we first arrive at Joe and Sara’s party, bottom-up processing would have it that our minds are completely taken up deciphering the immediate and are free from any preconceived ideas as to what might be taking place behind their front door. When we get inside we look about us, assembling information, and finally conclude that there are a lot of folk dressed up to the festive nines, clutching drinks and behaving tremendously brightly. ‘Aha!’ we say ‘a party!’ (Had the house contained sheep we would presumably not have been much surprised as we had no pre-formed views on this or any other matter.)

Top-down processing theory, on the other hand, claims that as we come through the front door we expect to see people roaring at each other across many decibels of heavy rock. We gather just enough evidence to confirm this hypothesis (which, given the size and number of Joe and Sara’s loudspeakers, doesn’t take long) and can immediately begin searching for more data to take us forward, still within our hypothesis, for example for either Joe or Sara (we have to give our bottle of wine to someone).

Boringly enough, the truth probably lies somewhere in between and, as usual, it probably all depends... We probably use a variable mix of top-down and bottom-up processing. Some situations are highly predictable (Sergeant Pepper, for example, or getting up in the morning). In predictable, well-known situations like these a skeleton minimum of actual data will suffice such that we get the experience experienced or the job done. Other, more novel, situations will demand that better attention be paid to incoming data, to new information. Had Joe and Sara’s house actually been full of sheep, for example, we would have had to fundamentally revise our hypothesis and search for a new one, and we would have had to deploy a much more thorough examination of the evidence.

It is often a question of novelty versus familiarity. Reading a romance or whodunit, for example, demands relatively little attention. There are relatively few new ideas or details. You’ve seen (or read) it all before. You can read it very fast and very well with little application and intermittent attention. This is why such reading suits airport lounges and railway carriages. Contrast reading a paper on, say, finding engineering solutions to the stresses on high bridges from intermittent and variable wind forces. This will need much more careful reading and much more bottom-up data collection. Skimming and inattention will only result in failure. We will need every fact, every turn of the argument, to understand. We cannot use knowledge already in our heads to guide us, as there is very little. (If you are a civil engineer, I abjectly apologise.)

We will return to this aspect of our minds but now we need to consider quite another: cascading analysis. How do we decipher the world or, in the context of a book on literacy, the written word?