Dialects or language varieties are a result of systematic, internal linguistic changes that occur within a language. Unlike accents, in which linguistic changes occur mainly at the phonological level, dialects reflect structural changes in phonology, morphology, and syntax, as well as lexical or semantic changes. The degree of mutual intelligibility that a speaker's language has with a designated standard linguistic system is often used to distinguish dialect from language. Mutual intelligibility means that speakers of one dialect can understand speakers of another dialect.

Although the construct of mutual intelligibility is frequently employed to differentiate dialect from language, there are counterexamples. On one hand, speakers may have the same language, but the dialects may not be mutually intelligible. For example, Chinese has a number of dialects, such as Cantonese and Mandarin, each spoken in different geographical regions. Although Cantonese and Mandarin speakers consider these dialects, the two lack mutual intelligibility since those who speak only Cantonese do not easily understand those who speak only Mandarin, and vice versa. On the other hand, speakers may produce different languages but have mutual intelligibility. For instance, Norwegian, Swedish, and Danish are thought of as different languages, yet speakers of these languages can easily understand one another.

A dialect continuum or dialect continua may account for lack of mutual intelligibility in a large territory. A dialect continuum refers to a distribution of sequentially arranged dialects that progressively change speech or linguistic forms across a broad geographical area. Some speech shifts may be subtle, others may be more dramatic. Assume widely dispersed territories are labeled towns A, B, C, D, and E and are serially adjacent to one another, thereby creating a continuum. B is adjacent to A and C, C is adjacent to B and D, and so on. There will be mutual intelligibility between dialects spoken in A and B, between B and C, between C and D, and between D and E. However, the dialects of the two towns at the extremes, A and E, may not be mutually intelligible, owing to the continuous speech and language shifts that have occurred across the region. It is also possible that some of the intermediate dialects, such as B and D, may not be mutually intelligible.

Because different conditions influence dialects, it is not easy to discriminate dialect precisely from language. Using mutual intelligibility as a primary marker of distinction should be considered relative to the territories of interest. For example, in the United States, the concept of mutual intelligibility appears valid, whereas it is not completely valid in many other countries.

Dialects exist in all languages and are often discussed in terms of social or regional varieties. Social dialects represent a speaker's social stratification within a given society or cultural group. Regional dialects are associated with geographical location or where speakers live. Regional and social dialects may co-occur within language patterns of the same speaker. In other words, social and regional dialects are not mutually exclusive.

Regional dialects constitute a unique cluster of language characteristics that are distributed across a specified geographical area. Exploration of regional dialectal systems is referred to as dialectology, dialect geography, or linguistic geography. For many years, dialects spoken in cities were thought of as prestigious. Therefore, in traditional dialect studies, data were mainly collected in rural areas. Surveys, questionnaires, and interview techniques were used as primary mechanisms of data collection. A field worker would visit an area and talk to residents using predetermined elicitation techniques that would encourage the speaker to produce the distinctive items of interest. The field worker would then manually note whether the individual's speech contained the distinctive linguistic features of interest. These methods generated a number of linguistic atlases that contained linguistic maps displaying geographical distributions of language characteristics.

Data were used to determine where a selected set of features was produced and where people stopped using this same set of features. The selected features could include vocabulary, specific sounds, or grammatical forms. Lines, or isoglosses, were drawn on a map to indicate the existence of specific features. When multiple isoglosses, or a bundle of isoglosses, surround a specific region, this is used to designate dialect boundaries. The bundle of isoglosses on a linguistic map would indicate that people on one side produced a number of lexical items and grammatical forms that were different from the speech of those who lived outside the boundary. Theoretically, the dialect was more distinctive, with a greater amount of bundling.

After the 1950s, audio and, eventually, video recordings were made of speakers in designated regions. Recordings allow a greater depth of analysis because they can be repeatedly replayed. Concurrent with these technological advances, there was increased interest in urban dialects, and investigators began to explore diversity of dialects within large cities, such as Boston, Detroit, New York, and London. Technological developments also led to more quantitative studies. Strong statistical analysis (dialectometry) has evolved since the 1970s and allows the investigator to explore large data sets with large numbers of contrasts.

Several factors contribute to the formation of regional dialects. Among these factors are settlement and migration patterns. For instance, regional English varieties began to appear in the United States as speakers immigrated from different parts of Britain. Speakers from the eastern region of England settled in New England, and those from Ulster settled in western New England and in Appalachia. Each contributed different variations to the region in which they settled.

Regional dialect formation may also result from the presence of natural boundaries such as mountains, rivers, and swamps. Because it was extremely difficult to traverse the Appalachian mountain range, inhabitants of the mountains were isolated and retained older English forms that contributes some of the unique characteristics of Appalachian English. For example, the morphological a-prefix in utterances such as “He come a-running” or “She was a-tellin' the story” appears to be a retention from older forms of English that were prevalent in the seventeenth century.

Commerce and culture also play important roles in influencing regional dialects, as can be observed in the unique dialect of people in Baltimore, Maryland. Speakers of “Bawlamerese” live in “Merlin” (Maryland), whose state capitol is “Napolis” (Annapolis), located next to “Warshnin” (Washington, D.C.), and refer to Bethlehem Steel as “Bethlum.” Because the “Bethlum” mill, located in Fells Point, has been a primary employer of many individuals, language has evolved to discuss employment. Many will say they work “down a point” or “down a mill,” where the boss will “har and far” (hire and fire) people. While most people working “down a point” live in “Dundock” (Dundalk), some may live as far away as “Norf Abnew” (North Avenue), “Habberdy Grace” (Harve de Grace), or even “Klumya” (Columbia).

Two other types of geolinguistic variables are often associated with regional dialects. One variable is a set of linguistic characteristics that are unique to a geographical area or that occur only in that area. For instance, unique to western Pennsylvania, speakers say “youse” (you singular), “yens” (you plural) and “yens boomers” (a group of people). The second variable is the frequency of occurrence of regional linguistic characteristics in a specific geographic area. For example, the expression “take 'er easy” is known throughout the United States, but mainly used in central and western Pennsylvania.