Abstract
This article analyses longitudinal diary data from one infant acquiring German to seek a better understanding of the role of onomatopoeia in early language development. Onomatopoeic words (OWs) are traced over time in relation to their corresponding conventional forms (CWs), and an analysis of their phonological transitions is considered in relation to the infant’s wider language development. OWs are found to serve as a bootstrapping mechanism on a lexical as well as a phonological level, and a lexical OW ‘template’ is proposed, which is found to support phonological development throughout the infant’s output.
Onomatopoeia occur in infant speech across many languages (Menn & Vihman, 2011), becoming less prominent in the lexicon throughout early development (Kauschke & Hofmeister, 2002). Despite evidence showing the prolific use of onomatopoeia in infants’ early words, this aspect of the vocabulary is disregarded in many studies of phonological and lexical acquisition. Behrens (2007), for example, sidelines onomatopoeic forms along with ‘hesitation markers … nonce words, and noninterpretable words’ (p. 15): it seems that onomatopoeia pose a difficult problem in the study of infant speech, and it is unclear whether these forms constitute a relevant component of the early lexicon, or whether they are superfluous to language development. In this article, a detailed phonological account of one infant’s onomatopoeia will attempt to identify a role for these words in the early output.
An overview of the infant’s onomatopoeia will demonstrate the acquisition of these forms over time in relation to non-onomatopoeic words. Following this, a phonetic account of her output will suggest a functional role for onomatopoeia in early development, which separates these forms phonetically from the rest of the lexicon. A wider phonological account will then demonstrate the systematic use of onomatopoeia as a lexical ‘template’, which facilitates production despite phonological deficiencies. Throughout the analysis a specific timeframe can be identified, where changes take place throughout the output as the infant’s phonetic, phonological and lexical abilities combine to bring about further language development.
Failure to acknowledge the role of onomatopoeic words is contradictory to their prominence in early word production. Menn and Vihman (2011) list the first five words produced by 48 infants acquiring a range of languages, and record 40 onomatopoeic words in total (20% of the full sample), equating to almost one per child. Furthermore, a study by Kern (2010) found that words classed on the Communicative Development Inventory (CDI) as ‘sound effects and animal sounds’ were the most common in the productive repertoire of infants from 8- right up to 20-months old, and, in fact, this category contributed to over a third of the entire output of infants between the ages of 8 and 16 months. It is unsurprising, however, that the literature on onomatopoeia in early language production is scarce, as authors who choose to include these forms in their analyses have had to contend with some puzzling findings. In a study on syllabification in Finnish infants’ lexical development, Kunnari (2002) found that onomatopoeia had a marked influence on her analysis, and so had to consider these words separately from the rest of her data. Similarly, Savinainen-Makkonen (2007) found that certain consonants occurred exclusively in the production of onomatopoeia, including clusters which were not typical of Finnish. The presence of onomatopoeia in infant data may pose a problem in generating interpretable results, yet the lack of understanding of these word forms only exacerbates the issue that researchers face in their analyses.
To paint an authentic picture of lexical acquisition, onomatopoeia must be considered as a relevant part of early infant data. Aside from their traditional role as phonetic imitations of sounds in the environment, infants often allocate extra referential meaning to onomatopoeia – for example, Hildegard Leopold’s use of moo to represent COW, as well as to refer to the sound that a cow makes (Leopold, 1970). In this sense onomatopoeic words serve to convey added referential meaning as they are extended to depict objects as well as the sounds that those objects make. Further evidence of this will be seen in this analysis.
