What then are we to make of home, but some and come; tone and tonic ('phone and phonic, for that matter!); wine vinegar; definitely finite; stove and glove; driver and river; tidy and city; likened and literacy; title and titular; pane and panel; potable potato potage; patron and patrol; mate, material, matriarch and maternal; idiomatic ideas and isolated islands; precious and previous; medicate and mediate; definitive and devious; secretary and secretory; calamity, celibacy and celebration; critical criterion; lemon, seven, and evening; metre and meter; idiotic and hideous idolatry; and on and on? (Consider the word ‘inevitability’ for a moment!) To reach pronunciation in the face of such irregularly variable phonic irregularity, we need, as Smith says, to ‘read from right to left’. By this he means that we often have to have read the whole word to meaning before being able to utter it correctly. We must already have reached the ‘identity’ of the word before we could pronounce it. In other words we must already have read it. We must have read it, visually, all the way to its meaning before we were able to access its proper phonemic representation, its pronunciation.

Berdiansky et al (1969) are cited by Frank Smith (2004 pp.143-146). They made a study of the vocabulary correctly spelled by a group of children all around the age of ten. Berdiansky and his collaborators attempted to formulate a set of ‘phonic rules’ which might be being used by the children to achieve the spellings. They found that the children had a correct spelling vocabulary of 6,062 words of two syllables or less, and some 3,000 words of three syllables or more. They found enormous difficulties in envisaging a feasible set of sound-based rules for spelling the 6,062 short words, and those rules they did come up with were laborious, complicated and implausible. They were obliged to give up altogether on the 3,000 or so words of three or more syllables as they added so horrendously to the complications. (The children, of course, were able to manage these words perfectly satisfactorily.)

In the end, they found an astounding difficulty in contriving a spelling system which was based on sound-letter correspondences if it were to manage even simple spellings of words of only one or two syllables. They found there were 69 ‘grapheme correspondences’ involved in over 200 ‘spelling-sound correspondences’. They found 166 rules, but also 45 major exceptions involving some 10% of words - mostly the commonest words, of course. No fewer than 73 rules were found necessary to cope with the behaviour of vowels alone. (Remember we are only talking about very simple words of two syllables or less!) They could not identify reliable rules enabling any decision to be made as to when a word is to be spelled by ‘phonic rules’ or when it is an exception which must be managed as a separate item altogether. Frank Smith's conclusion, naturally enough, after surveying this wreckage, was that the phonic rules paradigm had ‘limited effectiveness at great cost to memory’. (I discuss rules per se sceptically, in the notes to chapter five.)

Finding examples (and not exclusively in the English language) to demonstrate the unreliability of sound as a guide to the appearance of a word, or vice versa, is a little like shooting the proverbial fish in the proverbial barrel. Spotting these ‘absurdities’ is an old game we all enjoy and could play for hours without repetition. Actually, though, these examples are only felt to be absurd because we insist on considering the English spelling system as one which is solely, or somehow ought to be solely, phonically organised. When we find a spelling which is not clearly transparent in its letter/sound correspondences we upbraid it. Our spelling system, however, does a lot more than merely indicate the sounds of the language and it is perverse of us to reproach it for not doing so exactly at all times. We should rather be grateful for all the other information it gives us, making reading English so much easier (yes, easier!) and so much more interesting, at least for the fluent reader. There is logic, linkage, meaning and history in there, and it helps a good deal (And see chapter five defending English spelling from a historical perspective). Our spelling system is a visual signalling system signalling to a visual recognition system and, where necessary, it sacrifices mere phonetic regularity in favour of meaning or relationship. Would you really find English easier to read if it were spelt phonetically? Don’t you like the connection between would, could and should to be clear? And the difference between would and wood to be explicit - leap off the page at you? Would spelling workt, calld and landid help your search for meaning? Would you instantly know they were all past tense verb forms? What if you read landid and candid? Fancid and rancid? Mist and mist? Drest and jest. Past and past. Would you know that no and nollij were related? Sine and signal? How would you distinguish torque and talk? Week and weak? How would you spell work and walk? I could go on…