Ben Yagoda (Friend
of SbaCL and Not One-Off Britishisms blogger), who had recently noticed a US journalist saying learnt instead of
learned, asked whether I'd covered the
‑ed/‑t alternation. It's one of those things that I've been putting off
for a long time because it would be a very long post. Now I've been shamed out of my laziness.
In order to do this in any kind of sensible way, I feel like I need to explain some things about the past tense in English. I'll try to introduce terms gently, with links to sites with deeper
explanations. At points I will be a bit sloppy and use more familiar (and less
precise) terms (like past-tense). And I'm going to be very sloppy about
phonetic spelling, both because not everyone knows the
International Phonetic Alphabet (IPA)
and because if I tried to use the IPA we'd have to get into
not-especially-relevant differences in pronunciation of many of these
words.
The origins of ‑ed
Let's start by thinking a bit more about ‑ed. Old English had different
categories of verbs that were put into past tense (preterite) in various ways. The so-called strong verbs were those that
changed their internal vowel. Some of those are still 'strong' in Modern
English, like drink/drank and write/wrote.
Those that ended with the (then multiple) suffixes that would eventually
become ‑ed are
weak verbs. They don't
undergo an internal change to make past tense; a suffix is just stuck on the
end.
Nowadays, we think of strong verbs as "irregular verbs" and
‑ed verbs as "regular" verbs, but back in Old English the verbs
that we now think of as "irregular" fell into regular patterns in a more complex system.
For centuries, English has been bending toward verb weakness. Many Old English
"strong" verbs are now made past-tense with ‑ed, like
starved (rather than something like storve) and
baked (not boke).
But ‑ed is only the spelling of the past-tense suffix
We tend to think of
‑ed as the past-tense suffix because it's how
it tends to be represented in spelling. That spelling makes it look like it
has two sounds, but a common lesson in English Linguistics
101 is
that spelling is misleading. Notice how we pronounce
‑ed in the
following words:
-
stopped, stoked, passed, slashed,
torched = "stopt", "stokt", "past", "slasht", "torcht"
-
strobed, flogged, buzzed, judged,
blamed, pinged = "strobd", "flogd", "buzzd", "judjd",
"blamd", "pingd"
That is, each of these past tense forms is pronounced with one syllable.
The ‑ed does not represent a vowel+consonant combination.
Buzzed isn't "buzz-ed", it's "buzzd".
If you don't hear the difference between those, think about
learned in these two contexts:
I learned a fact versus
a learned scholar
The first has one syllable ("lernd"), the second as two distinct
syllables with a distinct vowel in the ‑ed. That two-syllable learnéd (sometimes spelled/spelt with the accent mark) is a
special case; it's an adjective, rather than a verb. We're going to stick to verbs, not adjectives in this post, but that adjective is handy for illustrating what we're not doing in words like
buzzed. We're not pronouncing a vowel in ‑ed.
Some other ‑ed verbs do have a pronounced vowel in ‑ed:
-
tasted, boarded, dated, padded, minded:
each has two syllables.
If you start from the spelling, you might think that buzzed is buzz+ed and the E has got(ten) lost. But language doesn't start from spelling, it starts from
sounds. Instead of the suffix being ‑ed, with some weird places where the
vowel is dropped, it makes more linguistic sense to see the suffix as ‑d and to observe that we have rules for what to do when that [d] rubs up against other sounds in pronunciation. The rules are:
-
The [voiced] -d becomes
[voiceless] ‑t when it follows a
voiceless consonant sound. (We say it assimilates to voicelessness.
Assimilation makes things easier to say quickly.)
-
A vowel is inserted (epenthesized) when we try to attach the suffix
‑d to a /t/ or a /d/ sound. These consonants are pronounced by tapping the gum ridge behind the teeth with the tip of your tongue (they're alveolar plosives). and if we tried to pronounce them together, you'd not be able to hear them
both. (In English, we would pronounce padd the same as pad.) So, inserting the
vowel makes the doubled alveolar consonants pronounceable for the speaker and hearable for the listener.
- In all other cases, the suffix remains ‑d in pronunciation.
