The most simple form of regularity involves a single class of verbs, a single principal part (the root or a conjugated form in a given person, number, tense, aspect, mood, etc.), and a set of unique rules to produce each form in the verb paradigm. More complex regular patterns may have several verb classes (e. g. distinguished by their infinitive ending), more than one principal part (e. g. the infinitive and the first person singular, present tense, indicative mood), and more than one type of rule (e. g. rules that add suffixes and other rules that change the vowel in the root).
Sometimes it is highly subjective to state whether a verb is regular or not. For example, if a language has ten different conjugation patterns and two of them only comprise five or six verbs each while the rest are much more populated, it is a matter of choice to call the verbs in the smaller groups "irregular".
The concept of regular and irregular verbs belongs mainly in the context of second language acquisition, where the defining of rules and listing of exceptions is an important part of foreign language learning. The concepts can also be useful in psycholinguistics, where the ways in which the human mind processes irregularities may be of interest. However, most other branches of linguistics do not use these categories; historical/comparative linguistics is more interested in categories such as strong and weak.
Irregularity can be caused by a regular sound change in the history of a language which occurs only in certain phonetic environments and thus affects only parts of the verbal paradigm. An example is Grammatischer Wechsel, the consonant change between the forms was and were.
Irregularity can be caused by various forms of assimilation, which again may occur in only one part of a verb. An example is English think - thought, which displays both a vowel assimilation known as Rückumlaut and an entirely separate consonant assimilation known as the Germanic spirant law.
Irregularity can be caused by a shift of stress. For example, Latin verbs usually have their stress on the penultimate syllable. Since most present tense forms have a single-syllable inflection, this means the stress is on the stem (pósso, 'I can'), but the first and second persons plural have a two-syllable inflection, which meant the stress was on the inflection (possémus, 'we can'). In French a process of levelling means that stress is always on the stem. The vowel peculiarities typical of the nous and vous forms of French irregular verbs result from the fact that only these parts experienced a shift of stress: je peux, nous pouvons.
Sometimes a regular system may fall into relative disuse, so that its remnants are seen as irregular. An example is the Germanic strong verb, such as English speak. (But beware: the weak-strong distinction in English and other Germanic languages does not co-incide with the regular-irregular distinction!)
Sometimes a perfect stem may come to be used as a present. (For example, the past tense of a verb meaning "to see" may come to be the present tense of a verb meaning "to know" - I know something if I have seen it.) The resulting verb then has past-tense inflections throughout its present tense, and may have gained a new (regular or irregular) past tense. Verbs of this sort in English are called preterite-presents; examples are can, may, must.
Almost all irregularities in verbs can be explained as part of the natural historical development of language, and grow out of some originally regular phenomenon.
A common error for small children is to conjugate irregular verbs as though they were regular. This is regarded as evidence that we learn and process our native language partly by the application of rules, rather than, as some earlier scholarship had postulated, solely by learning the forms. In fact, children often use the most common irregular verbs correctly in their earliest utterances but then switch to incorrect regular forms for a time when they begin to operate systematically. This allows a fairly precise analysis of the phases of this aspect of first language acquisition.