The goal of the current project is to determine what environments within the word affect the spelling of the vowels of English RP.
The vocabulary base over which the statistics were computed is intended to be monomorphemic words that would generally be familiar to secondary education level students in England. The current word list is approximate but probably close enough for current purposes. A first pass considered all the words that have a frequency greater than 0 in Unisyn, but which do not have a per-grade frequency breakdown in Zeno et al. Marie looked at these words (exhaustively for those with frequency less than 21) and rejected those that she did not judge to be familiar to students. Next, the lemma-level morphological and phonological data were read from Celex, ed. 2. If the Cobuild frequency recorded in Celex was less than 1, the word was rejected unless it was among those Marie had studied exhaustively and did not reject. Otherwise, the word was accepted if its morphological status code was M for monomorphemic, and none of the following was true: its spelling contained an uppercase letter or any nonletter; it contained only 1 letter and was not "a"; its pronunciation contained foreign phonemes; it had a language/dialect code other than B for British; it was among the words already explicitly rejected by Marie. Marie inspected this output as well, and tagged further words that she judged to be unfamiliar at the secondary school level.
The pronunciations were taken from Celex, ed. 2. These are fairly formal, standard, general Received Pronunciation forms. The spellings are also standard British English. Spellings and pronunciations were aligned automatically. Alignment followed the general principles that
The general philosophy for exploring contexts was that
Auxiliary files:
Exclusions 1: a list of words with positive frequency in Unisyn but no per-grade frequencies in Zeno. Frequencies and pronunciation, sorted by descending frequency. Those rejected by Marie are followed by an " x". A long file.
Exclusions 2: a list of words whose Cobuild frequency and other indicators in Celex indicate would be a good candidate for the word list, but which were rejected by Marie.
extract.pl: a program that extracted candidate words from Celex, excluded above-listed exclusions, then aligned spellings with pronunciations.
reassignE.pl: a program that tagged (with the mark _e) vowels whose spelling relies in part on silent final e.
monomorphs-listE: the word list generated by the above programs. Each word is on a separate line, and contains a spelling, a pronunciation, and finally an alignment between letters and phonemes. 6,129 words.
analysis5, an automatic analysis of the influence of environmental contexts on vowel spellings. Not particularly perspicuous in itself.
Statistics for each vowel. These reflect the application of some human judgement as to which environments are most productive and salient. For each vowel the spellings are given for the typical, default, environment, as well as for each environment that significantly alters the spelling of the vowel. Tables include all counts, consistency measures, and a complete list of words fitting the patterns.
The designation <_e> attached to a vowel letter, e.g., <i_e>, means that the spelling requires a final silent <e> at the end of the word. Strictly speaking, inflectional endings may follow the final <e>, and the letter may show up in compounds; but this word list is monomorphemic. It may also be noted that the final <e> may also be required for other reasons; in particular, certain consonants need to be followed by <e> at the end of the word. There can occasionally be uncertainties as to whether a word has a <e> because of the vowel, because of the consonant, or for both reasons simultaneously.
The following rules apply to the use of silent final <e>:
Another feature that indirectly indicates vowel quality is the doubling of the consonant that follows the vowel. The general rules for this are:
Because the rules for silent <e> and for consonant doubling apply broadly across 6 vowel letters, there may be some usefulness in seeing whether participants apply the rules correctly in the abstract. At a graphotactic level, do they recognize that consonant doubling and silent <e> are applicable only after one-letter vowel spellings followed by a single consonant? That consonant doubling applies only before another vowel (with the exception of <f>, etc.), and that silent <e> applies only in final closed syllables? Do they know which consonants can and cannot double, and which ones have special forms (like <ck>)?
At the level of individual vowels, the use of doubling and silent <e> can be crucial for indicating what sound a vowel letter is intended to spell. However, the stimuli must be crafted to ensure that a silent <e> would not be inserted for some extraneous reason, such as to support particular pronunciations of consonants. For example, final /s/, /z/, /v/, /ð/, /dʒ/, /ʒ/ often require <e>; /θ/, /ʃ/, /tʃ/, /ŋ/, to some extent /ɡ/, and most consonant clusters cannot follow historically long vowels in final position and so are not conducive to such tests (e.g., a syllable like */paɪʃ/ sounds un-English, so there may be little point in seeing whether a person attempts to put a silent <e> at the end of the word). Further, one should take care that /s/, /z/, /t/, and /d/ can be taken as inflectional endings at the end of the word, which would exceptionally be spelt after silent <e>.
<i> normally;
<i_e> before a single final consonant spelt with one letter or <th>;
<y> word-finally, normally;
<ie> word-finally, in a lexical word whose spelling would otherwise have fewer than 3 letters;
<y> after word-initial /h/ in polysyllables.
Statistics. Note that the diphthong /aɪ/ is here treated as a unit, under the assumption that the speller will recognize it as a phoneme, and not as a sequence of, perhaps, /æ/ plus /ɪ/. Note further that the triphthong /aɪə/ is treated as a separate phoneme, below.
<i> is the default spelling in the sense that it applies in a variety of miscellaneous contexts, though it is not actually the most common spelling of /aɪ/ in monomorphs. Spellers may be expected potentially to pick up on certain common contexts, especially those where the vowel /aɪ/ is followed by a single consonant and then a vowel (open syllables, e.g., bison); a consonant and then <le>, spelling /l̩/ (bible); and <ld> or <nd> in monosyllables (wild, bind). But these patterns are all redundant given that <i> is the default spelling. By far the most serious competitor to <i> as a regular spelling of /aɪ/ is <y>, at 1/8 of the frequency. In practice <y> is most common in Greek words, but such a pattern does not really show up in this dataset.
<i_e> is actually the most common spelling of /aɪ/ in monomorphs, because this environment is so frequent. Unlike some other vowels, silent <e> can be used after <th> (blithe, tithe), and the silent <e> can be shared with <le> (bridle).
The minority spelling <y_e> is analogous: in the same environment where <i_e> is required, <y> is not a possible spelling: <y_e> must be used instead (rhyme, type).
Note the spellings ending in <s> (<ais>, <is>, aisle, isle, lisle) occur only before <le>=/l/; but normal spellings are far in the majority in this environment (15 monosyllables end in /aɪl/, including bile, file, while, etc.).
The spellings ending in <gh> (<igh>, <eigh>) occur nonfinally only in monosyllables before /t/. If this context is used (N=29), the spelling <igh> becomes as frequent as the regular <i_e>. It does not have enough of a plurality to justify setting it up as a special environment; words like sight number 15, to 13 words like site.
The status of words like title is ambiguous. Here we consider the <le> to be the spelling of a syllabic, analogous to a vowel, and therefore not to be part of an <i_e> spelling. Either analysis leads to the same considerations, so the distinction is not important. Similarly fibre, where <re> spells a vowel, /ə/; and aisle, isle, lisle, where <le> is considered a special spelling of /l/ that is used after (silent) consonant letters.