It is plausible that infants should be able to associate an onomatopoeic word with its referent more easily due to the natural sound–meaning link, or ‘iconicity’, that onomatopoeia are posited to possess (Jespersen, 1950; Nygaard, Cook, & Namy, 2009). Indeed, it has been shown that iconicity can facilitate language acquisition in both adults (Nygaard et al., 2009) and young infants (Imai, Kita, Nagumo, & Okada, 2008), and it may be presumed that the iconicity of onomatopoeia causes a bias towards these forms in word learning, as infants are thought to ‘strive for … transparency of meaning–form correspondences’ (Slobin, 1980, p. 231). Evidence for an iconic advantage in form–referent mapping can be found in the literature on gesture, which has been shown to facilitate the comprehension of words amongst pre-linguistic infants (Caselli, Rinaldi, Stefanini, & Volterra, 2012). Indeed, Iverson and Goldin-Meadow (2005) show that the gestures used by pre-linguistic infants predict the words that they later go on to produce, suggesting an advantage for iconic form–meaning mappings even before the onset of word production. Sign language provides a rich dataset of iconic symbols that can help us to address the question of iconicity in language learning, but evidence suggests that iconicity does not play a role here. Orlansky and Bonvillian (1984) showed that deaf infants made less use of iconic signs than non-iconic signs in early sign learning, while Meier, Mauk, Cheek and Moreland (2008) found that deaf infants did not draw upon the pictorial advantage of sign language in the first stages of acquisition. Furthermore, the reality of iconicity in onomatopoeia must be called into question: these forms are largely conventionalised, and so we cannot be certain as to what extent their form–meaning link is truly iconic. It is likely, after all, that most infants learn onomatopoeia from the speech of the caregiver, as they would any other word form.
A phonological account of onomatopoeia in early word production will provide a bottom-up view of these forms, tracing their frequency and decline with reference to the phonological changes taking place in the infant’s output. This article seeks to uncover a role for onomatopoeia in phonological development, hypothesising that these forms facilitate production through their simple structures and iconic properties. It is proposed that these simple structures allow a more flexible approach to the production of onomatopoeia, making room for a lack of motoric precision in the early output, while their iconic properties are posited to make them easier to recall from memory. These features are thought to provide infants with a set of words which are both easier to remember and to articulate, despite limited capacity in both of these domains.
Elsen’s (1991) diary account of her daughter Annalena provides ample evidence of the use of onomatopoeia in the early output. It will be drawn upon here to analyse the production of onomatopoeic words (OWs) from the very first word until the 500-word point (the point at which she has 500 word forms in her lexicon) at 1;6. This diary account provides a detailed picture of the progression of OWs, from the use of rudimentary vocal gestures to the eventual production of the full adult form. The forms that Elsen describes include standard onomatopoeic forms that are commonly found in many languages, as well as highly idiosyncratic onomatopoeically derived forms which are far removed from the adult language.
Methodology
Participant
Elsen (1991) provides a highly detailed chronological account of the language development of her daughter, Annalena, an only child acquiring German in Munich, Germany. Her observations begin shortly after Annalena’s birth, and continue with a constant and thorough account of her language development until Annalena reaches 2;5. Elsen acknowledges the precocious nature of her daughter’s speech, and reports that this was also noted by doctors; indeed, Annalena had already produced five different words by 0;9.
Data
Elsen’s (1991) diary account was recorded on paper and included some audio back-up, though this was not available for this analysis. Exact dates are given and variability within word forms is highly detailed. Elsen also provides notes for most words, outlining whether words were produced spontaneously or whether they were elicited, and in which context the words appeared.
Analysis
Annalena’s word production was analysed from the very first word until the 500-word point and onomatopoeic words (OWs) were identified in this dataset. A word was classed as onomatopoeic if it attempted to imitate a sound from the environment; this definition therefore includes commonly recognised OWs such as woof or vroom as well as those which have been adapted by Annalena such as [lala] for Musik ‘music’. Forms carrying referential meaning in Annalena’s output – ‘functional’ onomatopoeia, such as [vavaɯ] meaning Hund ‘dog’ – and those which are merely imitations of sounds from the environment – ‘standard’ onomatopoeia, as in [bɯma] bum ‘boom’, to imitate a loud noise – are considered in the initial analysis, before a full investigation of Annalena’s functional OWs takes place. Naming conventions provided in the original analysis are adhered to here (e.g. [vavaɯ] appears under the target Hund ‘dog’, even though it can be translated literally as woof or bow-wow), and to avoid discrepancies in naming, the target word is considered to be the full conventional word, and will appear in capital letters in English (e.g. DOG). While in many cases the OW will remain in Annalena’s lexicon as part of the adult language, this analysis considers the lexical and phonological transition from OW to conventional word (CW), and so DOG relates to all word forms used to express the meaning dog throughout the recording period.
