If literacy is not embedded in real life, in genuine meaning, it becomes dull, distant, difficult and daunting. Under such a circumstance we all know what happens to pleasure, excitement, motivation and confidence.

The reading of any text at all, in real life, involves us very considerably. It is never passive, nor is it meaningless. Reading is an active process. In most circumstances we bring at least as much to the text as we will take from it. Smith (2004) usefully divides ‘information’ within and around reading into visual information, which the text provides, and non-visual information, which we provide. He points out the difference thus: Visual information is what you suddenly don’t have when the lights go out, whereas even in the darkness you still have a lot of relevant information in your head - including your prior knowledge, your aptitudes, attitudes and affective states and your knowledge of the text already read. This is the non-visual information you are bringing to the reading. It is prior to the text and essential to making sense of it. Visual information - the letters, letter patterns and words on the page - may seem like the senior partners in the practice of reading, but the plethora of expectations and prior information that a reader brings to reading may be at least as important to its success.

Reading, then, depends on two things. It depends on the text - its content, level, interest, clarity and so on - but it also depends on the expertise and expectations of the reader. For example, some readers keep themselves so completely informed about the doings of celebrities that the details of the wording in that copy of Hello magazine barely matter. There is relatively little text in that journal, and it is constrained to a few rather particular observations expressed in rather particular language, so even I would not need to pay very close attention to it in order to get the gist. We all know roughly how the Hello genre goes and what to expect from it. If we had picked up a copy of a medical textbook on the innervation and anaesthesia of teeth, inadvertently left on the waiting room table, though, most of us would have found the text much more difficult and the words themselves much more important to our reading. We are probably not used to this genre. It is unfamiliar and less ‘natural’ to us precisely because we bring less understanding to it (understanding of the conventions of the writing genre as well as of the knowledge and assumptions underlying the subject matter). We would have to read this book in a very different way than the way we read Hello to reach a similar degree of understanding. Reading, in other words, is always a trade-off between what is actually on the page and what we bring with us to the reading.

Interactive-compensatory reading: Smith initially overestimated the degree to which readers’ prior knowledge, and the context within the reading itself, enable that reading. (He is an enthusiast.) It seems that fluent readers reading texts which are not too difficult, or which do not contain too many new facts, names or ideas, read so fast that they are probably barely using contextual support at all. There is so little time. They are probably, in these circumstances, reading by fully automatic and direct word recognition - reading almost solely bottom-up in fact. (Even so, the fact that the text is easy enough for them to be able to do this implies that it is supported by a great deal of prior understanding and accurate expectation.) A less fluent reader, or perhaps the fluent reader who has decided to try reading that book about the innervation of her teeth, must attend more carefully to context clues. She may only reach full understanding of many of the words in such unfamiliar text if she uses support from the context in which they lie. This trade-off between ‘context effects’ (readers’ grasp of context and their use of it to support their reading) and direct and automatic word recognition is more elegantly named the ‘interactive-compensatory model’ (Rayner & Pollatsek 1989, Reimer 2006, Stanovich 2000, Underwood & Batt 1996).

According to the interactive-compensatory model, then, word recognition is often fully automatic, assuming simple text and a confidently fluent reader. Words are recognised directly, without reference to context. The assembly of meaning takes place beyond basic word recognition and does not influence it. However, this applies only when the reading really is fluent and there are no surprises. Where a text is difficult, or perhaps written badly, then even a fluent reader may be driven to using what they already know, and what they have already read, to support, perhaps even to provide, word recognition. (Perhaps, for example, the constituent orthographic elements from which it is intended that textual comprehension should emerge are configured in so intricately convoluted a manner, the ideation enunciated in a vocabulary so inordinately and unnecessarily labyrinthine, that felicity, lucidity, coherence and comprehensibility are all but entirely obliterated and the salience of contextual environment exponentially increases.)