This rule for final <y> is a very general graphotactic rule whereby <y> is substituted for <i> at the end of words. The main exception is that <i> is retained in exact borrowings from foreign languages; here we see a minor pattern in that Greek letter names use <i> for final /aɪ/ (chi, phi, pi, (xi, psi)).
English has a general graphotactic rule by which full, lexical words must be spelt with at least 3 letters. The regular solution is illustrated by lie, but spellings like high and lye are each about half as common. Collectively, there is strong evidence that a word like *ly would be illegal, except in grammatical words (by, my).
The fact that <y> is so common after word-initial /h/ in polysyllabic monosyllable stems from the fact that most of them are Greek in origin (hybrid, hyphen, etc.).
When the single letters <i> or <y> spell /aɪ/, they are treated as long vowel spellings for the purpose of consonant doubling (they reject it: *climmax, *briddle, *ickon) and licensing of silent <e> (bite, bridle, blithe).
The clearest contrast is the one involving final <_e>. Stimuli should end in /aɪ/ plus a single consonant phoneme other than /ʃ/, /tʃ/, /θ/, /ŋ/, /ʒ/, or of course /w/, /j/, /h/. These can be monosyllables or polysyllables. Scoring: if the subject spells the vowel simply with <i> or <y> (which should be the typical responses), then the consonant should be spelt with a single letter (no doubling), or <th>, and <_e> added. Fine points: there is some small support for spelling /aIm/ as <imb> (climb). Also, one could make a case for spelling /aɪf/ with <phe> and /aɪk/ with <iche> by analogy with other forms, but these should be very rare.
One could also give 2-syllable stimuli where the /aɪ/ is followed by a stop consonant plus syllabic /l̩/. Scoring: if the subject fails to add final <_e>, then the vowel is misspelt as well as the /l̩/.
Conversely, one could give monosyllabic stimuli with /aɪ/ followed by two consonants. The only very common cases in English involve /nd/ (mind) and /ld/ (wild), although we also find the isolated cases pint, whilst, (Christ). Probably other clusters should be avoided as unnatural. Scoring: subject should spell with <i> alone, no final <_e>.
For more typical stimuli, when /aɪ/ appears in a nonfinal syllable, don't allow that vowel to be followed by more than one consonant. Scoring: normally <i>, perhaps <y>, other spellings should be rare.
One could give stimuli with word-final /aɪ/. Best to make it monosyllabic or final-stressed, so as not to suggest the Latin inflexional -i. Ensure that the spelling of the rest of the word will require at least 2 letters. Scoring: subject should use <y> or perhaps <igh>, not <i>, <ie>, etc.
One could also give monosyllabic stimuli with word-final /aɪ/. Select onsets that would normally be spelt with one letter. Scoring: subject should use <ie>, <igh>, or <ye>, not e.g. <i>, <y>.
If testing the last rule, pick polysyllabic stimuli that begin with /haɪ/. Try to select words that sound like they could be Greek. Scoring: expect <hy->, although there is nothing technically wrong with <hi->.
<ia> normally;
<ire> at the end of words;
<i> before /r/.
Statistics. Note that the triphthong /aɪə/ is here treated as a unit, under the assumption that the speller will recognize it as a phoneme, and not as a sequence of, perhaps, /aɪ/ plus /ə/. This is not a foregone conclusion, because the spelling, in general, does treat the triphthong as a sequence; e.g., bias can be analysed as <i> spelling /aɪ/ and <a> spelling /ə/.
The normal case treats the triphthong as a combination of /aɪ/ and /ə/, using spellings typical of each of those two phonemes. In particular, only vowel letters are used.
At the end of words, the same class of spellings can be used; but in ordinary English words, an <r> is used. Graphotactically, the spellings behave as if the triphthong is a sequence of /aɪ/ followed by a consonant /r/ (fire). Consequently, the /aɪ/ is spelt as it would be before a single final consonant, with final silent e.
Before an actual /r/, the spellings of group 1 are possible, but more often the triphthong is spelt the same as /aɪ/ would be (pirate).
The phonology of this triphthong is complicated. Care should be taken to present it in environments that really are triphthongal for the participants. For example, it is not uncommon for speakers to have simple /aɪ/ before /r/ (e.g., pirate, siren) or /l/ (dial, vial), or for this triphthong to be pronounced the same as /ɑː/ (fire = far). Unless the participant pool have fairly homogeneous pronunciations, it may be safer to avoid this phoneme entirely.
That said, the clearest contextual sensitivity to check may be the spelling of /aɪə/ before a consonant, as in lion. Spellings with <r> should not be used in this environment: *liarn. Avoid the endings /z/ and /d/, because they can be taken as inflectional endings (cf. fires, fired).
<ou> normally;
<ow> before final /l/ or /n/;
<ow> finally
Statistics. Note that the diphthong /aʊ/ is here treated as a unit, under the assumption that the speller will recognize it as a phoneme, and not as a sequence of, perhaps, /æ/ plus /ʊ/. Note further that the triphthong /aʊə/ is treated as a separate phoneme, below.
This context is essentially that before any sound except final /l/ or /n/. Technically vowels should not be included here (cf. prowess) but there are not enough monomorphemic instances to infer that rule (most such vowels are pronounced /ə/ and therefore form the triphthong /aʊə/). Neither consonant doubling (*cloudd, *coutch) nor final silent e (*cloude) are permitted after <ou>, although the latter is permitted if present for other reasons (gouge, spouse). <ow> is used in a few words.
This is a very general pattern for phonemes spelt with a digraoh ending in <u> or <w>.
<ough> is used in a few words.
The most clear case is word-final position: /aʊ/ is virtually never spelt <ou> at the end of the word (only in the archaic grammatical word thou), and so sensitive spellers should not produce that spelling in that position. However, the other contexts are all fairly strong--only foul and noun have <ou> before /l/ or /n/--so these may also make acceptable tests of context sensitivity.
If using stimuli ending in /aʊl/, it may pay to make sure that the speakers do not rime fowl with towel, or at least make provisions for scoring accordingly.
<owe> normally;
<our> finally;
Statistics. Note that the triphthong /aʊə/ is here treated as a unit, under the assumption that the speller will recognize it as a phoneme, and not as a sequence of, perhaps, /aʊ/ plus /ə/. This is not a foregone conclusion, because the spelling, in general, does treat the triphthong as a sequence; e.g., towel can be analysed as <ow> spelling /aʊ/ and <e> spelling /ə/.
Statistics on this triphthong should be taken with a grain of salt. There would seem to be many words not included in this list: bower, cower, flower, glower, lower, power, shower, tower come readily to mind; for some reason Celex did not classify these as monomorphemic (mistaking the -er for a suffix?). When all is said and done, <ower> may actually outnumber <our> as the more typical word-final ending.