Results
Lexical acquisition
MOUSE [pɪˑpɪˑpɪˑ], at 0;10, is Annalena’s very first OW, occurring 37 days after the onset of word production. From here, her lexicon increases rapidly in terms of both OWs and CWs. The most striking observation to be made regarding the presence of OWs in Annalena’s output is the quantity of these forms that are produced in the first months of word use. Figure 1 plots the acquisition of OWs against CWs on a monthly basis. It is clear that OWs constitute an important portion of Annalena’s total lexicon between 0;10 and 1;0, peaking at 0;11, when the acquisition of CWs starts to increase (see Figure 2). At 0;11 Annalena’s lexicon can be seen to increase by 155%, with the acquisition of 11 new OWs. By 1;0 Annalena has 50 words in her lexicon, and the rate of acquisition of OWs remains constant at an average of just under six new OWs per month until 1;5. The acquisition of new OWs appears to stop abruptly at 1;6 – the month in which Annalena reaches her 500-word point.
Monthly acquisition of OWs and CWs.
Percentage of OWs in relation to vocabulary size.
In Figure 2 the proportion of OWs in Annalena’s output is plotted as a percentage of the cumulative total of new word forms (OW and CW) acquired per month. The steady tailing-off of OW acquisition over this period is equally apparent.
As can be seen here, these forms initially account for a substantial portion of the total vocabulary (almost 40% at 0;11), before CWs take over. From this analysis it is clear that the production of OWs contributes considerably to the early output, but once word learning takes off these forms no longer play an integral role in promoting word production. A closer look at the specific OWs in Annalena’s output reveals some interesting features that are specific to these forms.
Functional vs standard onomatopoeia
When the entire set of OWs is analysed, it is clear that Annalena’s production of OWs serves as more than to verbally imitate natural sounds. Indeed, extensive evidence can be found in the dataset of the use of OWs to refer to objects as well as sounds (for example [baɡbaɡba] ‘quack quack’ to express Ente ‘duck’). These OWs are termed ‘functional’ OWs in this analysis. In line with the traditional interpretation of onomatopoeia (e.g. Sapir, 1921), some of Annalena’s OWs serve only to imitate sounds (for example, tatütata ‘nee-naw’ (siren noise) [tɨtatɨta], kikeriki ‘cock-a-doodle-doo’ [kɪːkɪːç]), termed here ‘standard’ OWs. This dichotomous approach to OW production highlights the presence of two systems at work in Annalena’s output: OWs which represent non-human sounds, and OWs with further referential meaning, which name objects as well as refer to the sound that the objects make. In many cases Annalena creates an idiosyncratic OW form based on her interpretation of the sound that is produced by a referent; these forms are used as referents when Annalena cannot yet produce the CW form.
Elsen (1991, 1994) refers to these idiosyncratic OWs in her analyses, and provides evidence towards the influence of the input on Annalena’s OW production, stating that the form wauwau ‘woof’ was used only by Annalena’s grandmother, and never by Elsen herself. Annalena’s use of quak ‘quack’, piepiep ‘tweet’ and her own interpretation of a crow [bɔa] are declared to be spontaneous and unprompted on the infant’s part. In reference to onomatopoeia, Elsen (1994) states that Annalena ‘imitated forms deliberately’ (p. 311), suggesting an infant-initiated approach to language learning whereby OWs were used to facilitate production. Indeed, Annalena’s functional OWs are often further removed from the adult language than her standard OWs, produced with extra-phonetic ‘wild’ features (Laing, 2014; Rhodes, 1994) which bring the sound and the meaning of these words closer together.
