Issues in Transcription: Phonetics and Phonemics in Alphabets and Writing Systems

Though an (alphabetic/syllabic) written language gives a good idea of the phonology of a language as implicit part of the system, it does not usually accurately nor completely reflect the phonetic system of sounds itself nor the mechanical representation of the sounds uttered by speakers of the language. Linguists have created tools for accurately transcribing the physical utterances since the realization of this fact, from a rudimentary system invented by Bell in the 1860s to the International Phonetic Alphabet that is in common use today. Also included in this paper are psychological reasons for this and examples of written systems invented by various cultures to represent their languages in writing.

Getting a good paradigm for the spoken phonemes in the vocabulary of and in the regular systems of a language is important. In Sapir's case, it was a matter of understanding and describing a language completely in the annals of history before it disappeared. The understanding of a phonology of a language is necessary to non-native speakers of the language, the phonemic paradigms may be critically different in either tongue and result in the second-language learner to pronounce things in an unintelligible 'accent' in the other language. The diachronic study of sounds in a language has been a major focus since the beginnings of linguistics and a historical device in determining language relationships.

Sapir observes in The Psychological Reality of Phonemes a certain phenomenon in his Southern Paiute interpreter: what at first he interprets as an error on the part of the transcription actually reflects a dichotomy between a psychological reality and a physical one; phonetics v. phonemics. While Sapir transcribed the phrase “at the water” as /pá∙βa/, his native speaker syllabified it as /pa∙/-/pa'/ (51). At the time, comments Sapir, “my point of view at that time stressed phonetic accuracy rather than phonemic adequacy” and this explains why he was so stricken with the difference between what was said and what was heard. The style of transcription in which the Paiute speaker wrote is well documented and has a developmental reason for its occurrence (de Gelder 20). Speech is the primary means of language acquisition, and an ear trained to the certain phonemic system of a language also has a mind tuned to it (84-5). The act of writing is secondary to speech (28) and therefore acts as a mirror to this internal phonemic mapping, a /p/→/β/ transformation in a multisyllabic environment will still be written as the base phoneme that was mapped to the character 'p'; to write the mechanical sounds the ear has to be trained to be language-neutral, slightly ignorant of the context of the sounds.

Many languages have adapted a phonemic system of spelling for written language, as opposed to an ideographic, more abstract system. Most widely spoken languages with a literary history use words as a partial description of the sound, with the notable exceptions of the Chinese languages and Japanese (which uses a mixed phonemic/ideographic system, borrowing characters from Chinese). The granularity of each written unit varies from system to system, from syllabaries (Japanese has relatively few syllabic possibilities and can essentially encompass them all in 46 phonetic characters, two voicing markers (ex: ほ/ho/ → ぼ/bo/ and ぽ/po/), and three variant characters for composing glides (ex: き/ki/ → きゃ/kja/]) to component parts, such as the Latin alphabet which allows for complex strings of consonants and vowels or the more fixed Bopomofo, a (recent) phonetic system for writing Chinese words. And in-between systems exist as well, Tamil exists as a set of consonant graphs which change shape/gain markings to indicate the subsequent vowel sound. The Korean Hangul system is composed by arranging smaller consonant and vowel symbols into a single glyph representing a syllable.

The Roman alphabet evolved essentially as a descriptive phonetic system from the Greek; but even in Latin it reflects a slight dichotomy between what is said and what is heard. Only seven vowels (five monophthongs and two diphthongs written as digraph ligatures) are in the Roman alphabet, with no visual indicators for short or long vowels, nor for syllabic stress. For example, the word written as 'puella' could have two meanings depending on the pronunciation of the inflection: with a short /ə/, puella is the singular nominative form of the noun. However, with a long /ɑ/, puellā becomes the dative form of the noun. Languages that uses the Roman alphabet have adapted it to make up for the shortcomings when used in its own phonetic system. Such changes include accents, diacritic and nasalization marks and umlauts, as well as the creation of new glyphs (j as a variant form of i, G as a variant of C, u and v splitting into two distinctive graphs, w for the English semivowel, the Catalan ŀl, etc), or using provided characters for new purposes (Spanish ll to represent /ʎ/). Ultimately, though, the system has not been completely phonetically descriptive in any language (Castillan Spanish uses s for /s/ and /θ/, which are typological differences of an identical phoneme; variant forms of each other; compare spelled American English butter to /bʌdɚ/ instead of a more phonetically faithful /bʊtʼeɹ/, pronunciation of (bad, bade) to (/bæd/, /beːd/) ).