The phonology of this triphthong is complicated. Care should be taken to present it in environments that really are triphthongal for the participants. For example, it is not uncommon for speakers to have simple /aʊ/ before /l/ (towel, vowel), or for this triphthong to be pronounced the same as /ɑː/ (our = are). Unless the participant pool have fairly homogeneous pronunciations, it may be safer to avoid this phoneme entirely.
That said, the clearest contextual sensitivity to check may be the spelling of /aʊə/ before a consonant, as in towel. Spellings with <r> should not be used in this environment: *tourl, *towerl. Avoid the endings /z/ and /d/, because they can be taken as inflectional endings (cf. towels, towelled).
<ar> normally;
<a> before /f/, /s/ (not followed by /(ə)n/) , <th>, or nonfinal /n/ in the same syllable
<a> before /m/ in nonfinal syllables;
<a> in stressed non-initial syllables, other than as word-final phoneme.
Statistics. This vowel is pronounced more like [aː] than [ɑː] in most places in the north.
<ar> (car) occurs in a wide variety of contexts, including both open and free syllables, both word-internally and word-finally. It is, however, rare or nonexistent in doubly-closed syllables (*drarft).
A sizable number of words, mostly interjections or foreign words, have simple <a> (guava, pa). Several words have <al> before final <m> (palm).
Note that in most of these words spelt with <a> (no <r>), most Liverpudlians have the same phoneme as in cat (see /æ/ below). For these speakers, this case would completely disappear as a special environment; all words like raft, ask, bath, and dance, also aunt, would be assigned to the vowel /æ/; the remaining words (like half, farce, hearth, also rather) would be reassigned to the Normal environment (1 above), and the consistencies for this vowel would, in general, rise.
This may be an adventitious group that would disappear with more data; in any event some of the words are not exactly core English.
Presence of /ɑː/ in stressed non-initial syllables is a good marker of foreign provenance (avocado). Nothing rules out <ar> in the source languages, but logically the simple vowel is much more common.
If the participants do not reliably have case 2, i.e., a long vowel in words like raft, etc., the safest thing to test may be whether they double consonants or add final <e> for stimuli like park (vs. *parke, *parck, *parcke).
The last case, foreign-like words, is also a strong one. There is, however, a danger, in that applying this rule relies implicitly on identifying words as being of foreign origin. Normally one would wish to discourage such a perception in spelling experiments, so that the participants will more faithfully apply regular rules of English and not make up spellings that seem exotic.
If the speakers do have case 2, the most reliable experiment is to give stimuli that end in two consonants, like ask or branch. Words of that structure cannot be spelt with an <r>.
<o> normally;
<a> after /w/
The spelling of this vowel is very reliable. Even the spelling <ou> in cough, trough can be considered a form of <o>, in that there is a pattern whereby <gh> has to be preceded by a high vowel: <u> or <i>.
The spelling as <a> after /w/ is also very reliable. There is a minor pattern that excepts the vowel when it is followed by a velar consonant (golliwog, wog, (wok)), but there is not enough data to induce that rule here. People may be more likely to conclude that <a> is excluded in extremely informal or offensive words.
Whether spelt with <o> or <a>, the vowel is considered short, and so employs doubling (bobbin, cockle, doddle, coffee, collar, comma, bonnet, moppet, borrow, fossil, grotto, nozzle, scallop, squabble) and does not license silent <e>. (Exceptionally, one, typically pronounced /wɒn/ in Liverpool, is spelt as if containing a long vowel.)
The pattern of using <a> after /w/ is rather strong in the dialects that have this rule. Avoid following the vowel with /k/, /ɡ/, or /ŋ/, consonants that normally should require <o>. (On the other hand, that exception itself is so weak that it is probably not worthwhile testing it.)
<u> normally;
<o> before <v>.
The biggest competitor is <o>. This spelling is particularly common next to an <n> or an <m>--or especially between them, as in mongrel, monk, monkey, month, none--but then there are many counterexamples, such as mumble, which prevents establishing any clear environments for an <o> spelling (other than the below). <ou> is also fairly common (double).
Words that are spelled with <o> in this environment usually do not double following consonants (borough, money, stomach); an exception is worry. Sometimes final <e> is added: come, none, some, sometimes not: son, ton.
This pattern has only one exception, the informal and semijocular luv. In accordance with the general rules for <v>, that consonant is not doubled after the <o>, and final silent <e> is added when the <v> would otherwise be word final (shove).
When /ʌ/ is spelt <u>, it is clearly treated as a historically short vowel with respect to doubling (bubble, bucket, cuddle, duffle, gullet, mummy, tunnel, supple, surrey, tussle, utter, puzzle) and non-licensing of silent <e>.
When /ʌ/ is spelt <o>, it is treated like a short vowel in some words and as a long vowel in others. It takes silent <e> in some words (come, none, one, some) and rejects it in others (son, ton). The only example of doubling is worry. In the most regular environment, the case is completely indeterminate, since <v> cannot double and is always followed by <e>.
For many speakers, /ʌ/ as such will not be in their core accent; instead, almost all of these words will be pronounced with /ʊ/, the same vowel as in pull. It may be advisable when testing in northern areas to omit this vowel entirely, since, e.g., an RP /ʌ/ could be confused with a northern /a/ (cut coming across as cat and spelt accordingly). Alternatively, this vowel could be pronounced as /ʊ/ in the stimuli, if the participants are comfortable hearing such a pronunciation in formal settings. If this phoneme is presented, it may be advisable to check below for the treatment of /ʊ/; one may get some unexpected spellings in <oo>. It would probably be best to avoid stimuli ending in /k/.
Avoidance of final <e> can be tested straightforwardly, if one avoids stimuli where the consonant spelling itself calls for <e>. Also, if the participant spells the vowel itself with <o>, then arguably a final <e> is not wrong (come, etc.).
One can easily test doubling of final consonants ck, dge, ll, ff, ss at the end of monosyllables. There must not be any other consonant in the rime. The stimuli must have an onset, or else doubling would be expected by the three-letter rule. The vowel spelling <o> would not be expected in this context.
One would expect doubling of single consonants intervocalically, provided the vowel is spelt <u>. Do not use fricatives as the intervocalic letter, because their doubling patterns are complicated.
The pattern of using <o> before <v> is rather strong and worthy of being tested.
<a>
Statistics. Typically pronounced [a] in Liverpool.
The pattern is almost perfectly regular. The only exceptions are plaid, plait, salmon, and not everyone pronounces plait with this vowel.
The spelling <a> is treated as a short vowel, licensing doubling (abbot, bracket, haddock, raffle, rally, stammer, channel, apple, arrow, asset, attic, dazzle) and not licensing silent <e>.
