Heiner Kleine


Description and Statistical Analysis of the

Acquisition of German Phonology. A Case Report



This is a shortened version of the paper. If you are interested in receiving the full text with the complete sets of rules and all tables and figures, please send an inquiry via the Kleine Language Services start page to let me know. The paper is also available in German language.























Author's address

Heiner Kleine

Maarbachstr. 107

53347 Alfter


The present case report describes the acquisition of phonology by two children, who were closely followed for about one year after their "first word." For two defined times, the phonologic data are translated into rules on the basis of both the German standard and the distinctive features. The first type of rules provides a complete description of the children's phonological systems by tagms. The second set of rules, besides being a description by features, is used for statistical evaluations to assess the differences and similarities between the two children's phonological systems. The rules are further analyzed to determine the effort necessary for each child to approach the German standard system and to single out the features that are not constitutive for the structure of the children's systems. The analysis of the differences in phonological development shows that two different strategies of acquisition are likely to be responsible for many of those differences.



Language acquisition, phonology, distinctive features, statistical evaluation

1     Introduction


Weighty—not only by number—are studies of babbling and infants' earliest speech production. But however important these publications may be, in particular with reference to the development of language acquisition models (e.g., WRAPSA in Jusczyk 1993; investigations of language acquisition in Optimality Theory (OT) in Kager 2004), a review of the literature shows that there is only a relatively small number of studies that deal with the development of phonological systems as the lexicon is growing; the publications by Vihman (e.g. Vihman et al. 1985) need to be mentioned in this category. If the material treating phonological systems is fairly scarce, there is virtually none for those interested not only in finding information on phonological systems, but also in getting an overall view of the phonotactic developments and inventories in the course of language acquisition.

The present case study of two children was therefore conducted as a continuous registration of phonological elements and phonological change over quite a long period of observation so as to allow interindividual and diachronic comparisons that would reveal common and idiosyncratic features of phonology and phonotactics.

2     Subjects

2.1     Demographic data

The two subjects were N., female, born […], and A., male, born [….] Either child was born at term in the 42nd week of gestation after a normal course of pregnancy. Neither child presents with sensory, in particular auditory, motor or cognitive impairments.

2.2     Speakers in the children's environment[1]


3     Methods

3.1     Data collection

[…] the linguistic data were recorded from the first utterance that was definitely meaningful, which means that the examiner could trace it back to its original situational context. This criterion was used to prevent the evaluation from being distorted by words of unclear meaning that did not recur and therefore could not reliably be related to their phonetic source, i.e. the standard-system word imitated by the child.[2] Likewise excluded were protoverbal utterances of affect (Vihman and Miller 1988), such as were used, at 10½ months, by both N. (as [gai]) and A. (in the similar forms of [g̊ai] or [g̊aʲg̊]) to express pleasant sensations.

All spontaneous utterances were immediately analyzed for agreement with earlier utterances. When an utterance was found to be the first instance of a new word and met the aforesaid criteria, it was transcribed according to IPA; the same was done for second or any later instances of words if they showed phonetic deviations from earlier uses that exceeded the usual degree of variability.[3] Contrary to, e.g., Ferguson and Farwell (1975), nonspontaneous utterances, e.g., the resumption of words of a question in the child's answer, were never included in the evaluation to eliminate any bias by immediate imitation to the best possible extent.


N.'s first clearly identifiable word developed from [ai] and [hai] to [hais] "heiß" 'hot' within few days in late September / early October 1996, i.e., when she was almost 12 months old. A.'s first word was onomatopoetic: [ːt] "heruntergefallen" (from bumm 'boom') appeared when he was a little over 11 months old.  His first non-onomatopoetic word was [ːə] "Tür" 'door' and emerged at the age of somewhat over 13½ months.

For both children, the period of observation ended with the first word whose use could not be traced back to a specific situation or did otherwise not seem clearly motivated to the examiner, since this was the time when it was no longer possible to determine whether a word was used for the first time or might have been used earlier in some as then unidentifiable form and when, consequently, the chronology of language acquisition began to become inextricable.

The period of observation so delimited lasted from early October 1996 to early August 1997 for N. and yielded 280 words; for A., it lasted from mid-June 1999 to mid-April 2000 and yielded 357 words. All at first "unidentifiable referents" (Vihman, Macken, Miller et al., 1985) could eventually be identified and are thus included in these figures. The sizes of the subjects' lexicons correspond to the values found by other researchers.[4]

3.2     Evaluation of the data

The data were evaluated on two different times. The first time of evaluation was the 50-word stage, which occurred in either child at the age of 17 months, 1 week, i.e. in early March 1997 for N. and in early December 1999 for A. The reason for using a lexical threshold as a cut in a phonological study was that this time was found to be characterized not only by a marked increase in the rate of expansion of the lexicon, but also in the rate of approximation to the phonetic standard. […]


The second evaluation was done at the end of the data collection and was based on the whole material the child used at that time.

A specific problem resulting from small lexicons has to be considered in particular with reference to the first evaluation. If, as commonly, a phoneme is defined as the smallest unit allowing the differentiation of meaning, the question arises how phonemes can be delimited in extremely small lexicons. The substitution test proves impracticable in such a case since there may often not be two words differing by one sound only. […]

In the present study, phonemes were therefore derived from word-specific variants or phone clusters [5]: If, for instance, there was free variation in a word between [e], [i] and the phones between them, and this phenomenon was also observed in other words the child used, while other sounds were not included in this variation, the assumption of a phoneme |i‑e| would be well-founded, even though no strict demonstration by the usual tests were possible.[6] For lack of a better procedure, the existence of a "broad" phoneme, which might be represented by |ɪ|, would therefore be assumed in this case. The transcription depends upon the case; the phones covered by a specific phoneme result from the distinctive features used in the rules.

3.3     Description of the data

3.3.1     Phonological systems at the first time of evaluation (17 months)

The description of the phonological systems that the two children used at the age of 17 months begins with a comparison with Standard German, which means that for either child information is given which word-onset consonants, word-coda consonants, intervocalic word-internal consonants (if existent) and vowels were present in the language and how they were correlated with the standard language. This data is then taken as a basis to formulate the rules of phonotagmic substitution that a child used consistently. Finally, a comprehensive description of the two children's phonological systems observed at the first time of evaluation is given on the basis of feature-based rules.

