Tuesday, February 16, 2010

The secret lives of irregular verbs

From A New Introduction to the Italian Language, by Henry Marius Tourner (Edinburgh: Neill & Co., 1794)Ah, irregular verbs. They are the bane of first-year language students everywhere. The grammars of most languages are fairly regular — sure, nouns and adjectives offer occasional surprises (e.g., the gender-switching plurals of Italian, on which I am preparing a separate post), but is there any feature of grammar that bedevils students more than the irregular verb? I don’t care if we’re talking about English, Italian, German, Sanskrit, or Ancient Greek — just about all natural languages have them. (It’s been claimed that a few, like Turkish and Chinese, have very few irregular verbs or even none at all; and of course, artificial languages, like Esperanto, are designed quite deliberately to have none.)

Students find them intimidating, nothing more than columns of more or less random variants to be stumbled through and memorized. But all words have etymologies, even the irregular forms of verbs. Etymology is really nothing more than asking how a word became the way it is. The inquiry is normally confined to root words (nouns, adjectives, infinitive verbs, and so on), but the question may be asked of particular forms too. In fact, it should be asked of particular forms, especially those for which no intuitive explanation comes readily to mind. When we do so, we find that a lot of the so-called irregular forms aren’t really so irregular after all. Let’s look at some examples.

In the Germanic languages

English has a large number of so-called irregular verbs, close to 200 of them. Moreover, these are among the most commonly used of all English verbs. It’s been estimated that when you use a verb in everyday English conversation, something like 70% of the time it’s one of the “irregular” ones. But most of these are not really (or were not originally) irregular. Rather, most of them followed very specific, if fairly complex, rules for changing their internal vowel. This process is called ablaut, and it runs the whole gamut of the Germanic language family, from the earliest recorded languages to the present day. Students of Old English often encounter the terminology of “strong” and “weak” verbs — which is another way of saying the same thing.

Even today, when we have lost touch with the rules that made these verbs more regular, we can still detect patterns. Consider just a couple of the representative classes: (1) ring, rang, rung; sing, sang, sung; spring, sprang, sprung; (2) blow, blew, blown; grow, grew, grown; know, knew, known; show, shew, shown; throw, threw, thrown. Notice shew, the original past tense of show, unfamiliar to most of us today but still in use as recently as 1915. This form has now been supplanted by a seemingly more “regular” form, showed. In reality, though, the attempt to regularize the verb has really made it less consistent with other verbs of its class. And this process of “regularization” is gaining momentum. Consider: hang, hung, but also hanged; plead, pled, but also pleaded.

Some verbs are irregular in English for a different reason. Ever wonder why the past tense of go is went? Or why we have to be, I am, I was? Phonologically, these forms have nothing whatsoever in common. As it happens, went is the past tense of wend, a verb synonymous with to go but almost extinct in Modern English. It clings to life in a few phrases such as “to wend one’s way”, and in computer programming constructions such as WHILE … WEND. But while wend has all but wended away, its past tense, went, has stuck around, stuck to a different verb! Lingusts call this process suppletion. In the case of to be, we again have forms from two different Old English verbs: beón and wesan. As a side note, in Old English, gán “to go” had yet a different irregular past tense: ic gá “I go”, but ic éode “I went”. But I digress.

In the Romance languages

A major class of Romance verbs represents the case of simple erosion, a process occurring in all languages all the time. Erosion occurs for many reasons, from easing pronunciation to nothing more than laziness. There are innumerable examples in English — e.g., every, pronounced “evry”, probably pronounced “probly” (even “prolly”), walked pronounced “wakd”. And how about one from French — est-ce que pronounced “esque”. Eventually, the spelling of these works is likely to change. But to return to verbs, where this kind of erosion or contraction has occurred, traces of the older, “uneroded” verb form survive in different forms or tenses of the modern verb, right alongside newer forms.

Think about the following Italian verbs: bere “to drink”, versus bevo “I drink”; dire “to say”, versus dico “I say”; fare “to do”, versus faccio “I do”. Where did these consonant changes come from? In turns out that bere is contracted from an older form, bevere, dire from dicere, fare from facere. These earlier forms may be found in the works of Dante and Boccaccio. And seen in this new light, the forms bevo, dico, and faccio look much more regular. We find the same phenomenon in French, where the verb boire “to drink” has “irregular” plural forms buvons, buvez, boivent. Why? Again, because of older form(s) of the infinitive, beivre, boivre > boire. We find boivre and beivre, in illo ordine, attested in the twelfth-century Roman de Renart and the thirteenth-century Fables of Marie de France.

Another class of Romance verbs appear to be irregular only because two verbs coalesced into one — suppletion again. This explains the case of Italian andare “to go”, versus vado “I go”, French aller / vais, Spanish ir / voy, Portuguese ir / vou, Catalán anar / vaig, and so on. What we have here is a pair of Latin verbs, īre and vādere, both meaning “to go”. Rather than one verb winning out over the other entirely, we ended up with surviving forms from both. Why? So far as I know, that’s a mystery. But it’s a mystery taken up with remarkable consistency into Latin’s many daughter-languages! Not quite so irregular as things might first have appeared, eh?