Northerners in general pronounce this vowel lower than in RP, but that fact in itself shouldn't affect stimuli selection. A more significant dialectal difference is that most northerners will pronounce many of the words listed under RP /ɑː/ with this vowel instead: e.g. after, class, branch. From a structural point of view, that too should have little effect: most if not all of those words are spelt with <a> and so their spelling just reinforces the rule for spelling this phoneme.
One can easily test doubling of final consonants ck, dge, ff, ss at the end of monosyllables. In RP, stimuli with /æ/ before /f/ or /s/ would sound unusual, but not necessarily illegal; in northern speech they should be just fine. There must not be any other consonant in the rime. The stimuli must have an onset, or else doubling would be expected by the three-letter rule for many other letters as well.
<e> normally;
<ea> between /r/ and /d/.
Statistics. Note that this phoneme is here treated as different from the diphthong /ɛə/.
The only serious competitor to <e> is <ea>. There are hints of a pattern--e.g., <ea> is common before /d/ as in head--but there are in general at least as many counterexamples, such as bed, so the patterns cannot be elevated to a general rule (except as below).
E.g., bread. There are 6 cases of this pattern, and 3 counterexamples, making it a very weak pattern.
When this vowel is spelt <e>, it is treated as an historically short vowel, licensing doubling (pebble, reckon, meddle, effort, fellow, dilemma, fennel, zeppelin, berry, essay, fettle) and not licensing silent <e>.
The patterns for the use of <ea> are weak enough that one would not really count any use of <e> in a nonword wrong. But if one does wish to look for such a sensitivity, the best stimuli to pick would be like /ɡrɛd/, which is maximally similar to bread, dread, spread, thread, tread.
<are> normally;
<a> before /r/.
Statistics. Note that this diphthong is here treated as a phoneme, and not as a sequence of the phonemes /ɛ/ and /ə/. In practice, a speaker may well see it as a sequence, a view that is supported by the spelling in most words.
Note that there is a great deal of variation in how this sound is treated and phonemicized. For many speakers, the sound before /r/ will be perceived as a separate phoneme, perhaps /ɛ/ or /e/, with only a minimal or incidental trace of a schwa. For many Liverpudlians, /ɛə/ in general will be conflated with /ɜː/, so that fare and fur are homophonous. Under such a system, the rules for spelling the merged vowel will be different from those presented here.
<are> (care) is the most common treatment by a wide margin, but there are also many cases of <air> (fair) and even several <ear> (bear) and <ere>, although the grammatical words there and where are the only common cases of the last. It is very rare for /ɛə/ to be followed by a consonant (other than /r/, discussed below); when it is, <air> is used (cairn) and there is no doubling of consonants (*cairnn) or final <e> (*cairne) unless the consonant itself calls for it (scarce).
For people who have /ɛə/ before /r/, the diphthong is spelt as if representing /ɛɪ/.
This phoneme is treated graphotactically as if a sequence of a long vowel /ɛɪ/ followed by a consonant /r/; hence the final <e> (graphotactically a silent <e>) in many of the spellings (bare, there). Similarly, when the phoneme appears before an actual /r/, it is treated as if a long vowel; hence no doubling in area, bolero, etc.
There would seem to be few interesting contextual effects to study here, since the phoneme occurs almost entirely in one position: word final. The other position---before /r/---is rather unreliable across speakers.
The most interesting effects emerge if the speakers do not actually phonologize this vowel as intended. If one perceives it as /ɛ/ + /ə/, for example, then one can rephrase the question to more interesting contextual ones, such as, How is /ɛ/ spelt before /ə/?, which answer will be quite different from how /ɛ/ is spelt before consonants.
In a typical Scouse accent where /ɛə/ and /ɜː/ are merged (approximately [ɛː]), the combined phoneme has an interesting context-sensitive distribution. Word finally, as in care, it will normally be spelt <are> or <air>, as here; before a consonant in the same morpheme, as in nurse, it will normally be spelt <ur>, <er>, or <ir>. However, this may be difficult to test in a classroom setting, where students may not expect to hear these phonemes equated.
<a> normally;
<ai> in a doubly closed syllable;
<ai> before /l/ or /n/ in the same syllable;
<a_e> before final /st/;
<a_e> before a single final consonant spelt with one letter;
<ay> when stressed and word final;
<et> when unstressed and word final.
Statistics. Often pronounced more like /eɪ/ in Liverpool.
<a> is the default spelling in the sense that it applies in a variety of miscellaneous contexts, though it is not actually the most common spelling of /ɛɪ/ in monomorphs. Spellers may be expected potentially to pick up on certain common contexts, especially those where the vowel /ɛɪ/ is followed by a single consonant and then a vowel (open syllables, e.g., baby); a consonant and then <le>, spelling /l̩/ (cable); and <nge> (strange). But these patterns are all redundant given that <a> is the default spelling. Note there is no doubling of consonants after this sound. Other spellings are fairly unusual in the default environment. <ai> is regular before /θ/, but there are only two examples; in other cases the spelling is either foreign (liaison, raisin) or etymological (bailiff, daisy).
More specifically, <ai> is used when /ɛɪ/ is followed by two or more consonant letters (faint, traipse, faith, aitch), other than <st> (paste) and <ng> (strange); but there is perhaps not enough evidence to induce this complete pattern.
Both cases have many exceptions, especially the more general spelling <a_e> (below).
/ɛɪ/ is exceptional in permitting the use of silent final <e> across two consonants; but this applies unambiguously only to <st>. It is also possible to interpret the spellings in -ange /ɛɪndʒ/ (range) as sharing a final <e> between the consonant and the vowel, although we do not do so here.
This context includes <th> words (bathe, cf. lithe). In principle ache could be considered analogous to <th> and treated here, but instead we classify it as exceptional, falling under the default environment.
Much less common alternatives include <ay>, <ey>, and <eigh>.
These are almost entirely words of recent French origin (cabaret). Another important spelling is <e>, or rather <é> (café).
The single-letter spelling <a> is treated as an historically long vowel, and hence avoids doubling and licenses silent <e> (bacon, bake). Arguably the spelling <e> is also treated as a long vowel, and the final in crepe, suede may be taken as part of the vowel spelling, since otherwise the pronunciation would surely be /ɛ/.
Spellings ending in <gh> are legal only word-finally or before <t> (eight, straight, weigh).
One of the most distinctive environments is where the /ɛɪ/ is followed by two consonants within the syllable. Before /ndʒ/ or /st/ one expects <a> with silent <e>; before /nt/ one expects <ai>.
Technically <ai> and <ei> are legal, if not always common, in most environments. But word-finally only <ay> and <ey> should show up, possibly <eigh>, but not <ai> or <ei>.
<ur> normally;
<er> in any syllable after the first;
<er> after /v/;
<or> after initial /w/;
<ir> after /w/ in a cluster;
<ir> before /k/;
<er> before /m/;
<ir> before /t/ in monosyllables.