It is in the discrepancy between functional and standard OWs that we begin to see a role for onomatopoeia in Annalena’s word production. Functional OWs provide Annalena with a comprehensible sub-lexicon with which she can communicate and express meaning effectively, despite not having acquired the appropriate CWs. From the dataset of Annalena’s first 500 words, 52 are judged as OWs, and 16 of these are considered to be functional. These are detailed in Table 1, along with an account of the ‘wild’ features realised by Annalena in their production. An OW is classed as being functional if it serves to define more than merely onomatopoeic expression in Annalena’s output – OWs which are adapted referentially as ‘names for things’ (e.g. [baɡbaɡba] for DUCK).
Functional OWs in Annalena’s Output.
Note: Standard German pronunciation of Pferd would be [pfeːɐ̯t], but the mother’s dialect omits initial /p/ and so the target pronunciation is [feːɐ̯t].
As is apparent from Table 1, these forms include a range of extra-phonetic, or ‘wild’, features, produced in order to imitate the sound representing the word in question. These wild features include segments that are not found in the native language phonology, unusual pitch and intonation changes, gestural accompaniments, variability and the unconventional production of familiar phonemes. The examples in Table 1 illustrate the idiosyncratic nature of these OWs in Annalena’s output, and the various ways in which meaning can be conveyed through iconicity in the production of OWs, despite these wild forms being far removed from the phonological conventions of the adult language. The invention of words through the imitation of natural sounds has been reported in many accounts of early infant word production (Jespersen, 1950; Werner & Kaplan, 1963), and these forms are reported to ‘generally disappear quickly in favour of the traditional forms’ (Jespersen, 1950, p. 151). However, Annalena often produces these idiosyncratic functional forms for some months before acquiring the full adult form. It could be assumed that Annalena will acquire CWs over time as she develops a more complete phonological inventory, but as Table 1 shows, she continues with the OW production of SHEEP until age 1;8, and produces OW CROW until after her second birthday. It is evident that a complex system of language learning is at work here, and this will be explored in detail in this analysis.
In Figure 3 we see the transition from OWs to CWs in Annalena’s output over time. It can be seen that OWs appear well before the CW is first produced in all instances, showing the trajectory of lexical development as OWs are initially used functionally, representing the CW equivalent that Annalena will eventually move towards in production. Many forms undergo a period of alternation, as Annalena produces both the OW and the CW in tandem to express the same meaning, eventually moving towards the CW, at which point the OW no longer serves to carry referential meaning. During this transitional period Annalena produces both individual tokens of OWs and CWs (for example, at 1;0 she produces DUCK as both [baɡba] and [nεndə]), as well as combinations of the two, as in COW [mːkɯç] at 1;5. This transitional phase marks an incremental shift in representation from the OW to the CW form.
Progression of Annalena’s OWs over time.
Fifty percent of OWs alternate with the CW in a transitional phase before the CW fully takes over. With the exception of CAT and CROW, all of the forms which undergo a period of transition are acquired between 3 and 5 months after first word production: this corresponds to the period of 0;11 to 1;1, just before lexical acquisition takes off at 1;2 (see Figure 2). It is unsurprising that we find evidence of shifting representations here, as Annalena’s lexical knowledge begins to increase rapidly as new word forms are acquired, and thus new forms are likely to be less stably represented in the output. Most of those forms which do not undergo a period of transition are acquired either in the earliest stages of word learning, and are thus more stably represented in Annalena’s lexicon, or are acquired later, when Annalena’s production experience is much more established. It seems that this period of transition is not limited to her OWs, but is taking place across the board in her language development: new linguistic knowledge is being acquired, causing Annalena’s lexical and phonological systems to become less stable in terms of both internal representation and output.
Fundamentally, the transition from OW to CW does not take place across the board but happens over time, on a word-by-word basis. However, 69% of her OWs make the transition to CW after 1;2 – coinciding with the steep increase in her lexical knowledge shown in Figures 1 and 2. Changes taking place elsewhere in Annalena’s phonological development can help to account for these lexical advances; the following analyses provide evidence for both a phonetic and a phonological shift in Annalena’s language development, which coincides with the period of transition noted in Figure 3, paving the way for the steep lexical advance shown in Figures 1 and 2.
