Scholes (1998 p. 187) says

‘There are no truly phonetic writing systems, nor should there be. The purpose of an orthography is different from that of a phonetic system. Orthographies are meant to convey meaning and they succeed in doing so by ignoring variations of dialect and idiosyncratic speech.’

Smith (1985) says that ‘We learn to read by reading.’ but how might we do that? Researchers, when not participating in the reading wars, appear to have reached a sort of consensus that reading is, right from the start, primarily a visual and orthographically mediated activity. Could it be otherwise? ‘… this is not a viable possibility for a written language such as English in which the relationships between letter groups and pronunciation is inconsistent and subject to higher-level influences.’ (Seymour, in Beech and Colley 1987 p. 32). And ‘We conclude that even beginning readers can, and often do, read whole words without analysing the grapheme-phoneme relations in these words … young children find it easy to take the logographic approach.’ (Goswami and Bryant 1990 p. 34) and ‘… there is very little direct evidence that children who are learning to read rely on letter-sound relationships … there is a great deal of evidence that … they adopt a global strategy, which means that they either recognise the word as a pattern or remember it as a sequence of letters.’ (ibid. p. 46)

Reading is not primarily a phonological activity. Visual decoding and speed are of the essence and should be specifically encouraged. So should the use of context. (All reading should be specifically aimed at meaning; I hope I can add ‘of course’.) Stanovich, West and Freeman (1981) have shown that fluent readers reading simple texts are probably going too fast to make much use of context when decoding text and are, in fact, reading pretty well exclusively bottom-up - are accessing meaning solely from the text on the page. An ‘interactive compensatory’ but primarily visually mediated model probably applies. In the early stages of reading, or when reading difficult text, meaning may be accessed using a mix of text and context, the precise proportions of mix being decided on a moment-to-moment basis by the mind. Bottom-up reading certainly frees top-down processing capacity for comprehension and assimilation at speed. It amounts, perhaps, to the difference between word recognition and reading. Reading is reaching meaning, a habit which tuition should seek to establish early. (Harrison and Coles 1992, Oakhill, in Beard 1993, Oakhill and Yuill in Funnel and Stuart 1995)

Some less technical evidence already touched upon (and see notes to this chapter) that we read primarily visually is our ability to produce the correct meaning for words which are spelled differently yet pronounced the same (Heterographic homophones); for example were/whirr; we’re/weir; see/sea; made/maid; pray/prey; lee/lea; road/rode; no/know; need/knead, bow/bough, ruff/rough, missed/mist, way/weigh, saw/sore and on and on. If we were reading to meaning purely, or even mainly, by sound, these heterographic homophones would be problematic. We would experience a falter when reading them. We would be able to find the sound of the words, but how would we know which meaning to allocate to them? Between ‘they’re’, ‘there’ and ‘their’ is but a visual difference, after all. No; when we read one of these words to meaning correctly we must have accessed the semantic system, initially, without recourse to the sound of the word at all. Indeed, accessing the sound would confuse, exactly as RED written in green did when we looked at the Stroop effect.

Similarly, we can ‘disambiguate’ words which are spelled the same but pronounced differently, (heterophonic homographs), for example wind/wind; present/present; live/live; read/read; contract/contract; deliberate/deliberate; rebel/rebel; content/content row/row and on and on. Only once we had the meaning could we find the pronunciation. Clearly we are only locating the phonetic version of the words we read after the reading has already happened. This is almost certainly what is happening even if we hear an ‘inner voice’ while reading, it is simply the onward processing of information which is already safely in the semantic system anyway; simply inevitable spreading activation reverberating round the systems of language management. It is almost certainly nothing more than a ‘useful second pass’ which may enable us either to double check our reading or to hold on to what we have just read for long enough to grasp the meaning of longwinded sentences which take a relatively long time to read. It is not, most of the time, the reading itself.