Statistics. As noted above, /ɜː/ totally merges with /ɛə/ (as approximately [ɛː]) in a typical Liverpool accent. Note that /ɛə/ occurs mostly at the end of a word, and /ɜː/ occurs mostly before consonants; so for these speakers, the combined rule set for the merged phoneme can easily be constructed.
When none of the following environments apply, <ur> is used over half of the time (burn), but <er>, <ir>, and <ear> are all reasonably common (herb, bird, earth).
These are mostly Latinate words (assert, reserve); the spelling <er> reflects the fact that <er> is much more common in Latin than <ur> or <ir>.
These words too are mostly Latinate (in fact aver, avert, overt overlap with the preceding rule).
The <o> substitutes for <u>, which is strongly avoided next to <w> (word, worm).
These are mostly "expressive" words, and may have influenced each other to some extent: quirk, squirm, squirt, swirl, twirl.
This rule may emerge as such primarily because a lot of words from the same root were classified as monomorphemic in Celex (circa, circle, circuit, circus, cirque).
Words with /ɜːm/ tend to be Latinate (determine, sermon, vermin).
The spelling of /ɜːt/ is a marginal pattern that affects only a handful of words.
This phoneme is mostly spelt as if a single, short vowel plus the consonant /r/. Thus none of the spellings end in <e>. There is, however, no opportunity for doubling, because the sound does not occur before a vowel morpheme internally.
As noted earlier, Liverpudlians tend to confound this vowel with /ɛə/, so some care should be taken in selecting or scoring stimuli. However, in nonfinal position, /ɛə/ is fairly rare, so one should expect little interference in testing most of the patterns given here.
If one is interested in differentiating performance in different environments, it should be noted that many of these environments overlap. Thus a stimulus like /əˈwɜːk/ is in a noninitial syllable (predicting <er>), follows /w/ (predicting <or>), and precedes /k/ (predicting <ir>). Cleanest results would be obtained by selecting stimuli that do not overlap categories. On the other hand, it could be interesting to quantify what happens when environments conflict, as in this example, or reinforce each other, as in, e.g., overt.
<a> normally;
<u> after initial /s/;
<o> after initial /k/;
<o> before /n/;
<o> in medial syllables before /l/;
<e> in medial syllables before /r/;
<e> before final /nt/;
<o> before final /k/, /p/, /t/, or /m/;
Unspelt in final /zəm/;
<u> before final <s>;
<e> before final /l/;
<ar> before final /d/;
<er> after /t/, when word-final or preceding /n/;
<er> word-finally after /d/;
<er> word-finally after /ð/;
<er> word-finally after /p/.
Statistics. Schwa appears only in unstressed syllables, but is very frequent there. Some care should be taken here, in that northern speech often has a full vowel where RP, and therefore this list, has a schwa.
<a> is by far the most common spelling for /ə/ in the general case (saliva), but 8 or 9 other spellings are also quite common, including <o> (daffodil), <er> (clever), <e> (shellac), <or> (forbid), <u> (triumph), <ar> (cellar), <re> (fibre), <ur> (pursue).
This case is due mostly to forms of the Latin prefix sub-. These forms have a tendency to be pronounced with /ʊ/ in the north. A few words in <a> (saloon) and <o> (solicit) are also found.
This case is due mostly to forms of the Latin prefix con- (commence, condense). These forms have a tendency to be pronounced with /ɒ/ in the north. In other forms, the spelling <a> is most common (casino).
This pattern emerges in part beceause it recapitulates the con- rules above (confide), in part because final /ən/ is disproportionately common (apron). This is only partly due to the Greek ending -on (rhododendron). The spellings <a> (pagan), <e> (token), and <er> (lantern) are also quite common. Note that no particular effort has been made here to distinguish the sequence /ən/ from syllabic /n̩/.
This pattern is due mostly to an Italian diminutive pattern (tremolo); otherwise <a> (buffalo), and <e> (procelain) are quite common as well. Note that it is not unusual for vowels to disappear entirely in such words.
This pattern depends to a large extent on the fact that Latin mostly had <e> in this position (camera, viscera). Otherwise <a> (minaret), and <o> (calorie) are quite common as well.
This depends on a common Latin pattern (aliment, silent), but there is also a prolific competing French pattern in -ant (tenant).
In general, <o> is used before word-final voiceless or nasal stops. This rule doesn't generalize to voiced stops, however; in particular, note the rule for final /d/ below. Other spellings are possible, but the default spelling <a> is surprisingly rare before voiceless stops.
An exception to the general rule for final /m/ (prism, spasm).
This pattern is due to the many Latin loans in -us. A more typical English spelling is with <a> (carcass, compass, terrace).
-el is a common native ending, but there are also many words in Latinate -al.
A few words use <o> or <a>.
Although <ter> (bitter) and <tern> (lantern) are common patterns, there are also many words in <ta> (data), <tan> (titan), <tre> (centre), <tar> (altar), <ton> (carton), and <tor> (motor).
<der> is most common in native words (murder), butthere are many Latin and Romance words in <da> (agenda, armada).
This is a particularly reliable pattern in that there are a dozen words meeting it, and no exceptions. On the other hand, most of the words are like either in being grammatical.
There are only a handful of exceptions, like pupa.
In general, schwa behaves more like a short vowel than a long one. It rarely is found with silent <e>, and occasionally is found before doubled consonants.
Schwa is interesting because there are many patterns in the statistics. But one drawback is that few of the patterns are really absolute. Most are like the case of medial /l/ and /r/: <o> is twice as likely as <e> before /l/, and <e> is twice as likely as <o> before /r/. The patterns are useful in a wide-ranging test of general sensitivity to context, but in an absolute sense, neither answer is truly wrong in either context. Among the few absolutes is that spellings with <r> are not found in doubly closed syllables; thus a spelling like *talernt would be impossible.
The other issue is that of speaker variation, include local accent. It is possible that entire patterns rest on pronunciations not common in the north. For example, the rule that schwa is used after initial /k/ depends on schwa pronunciations of the Latin prefix con-, which in fact is typically pronounced /ɒ/ in the north (consider) even in near-RP usage.
<o> normally;
<oa> in a closed syllable, except before /ld/ or /lt/;
<o_e> before a single final phoneme spelt with 1 letter;
<ow> at the end of a one-syllable word.
Statistics. Also pronounced [oʊ].
<o> is by far the most common spelling in open syllables, but a prominent subtype is words that end in /əʊld/ or /əʊlt/ (bold, bolt). This environment includes word-final position except in one-syllable words. Most such words are loanwords, spelt with <o> (buffalo), but there are many native words as well, which are spelt with <ow> (sparrow).