Phonetic influence of OWs
An important development takes place in the first half of Annalena’s second year which enables the eventual change in representation from OW to CW for many of the forms detailed in Table 1. Phonologically derived motivations for the segmental differences between the OWs and their CW counterpart can be identified when Annalena’s full output is examined in this period; the developments in her consonant inventory between the ages of 1;0 and 1;2 (Figures 4a and 4b) show a marked shift in her phonetic capacity which takes place shortly after her first birthday.
Annalena’s consonant inventory at 1;0.
Annalena’s consonant inventory at 1;2.
Those segments which had appeared only in OWs at 1;0 have stabilised in CWs by 1;2, while Annalena has acquired almost a full set of fricatives during this period. The changes seen across these two consonant inventories can account in part for the shift from OWs to CWs that begins at 1;2.
Figure 4a shows how certain segments were produced exclusively in OWs, reflecting the unconventional characteristics of these forms which do not necessarily match the phonological framework of the ambient language. Before 1;0 Annalena produces the bilabial trill /ʙ/ exclusively in CAR (see Table 1), and the approximate /w/ only in DOG (see Table 1): neither are typical phonemes of German. A predominance of labial consonants is reported in Annalena’s babbling phase, including /w/, /b/, /bv/ and /ʙ/, which accounts for the continuing use of these segments in OWs; these segments enable Annalena to represent words using the well-rehearsed phonemes that are available to her. These examples demonstrate how OWs may work functionally in early word production; despite Annalena’s inability to consistently produce the segments necessary for the established German OW forms (brumm and wauwau, respectively), her own forms can still be understood in communication with the caregiver. Furthermore, the use of non-native phonemes in these OWs appears to facilitate further word production, as Annalena later goes on to use these segments in the production of CWs by adapting word forms with these consonants. For example, Affe ‘ape’ [ɔwa], Auge ‘eye’ [awa], Opa ‘grandfather’ [ɔwa], zwei ‘two’ [waɪ] and Vorsicht ‘look out’ [was:ɪ:ç]. These early consonants facilitate word production over time, as Annalena practises and stabilises them with the production of OWs, and later acquires new words by adapting them to fit her own consonant inventory.
As well as these non-native phonemes, Annalena produces a number of German segments in OWs which are avoided in CWs: /k/ is produced in kikeriki ‘cock-a-doodle-doo’ [kɪːkɪːç] and killekille ‘tickle tickle’ [kɪɫˑəkɪɫˑə] but is avoided in kalt ‘cold’ [malː]; /ç/ also appears in kikeriki whilst being dropped from the cluster in Milch ‘milk’ [malː]; finally, /v/ is produced in the OWs DOG and CAR at 0;11 (see Table 1), but is not used consistently in a CW until 1;4 with wischen ‘wipe’ [vɪs̪:ə, vɪs̪:əˑn]. Again, all of these consonants are reported to first occur in Annalena’s babbling phase, but yet she does not apply them throughout her early word forms, nor does she make use of other consonants from her babble that could enable more accurate or varied word production at this point. A similar account of unprecedented accuracy in OWs is reported in a study by Savinainen-Makkonen (1998, cited in Kunnari, 2002), which supports the evidence here showing that OWs do not correspond to the phonetic limitations which are otherwise apparent in Annalena’s lexicon. It seems that OWs (both functional and standard) provide a platform from which Annalena can experiment with the production of difficult segments, using features typical of OW structures such as reduplication, vowel lengthening and variability to facilitate the production of OWs. This can ease the demands of memory and planning that are consistent with phonologically more complex CWs as she does so.
