§39. [r] is produced by raising the body of the tongue so that the sides of it press against the upper teeth, tilting the point of the tongue so that it just barely touches the bony ridge of the gums, and allowing the breath to escape with a distinctly audible friction over the point of the tongue and between the teeth, which are slightly open. It may be described, therefore, as an alveolar r, with reference to the position of the tongue. The vocal chords are in vibration and the lips are slightly drawn back. The breath escapes with considerable force between the up-tilted point of the tongue and the alveoli, and it is here that the consonantal friction in [r] is produced, not as the air passes between the teeth. The tongue positions for [r] are somewhat similar to those for [ʃ], [ʒ], but in [r] the teeth are open, in [ʃ], [ʒ] the upper and lower teeth are generally in close contact.
§40. This is the sound commonly heard in American speech for r initially, as in red [rɛd], between vowels, as in very [ˈvɛrɪ], and after consonants, as in dress [drɛs]. Before proceeding further with the consideration of various other kinds of r, the student is advised to observe extensively the occurrence of [r] in the three positions just mentioned in different words, and to study the sound itself so as to be able to distinguish the consonantal from the vocalic element in it. A voiceless r, which is merely frictional, should be compared with the voiced fricative [r], as for example the tr of tread pronounced separately without voicing as compared with red, read (preterite of the verb), pronounced [rɛd] with voicing.
§41. A trilled or rolled r, though not very common in American speech, is sometimes heard, especially for r between vowels, as in very, hurry, etc. It is commonly cultivated in stage pronunciation on the ground that it carries better than the fricative r. It is also cultivated by telephone operators in the pronunciation of three. It is formed by causing the point of the tongue to tap or vibrate against the gums, once or more, and in highly-developed forms of trilled r, a considerable number of times. This sound is so rare in American speech that it has not seemed necessary to provide a special symbol for it. It may be regarded as a variant form of [r].
§42. [ɹ] is also a voiced sound, but the friction accompanying the vocalic element is so slight that one might hesitate to group it with the fricatives, or with the consonants at all. Its orthographic representation, however, is r, and it is commonly thought of as being a variety of this sound.
In pronouncing [ɹ] the point of the tongue is not tilted as high as in [r], but if it were permitted to touch the roof of the mouth, which it does not do, it would strike the region just back of the upper teeth and in front of the place where the concavity of the roof of the mouth begins. It is commonly heard in American speech before consonants and finally, as in part [pɑːɹt], hard [hɑːɹd], heard [hɚɹd], cord [kɔɹd], fir, fur [fɚɹ], demur [dɪˈmɚɹ], car [kɑːɹ], dinner [ˈdɪnɚɹ], color [ˈkʌlɚɹ], never [ˈnɛvɚɹ], etc. There is less friction in the pronunciation of [ɹ] than of [r], the space between the tongue and the roof of the mouth being greater, and some phoneticians do not recognize any consonantal value for orthographic r before consonants and finally. It is true that [ɹ] is regularly omitted by some speakers, especially in the East and South in America, when it is final or stands before another consonant, the difference between taw and tore, pot and part, so far as there is one with such speakers, being altogether a difference of vowel quality or length. In unstressed position, as in never, the word ends, in this manner of speech, with the vowel [ə], as in [ˈnɛvə]. A word like part consists, in this pronunciation, of only three elements, [p], [ɑː] and [t] giving [pɑːt]. But in all regions of the United States, especially away from the Atlantic seaboard, an orthographic r commonly has phonetic value before consonants and when final. Whether one calls this sound which is heard a consonant or not is of little importance, provided the existence and quality of the sound itself are recognized. Its presence can be easily demonstrated by observing the tongue positions in pronouncing a word like part. This word, in American pronunciation which is not typically Eastern, contains four elements, the first and last being stop consonants, the second and third resulting from a shifting of the tongue from mid to high position accompanied by curving or tilting of the point. In never, when the word ends only in a vowel, as in Eastern American pronunciation, the tongue position at the conclusion of the word is that of [ə], that is mid position, with the point of the tongue touching or on a level with the lower front teeth. With those, however, who are said to pronounce their r's, the word ends with the tongue in high position and the tip of the tongue on a level with the roots of the upper teeth, giving [ˈnɛvɚɹ]. The difference between [r] and [ɹ] may be tested by pronouncing the word never by itself, and then by letting it be followed by rains, as in it never rains. Of course if one has no final r, this would be simply [ɪt ˈnɛvə reːnz]. But if one pronounces final r's, the final consonant of never cannot simply be carried over like a long consonant*, to satisfy the demand for the initial consonant of rains. A slight modification in articulation is observable in pronouncing the two r's, which is adequately represented, however it be named, by the two phonetic symbols [ɹ] and [r].
§43. Some speakers, especially those of an unenergetic habit of enunciation, pronounce [ɹ] for [r] even in the stressed initial position, between vowels, and after consonants. The pronunciation of [r] for [ɹ], that is a strongly fricative consonant finally and before other consonants, as in [ˈnɛvɚr], [pɑːrt], etc., is current in localities, but is not general in standard American English.
§44. Another variety of r is heard, especially in the North Central states and in the Middle West, which is produced by bending back the point of the tongue so far that if it actually came into contact with the roof of the mouth, it would strike about the middle of the hard-palate. This is often spoken of popularly as 'guttural r,' though it would be truer to the facts to call it a hard-palate r, or simply, back r. Dialect story writers usually represent it by doubling the spelling, as in corrn, farrm, etc. The sound is often so marked in the regions in which it occurs as to constitute as distinct a dialect feature as the loss of [ɹ] before consonants and finally is for the Atlantic seaboard. Speakers who have this back [r] are often said to 'roll their r's', though as a matter of fact there is no more rolling or tapping of any of the organs of speech in pronouncing this r than there is in pronouncing the common [r], [ɹ]. It is, however, sometimes prolonged. Englishmen and Eastern Americans often find this sound offensive.
*the text cross-references a different section that talks about "long consonants" in words and phrases like pen-knife, mad dog, egg glass