3.3.2     Phonological systems at the second time of evaluation (end of study)

The situation at the end of the study is described by first listing N.'s and A.'s word onsets, word codas, intervocalic consonants, and nuclei as functions of the Standard German phonotagms. The rules given in these listings offer a comprehensive survey of all phonotagms of the standard language and their equivalents, if realized, in the subjects' systems. The phonological systems are then again described by feature-based rules.

3.3.3     Phonological distances

The feature-based rules are further used to calculate phonological distances between the children's systems and the standard language. The distances serve, on the one hand, to measure the phonological development; this is done by comparing the distances of the overall systems and of the subsystems of the first and second times of evaluation. On the other hand, the distances are used for a synchronous, interpersonal comparison, which reveals differences of language development.

3.3.4     Legend

[ ]         Phone, distinctive feature           | |          Phoneme

[ ]         Morph                                      Ɨ           Morpheme boundary (by substitution: +)

[ ]         Lex                                          ǂ          Lexeme boundary (by substitution: )

[ ]        Phonotagm                               / /         Phonemotagm


R          reference system

[a]         earlier tagm that had been abandoned by the time of examination


phonological system used by N. or A. at the first time of evaluation

phonological system used by N. or A. at the second time of evaluation


used for the first time within two weeks of the time of evaluation (of rule 1 / by subject X)

=          concatenated to

used for the first time within two months of the time of evaluation (of rule 1 / by subject X)

         originating in

         is replaced by, yields

a ↦ b    a replaced by b within the following two months

(V)         vowel

(C)         consonant

          tagms / realizations used in free variation

      no less than x consonants, e.g.,

      no more than y consonants, e.g.,

Ø         element inexistent

      between x and y consonants inclusive, e.g.,

?        element not observed

(U)       environment

      exactly x consonants, e.g.,

/ ___ A            before A

/ A ___                                                           after A

/ A ___ B                                               between A and B

/(U1)/   phonotactic environment 1, i.e. Ɨ

/(U2)/   phonotactic environment 2, i.e. Ɨ

{[a] : [ b]  /___ /(U1)/ : /(U2)/}       phonotagm [a] is used in /(U1)/, phonotagm [ b] is used in /(U2)/

4     Results

4.1     Phonological systems at the first time of evaluation (17 months)

4.1.1     N.     Word onset

At 17 months, N. used the following word-onset inventory:


[ p]

[ b]

[ t]

[ d]




[ f] ([ɸ] [w̥] [hw̥])

[ hw]

[ s]

[ z]





[ m]


[ n]






[ h]


A striking feature of N.'s word-onset system is its high degree of regularity. Except for [h], all her word-onset consonants at this age were characterized by being [-compact]. According to the feature [acute], they fell into two groups, which corresponded to two different places of articulation: labial and alveolar. No velar consonants existed at this time.


The diphonic word-onset tagm [ts] occurred in the word [tsʊ̥] "zu" from age 13½ months, but was abandoned at the age of about 15 months.

[…]     Word coda

Immediately before the word boundary, N. at the age of 17 months used only two phonotagms, [f]—also realized as [ɸ]—and [s]; the word-coda inventory was completed by [j] and [w], which occurred before vowels only. [f] and [s] were not, however, distributed according to the standard pattern: there was rather a set of specific rules to govern substitution in function of the preceding nucleus. The following rules can be derived from the paradigms:


[…] Standard German [f], [s] and—generally speaking—polyphonic word-coda tagms with |s| as the final phoneme are replaced by [f] when they follow [a] or [u]. The coda tagms of the other Standard German words that N. used at the time of the first evaluation, i.e. [d], [l], [s] and [ʃ], and the syllable den of baden 'take a bath,' were replaced by [s], if necessary with the insertion of [ɪ] [….] The determinant was the nucleus, as can be shown by [buf] "Bus." The analysis of the rules under the aspect of distinctive features shows that the value of the feature [acute] of the word-coda consonant always agreed with the value of that feature in the preceding nucleus.

That these rules are no artifacts resulting from N.'s lexicon, is verified by rule 2.2, which was first used for substitutions after [a] and, at the age of nearly 20 months, also spread to the environment [u]__, where it was to be effective for several weeks: [f] in Bus and other words had by that age been corrected to result in the coda [s], but [buf] had now turned into [buɪs]. This means that the assimilation of vowel and consonant tagms according to the value of [acute] was not changed by rule 2.2, but the nucleus ceased to be the determinant and was replaced in that function by the coda.[7]

[…]     Nucleus and intervocalic word-internal consonants

The method described above of delimiting phone clusters allows the following vowel tagms to be identified in N.'s language on the first evaluation:

·        With [-long]: ɪ], [a], [ɔ], and [u].[8] [e] was used, but was not reliably separated from [ɪ] so that a phoneme |ɪ| extending from [i] through [ɪ] to [e] was stipulated.

·        With [+long]: ɪː], [æː], and [aː]. N. had a tendency of realizing ɪː] as a diphonic tagm, i.e. [ei].

·        Diphonic nuclei: [] and [au]; lengthening of the first element was possible.

The analysis of the vowel tagms by their source forms in the standard language shows that phonotagms that contained no elements with the feature value [+acute] always originated in tagms that differed from N.'s phonotagms only in the value of [open]. This means that under the aspect of N.'s phonological system, the standard was imported largely one to one.

The situation was different for vowel tagms that contained an element with the value [+acute]. [+acute] seems to have inhibited the differentiation of the two values of [compact] so that |ɪ| and |ɪː| represented very "wide" phonemes, with the long nucleus even including the diphonic realization [ei]—under the aspect of N.'s phonological system, the latter can be seen as a glide covering the whole range of the phoneme.

[…]     Description of the phonological system by a set of rules

The following rules comprehensively describe N.'s phonological system at the age of 17 months:

Word onset

Number of elements

N1-01 = / ¯¯/ / __ [-cns]

Inhibition and selection


N1-03 ® Ø / ___ [-cns]

N1-04 ® Ø / ___ [-cns]

N1-05 ® Ø / ___ [-cns]

N1-06 ® Ø / ǂ ___


N1-07 [+cns] + Ø =




Word coda

Number of elements

N1-10 = / ¯¯/ / [-cns] ___

Inhibition and selection


N1-12 ® Ø /[-cns] ___

N1-13 ® Ø /[-cns] ___

N1-14 ® Ø /[-cns] ___

N1-15 ® Ø /[-cns] ___

N1-16 ® Ø / ___ ǂ


N1-17 [+cns] + Ø =




Number of elements

N1-19 = / ¯¯¯/ /

Inhibition and selection


N1-21 ® Ø 

N1-22 ® Ø

N1-23 ® Ø

N1-24 ® Ø


N1-25 [-cns] + Ø =





N1-29 / ___ [-cns]


4.1.2     A.     Word onset

A., when 17 months old, used the following inventory of word-onset tagms:


[ p]

[ b]

[ t]

[ d]

[ k]

[ g]


[ ts]



[ m]


[ n]








[ h]


Except for the diphonic tagm [ts], the system was very regular, and except [h], all word-onset tagms were either [+cns, +abr] or [+cns, +voc], i.e., the fricatives of the standard were almost completely absent.