Because we follow rules when we pronounce all those variants of -(e)d and nothing else changes, those are very regular verb endings. Notice that nothing major changes in the verb root. The a in taste is the same as the a in tasted, and the o in stop is the same as the o in stopped, etc. In the irregular verbs discussed below, that's not always the case.
This all means means that the difference between learnt and learned is very small: just the difference between saying the [t] sound and saying the [d] sound. We're not saying more sounds if we say the version that's got more letters.
Late addition: Marianne Hundt reminds me that things are not always straightforward—there can be back and forth between regularization and irregularization in the timeline. What follows us just about where we are now.
t/d variation
Now we move to the ones that seem
irregular in Modern English and whether they are the same in British
and American English.
In each case, I've had a look at the
Corpus of Global Web-Based English to see what percentage of the BrE/AmE usage is in the irregular form. So, where it says
98% in the first table for bent, it means that 98% of the examples are bent and 2% are bended. I've rounded all the percentages to the nearest whole number.
final d > t (no vowel change)
British and American English don't differ in using these irregulars:
| Base form |
Past form |
AmE % |
BrE % |
| bend |
bent | 98* | 98* |
| lend | lent | 100 |
100 |
| send | sent | 100 |
100 |
| spend | spent | 100 |
100 |
| | |
|
While we have a pattern here of end>ent, it's not a regularity. No one says tent as the past tense of tend, or ent as the past of end. I haven't tried searching for rend/rent because I'd be overwhelmed by the 'lease' meaning of rent.
*Bended is like learnéd, in that it's used as a participial adjective (as in on bended knee). So, the 2% or so of bended are a different thing. As a verb, everyone's saying bent: I bent the rules, not I bended the rules.
-pt versus -ped with vowel change
Here we see AmE moving toward regularization for creep and leap, but not other rhyming verbs. Irregularity is easier to maintain in much-used verbs—we learn the irregular form because we hear it. When we go to make a past-tense for a verb we've heard less, we often have to make up a past-tense form on the spot, and that is most easily done with -ed. It's a bit surprising that wept is still so strong, considering it's the least-used of any of this set.
| Base form | Past form | AmE % | BrE % |
|---|
| creep | crept | 62 |
92 |
| leap | leapt | 52 | 79 |
| sleep | slept | 100 | 100 |
| sweep | swept | 100 | 100 |
| weep | wept | 99 | 98 |
| | |
|
These irregulars all have a vowel change in common: the -pt version has a "short E", while its -ed counterpart (creeped, sweeped) has a "long E"—even leapt, whose spelling seems to indicate otherwise.
This case is different from other possible -pt endings, like slipt and stript. Since slipt is how slipped is actually pronounced (see above), slipt/slipped is just a spelling difference, not an irregular verb issue. (They are also spelled/spelt with a 'd: slipp'd and stripp'd.) The numbers for these are so low that they would show up as 0 in the table, but there's an interesting detail about those tiny numbers: slipt is only present in the GB corpus (6 times), and stript is only in the US corpus (10 times).
-Nt versus -ned with vowel change
In these ones, a final nasal consonant is followed by the -t suffix. The irregular forms also have a vowel change: the -Nt version has a "short E", while its -ed counterpart (leaned) has a "long E".
AmE uses regulari{s/z}ed leaned, while BrE still mostly uses leant, but both have mostly regulari{s/z}ed dreamed, and no one is saying meaned.
| Base form | Past form | AmE % | BrE % |
|---|
| dream | dreamt | 16 | 33 |
| lean | leant | 3 | 75 |
| mean | meant | 100 | 100 |
I have to wonder if the loss of leant is related to its having homophones: lent, as a past tense of lend.
-rnt versus -rned
These have no vowel change. So, in spoken language, the difference is between saying burnt and burnd.
| Base form | Past form | AmE % | BrE % |
|---|
| burn | burnt | 23 | 42 |
| earn | earnt | 0 | 3 |
| learn | learnt | 4 | 44 |
These are a little tricky because burnt is more common than burned as an adjective (e.g. burnt offerings), and as we've already seen, there are some funny things going on with learned as an adjective. But it's hard to trust that automatic processes for the corpus have accurately tagged the adjective use, so I haven't used that tagging to come to the numbers above. They include everything.
