Kenneth J. Logan, Ph.D., CCC-SLP
Dept. of Communication Sciences and Disorders

1. Write down everything that each speaker says. Before conducting a speech or language analysis, it is helpful to first transcribe what your subject has said. It is also quite helpful to transcribe the speech of your subject’s conversational partner as well. In that way, you will have a clearer sense of the context within which particular remarks were made. For the most part, your speech or language analysis can then be based upon the written transcript and you will not have to repeatedly return the audio or video tape.
 
 

2. Organize the transcript according to utterances. For our purposes, an utterance will be defined as a string of words or clauses that (a) communicates an idea, (b) is set apart by pauses, and (c) is bound by a single intonational contour (Meyers & Freeman, 1985). In many cases an utterance will simply be a sentence. For example, each of the two strings of italicized words below would be regarded as an utterance.
 
 

Mother: What is that guy called? [One utterance with 5 words in it.]

Child: He is an alien. [One utterance with 4 words in it.]
 
 

3. Reference utterances within the transcript. Because you may want to make reference to specific utterances in your analysis, you should "identify" each utterance in some way. Because you generally will need to elicit between 50 and 100 utterances from a child, it is probably best to reference the child’s utterances separately from the mother’s (or other partner’s) utterances. Keep your referencing system simple. The most common way to reference utterances in a transcript is to assign a letter to each speaker (e.g., M = mother; C = child) and then follow the letter with number that identifies the particular utterance. For instance,
 
 

M1: What is that guy called?

C1: He is an alien.

M2: Oh, I thought he was a soldier.

C2: No, he’s really an alien.
 
 

4. Don’t confuse "utterance" with "speaking turn." In the examples above, both the mother’s and child’s speaking turns consist of only one utterance. Often, however, a speaker will produce two or more utterances within a speaking turn. When this happens, put each separate utterance on it’s own line. For instance,
 
 

M1: He’s climbing way up.

C1: He’s probably dizzy.

C2: Let’s pretend that he is dizzy.

C3: I think he needs a doctor.

M2: Who should be the doctor?
 
 

5. "Utterance" is not synonymous with "sentence." Although many utterances in your transcript will be "complete sentences" (i.e., they will contain a noun and a verb in a subject-predicate relationship), not all utterances are "complete sentences." In some cases, a speaker will delete redundant information in his or her speech. The process of deleting such information is called ellipsis. Consider the following example (C25) of an elliptical utterance produced by a child:
 
 

M25: When do you think he will be ready?

C25: In a little while. [pause]

C26: Hey, let’s get this guy ready too!

In this case, the child’s string of words (In a little while) does not contain a verb or a subject, yet the string of words is entirely appropriate in the context. In fact, it would probably seem somewhat stilted if the child had said, "I think he will be ready in a while." Further, this particular string of words does meet our criteria for an utterance. That is, it (a) communicates an idea, (b) is set apart by a pause from the child’s next utterance, and (c) it is bound by an intonational contour (i.e., in a while would likely have a falling intonation pattern associated with it and the following utterance (Hey, ...) would start out with a different intonational pattern.
 
 

6. Utterances should be transcribed verbatim. Speakers often produce sounds, words and phrases that ultimately have nothing to do with the final meaning of an utterance. Such vocalizations are called disfluencies. Consider the following examples:
 
 

C34: [This- ]this guy lives over here.

C45: Mom, [can we ]can we go home yet?

C52: [I want-] She’s going to go up in the plane.

C58: [W-w-]we like to see grandpa.
 
 

In example C34, the child repeats the word this; in C45, he repeats the words can we; and in C58, he repeats the consonant [w]. In C52, the child has apparently "abandoned" his initial attempt at the utterance. None of the repetitions or abandoned segments add meaning to the sentence, and so they are regarded as instances of disfluency. Such instances should be separated from the rest of the sentence with square brackets. While words, phrases, etc. within the brackets are not typically included in measures of speech or language complexity (e.g., number of syllables or words per utterance), it may be interesting to report how many disfluencies occur within a particular sample.
 
 

7. Use standard notation to transcribe speech disfluencies. A carefully developed written transcript can provide a lot of information about a child’s speech fluency. Generally speaking, it is best to use standard notation that captures the child’s disfluent speech behavior. In other words, try to record all instances of sound prolongation, sound or word repetition, revision, etc., in a similar manner. In cases of repetition, try to capture how many times the sound, syllable, or word has been repeated. In cases of prolongation, try to roughly indicate the length of the prolongation but typing either a few (for a brief prolongation) or many (for a long prolonation) characters. Examples are given below.