The analysis of the Standard German sources of A.'s word-onset tagms revealed that in almost all cases, his [‑abrupt] onset tagms—including diphonic [ts], where this feature value was present in the second element—originated in the same sounds of the standard.


[…] Particularly noteworthy is the word-onset tagm [g], which never derived from Standard German [g], but from [ʀ], [d] or [ʃn]. Words with Standard German [g] did not exist in A.'s lexicon at the time of evaluation.

The word onset Ø always originated in Ø, but about half of the cases with a Standard German Ø onset were supplemented with a word-onset consonant.     Word coda

At the age of 17 months, A. used fairly many word-coda tagms, which had no regular relation to the source tagms of the standard. The following word-coda tagms were observed (the tagms marked with an asterisk occurred for the first time shortly after the evaluation):



[ b]

[ t]

[ d] *

[ k]




[ pf]


[ f]

[ w]

[ s]


[ x]

[ j]







[ n]










[ ʀ] *








[ h]


The word-coda tagms obviously formed no coherent system at this time. The analysis of the underlying tagms of the standard language rather suggests that A.'s inventory of word-coda tagms was augmented each time a newly acquired word made this necessary. Diphonic word-coda tagms were often rendered by their abrupt element: [k], for instance, was used for [ks]. The value of [voiced] was disregarded in the process, as is shown by [k] from [ŋk] and [b] from [mp] ([gákèː] "danke" 'thank you,' [ː̀ː] "Lampe"); since the coda tagm was followed by a vowel in both cases, "final devoicing" can be ruled out as an explanation of the phenomenon.

The diphonic word-coda tagm [lt] was, however, not replaced by [t], but by [j]—as was monophonic [l], which was replaced by [w] in another case, though. As [x] could also originate in [t] ([bɔx], later [bʶɔx] "Brot" 'bread') and [k] could originate in [d] ([gákɔ̀ː] "Radio")—although this was an instance of a word-internal consonant—it becomes evident that the approach of tracing A.'s word-coda tagms back to their Standard German sources is not very fruitful.


[…] [s] is maintained when it follows [ɪ] after any substitution concerning vowel tagms has taken place. There is further the substitution of [s] by [f] after [u] (in [hauf] "aus" 'out') [….] The rule [-cns] [‑cns]+[ɪ] / __ [s] that emerged in the further course of language development suggests, however, that this example was not accidental, but that [s] indeed required a preceding vowel with the feature value [+acute] or—to put it differently—that [s] could not occur after a [‑acute] vowel. When A. was 17 months old, his lexicon included no words in which [s] occurred after other nuclei than those mentioned above.

[…]     Nucleus and intervocalic word-internal consonants

The following tagms were identified in A.'s nuclei:

·        Monophonic tagms: [ɪ], [e], [a], [ɔ], and [u]. There had been no clear separation between [ɪ] and [e], on the one hand, and [ɔ] and [u], on the other hand for a fairly long time, but by the time of the first evaluation, the separation had become stable. At the very beginning of language acquisition, [ʏː] (in [ːə] "Tür" 'door') had also been used, but was abandoned before long.

·        Diphonic tagms: [aɪ] and [au].

The value of the feature [long] of monophonic vowel tagms was in general dependent on whether or not the syllable was closed, i.e., the vowel was short in closed and long in open syllables. The exceptions to this rule cannot be systematized phonologically, but were largely determined by situational factors. The urgent request, for instance, to be lifted could be realized as [ːx] 'up,' in contrast to usual [hɔx]. The feature [long] therefore lacked phonematic relevance.

[…]     Description of the phonological system by a set of rules


4.2     Phonological systems at the second time of evaluation

4.2.1     Preliminary notes

The first part of this chapter deals with the rules that the children used to substitute their tagms for the standard tagms. It is a survey of the complete inventory of N.'s and A.'s word-onset, word-coda and intervocalic word-internal consonants and their nuclei.


The rule numbers and tagms are followed by a code of use: A stroke marks the absence of a tagm, while the codes N and A show that the tagm was present in N.'s and A.'s systems, resp. Since […] the time of evaluation was chosen for reasons outside the phonological system and is therefore arbitrary under the aspect of phonology, the tagms registered for the first time within two weeks and two months of the time of evaluation are also marked with the code, which is then supplemented with a single and double asterisk, resp. Such later substitution rules will, however, be disregarded for the formulation of the rules of the phonological systems.

A lean-faced code, followed by a number, means that the tagm was realized in the same way as in the standard and that this is evidenced by the given number of paradigms. If secondary forms existed for any of the paradigms or if other notes concerning the rules are necessary, they will be given in footnotes.

When substitutions by other than the Standard German tagms were observed, N or A is set in boldface; this refers to detailed information in the sections 4.2.2 and 4.2.3, where complete words can also be found.

Word onset (cf., p. 12 and, p. 15)

1. [ b]            N         A, 4

2. [ d]            N, 3     A

3. [ f]             N, 3     A, 5

4. [ g]            N, 3     A

5. [ h]            N, 4     A, 6

6. [ j]             N, 3     A

7. [ k]            N, 3     A, 6

8. [ l]             N         A

9. [ m]           N, 3     A, 5

10. [ n]          N, 2     A, 6

11. [ p]          N         A

12/13. [ ʀ]     N         A, 4

14. [ z]          N         A, 5

15. [ ʃ]           N         A

16. [ t]           N, 4     A

17. [ v]          N         A

18. [ pf]         N         A

19. [ sk]         -           -

20. [ bl]         N         A, 3

21. [ fl]          N         A

22. [ gl]         N         A

23. [ kl]          N         A

24. [ pl]         N         A, 1

25. [ ʃl]           N         A*

26. [ ʃm]         N         A

27. [ gn]     -           -

28. [ kn]        N         A

29. [ ʃn]         N*        A

30. [ ʃp]         N         A

31. [ bʀ]        N         A

32. [ dʀ]        N         A

33. [ fʀ]         N         A**, 1

34. [ gʀ]        N         A

35. [ kʀ]         N         A, 2

36. [ pʀ]        -           -

37. [ ʃʀ]          N*        A

38. [ tʀ]         N         A

39. [ vʀ]         -           -

40-48 = 31-39           

49. [ ts]         N         A, 4

50. [ tʃ]          N         A

51. [ ʃt]          N         A

52. [ kv]         N**       -

53. [ ʃv]          N         A

54. [ pfl]        N*        A**

55. [ skl]        -           -

56. [ ʃpl]         -           -

57. [ pfʀ]       -           -

58. [ ʃpʀ]        N*        A

59. [ ʃtʀ]         N         A

60-62 = 57-59         

63. [ tsv]        N*        A


Word coda (cf., p. 14 and, p. 15)


