1.2 Regular and Irregular Forms in English

English inflections are relatively impoverished and the distinction between regular and irregular inflections only extensively exists for the past tense forms of verbs and plural forms of nouns. However, the generation procedures of regular and irregular forms, especially the past tense forms, offer an unusually sharp contrast between a highly regular procedure and a highly irregular and idiosyncratic set of exceptions.

Regular English past tense verbal forms are formed by the rule of adding the past tense suffix-ed to an unchanged stem verb, with three different phonologically determined forms[t], [d]and[id]depending on the final segment of the stem. This is an apparently paradigmatic example of a rule-based process. By contrast, irregular English past tense verbal forms are generally idiosyncratic and phonologically unpredictable, although they show a variety of irregularity within a continuum. Some undergo none overt changes (e.g.,rid-rid, hit-hit). That is, their past tense forms are identical to the stem verbs. Interestingly, all the English verbs whose past tense forms undergo none overt changes end with t or d. Some are generated only by changing the final consonant into[t]or[d] (e.g.,bend-bent, make-made). Also some undergo a stem vowel change but end in t or d (e.g.,sleep-slept, feel-felt). About half of the irregular English verbs end in t or d, because they originally took some version of the regular-ed suffix. Many more are formed by a stem vowel change (e.g.,sing-sang, wear-wore). Finally, there are some suppletive past tense verbal forms with highest degree of irregularity (e.g.,is-was, go-went), which involve a completely arbitrary relation between the stem and its past tense form.

Many past tense forms of irregular verbs, however, do not show total arbitrariness in that the phonological content of their stem is largely preserved, and some patterns can be easily recognized with similarities among clusters of irregular verbs in their stems and past tense forms, such as feed-fed, speed-sped, bleed-bled, breed-bred; keep-kept, weep-wept, sleep-slept, sweep-swept. Therefore, many irregulars are not just a set of arbitrary, individually memorized by rote memory and idiosyncratic items, but quasi-regular and cannot be simply attributed to a lexicon of stored exceptions. But on the other hand, nearly all the family members of verbs with phonological similarity have exceptions. For example, the past tense forms of bring and sing are not brung and sung, although the past tense forms of cling, fling, sling, sting, wring and swing are clung, flung, slung, stung, wrung and swung. Also, the past tense forms of drink, sink and shrink are drank, sank and shrank, but the past tense forms of think and slink are thought and slunk. Even trickier, blink and wink are regular verbs and their past tense forms are blinked and winked.

In a very similar way, regular English noun plurals are formed by adding the suffix-s to a countable noun, with three allophones[s], [z]and[iz], whereas there are several classes of irregular nouns with different levels of regularity. One class has a zero plural, for example, sheep and deer. That is, their singular and plural forms are identical. Another class involves the voicing of the final consonant, such as knife-knives, wolf-wolves. The third class is formed by adding the Old English noun plural suffix-en, such as child-children, ox-oxen. The rest native irregular English noun plurals involve a stem vowel change, like mouse-mice, louse-lice, foot-feet, tooth-teeth. Finally, the plurals of the nouns borrowed from Latin and Greek are formed generally by changing the word endings, for example phenomenon-phenomena, criterion-criteria, stimulus-stimuli. Again, many irregular nouns show the phonological similarity effect and form some patterns. For instance, the pairs of foot-feet, tooth-teeth and goose-geese as well as knife-knives, life-lives, wife-wives undergo the same stem vowel change to form the plurals.

Another area of regular and irregular nouns concerns regularity within noun-noun compounds. According to Pinker (1998: 231) and Pinker and Ullman (2002: 459), the first noun is generally a modifier and the second noun is the head, and the new compounds inherit the properties of the rightmost morphemes, which is sometimes called the right-hand head rule. According to Gordon (1985) and Pinker (1999: 207), the first noun of English noun-noun compounds must be singular if it is regular and countable, but it could be plural if it is irregular and countable. For instance, rats eater and standards setter are not acceptable, while mice eater and criteria setter are far more acceptable.

Suffixation of regular inflections is highly productive in the sense that generally there are far more regular words than irregular words in languages, and that the suffixation applies predictably to thousands of words and is generalized to almost all the new words in the language as the default treatment. For example, there are thousands of existing regular English verbs and hundreds of new ones are added all the time, but the number of irregular verbs is only about 180 (Pinker, 1998: 222). Additionally, when a new word like clone is coined, it will be treated as a regular verb and the past tense suffix-ed is added to it to form its past tense form. Likewise, when a foreign word comes into English, it will generally receive the default treatment, even if it has a similar pattern to the existing irregular English words. For example, the plural forms of chief and gulf, which were borrowed from French, are not chieves or gulves, but chiefs and gulfs; the past tense form of deride, which was borrowed from Latin, is not derode, but derided. Sometimes, there is a transitional period of regularization for the irregular foreign words. For instance, when the irregular Latin noun focus was borrowed into English, it kept its original plural form foci. Later it underwent regularization in English as focuses, and now both foci and focuses are acceptable. It can be predicted that the regular form focuses will win over the irregular form foci in the future. Furthermore, the default treatment is also applied to nearly all the novel verbs. For instance, Berko (1958) found that even preschool children could create the past tense forms for the invented words like rick by adding the regular suffix-ed. Moreover, these patterns are even occasionally generalized by human speakers. For example, Bybee and Slobin (1982) found that children occasionally produced novel past tense forms like brang for bring based on sing and spring. According to Xu and Pinker (1995), although these errors are rare, all children make them. And Bybee and Moder (1983) also found that most college students offered splang or splung when they were given the novel verb spling and asked to guess its past tense form.

Irregular forms have a distinctive feature of high frequency. That is, most of the irregulars are among the most common words and used at high frequencies. According to Francis and Kucera (1982), the 13 most frequently used verbs in English-be, have, do, say, make, go, take, come, see, get, know, give, find -are all irregular. And according to Marcus et al. (1995), about 14% out of the 1, 000 most frequent verbs are irregular. Based on Bybee (1985), Old English had about twice as many irregular verbs as Modern English like cleave-clove, abide-abode. These have been regularized over the centuries because of their low frequencies, unless they fall into certain patterns whose family members may strengthen memory traces for each other. One explanation is that if irregular forms are not used frequently, children would not hear these verbs often enough and they might disappear from the language (Marcus et al.,1992). Another explanation from a diachronic perspective for this is that people cannot remember them firmly, and consequently they will be gradually treated as a regular form, the so-called default treatment, and become regular forms (Pinker, 1999).

Another distinctive feature of the irregulars is that all of them are monosyllables or prefixed monosyllables (at least in English) like oversleep and become, and follow the canonical sound pattern for simple English words. Therefore, there are no phonologically unwieldy English irregulars, although some English regulars like hatched, months are less pronounceable.

Up to the present, most of the studies on the processing of regular and irregular forms have focused on English past tense verbal forms, for regular and irregular English verbs are equated for meaning (referring to past), complexity (one word), syntax (tensed) and could be matched on the syllable structure and frequency. As Pinker (1991: 531) stated, past tense inflection is an isolable subsystem in which grammatical mechanisms can be studied in detail, without complex interactions with the rest of the language, and past tense marking-ed is also insensitive to lexical semantics in that both regular and irregular past verbs express past events or states and that the regular-irregular distinction does not correlate with any feature of verb meaning.