Utterance Comments

• [um, um]he went shopping Child produces interjection "um" two times
• [My unc-/]My aunt lives here. Child abandons original attempt, cutting off what seems to be the word "uncle" and then "repairing" the error.
• [She eats]She eats candy. Child says the words "she eats" then restarts the sentence
• [S-S-]Sam bought a bike. Child says /s/, then stops and restarts again, before finally saying the entire word
• [SSSS]Steve bought a car. Child prolongs the /s/ in Steve for longer than what’ expected. (4 /s/’s tell us the prolongation was fairly lengthy).
• [a-----]I like to skydive. Child prolongs the posture associated with /ai/ in such a manner that no sound is forthcoming
• [I, I, I]I like that hat. Child say the word "I" then stops and restarts the utterance. This behavior is done three times before the sentence is finally produced fluently.
• [Be-be-because]she can’t go. Child produces the first syllable in "because," then stops and restarts the words. This is done twice before the word is spoken fluently.
 
 

[I, I]You can’t [d-]do that. One revision, one sound repetition.
 
 

In some cases, speakers will embed one disfluency within another. In such cases, bracket all disfluencies together (i.e., as one disfluency) and make note that several disfluencies are clustered together. Examples are listed below.
 
 

Utterance Comments

[She/he, he, h-he,]you can’t. "she and "he" both subsequently revised, sound/syll repetitions on "he"
[yyyyou]you can’t go there. Sound prolongation occurs on syllable that is repeated.
 

7. "And then, and then, and then": Parsing long sentences. Young children often rely upon the conjunction and to link together sentences. In some cases, as many four or five sentences may be linked together at once. For example,
 
 

C61: We went to McDonald’s and we had a Happy Meal.

Beyond the conjunction and, however, all multi-clausal productions that meet the utterance criteria specified above should be treated as one clause. For example,
 
 

9. Selecting a starting point for your transcript. Small children often take a few minutes to "warm up" when they are being observed. Thus, the first few minutes of any interaction may consist of numerous one word utterances (e.g, "Yep," "Nope" and the like). Accordingly, it is generally wise to start your transcript about 5 or 10 minutes into the audio or video tape. In this way, you are more likely to end up with a representative sample of the child’s speech and language abilities.
 
 

10. Unintelligible speech: Sometimes the child’s speech will be hard to understand due to background noise, imprecise articulation on the part of the child, etc. In such cases, mark unintelligible speech with a string of two capital X’s (e.g., XX). It is usually possible to hear how many syllables the child was attempting to say. Thus, if possible try to mark how may syllables were unintelligible. For example:
 
 

Intended utterance: He’s a good baseball player.

What you heard: He’s a good ????????????

Transcribe: He’s a good XX XX XX XX
 
 

Although unintelligible utterances should be included in your transcript, do not attempt analyze the grammatical, semantic, pragmatic, or speech fluency properties of such utterances. If the unintelligibility stems from the child’s speech or language skills (and not from background noise), you may choose to report statistics such as "the number of utterances containing unintelligible speech" or "the percent of unintelligible syllables" in your final report. Such measures give additional clues to the examiner about the child’s speech and language development.
 
 

11. Errors in articulation/phonology. Young children do not always pronounce all attempted words in an adult-like manner. Although information about the child’s phonology is certainly important, it is not generally necessary to include such information in your transcript. Instead, write down the adult gloss of child’s attempted word. For example,
 
 

Child says: He "cwimbed" way up the "twee" (child substitutes /w/ for /l, r/)

Child says: He has two car. (Child deletes plural /s/ ending)

Your transcribe: He has two car. (sic)
 
 

Child says: She swimming in the pool. (Child deletes auxiliary verb is)

Your transcribe: She swimming in the pool. (sic)
 
 

It is advisable to follow such grammatical, morphological, and lexical errors with the Latin expression sic. In doing so, you tell the reader that the omitted or incorrect word on the transcript was produced by the speaker – and not by the typist. Notation such as this is invaluable in cases when you might be analyzing a transcript over the course of several days and can’t recall exactly what the child said.