Phonological influence of OWs
Alongside this apparent phonetic bias we can see more consistent trends throughout Annalena’s OWs which point to a phonological role for these forms. In a paper investigating the phonological constraints in Annalena’s output, Elsen (1994) describes her daughter’s tendency to avoid certain forms, including a refusal to produce the word Hund ‘dog’, despite being presented with plenty of opportunities to imitate this form from those around her, and despite a clear understanding of the CW. Elsen goes on to discuss Annalena’s approach to word production in terms of the ‘Principle of Avoidance’, whereby Annalena is reluctant ‘to produce sound combinations beyond her phonological level’ (1994, p. 311). This account suggests that a phonological limitation is prompting Annalena’s use of OW DOG, which enables communication while her phonological skills catch up with her wish to express certain word forms. Ferguson and Farwell (1975) describe infants’ process of avoiding specific words and phonemes as one of ‘great selectivity’ (p. 433), which relates to a more recent proposition from Vihman and colleagues (Menn & Vihman, 2011; Vihman & Croft, 2007) regarding the ‘selection’ of word forms based on infants’ specific production abilities. In this case, Annalena can be seen to select the OW equivalent of Hund, and by doing so is matching her own vocal capabilities to specific lexical forms.
A thorough phonological analysis of the changes taking place in this period can demonstrate how Annalena makes used of OWs to overcome the phonological limitations presented by the CW forms. This can be seen across her lexicon, and takes into account the phonetic changes observed between Figure 4a and 4b, as a phonological shift prompts a transition to the CW forms, as well as the acquisition of many new words. An account of the phonological limitations in Annalena’s output in relation to her functional OWs demonstrates this point.
Fricatives
As shown in Figure 4a, fricatives are highly restricted in Annalena’s early words, and in fact are only produced in OWs up until 1;0. They begin to make their way into Annalena’s output between 1;1 and 1;2, both through the acquisition of new words and through the transition from OW to CW. Of the 24 target words containing fricatives that were attempted at 1;1, only 25% were realised with a fricative, while at 1;2 64% were attempted with a fricative (compare Kissen ‘cusion’ at 1;1 [ɡεjə], and 1;3 [kɪs:], [kɪsʃə], [kɪ:ça], by which time /k/ had also been acquired – see below). This highlights a change in Annalena’s approach to word production over this short timespan – until this point, fricatives have only been produced in OWs; the shift at 1;2 finally sees Annalena attempting to produce (and often succeeding in producing) fricatives in CWs, and in turn this leads to the transition of many OWs to the CW equivalent.
BIRD presents the first example of fricative avoidance in Annalena’s OWs. Word-initial /f/ is not reported to stabilise until 1;5, and we don’t see any correct attempts at producing this segment until 1;3, at which point CW Vogel is attempted as [vɯɡ╎l] (with high variability), alongside vier ‘four’ [fɪa, vɪa], Fuchs ‘fox’ [vɯɡ, ɯkz] and Fuß ‘foot’ [vɯs:]. In total, 19 new /f/-initial words are first attempted at 1;3, amongst which 14 include a word-initial /f/ or /v/ in at least one token.
CAT, DONKEY, EAT, HARE and MUSIC all contain a word-medial /s/ or /z/ in the CW, and Elsen refers to the Principle of Avoidance in terms of these segments. Indeed, aside from two exceptions she claims that Annalena ‘ignores’ /z/ until 1;3 (1991, p. 84), and accordingly, Annalena transitions from OW to CW in all five of these forms after 1;3. Word-final /s/ in CW MOUSE exhibits similar restrictions, and Annalena acquires this form alongside the others at 1;3, with additional /s/-final words appearing in her lexicon during the same period (e.g. Bus ‘bus’ [bɯθ, bɯs̪], Netz ‘web’ [dεts:]). However, we do see final /s/ as early as 1;1 in heiß ‘hot’ [ais] and tschüß ‘bye’ [tɨs̪, tsɨs], though Elsen reports an unstable production of this segment, which is not truly mastered until 1;5.
The initial /ʃ/ in SHEEP is highly problematic for Annalena, and she can be found to avoid this segment throughout her lexicon. Indeed, she does not produce word-initial /ʃ/ correctly until 1;5 with schau mal ‘look’ [ʃç̚aɯmaˡ], but even then this is one correct attempt amongst a range of tokens. Out of the 45 /ʃ/-initial words that are attempted by Annalena in her first 18 months, only five forms contain a /ʃ/ segment, and including schau mal only two of these are realised in the correct position – this segment is highly limited in Annalena’s output.
