Although many different theories exist, the following two major theories of word recognition are of particular interest:
- the theory of phonic interpretation (letter-by-letter or bottom-up processing);
- the theory of top-down processing (where words are recognized as wholes).
Both offer extensive evidence to support their assumptions, but it seems that humans make use of both strategies depending on their reading proficiency, type of text and the reader's background knowledge of the field (this may, perhaps, explain why reading on familiar topics with familiar terminology is so much faster), language, reading situation, etc.
For example, reading a text with few low-frequency words is likely to involve high percentage of words recognized as wholes and guessing based on expectations5; bottom-up interpretation may be used to recognize words whose exact graphological representations are not readily available in the person's lexicon. It must be noted though, that the view presented here is a simplified one: the amount of sub-theories and (controversial) findings described in literature makes it impossible to discuss or even list them here.
Footnotes:
5Guessing may account for the so-called “proof-reader’s problem” (when a person does not notice errors) and also reading errors — a person perceives what he/she expects (or what he/she has recently thought about, i.e. what has recently been activated) rather than what is actually written. From our own reading experience we may also see that, having reached the end of a page, we can quite often predict what the next word will be on the next page (sense relations (e.g., collocations), priming, sentence structure and experience seem to be the possible sources of such predictions).