Note that the ideographic Chinese languages have a slight tendency to make homophones of a certain high frequency of use, visually similar characters (de Gelder 148), and usually share radicals (ex: 芳 and 方, both pronounced as fong in Mandarin, mean “aromatic” and “square,” respectively. 149) that suggests a tendency to slight clues as to a word's pronunciation, sort of a hidden, indirect phonetic 'spelling'. As all these adaptations and tendency from a purely phonetic system show that the Roman alphabet in its original form is inappropriate for describing mechanical sounds and extending beyond its original phonetic sounds, but keep in mind that the system is open and extensible.

In 1876 Alexander Melville Bell published a proposed system for descriptive phonetics in his Visible Speech. It was an interesting system, if nothing else, using symbols to represent the physical shape and articulatory details of a sound (as a site note, the system's constructs and visual style are coincidentally quite similar to Tamil).

Bell, at the time, tried to make his system as universal as possible. There are several chapters trying to reconcile the use of clicks in African languages and the tonal qualities of Chinese syllables with additional diacritic marks (Bell 70-2). In many aspects this system is very complete and addresses many of the phonetic issues the IPA also does. However, the points of articulation are fixed in the visual speech. The languages to be documented in the coming decades do not fit in the glyphs in many situations. Each glyph was attributive, a combination of visual cues (Bell 50) for the articulation of a sound (voicing, aspiration, tongue position, etc). The system itself requires a fair amount of training to begin reading functionally as well, compare IPA /vɪʒjʊəl/ to Visual Speech /31M1BWE/ for the word visual. This effort, which is an impressive piece of work for a century and a half old piece of descriptive linguistics, still falls short and has never been in common usage.

The International Phonetic Alphabet is developed by the International Phonetic Association, an organization that has been around since 1886. It is based on the Roman alphabet, making it more familiar to most linguists (IPA 20). The introduction of a unique symbol for each possible manufactured sound prevents the IPA from falling into the same trap Visual Speech did: the attributes are not fixed in the glyph, so to extend the IPA for an as-of-yet unknown phoneme all that needs be done is have a new arbitrary glyph added to the alphabet. The modifying markers and diacritics in IPA are optional, allowing for a 'narrowed' transcription ( English /haʊ ar juː/ does not need the same diacritics as Cantonese /jɐu|̷ jɐt˥ tsʰi˧/ (IPA 60)). Edward Sapir uses a version of the IPA for transcribing Paiute in Reality, as well as his other Native American dictionaries.

Computers have no problem with phonetics, understanding each phoneme as its true form regardless of the relationships between them in a language. Computers as a whole tend to have a different set of issues in transcription (Henisz 25). The pitch of the speaker's voice is a larger obstacle to its transcribing accurate phonemes than the system of speech or its 'understanding' of the language. There exist problems with many applications and their approach to parsing audio, the acoustic environment of a speaker must be much less noisy for them to be recognized than a human listener needs. Rudimentary speech-to-text software exists in commercial software today, and is functional in limited environments according to acoustic noise and the speaker's voice and speech patterns. The need to understand a sound system in a language before moving on to morphology, syntax and other larger issues has resulted in the invention of systems for representing the physical sounds represented and more than the existent writing systems provided. Interestingly enough, the prevalent system is merely an extension of an existing one which has been extended in many ways since its inception; the Roman alphabet.


Bell, Alexander Melville. Visible Speech: The Science of Universal Alphabetics. N. Trübner Co.: New York. 1867.

De Gelder, Beatrice, and José Morais, editors. Speech and Reading: A Comparative Approach.

Henisz-Dosert, Bożena, R. Ross Macdonald and Michael Zarechnak. Machine Translation. Mouton: New York. 1979.

Handbook of the International Phonetic Association. Cambridge Press: Cambridge, UK. 1999.

Ladefoged, Peter. A Course in Phonetics. Harcourt Brace Jovanovich: San Diego, CA. 1982.

Pike, Kenneth L. Phonemics: A Technique for Reducing Languages to Writing. Ann Arbor, MI: U. of Michigan Press. 1954.

Sapir, Edward. “La Réalité Psycholique Des Phonèmes,” Journal de Psychologie Normale et Pathologique. 30:247-265. 1933.