Nucleus and intervocal word-internal consonants (cf., p. 14 and, p. 15)



Intervocalic word-internal consonants


4.2.2     N. (age 22 months)     General information

Deviations from the standard were found for the following substitution rules that N. used at the age of 22 months. Paradigms for which more than one rule is applicable are italicized and cross-referenced to the other rule(s). As in the preceding list, the rules and paradigms that were added within the first two weeks of the time of evaluation are marked with one asterisk, while those added within the period from two weeks to two months after the evaluation are marked with two asterisks. The arrow between two pronunciations of one word means that the second pronunciation replaced the first one within the two months following the evaluation. A question mark ? in a rule with an environment specifies a tagm that was not observed in the environment thus marked. Two opposing arrows indicate that the respective tagms occurred in free variation.     Word onset

1.1       [ b] [ b]         Paradigms: backen, Bauch, bitte.

1.2       [ b] [ h]         Paradigm: Bäcker.

8.1       [ l] [ z]           Paradigms: lacht (cf. 8.4), Lampe, langsam [zaŋ], laut.

8.2       [ l] [ l]            Paradigm: Loch.

8.3       [ l] [ v]           Paradigm: Löffel.

8.4       [ l] [ j]            Paradigm: lacht [jax] (cf. 8.1).

11.1     [ p] [ p]         Paradigms: packen, Post.

11.2     [ p] [ b̥]         Paradigms: piekst [b̥̥eis], putzen.

12/13.  [ ʀ] [ h]         Paradigms: Rassel, Regen, rennen, runter.

14.       [ z] [ s]          Paradigms: sauber, sechs, Sonne.[9]

15.1     [ ʃ] [ s]           Paradigms: Schaukel, Schoß.

15.2     [ ʃ] [ ʃˢ] [10]       Paradigms: chic, scharf, Schiff, Schild.

17.1     [ v] [ v] [11]      Paradigms: warm, warten, was, Wäsche.

17.2     [ v] [ b]         Paradigm: Wasser.

18.       [ pf] [ f]         Paradigm: Pferd.

20.       [ bl] [ b]        Paradigm: Blume.

21.       [ fl] [ f]          Paradigms: Fläschchen, Fleisch, fliegen.

22.       [ gl] [ g]        Paradigm: gleich.

23.       [ kl] [ k]         Paradigms: klein, Klötzchen.

24.       [ pl] [ p]        Paradigms: Platz, Plätzchen.

25.1     [ ʃl] [ ʃˢ]          Paradigm: schlafen[12].

25.2     [ ʃl] [ z]          Paradigm: Schlüssel.

26.1     [ ʃm] [ p]        Paradigm: schmutzig[13].

26.2*    [ ʃm] [ m]       Paradigms: schmeckt, Schmetterling [mektaiŋ̩] [meta.liŋ], schmusen.

28.       [ kn] [ ŋ]       Paradigm: Knopf.

29.1*    [ ʃn] [ ʃˢ]         Paradigm: schneiden [ʃˢain̩] [hjain̩].

29.2**  [ ʃn] [ n]        Paradigm: Schnupfen.

30.       [ ʃp] [ p] [14]     Paradigms: Spiegel, spülen.

31.       [ ] [ b]       Paradigms: Brille, Brot.

32.       [ ] [ d]       Paradigms: dran, drauf, drei.

33.       [ ] [ f]         Paradigm: Frau.

34.       [ ] [ g]       Paradigm: groß.

35.       [ ] [ k]        Paradigm: Krümel.

37.*      [ ʃʀ] [ ʃˢ]         Paradigms: Hubschrauber [ʃˢaube], Schrank.

38.       [ ] [ t]         Paradigms: tragen, trinken, trocken.

49.1     [ ts] [ ts]        Paradigms: anziehen, zu.

49.2     [ ts] [ t]         Paradigms: Zaun, Zug.

50.       [ ] [ t]          Paradigm: tschüs [tis].

51.       [ ʃt] [ t]          Paradigms: einsteigen [aintain], Stein, Stufe, Stuhl.

52.**    [ kv] [ k]        Paradigm: bequem [keːm].

53.       [ ʃv] [ f]          Paradigms: schwer, verschwunden.

54.*      [ pfl] [ f]        Paradigm: Pflaume.

58.*      [ ʃpʀ] [ p]       Paradigm: gesprochen.

59.       [ ʃtʀ] [ t]         Paradigms: Straße, Strümpfe.

63.*      [ tsv] [ ts]      Paradigm: zwei.

Particularly striking among the substitution rules for monophonic word-onset tagms are the four rules for [ l]. When the onset of Löffel 'spoon,' which was certainly the product of assimilation in N.'s language, is disregarded, the word-onset tagm l] is characterized by the transition from the effort to attain the standard realization to the consolidation of that realization. The tagm [z], which did not compete with standard [z], because the latter was always replaced by [s], was still prevalent at this time; rule 8.2, however, testifies to the advance of the substitution that would produce the standard form. Another development occurring at the time of the second evaluation was the beginning of the differentiation between [s] and [ʃ].

The tagm [ʀ] was always replaced by [h] in the word onset.

The polyphonic word-onset tagms in the words used by N. were generally simplified to monophonic tagms by dropping the non-abrupt, non-nasal element. If the word onset included a nasal element, it was maintained, while the other elements were dropped; in the case of [kn], [k] and [n] were fused as it were (see rule 28) to become the word-onset tagm [ŋ], which does not exist in the standard.

