Rules for calculating MLU The text below sums up the procedure for calculating Mean Length of Utterance (in
morphemes, for English) as proposed by R. Brown (1973) A First Language: The Early Stages.
Cambridge, MA: Harvard University Press, p. 54:
Start with the second page of the transcription unless that page involves a recitation of
some kind. In this latter case, start with the first recitation free stretch. Count the first
100 utterances satisfying the following rules.
Only fully transcribed utterances are used; none with blanks. Portions of utterances,
entered in parentheses to indicate doubtful transcription, are used.
Include all exact utterance repetitions (marked with a plus sign in records). Stuttering is
marked as repeated efforts at a single word; count the word once in the most complete
form produced. In the few cases where a word is produced for emphasis or the like (no,
no, no) count each occurrence.
Do not count such fullers as mm or oh, but do count no, yeah, and hi.
All compound words (two or more free morphemes), proper names, and riualized
reduplications coun as single words. Examples: birthday, rakety-booom, choo-choo,
quack-quack, night-night, pocketbook, seesaw. Justification is that there is no evidence
that the constitutent morphemes functions as such for these children.
Count as one morpheme all irregular pasts of the verb (got, did, went, saw). Justification
is that there is no evidence that the child relates these to present forms.
Count as one morpheme all diminutives (doggie, mommie) because these children at
least to do not seem to use the suffix productively. Diminutives are the standard forms
used by the child.
Count as separate morphemes all auxiliaries (is, have, will, can, must, would). Also all
catenatives: gonna, wanna, hafta. These latter counted as single morphemes rather
than as going to or want to because evidence is that they function so for the children.
Count as separate morphemes all inflections, for example, possessive [s], plural [s],
third person singular [s], regular past [d], and progressive [ing].
The range count follows the above rules but is always calculated for the total
transcription rather than for 100 utterances.
|