Summing up

Learning how irregular verb forms originated makes remembering them much easier. Typically, there are three major explanations for irregular forms. (1) Ablaut, or the changing of internal vowels to convey grammatical information; these are the “strong” verbs of the Germanic languages. (2) The erosion or contraction of forms over time, as in the examples of Italian bere and dire. And (3) suppletion, where forms from independent (synonymous) verbs coalesce into a single verb in the modern language.

These are hardly the only reasons verbs become irregular — I haven’t touched on changes due to orthography, or brought about by historic vowel and consonant shifts — but if you take the time to learn how these three broad classes operate, you will be surprised how often those troublesome irregular verbs begin to fall in line.


  1. I think it's the verbs that tend to be irregular for the simple reason that most commonly learned languages have far more noun roots than verb roots, and therefore individual verbs are more frequent in running text than individual nouns. Since it's only the high-frequency irregulars that can avoid being leveled, we end up with lots of irregular verbs and only a few irregular nouns. Leaving out the Latin and Greek plurals, there are only about 40 irregular noun roots in English, compared to about 135 irregular verb roots (I count roots because some verbs have many more prefixes than others).

    A few points:

    Only about half the English irregular verb roots are strong (ablaut) verbs: the rest are weak verbs with -t instead of -ed, with internal loss, or other peculiarities.

    Show in fact began as a weak verb and acquired strong forms only later, a rare "wrong-way conversion": it now has the hybrid conjugation show, showed, shown, with occasional uses of showed as past participle. The 1915 use (Shaw, I presume) is the separate verb shew, the direct descendant of OE scéawan, which is now almost obsolete and has been pronounced the same as show for centuries. Show is from a variety of OE or EME that changed the falling diphthong into a rising one, sceáwan.

    WEND is of course a portmanteau form of "while+end".

  2. Thanks for the comment and links, John. A couple of further remarks. On the nouns: (1) In your first group of irregular nous, a number of these have been normalizing recently, e.g., hooves or hoofs, rooves or roofs (not on your list). (2) Your fourth group demonstrates an entire category of words irregular because of borrowing, a whole class I left out of my post, but worth remembering. This happens in other languages, too — e.g., Italian il jazz, il computer, il jeans, with plural forms the same as the singular, and all irregular. (3) You third group of nouns had ox/oxen, about which I have written previously.

    On the verbs: (1) “Only about half the English irregular verb roots are strong (ablaut) verbs: the rest are weak verbs with -t instead of -ed, with internal loss, or other peculiarities.” Are they weak? Most of them show systematic internal vowel changes. (2) Regarding shew, very interesting. I will have to look into that a bit more. At a glance, then, it looks like shew, show(ing) might be analogous to sew, sewing?

  3. I should have noted that my lists are drawn from the Cambridge Grammar of the English Language: I looked at them today and noticed that they aren't marked. Fixed now.

    Among the verbs, Group 1 is clearly weak; the vowel variations that some subgroups show are not of the ablaut type, but rather represent irregular shortening in ME. Some Group 4 words are the same: the rest represent cases in which an -i- was lost in Proto-West-Germanic except in the present stem, e.g. sell with i-mutation but sold without, and think with i-mutation and thought without.

    Here's what the OED3 has to say about show:

    A common WGer. weak verb: OE. scéawian = OFris. skawia, skowia, schoia, skua (WFris. skoaije, skôgje, skouje), OS. skawon (MLG. schowen), MDu. scauwen, schauwen (mod.Da. schouwen), OHG. scauwôn, scouwôn (MHG. schouwen, schawen, mod.G. schauen): — WGer. *skauwôjan, f. OTeut. *skau- to see, look, whence Goth. skaun-s beautiful (see SHEEN a.); other alleged cognates in Gothic are spurious. In all the continental WGer. langs., as in OE., the verb means ‘to look at’; the sudden change in Eng. (c 1200) from this to the causative sense ‘to cause to see, exhibit, manifest’, is difficult to account for. (The existence of the causative sense in OE. is not really proved by the rare áre gescéawian `to show mercy', griđ scéawian `to grant a safe-conduct', as these uses may be explained as developed from the sense ‘to look out, provide’.)

    From early ME. the verb has had a strong conjugation (after KNOW v., etc.) by the side of the original weak conjugation; in the pa. tense this survives only in dialects; but for the pa. pple. shown is now the usual form; the older showed is still sometimes used in the perfect tenses active (chiefly with material object), but in the passive it is obs. exc. as a deliberate archaism.

    The spelling shew, prevalent in the 18th c. and not uncommon in the first half of the 19th c., is now obs. exc. in legal documents. It represents the obsolete pronunciation (indicated by rhymes like view, true down to c 1700) normally descending from the OE. scéaw- with falling diphthong. The present pronunciation, to which the present spelling corresponds, represents an OE. (?dialectal) sceáw- with a rising diphthong.

  4. Thanks for that! Valuable information, and yet another example of why I really need to get OED access! *grumble*

  5. The "wend" in the "while... wend" programming example actually is a portmanteau of "while" and "end": the end of a "while" loop.

  6. Is it? I haven’t heard that, but of course, it’s certainly plausible.