This environment is for words that end in /əʊ/ in a closed syllable, where the graphotactics do not permit silent <e>: before two consonants (boast, coax) or consonants spelt with two letters (oath, coach). A few words have <o> before <st> (host).
<o_e> is usually employed where the graphotactics permit it. However, many words use <oa> instead (boat), and some use simple <o>, especially before /l/ (roll, control).
One-syllable words emerge as a special case simply because that is the most reliable simple test for native words. As mentioned above, most native words ending in /əʊ/ spell it <ow>. A handful of words end in <o> - mostly grammatical words (fro, go, no, so). A half dozen words end in <oe>. These are mostly lexical words that would be below the 3-letter minimum if spelt with simple <o> (doe, foe, hoe, roe, toe).
The spelling <o> is treated as the spelling of a long vowel, licensing silent <e> (dole) and blocking consonant doubling (bogus). However, in many words before /l/ in the coda, the vowel is treated as if short (scroll, control, old).
Spellings ending in <w> appear only finally or before <l> or <n> (yellow, bowl, own, sew); cf. the graphotactics of <ow> spelling /aʊ/. However, owe adds <e> to meet the 3-letter minimum.
Among the clearest cases where a particular spelling is pretty much required are the first syllable in polsyllables, which requires <o> (soda), and monosyllables ending in /tʃ/ and /ks/, which require <oa>.
Words like roll provide an interesting test of whether people add final <e> based on sounds alone (in which case *rolle would be expected, as a tense vowel followed by a single consonant phoneme) or whether they take the spelling into account (omitting the <e> because the /l/ is graphically doubled.) Of course, spellers are also free to spell as role.
<ea> normally;
<e> in a syllable followed by a vowel;
<i> in medial syllables before /n/;
<e> unstressed before final /z/;
<e> in grammatical words;
<i_e> in the last syllable of polsysllables, before a consonant;
<ee> before /d/ or /p/ not followed by a vowel letter;
<ee> when final, in lexical words.
Statistics. For some speakers' pronunciations, under some interpretations, the vowel of words like beer would be interpreted as being a sequence of /iː/ + /ə/, and therefore should be included here. Instead, they are listed as a separate phoneme, /ɪə/. Further, the final vowel of words like curry probably belong here. Instead, they are classified under /ɪ/, reflecting a more traditional RP pronunciation.
Contrary to what one might expect, the usual spelling of this vowel is not with the single letter <e>, but with a digraph. <e> is not often used with silent <e> to spell /iː/, except fairly rarely in Latinate words (scheme). Nor can <e> be used alone before a consonant plus <le> (*betle; cf. able, bible, noble, scruple). <ea> is favoured (bean), but <ee> is also very common (beet), and <ie> is a respectable third (priest). A few French (chic, clique) spellings are also found.
In words borrowed in the traditional way from Latin, the spelling <e> is by far most common. The simplest way to identify words as Latin from their phonology is that they are polysyllabic, hence this rule (create, credo, sequel). However, it also picks up many recent loans from foreign languages, where the spelling <i> predominates. Speakers who can identify words as being Spanish (tortilla), Italian (graffiti), etc., may pick these out more reliably. Much less common spellings in Latinate words include <ie> (rabies), <ae> (daemon), and <oe> (amoeba).
This rule is simply a fairly reliable sieve for Romance words, which are usually spelt with foreign <i> (Portuguese albino, Italian casino, etc.).
(Putatively) singular words ending in unstressed /iːz/ are almost always Latin words spelt <(i)es> (rabies, herpes).
Grammatical words are the only important class of native English words where /iː/ is spelt with a single <e> (be, me, he, she, we).
Polysyllabic words with /iː/ before a consonant in the final syllable tend to be French, and follow French orthography (fatigue). Most of the rest are Latinate and are therefore spelt <e_e> (compete). This filter also picks up several nativ/ized words with the normal native spellings (between, appeal).
Before /d/ and /p/, <ee> is favoured (greed, needle, beep, steeple), although <ea> is quite common (bead, beadle, cheap).
At the very end of the word, <ee> (bee) is more common than <ea> (pea). A few foreign words in <i> are also found (ski, kiwi).
Even though <e> is not often used with silent <e>, it is definitely treated as a long vowel, as it never licenses consonant doubling. The spelling <i> may also be considered a long vowel.
It should be noted that a good many of the words classified, conservatively, under /ɪ/ arguably belong here instead. In particular, the traditional unstressed /ɪ/ in open syllables should be thought of as /iː/ (curry, radio). Such an approach changes the statistics somewhat. For example, this suggests that the category "Final" is insufficient. One would distinguish "Final stressed" from "Final unstressed". The former would look similar to the current table and favour the spelling <ee>; the latter would look like the "Final" table under /ɪ/ and favour the spelling <y>. This factor should be kept in mind when designing and scoring stimuli that end in /iː/.
<i> normally;
<e> unstressed, before final /t/;
<y> word-finally;
<e> unstressed initial in open syllable;
<e> unstressed after initial /d/ or <s>;
<e> unstressed after /r/ in 1st syllable.
Statistics. In final (or rather, unchecked?) unstressed position, this represents a fairly old-fashioned pronunciation. Probably more accurate to consider these words as having the phoneme /iː/ in general Liverpool speech. Also, several words listed here as having unstressed /ɪ/ are shifting to /ə/ (e.g., eleven, orange).
<i> is by far the most common spelling when the vowel is stressed. The only serious contender is <y>, which is found in Greek words. When unstressed, the sound usually has the same spelling, but it can also be a relatively recent reduction of /ɛ/, sometimes /ɛɪ/ or /iː/, and consequently it often has spellings appropriate to those other sounds (molecule, hurricane, captain, college). The following situations sometimes help to differentiate original /ɪ/ from /ɛ/.
Usually -et (planet), but often -it (tacit), mostly in Latin words.
/ɪ/ appears at the end of the word when unstressed, where it is usually treated as being an original /ɪ/ and so spelt <y> (which substitutes for <i> by regular graphotactic rules; e.g. berry). Other, less common spellings are <ey> (chutney) and <ie> (movie). In quite a few foreign words, the original spellings are retained, most often <i> (confetti) or (especially Latin) <e> (acme).
Word-initially in open syllables, <e> has an advantage over still common <i>, perhaps because the Latin prefix ex- often reduces to a single vowel.
After initial /d/, <e> wins out, because of the frequency of the Latin prefix de-. Similarly, Latin se- accounts for most of the preference for <e> after <s>.
This rule is mostly due to the workings of Latin prefixes like re- and pre-.
When stressed, /ɪ/ spelt with single <i> or <y> is definitely a short vowel, licensing doubling (brittle) and not licensing silent <e> (bid). When unstressed and spelt with a single vowel letter, it may vary between short (viscid, albumen, amulet, vinyl, spinach, ellipse) and long (vermicelli, baby, elephant, platypus, engine, image, college, erect, decree, rebel, secede) treatments. As a rule of thumb, both doubling and silent <e> are avoided unless required by the source language.