The most important exceptions to these general rules were the rules 18 and 54, which are likely to be due to the father's influence because his language does not include [pf] in the word-onset, and the rules 49.1 and 63, where no further reduction of [ts] was seen.     Word coda


The word-coda particularities include [ŋ], which developed from -gen according to rule 69.3 and was pronounced as a syllabic consonant in these cases. […] As to [ l] (rule 71), an increasing approximation to, and adoption of, the standard pronunciation was found, as has already been described for the word onset in [….] The incipient differentiation between [s] und [ ʃ], which has already been described for the word onset, was also observed in the word coda (rule 80). [t] as the second component of word-coda tagms was maintained if the first component was voiced; it was, however, dropped if the first component was voiceless (rules 144 to 154). The regular diphonic word-coda tagms with [t] after voiceless consonants were adopted within few weeks of the time of evaluation; in these cases, the [t] was initially often realized like a whispered syllable that was clearly separated from the preceding portion of the word.     Nucleus and intervocalic word-internal consonants



In the vowel tagms, the differentiation of the length feature does not seem to have completely consolidated. Open and closed vowels were not differentiated. The round, acute vowels, the so-called umlauts, were in the phase of transition from approximate substitutions to Standard German realizations, although the evidence is limited to only one paradigm each for [øː] and [œ].

Intervocalic word-internal consonants


The intervocalic word-internal consonants correspond mostly to word-coda consonants, in few cases—e.g. rule 265/266—to the word onset.     Description of the phonological system by a set of rules

For the time of the second evaluation, the structure of N.'s phonological system—which underlies the substitution rules—can be described with the following rules:


4.2.3     A. (age 21½ months)     General information

When A. was 21½ months old, the following substitution rules of his deviated from the standard. (Information on the symbols used in the lists is given in section     Word onset


At the age of 21½ months, monophonic and diphonic tagms were present in A.'s phonological system. If substitutions concerning monophonic tagms did not agree with the standard, they were in most cases paradigm-specific. Systematic differences from the standard were found for [ j] (rule 6) and [ l] (rule 8), which had merged in [ z]; the beginning of their differentiation was observed at the time of evaluation, though. The word-onset tagms that are diphonic in Standard German were in a similar transitional stage: Some paradigms followed the general rule, i.e., omission of the non-abrupt, non-nasal component and realization of the resulting monophonic word onset, whereas other paradigms showed the Standard German diphonic realization. If the first component of a diphonic word-onset tagm was ʃ ], this sound was dropped, except in the Standard German word-onset tagms ʃn] and ʃʀ], where the elements [n] and [ʀ], resp., were dropped. […] A particularly striking feature was the use of ] for ] and of ] for ]. This type of substitution covered the triphonic word-onset tagm ʃtʀ], too, and A. used it for a long time. […]     Word coda


A striking substitution among the word-coda changes is [ŋ] from Standard German ‑gen; it was sometimes supplemented by [ne] (e.g. [liːŋne] "liegen" 'to lie (down)') so that the original number of syllables was maintained. The plosive was also dropped in -ben and -den; the nasal consonant, assimilated accordingly, was either pronounced syllabically or the number of syllables was maintained by the addition of [e]. […] When [ʀ] was the first component of polyphonic tagms, it was often dropped, but there were also diphonic tagms with [ʀ] or [ʶ]. […] Two weeks after the second evaluation, A. simultaneously adopted all the diphonic word-coda tagms with [t] after a voiceless consonant; the [t] was at first distinctly separated from the beginning of the word and realized like a whispered syllable.     Nucleus and intervocalic word-internal consonants



The length feature of the vowel tagms had not yet been very much consolidated at the time of evaluation. Although there was a well perceivable difference between long and short vowels, half-long variants were realized conspicuously often and in part occurred in free alternation with either the corresponding long or the corresponding short form. The half-long variants therefore do not have the status of phonemes. Long vowels were also observed as the first element of diphonic tagms (rule 212.2). At the age of 21½ months, A.'s vowel tagms also showed deviations from the standard in respect of openness, although free alternation between different degrees of openness was not usually observed.

Intervocalic word-internal consonants

[…]     Description of the phonological system by a set of rules

A.'s phonological system at the time of the second evaluation can be described by the following rules:


4.3     Comparison between the phonological systems used at the two times of evaluation

4.3.1     N.

The comparison of the systems N1 and N2 shows a slight increase in the total number of rules, which hides a much more drastic change in the structure of the rule set, though. When all onset, coda and nuclear rules are considered, a marked reduction of inhibition and selection rules is offset by a still greater increase in the number of concatenation and substitution rules.

The lower number of selection rules on the second evaluation is directly linked to the expansion of N.'s inventory of phonemes, because this resulted in a reduction of the necessity to block elements in the tables of phones, which are essentially adapted to the requirements of the standard language. The remaining selection rules have, however, become more complex. Whereas some groups of sounds were completely blocked in N1, e.g., all central consonants in the word onset, the inhibition in N2 was more differentiated and necessitated several, partly quite elaborate, rules. The greater number of concatenation rules in N2 reflects the increase in multi-component phonotagms, which had totalled only three in N1.

The word onset had been enlarged in N2 by the addition of several compact consonants, of which there had been only one in N1. The clear structure of N1 was kept during this enlargement, in particular if [ʃˢ] is considered the voiceless pendant of [j], which is an unforced assumption because [ç] and [x] did not exist in N2. Like N1, N2 needed only one rule to generate the single diphonic word-onset tagm—which was, however, different from the diphonic tagm of N1. The word onset substitution rules had the same function in N1 and N2, namely the modification of non-tense, voiceless consonants.

The greatest changes between N1 and N2 occurred in the word coda. After all consonants but four had been blocked in N1 and two of the four, [f] and [s], had been distributed in strict agreement with the preceding vowel tagm, the N2 system comprised nearly five times as many monophonic word-coda tagms and thus almost all of Standard German. The difference is even greater in respect of polyphonic word-coda tagms, which were completely lacking in N1. A more detailed analysis of the concatenation rules N2-16 to N2-20 shows that about three quarters of all polyphonic word-coda tagms in N2 terminated in [f] or [s], whereas this applies to only one quarter of the Standard German word codas. This high proportion, which was not due to the wrong pronunciation of other word-coda tagms of the standard language, but to N. "selecting"[15] her words, may be regarded as a reflex of the word coda in N1. A final devoicing rule[16], whose function had been fulfilled by an inhibition rule in N1, supplemented the word-coda substitution rules of N2.


The distances of the phonological systems N1 and N2 from the reference system of Standard German[17] reveal a slight approximation to the reference system. When the overall distance is analyzed by the position of tagms, it becomes apparent that the word coda had the greatest distance from the reference system, whereas the distance of the nucleus was smallest (cf. Table 1 and Fig. 1). This order did not change between N1 and N2, but the nucleus contributed least to the approximation to the reference system. The word onset showed a slight approximation to the reference system, which nearly paralleled that of the overall system. The word-coda distance was reduced a little more than the overall distance. This allows the conclusion that the reduction of the phonological distance to the reference system between N1 and N2 was mainly due to the changes in consonant tagms.