I had the feeling that these differ in preterit (
I learnt French) and
perfect (
I have learnt French) forms. So, I searched for these in the formula "PRONOUN [has/have/had] VERB+ed/t". The numbers for BrE irregulars go down in this condition (I tried it with other pronouns too), which tells us something, but I haven't got time to look into what it tells us. (Given that we no longer have the risk of errant adjectival
learneds, I expected the percentage to go up!)
| Past form |
AmE preterit |
AmE perfect |
BrE preterit |
BrE perfect |
| burnt | 17 |
21 |
33 |
39 |
| learnt |
3 |
6 |
31 |
36 |
So, I was right that there's more
-rnt in the perfect than in the preterite, but it's a smaller gap than I'd thought I'd find.
-led versus -lt
Finally, the Ls, one of which you've seen already in this post: spelled/spelt.
These fall into two categories, with and without vowel change.
The vowel-changing ones are solidly in the "irregular" category, with a bit of movement in the rarest of those, kneel>knelt.
| With vowel change | Past form | AmE % | BrE % |
|---|
| deal | dealt | 100 | 100 |
| feel | felt | 100 | 100 |
| kneel | knelt | 85 | 89 |
We see some of the biggest differences between AmE and BrE in the non-vowel-changing ones—with some caveats about homonyms below.
| Without vowel change | Past form | AmE % | BrE % |
|---|
| build | built | 100 | 100 |
| dwell | dwelt | 86 | 83 |
| smell | smelt | 13 | 48* |
| spell | spelt | 7 | 49† |
| spill | spilt | 11 | 38^ |
| spoil | spoilt | 5 | 51 |
*Smelt is a bit tricky because it can be a verb in its own right (smelting metal) and it's also a fish that's eaten in North America. The corpus, however, is bad at distinguishing these things. The majority of smelts in the results reported here are the past tense of smell, but it would be too much work to tell you exactly how many.
†Spelt is another problem one because it is the name of a grain. I tried sorting out the noun uses from teh verb ones, but it turns out that most of the ones tagged as "noun" in the corpus are, in fact, instances of the verb. So the numbers here include all spelts.
^In the case of spilt, I wondered how much adjectival use mattered, particularly in the phrase "cry over spilled/spilt milk". So, I searched for "spilled/spilt milk" and found that Americans are pretty evenly split on spilled versus spilt in the phrase (36 hits vs 32), whereas in British English it was 76 versus 18 hits. Those spilt milks account for 14–18% of the spilt percentages above (which is to say, that phrase isn't adding much to the AmE/BrE difference).
miscellaneous irregulars
There are a few more irregulars-ending-in-t; these ones end in fricative sounds. But it's not worth saying much about them, since they're much the same in British and American English.
leave>left: Everyone uses the irregular for this one. Where leaved happens, it has to do with leaves (like on a tree or a table), not leaving.
vex>vext: The -t version is still playable in Scrabble, but the corpus tells us no one's using it in UK or US. I'm not even bothering to look for other verbs ending in x.
dress>drest: No one's using this one either! But...
bless>blest: We find a bit more of this one, since old-fashioned spellings are common in religious language, either because they're quoted from long-ago translated scripture or because they're styled to sound like scripture. Still, only 2% of the AmE "past" forms are blest and only 1% of the BrE. (I say "past" because a lot of them are probably adjectives.)
The moral of the story is...
While some -t spellings are more common in current BrE than in current AmE, it would be wrong to call them "the British spelling", with one exception: leant. There we have clear evidence of a transatlantic divide where the -t version is the firm majority in the UK and the -ed version is much preferred in the US.
In the other cases, there may be more preference for one or the other in US or UK, but the same forms have the majority/minority in both countries (at least in this corpus, which was collected 12 years ago). That is to say, you're much more likely to see spelt from a British writer than an American one, but an awful lot of British writers are writing spelled. Learnt will tell you that a document is almost certainly not American, but learned will not tell you that the writer isn't British—and so forth.