Some of these patterns should be taken with a grain of salt. As noted, some of them seem to depend on the presence of Latin prefixes in this supposedly monomorphemic list. Their contribution will naturally vary depending on fairly subjective judgements as to whether the prefix in question is identifiably a morpheme in modern English. Further, as mentioned, whole patterns may become invalid if the local speakers sound a substantial portion of the words with some other vowel, most likely /iː/ or /ə/.
<ear> normally;
<e> when stressed before /r/;
<io> when unstressed and checked;
<ia> when unstressed and free.
Statistics. May be perceived as sequence of /ɪ/ plus /ə/ or, depending on the individual's pronunciation, /iː/ plus /ə/. Before /r/, may be simply /iː/ or /ɪ/.
Normally, when stressed, this phoneme is spelt as if a sequence of /iː/ and consonant /r/. The relative frequency of the spellings is similar to that of /iː/: <ear> (dear, cf. dean) followed by <eer> (deer, cf. deep) and more remotely by <ier> (pier, cf. thief). Sometimes <ere> is used, mostly in Latinate words (severe, cf. theme), and <ir> may be used in other foreign words (emir). Occasionally, stressed /ɪə/ is spelt as if a sequence of /iː/ plus /ə/, resulting in a plethora of spellings (peony, lyceum, pyorrhoea, vehicle).
For those speakers who have /ɪə/ before /r/, that vowel is spelled as if /iː/. Its source is always Latin (series) or foreign (lira).
When unstressed, the diphthong is spelt as if a combination of unstressed /ɪ/ and /ə/. The most common spelling is <io> (chariot), but there is a plethora of other spellings (alias, orient, pancreas, pantheon, etc.)
Among unstressed spellings, final position is different in that <o> doesn't spell schwa word-finally. This gives the advantage to <ia> (trivia) and <ea> (area).
When <r> is part of the spelling, a preceding single vowel is treated as long: no doubling, and silent <e> may be licensed (here).
From a strictly phonological point of view, one would expect the sequence /juː/ to be analyzed as /j/ + /uː/. It is here treated as a unit primarily because the spelling system does so in the overwhelming majority of cases. If one prefers to think of the units as separate, the analysis could fairly easily be reinterpreted as a specialized environment for the spelling of /uː/.
Statistics. For some speakers, a word like cure may be perceived as this phoneme (/juː/) followed by /ə/. However, such sequences are treated separately, below. Similarly, some speakers may perceive (or pronounce) words like formula as having a form of /juː/. Instead, they are here treated separately, below.
<u> normally;
<u_e> before a single final consonant;
<ew> at the end of monosyllables;
<ue> at the end of polysyllables.
/juː/ occurs mostly in Latin loans and is spelt in the Latin fashion (human). This default environment basically comprises nonfinal syllables. The spelling <eu> occurs in a few words, mostly of Greek origin (eunuch).
<u> plus silent <e> strongly predominates (fume), although one sees several other spellings, most notably <ew> (lewd) and <eu> (deuce).
At the end of monosyllables, <ew> predominates (dew), but a few words have <ue> (cue, due, hue). These last can be interpreted as spellings with <u> that have been extended both to meet the 3-letter minimum and to avoid ending a word with a <u>. It should also be noted that a very few monosyllables treat /juː/ as a sequence of consonant /j/ plus vowel /uː/: youth, yule, yew, you.
At the end of polysyllables, <ue> predominates (argue), arguably with <e> to avoid ending a word with a <u>. The spelling <ew> also exists, as in monosyllables; the relative frequencies are simply reversed.
When /juː/ is spelt as a single letter <u>, it is treated as a long vowel, avoiding consonant doubling and licensing silent <e>.
From a strictly phonological point of view, one would expect the sequence /jʊ/ to be analyzed as /j/ + /ʊ/. It is here treated as a unit primarily because the spelling system does so in the overwhelming number of cases. If one prefers to think of the units as separate, the analysis could fairly easily be reinterpreted as a specialized environment for the spelling of /ʊ/.
<u>.
Statistics. /jʊ/ occurs only in unstressed medial open syllables in Latinate words, and is always spelt with <u>, as if a long vowel. Many speakers may in fact categorize it as /juː/, in which case these words would all be treated as examples of words where /juː/ appears in the normal environment and has the modal spelling. For others, some of these words may be pronounced with /ə/. In this case, they would be transferred to the schwa environment "in medial syllable before /l/", where <u> now becomes the modal spelling, pushing <o> back into second place.
From a strictly phonological point of view, one would expect the sequence /jʊə/ to be analyzed as /j/ + /ʊə/. It is here treated as a unit primarily because the spelling system does so in the overwhelming number of cases. If one prefers to think of the units as separate, the analysis could fairly easily be reinterpreted as a specialized environment for the spelling of /ʊə/. Of course /ʊə/, in turn, could also be thought of as a sequence of /ʊ/ + /ə/ or even of /uː/ + /ə/, depending on one's pronunciation and the level of phonological abstraction.
<ure> normally;
<u> before /r/.
Statistics. The words in this group have a wide variety of pronunciations. Some people, for example, do not pronounce a /j/ in allure; some pronounce cure and pure with /jɔː/; some pronounce jaguar with 3 syllables; some do not have the /ə/ element before /r/. When all these exclusions are taken into account, there is really very little left to call a pattern.
<or> normally;
<a> before /l/;
<o> before /r/;
<au> before /n/ in the coda;
<ore> finally;
<ar> after /w/, except before /l/.
<or> is by far the most common spelling nonfinally (stork), but also very common are <au> (auto), and (especially before <n>) <aw> (spawn). Other minor spellings include <oar> (board), <our> (source), and, before /t/, <augh> (slaughter) and <ough> (thought).
Before coda /l/, spellings with <r> are not used. <a> is most common (call) but <aw> (sprawl) and <au> (vault) are also common.
/ɔː/ is found before /r/ primarily in Latinate words, where it is usually spelt <o> (thorax), sometimes <au> (aura).
Only <au> is found before /n/ (flaunt).
<r> spellings are common in word-final position, mostly <ore> (more), but <or> (for), <oar> (soar), and <oor> (door) are also fairly common. In addition, the spelling <aw> is very frequent in nativ/ized words (law).
By far the most frequent spelling after /w/ is <ar> (dwarf, quart). However, the rule about following /l/ trumps this rule (wall).
When /ɔː/ is spelt <a>, it is treated as a short vowel, with consonant doubling and no silent <e> (ball, swath). When spelt <o>, it is treated as a long vowel, with no consonant doubling (chorus).