The high degree of word-coda restructuring is also reflected by the fact that the N1-N2 distance for this subsystem is approximately the same as the respective R-N2 distance. The word onset and the nucleus, on the other hand, show distinctly smaller distances between N2 and N1 than between N2 and R, as does the system as a whole.


4.3.2     A.


The word onset in A1 was characterized by great regularity and simplicity. As far as the rules are concerned, the simplicity was reflected in six inhibition and selection rules to block all the phones that were not needed, while the regularity found its expression in the fact that only two concatenation rules and no substitution rules were necessary. The word-onset rules of A2 show a completely different picture. The lower number of inhibition and selection rules is explained by the greater inventory of phonemes. The monophonic word-onset tagms were fairly regular in A2, too, but the nine diphonic tagms were so diverse that—even though the second component was [+voc] in all cases but [ts]—four concatenation and one substitution rule were necessary for their complete description.


In agreement with this finding, the A1-A2 distances of the word-onset and nuclear subsystems showed values above those between R and A2. For the word coda and the overall system, A1‑A2 and R-A2 were approximately equal.

4.4     Similarities and differences in N.'s and A.'s language developments

4.4.1     Phonological systems on the first evaluation

The word onset was limited to few tagms in both N1 and A1. The rules N1-03 and N1-04 and A1-03 and A1-04 block similar groups of consonants; following the application of further inhibition rules, two quite regular word-onset subsystems result, with almost only non-compact consonants in N1 and mainly abrupt and vocalic consonants in A1. The regularity of either word-onset system is disturbed by one diphonic tagm each. Unlike A., N. seems to have selected her "active" words so as to have them fit her phonological system as well as possible[18], whereas A. used his restricted inventory of word-onset tagms to reproduce a large number of different tagms of the German standard.


4.4.2     Phonological systems on the second evaluation

N2 and A2 showed considerable differences in the word onset, which became also manifest in the about 50 % greater number of rules in the latter system. Although the monophonic tagms generated by either system were largely identical, N2 had only one diphonic word-onset tagm—[ts], which is also present in the reference system —whereas A2, on the other hand, comprised a total of nine diphonic tagms in the word onset. Except for one of these tagms, the second component was a [+voc] consonant, which is also the most frequent second component in the Standard German word-onset tagms. Both N2 and A2 contained variants in the form of voiceless lenes—not completely identical in the two systems—and there were two substitution rules in either system to describe them.

Textfeld: Fig. 2:  Angle b as a measure of approximation of a test system (T) to a reference system (R) between times 1 and 2 

4.4.3     Rates of approximation to the reference system

If a triangle is constructed according to Fig. 2 so that the three sides correspond to the distances between the systems, the angle b will be a measure of the rate at which the test system approaches the reference system. The evaluation of the phonological distances of N. and A. showed that, at an angle of b=52°, A.'s overall system approached the reference system faster than N.'s overall system, for which a value of b=62° was found. The differences were even greater in respect of the word onset and the nucleus; the calculation of the latter subsystem yielded an angle b of 33°, which indicates a particularly fast approximation to the reference system. For the word coda, on the other hand, the two systems showed virtually identical values of 57° and 58° (cf. Table 2).

5     Discussion

The description of early infant speech by means of phonological rules raises different questions. The first is in how far a set of rules like the one given above is suitable to describe infant sound sequences that are difficult to understand on first hearing—and not rarely on repeated hearing, too—and moreover seem rather inconsistent. Does such a set really describe phonological structures that are characteristic of a certain time of the child's development, or does it rather force into an arbitrary system what is actually marked by a more or less pronounced lack of rules?[19] The substitutions described above of tagms of the standard language do, however, show very clearly that N.'s and A.'s ways of producing words were all but ruleless.[20] The analysis of the substitution rules reveals that they can often be explained by the replacement or avoidance of certain distinctive features. At the first time of evaluation, for instance, A. used almost only abrupt and nasal consonants in the word onset, whereas N.'s word coda would be only [f] or [s] if no vowel followed. That the use of rules on the basis of distinctive features allows a concise description of the phonological structure in such cases is quite obvious.


When these considerations are applied to the analysis of the two children's phonological systems and their developments, fairly great differences become apparent. Whereas even before the first time of evaluation, N. had shown a strong tendency to adapt new words to structures she had developed before, A., when acquiring a new word, seemed to be tempted to expand or modify the structures he had used so far in order to assimilate the sound pattern he produced to the pattern he had heard. These tendencies were not distinct in the word onset because both children had developed quite regular inventories of consonants by 17 months. They were, however, very pronounced in the word coda. From her first word, heiß 'hot,' N. was inclined to have new words with acute vowel tagms terminate in [s], even if there was no such combination in the Standard German source. The [s] word coda was supplemented by an [f] coda after non-acute vowels when she was 15 months old. She had thus built a regular system that continued for quite a long time to be most important for her forming the word coda. A., on the other hand, restructured his word-coda system as needed for the acquisition of new words; between 14½ and 15 months of age, for instance, [k], [m], [x], [t], and [n] (in this order) were added. Since A.'s first word, Tür 'door,' does not contain a clearly noticeable word coda—the less so if the parents' pronunciation is also considered—this fact, i.e. the absence of a marked first impression, might be the explanation for the large inventory of word-coda consonants at the time of the first evaluation.[21] This could further mean that either child's first word had a decisive importance for the initial development of the phonological system.


The children showed marked differences in the rates of approximation of their phonological systems to the standard system. The rates of approximation, as measured by angle b (cf. p. 18), were virtually the same for N.'s and A.'s word codas, but A. approached the reference system much more directly than N. did, as far as the word-onset and nuclear subsystems and the overall system were concerned (cf. Fig. 4). As a result, A2—in respect of both the subsystems and the overall system—was closer to the reference system than N2, although part of the distances between the reference system and A1 had been greater than between the reference system and N1. If the phonological distances N1-N2 and A1-A2 are taken as a measure of the intensity of remodelling the phonological systems, N. is found to show almost the same intensity as A. does; only for the word onset was her intensity of remodelling only about two thirds of A.'s value. […]

The question therefore arises which factor was responsible for N.'s much slower approximation to the reference system, although she remodelled her phonological system similarly intensely as A. did. Among the factors that were different for N. and A., e.g., first word, size of the inventory of phonotagms, etc., the acquisition strategy seems to be the factor that best explains the different rates of approximation during the period of examination. As has already been mentioned above, the observations suggest that A. remodelled his system—not rarely ad hocas he learned new words. This means that A. had a tendency of not being particularly "careful" with any system he had developed by a specific time. This allowed him, on the other hand, to integrate influences emanating from the goal, i.e. the standard, fairly easily. N., however, remodelled her system rather by developing the rules she had found earlier. This probably limited the extent of changes by allowing only modifications that did not—or at least not too much—affect the existing system. The direction of changes thus depended on the prior system more than on the goal. Such an acquisition strategy can be expected to result in lower rates of approximation.

The rules can also answer the question how important specific distinctive features were at the two times of evaluation. […] Supposing that the rules have been formulated in a way adequate to the system described, i.e., that they are economical, the absence or disproportionately rare use of a distinctive feature in the rules could suggest that this feature had no importance for the child, although this would give no indication as to whether the lack of importance was more strongly related to perception or production.

Such absences of distinctive features were indeed found in the rules of N1 and A1. […]

Although at the first time of evaluation, the phonological systems of the two children examined in this study comprised phonemes with both positive and negative values of [intermediate] and [central], these features had no distinctive force because the other features of those phonemes were sufficient to set them off against other phonemes. In view of the differences between the subsystems both of either child and interpersonally, it is very striking to find that, on the first evaluation, the features [intermediate] und [central] had no (in N1) and almost no (in A1) function in the phonological systems. This suggests that these two features were not yet available for the production of sounds at that time.


6     References

BEST, Catherine T. (1993) "Emergence of language-specific constraints in perception of native and non-native speech. A window on early phonological development,” in Bénédicte de Boysson-Bardies, Scania de Schonen, Peter W. Jusczyk, Peter F. MacNeilage, and John Morton, eds., Developmental neurocognition: Speech and face processing during the first year of life, Kluwer, Dordrecht.

BLOOM, Lois (1993) The transition from infancy to language: acquiring the power of expression, Cambridge University Press, Cambridge (UK).

BOYSSON-BARDIES, Bénédicte de and Marilyn May VIHMAN (1991) "Adaptation to language: Evidence from babbling and first words in four languages," Language 67 (2), 297-319.

BRENT, Michael R. and Timothy A. CARTWRIGHT (1996) "Distributional regularity and phonotactic constraints are useful for segmentation," Cognition 61, 93-125.

CHURCH, Kenneth W. (1987) "Phonological parsing and lexical retrieval," Cognition 25, 53-69.

CUTLER, A. (1994) "Segmentation problems, rhythmic solutions," Lingua 29, 81-104.

EIMAS, Peter D. (1975) "Auditory and phonetic coding of the cues for speech. Discrimination of the /r-l/ distinction by young infants," Perception and Psychophysics 18, 341-347.

FERGUSON, Charles A., David B. PEIZER, and Thelma E. WEEKS (1973) "Model-and-replica phonological grammar of a child's first words," Lingua 31, 35-65.

FERGUSON, Charles A. and Carol B. FARWELL (1975) "Words and sounds in early language acquisition," Language 51 (2), 419-439.

GOAD, Heather and David INGRAM (1986) "Individual variation and its relevance to a theory of phonological acquisition," J Child Lang 14, 419-432.

GOLDIN-MEADOW, Susan, Martin P. SELIGMAN and Rochel GELMAN (1976) "Language in the two-year-old," Cognition 4, 189-202.

GRIESER, DiAnne and Patricia K. KUHL (1989) "Categorization of speech by infants: Supports for speech-sound prototypes," Developmental Psychology 25 (4), 577-588.

INGRAM, David (1978) "Sensorimotor intelligence and language development," in Andrew Lock, ed., Action, gesture and symbol. The emergence of language, p. 261-290, Academic Press, New York.

INGRAM, David (1981) Procedures for the phonological analysis of children's language, University Park Press, Baltimore.

INGRAM, David (1992) "Early phonological acquisition: A cross-linguistic perspective," in Charles A. Ferguson, Lise Menn, and Carol Stoel-Gammon, eds., Phonological development: models, research, implications, p. 423-438, York Press, Timonium.

JAKOBSON, Roman (1941) Kindersprache, Aphasie und allgemeine Lautgesetze, Almqvist & Wiksell, Uppsala.

JUSCZYK, Peter W. (1993) "From general to language-specific capacities: The WRAPSA model of how speech perception develops," Journal of Phonetics 21, 3-28.

JUSCZYK, Peter W. (1995) "Language acquisition: Speech sound and the beginnings of phonology," in Joanne L. Miller and Peter D. Eimas, eds., Speech, language, and communication, p. 263-301, Academic Press, London.

JUSCZYK, Peter W. and Carolyn DERRAH (1987) "Representation of speech sound by young infants," Developmental Psychology 23 (5), 648-654.

JUSCZYK, Peter W. and Paul A. LUCE (1994) "Infants' sensitivity to phonotactic patterns in the native language," Journal of Memory and Language 33, 630-645.

KAGER, René, Joe PATER, and Wim ZONNEVELD (2004) Constraints on Phonological Acquisition, Cambridge University Press, Cambridge.

KENT, Ray D. (1993a) "Infants and speech. Seeking patterns," Journal of Phonetics 21, 117-123.

KENT, Ray D. (1993b) "The biology of phonological development," in Charles A. Ferguson, Lise Menn, and Carol Stoel-Gammon, eds., Phonological development: models, research, implications, p. 65-91, York Press, Timonium.

KLANN-DELIUS, Gisela (1996) "Einige praktische Probleme von Langzeituntersuchungen und ihre methodologischen Implikationen," in Konrad Ehlich, ed., Kindliche Sprach­entwicklung, p. 17-29, Westdt. Verlag, Opladen.

KLEINE, Heinrich (1989) Phonologische und statistisch-dialektgeographische Untersuchun­gen an nordelsässischen Ortsdialekten, Franz Steiner, Stuttgart. (ZDL Beihefte 63)

KUHL, Patricia K. (1979) "Speech perception in early infancy. Perceptual constancy for spectrally dissimilar vowel categories," Journal of the Acoustical Society of America 66, 1668-1679.

KUHL, Patricia K. and Joanne D. MILLER (1982) "Speech perception by the chinchilla. Voice-voiceless distinction in alveolar plosive consonants," Science 190, 69-72.