Several general graphotactic rules apply here. One is the avoidance of <r> spellings before complex codas (flaunt, *flornt), before /r/ (chorus, *chorrus) and usually before /l/ (ball, *borl). Another is the use of <gh> spellings only at the end of a word or before <t> (ought, *broughn); <w> digraphs mostly at the end of a word or before final <l>, <n>, with <u> digraphs more common elsewhere and seriously avoided in word-final position.
Many if not most younger speakers also pronounce traditional /ʊə/ words like sure with this vowel, so expect also occasional spellings with <ure>.
<oi> normally;
<oy> at the end of a word.
Statistics. This diphthong could be interpreted as a sequence of phonemes, most likely /ɒ/ + /ɪ/, which is supported by the spelling; but we treat it here as a single phoneme. In addition, the rare sequence /ɔɪə/ is also treated as a single phoneme, below.
<oi> is overwhelming most common in nonfinal position. A few words have <oy>. In coyote and doyen the use of <y> is due to a graphotactic rule--<y> is used between two vowel letters-- but those are the only two examples here.
<oy> at the end of a word, by normal graphotactic rules, which subtitute <y> for <i> at the end of a word.
Statistics. There are only a handful of words that have this sequence, and the spelling and pronunciation vary widely. In the Celex database are found only coir and soya.
Participants would likely interpret final /ɔɪə/ as being morpheme-final /ɔɪ/ plus the morpheme /ə/, which would suggest the spelling <oyer>. Morpheme internally, the reasonable spelling would be <oy> followed by a non-r schwa spelling such as <a>. But there are few if any concrete examples to abstract a spelling rule from.
<oo> normally;
<u> in nonfinal syllables after consonants where /j/ is impossible;
<ue> word-finally, after consonants where /j/ is impossible.
Statistics. See also /juː/, /jʊ/, /jʊə/, /ʊ/, /ʊə/, all of which can be interpreted in whole or part as variants of this vowel.
<oo> is by far the most common spelling here, but one finds also the (predominantly French) spelling <ou> (soup, coupon). Particularly after sounds that cannot be followed by /j/ (viz., /l/, /r/, /ʃ/, /tʃ/, /dʒ/, /s/) but also in some foreign loans (barracuda, haiku), the spelling <u> may be used, which takes silent <e> (flute). A few words are spelt with <o> (lose) or <ui> (bruise).
In part because polysyllabic momonomorphemic words tend to be foreign or Latinate, <u> predominates in this environment (lucid), although a few French words have <ou> (routine).
Word-finally, it is again only in words where preceding /j/ is impossible that anything other than the default <oo> predominates, and then only barely. <ue> is most common (blue), followed closely by <ew> (brew), then <oo> (loo).
The unusual spelling of this vowel is due to the fact that it is a conflation of two vowels. One is /uː/ proper, which is normally spelt <oo> (roost, bazooka, woo), but also (particularly French) <ou> (group, soubriquet, caribou, silhouette), <o(_e)> (tomb, move, movie, who) and in a few words <oe> (canoe, shoe). In a few foreign words this can be spelt <u> (hula). This sound can occur in any environment. The other sound is spelt as if /juː/: <u> (gluten), <u_e> (brute), <ui> (bruise), <eu> (sleuth), or <ew> (shrewd). This spelling can occur only in contexts where /j/ could not legally precede the vowel by the rules of phonotactics. This distribution is complicated by the facts that: (1) /uː/ proper has no limits on its distribution (except that it does not normally appear in Latinate words), so no clean division can be made between the two cases; (2) A handful of words with <u>, like hula, actually fall into the first group; and (3) different accents have different conceptions of what consonants /j/ can come after, and so different words may fall in different groups. To my knowledge, though, Liverpool accents are not (much?) different from RP in this respect.
When this sound is spelt with a single vowel letter, whether <u> or <o>, it is treated as a long sound, eschewing consonant doubling (gluten) and licensing silent <e> (fluke, move).
As per usual rules, digraphs ending in <w> are rare word-internally, and those ending in <u> are almost nonexistent word-finally.
The easiest things to test for would be spellings after consonants where /j/ is permitted (in RP, the clearest cases are /p/, /b/, /f/, /v/, /m/, /t/, /d/, /n/, /k/, /h/). Except for clear loanwords, the <u> class of spellings are illegal in this context; one should expect one of the spellings that begins with <o>.
<u> normally;
<oo> before /k/.
Statistics. As mentioned above, broad northern speech does not distinguish this vowel from /ʌ/. Also, some speakers have /uː/ in at least some of the -ook words, such as book, cook.
The normal <u> is found in a small set of words where the vowel is preceded by a labial consonant and is typically followed by an /l/ or voiceless postalveolar (bull, butch, push, full). This spelling also includes a few words where the vowel is unstressed (rubella) or followed by another vowel (ruin, suicide); in these words, an alternative pronunciation /uː/ is common and perhaps predominant in careful speech. The default environment also has several words that are spelt with <oo> and a few with <o>.
Almost all words with /ʊk/ in RP spell the vowel as <oo> (book, cook, look).
In the core words where all speakers agree on the pronunciation /ʊ/, the spelling <u> is treated as short: it doubles consonants and fails to license silent <e> (bullet, butch, cuckoo). In those words whose pronunciation alternates with /uː/, the vowel is treated as long (suet, rubella).
A rather complex cross-cutting of phoneme sets across dialects makes this phoneme a bit difficult to test. For an RP speaker, one could, for example, see whether the speaker knows to spell /ʊ/ as <oo> before a /k/, a very robust pattern; <u> is arguably wrong there, being supported only by cuckoo. However, a Liverpudlian is likely to take /ʊk/ as the rime of muck, puck, suck, etc., under which circumstances <u> is not only legal but the most common spelling. It would be unclear, therefore, whether one would be testing knowledge of spelling patterns or control of southern accents.
A pattern that works cross-dialectically is that /ʊ/ cannot be spelt <u> after /w/, but the evidence set is small and the actual spellings rather mixed (wood, wool, whoosh, wolf, woman, would, northern worry).
<our> normally.
Statistics. Cf. also /jʊə/, for the case where this sound is preceded by /j/. This is a very small set of words (11 in Celex), for which patterns can only be drawn by analogy with other vowels. The issue is complicated by the fact that many younger people pronounce at least some of these words with /ɔː/ instead of /ʊə/ (tour, tourniquet), and others as /uː/ or /ʊ/ (bravura, lurid, cruel, gruel), or as a sequence of /uː/ or /ʊ/ plus schwa. A great deal of caution is therefore called for! In principle, this phoneme is spelt as if /uː/ (also /juː/ after consonants that forbid following /j/, e.g., lurid) plus a consonant /r/ (dour, lure, boor) unless the latter is actually present (bravura). Occasionally it is spelt as if /uː/ or /juː/ plus a schwa vowel (cruel).
When spelt as a single letter <u>, the treatment is as a long vowel.