LANDAUER, T. K. and L. A. STREETER (1973) "Structural differences between common and rare words: Failure of equivalence assumption for theories of word recognition," Journal of Verbal Learning and Verbal Behavior 12, 119-131.

LOCKE, John L. (1983) Phonological acquisition and change, Academic Press, New York.

LOCKE, John L. (1993) The child's path to spoken language, Harvard College, Cambridge, Mass.

MACKEN, Marlys A. (1992) "Where's phonology?" in Charles A. Ferguson, Lise Menn, and Carol Stoel-Gammon, eds., Phonological development: models, research, implications, p. 249-269, York Press, Timonium.

MACKEN, Marlys A. und Charles A. FERGUSON (1983) "Cognitive aspects of phonological development: Model, evidence and issues," in Keith E. Nelson, ed., Children's language, vol. 4, p. 256-282, Lawrence Erlbaum Associates, Hillsdale, NJ.

MENN, Lise (1979) "Towards a psychology of phonology: Child phonology as a first step," Proceedings of the Conference on Applications of Linguistic Theory in the Human Sciences, p. 138-179, Michigan State University, Lansing.

MESSER, Stanley (1967) "Implicit phonology in children," Journal of Verbal Learning and Verbal Behavior 6, 609-613.

MOSKOWITZ, Breyne Arlene (1980) "Idioms in phonology acquisition and phonological change," Journal of Phonetics 8, 69-83.

RATNER, Nan Bernstein (1994) "Phonological analysis of child speech," in Jeffrey L. Sokolov and Catherine E. Snow, eds., Handbook of research in language development using CHILDES, p. 324-366, Lawrence Erlbaum, Hillsdale, NJ.

SAFFRAN, Jenny R., Elizabeth K. JOHNSON, Richard N. ASLIN, and Elissa L. NEWPORT (1999) "Statistical learning of tone sequences by human infants and adults," Cognition 70, 27-52.

SCHLEIER, Inge (1996) "Phonologie und Kindersprachforschung," in Konrad Ehlich, ed., Kindliche Sprachentwicklung, p. 121-134, Westdt. Verlag, Opladen.

SCHWARTZ, Richard G. (1988) "Phonological factors in early lexical acquisition," in Michael D. Smith and John L. Locke, eds., The Emergent Lexicon. The child's development of a linguistic vocabulary, p. 185–222, Academic Press, Inc., San Diego.

SMITH, Michael D. (1988) "The meaning of reference in emergent lexicons," in Michael D. Smith and John L. Locke, eds., The Emergent Lexicon. The child's development of a linguistic vocabulary, p. 23-48, Academic Press, Inc., San Diego.

STEMBERGER, Joseph Paul (1988) "Between-word processes in child phonology," J Child Lang 15, 39-61.

STOEL-GAMMON, Carol (1985) "Phonetic inventories, 15-24 months: A longitudinal study," Journal of Speech and Hearing Research 28, 505-512.

STOEL-GAMMON, Carol and Judith A. COOPER (1984) "Patterns of early lexical and phonological development," J Child Lang 11, 247-271.

SUOMI, Kari (1993) "An outline of a model of adult phonological organization and behaviour," Journal of Phonetics 21, 29-60

VIHMAN, Marilyn May (1993) "Variable paths to early word production," Journal of Phonetics 21, 61-82.

VIHMAN, Marilyn May (1996) Phonological development: the origins of language in the child, Blackwell Publishers Ltd, Oxford.

VIHMAN, Marilyn May, Marlys A. MACKEN, Ruth MILLER, Hazel SIMMONS, and Jim MILLER (1985) "From babbling to speech: A re-assessment of the continuity issue," Language 61 (2), 397-445.

VIHMAN, Marilyn May and Ruth MILLER (1988) "Words and babble at the threshold of language acquisition," in Michael D. Smith and John L. Locke, eds., The Emergent Lexicon. The child's development of a linguistic vocabulary, p. 151-184, Academic Press, Inc., San Diego.

WERKER, Janet F. (1991) "The ontogeny of speech perception," in Ignatius G. Mattingly and Michael Studdert-Kennedy, eds., Modularity and the motor theory of speech perception, p. 91 – 109, Erlbaum, Hillsdale, NJ.

WINITZ, Harris and Orvis C. IRWIN (1958) "Syllabic and phonetic structure of infants' early words," Journal of Speech and Hearing Research 1,3: 250-256.


[1] This survey is only to point out possible explanations of conspicuous features. For the importance of other speakers' influences, cf. e.g. Schwartz (1988), p. 201.

[2] Cf. also Locke (1993), p. 363, and Vihman and Miller (1988), p. 173.

[3] Cf. Schwartz (1988), p. 205.

[4] Cf. Vihman (1996), p. 170.

[5] Cf. the procedure in Goad and Ingram (1986), p. 426.

[6] The phone cluster of a phoneme is largely coextensive with Ferguson and Farwell's (1975) 'phone class,' but the latter is significantly "broader" in some cases and allows greater intersections. ….

[7] Words with [f] after [i] or [ɪ] such as Brief {letter} or Schiff {ship} appeared in N.'s lexicon only after the word-coda inventory had already increased so that no rule parallel to 2.2 could be demonstrated for [f].

[8] The short vowels are the same as in Kent (1992), p. 73.

[9] From the age of 22½ months, the substitution [ z] [ z] prevailed.

[10] This sound was occasionally realized in more or less voiced variants.

[11] The realization [w] was also observed.

[12] N. adopted the form [laːfən] for this paradigm at the age of 23½ months.

[13] May have been influenced by putzen {to clean}.

[14] [] was also observed.

[15] More detailed discussions of this issue are found, e.g., in Ferguson and Farwell (1975), Stoel-Gammon and Cooper (1984), Schwartz (1988), and Vihman (1993).

[16] The rule 69.2 with [ g] : [ x] / ___/(U1)/:/(U2)/ in N2 was disregarded here because it affected only part of the paradigms; all those paradigms were also covered by the usual final devoicing rule 69.1, which corresponded to the standard system.

[17] The calculations were performed on the basis of Kleine (1989)—with some modifications to account for the great divergences between the systems. …

[18] Similar observations were made, for instance, by Stoel-Gammon and Cooper (1984), p. 265.

[19] A more comprehensive discussion of this point is found in Vihman (1996), p. 22-26.

[20] Cf. Vihman (1996), p. 4.

[21] Contrary to the vast majority of the inventories of American children that Stoel-Gammon (1985), p. 506, had examined and contrary to N1, A1 had more tagms in the word coda than in the word onset.