Finally, initial /ʃv/ in PIG presents an especially difficult problem for Annalena, and this cluster is not mastered in her output until 1;9, which is beyond the timeframe of this analysis. However, owing to the acquisition of fricatives, Annalena is able to transition to the CW here as early as 1;5, as she makes use of cluster reduction in order to facilitate production, as in [vaɪn].
Word-initial /k/
Until the end of 1;1 Annalena had only produced /k/ in OWs. Its distribution is highly limited, and again Elsen (1991) refers to Annalena’s approach here as another example of the principle of avoidance. Before 1;1 Annalena attempts only one /k/-initial word, the OW form kikeriki ‘cock-a-doodle-doo’. At 1;2 Annalena begins to produce initial /k/ correctly, in Kater ‘tomcat’ [kɑːtə] and Kuckuck ‘peek-a-boo’ [kɯkɯː], and acquires 10 new /k/-initial words between 1;2 and 1;3, of which eight have a correctly produced word-initial segment in at least one token.
CAT, COW and (to a certain extent, see below) CROW are all restricted by this phonological limitation in Annalena’s output, and accordingly none of these OWs transitions to the CW form until word-initial /k/ has stabilised in Annalena’s output.
/nt/
Annalena avoids this cluster almost completely, with only one target word containing /nt/ (Mund ‘mouth’, produced as [mama, maɪ, manθ]) until its first correct realisation at 1;2. CW DOG [ʔɯnt] is one of three new words that Annalena produces with this cluster in this month, as well as Kind ‘child’ [tɪntʰ] and Elefant ‘elephant’ [man tʰ]. Mund is now also produced with the correct cluster, as [mɔntʰ], and at 1;3 we see the transition of DUCK from OW to CW [æn:tθə, εndə].
Word-medial /n/
There appears to be some restriction on the production of /n/ in Annalena’s output, though this is surprising as nein ‘no’ [nan] appeared as her first word at 0;8. However, while this segment is stable word-initially, Annalena has problems producing it word-finally, for example Bein ‘leg’ [baɪ] and Stein ‘stone’ [daɪ, daja] at 1;5. There are exceptions to this restriction, as she can produce /n/ in an ‘on’ [ʔan] at 1;3, though Elsen acknowledges this anomaly herself, positing that it may be due to the lack of any other segments in this form.
There are few examples of word-medial /n/ in Annalena’s dataset, and these can often be found to exhibit consonant harmony: Kanne ‘can’ [nana] at 1;3 (with word-initial /k/ also presenting a problem here), Annalena [nana] from 0;10 until 1;5. Banane ‘banana’ is an exception, produced as [maɪna, mana] from as early as 0;11, but there is especially high variability in the realisation of this word.
BEE transitions from OW to CW form [bɪnə] at 1;4, and we see the acquisition of a number of words with medial /n/ in the same period: Honig ‘honey’ [ho:nɪs̪], Pony [pɔnɪ] and Spinne ‘spider’ [pɪnˑə] at 1;4, and Birne ‘pear’ [bεnˑɑ:t] and Sonne ‘sun’ [z̪ɔnˑə] at 1;5.
Word-medial /t/
/t/ appears in Teddy at 0;10, but bitte ‘please’ is the only target word with medial /t/ in Annalena’s output in this period, produced with reduplication as [bɪtˑəbɪt] at 0;10. Otherwise, the first target word with this segment is Kater ‘tomcat’ [tɑːtə, kɑːtə] at 1;1.
Annalena’s first CW attempt at CAR [ɑːtɔ] follows at 1;2, and in the same period eight other new word forms with medial /t/ are acquired, including Tomate ‘tomato’ [bɔmɑːtə] (1;2), and Schmetterling ‘butterfly’ [mεt:tin:] and warte ‘wait’ [watˑa] at 1;3.
/h/
/h/ only occurs at word-initial boundaries in German, and before 1;1 Annalena has attempted just two /h/-initial words, though she produces /h/ correctly in certain tokens of both: [halː, halɔ] hallo ‘hello’ at 1;0, and [haɪ, haɪç] heiß ‘hot’ at 0;11.
Between 1;1 and 1;2 she acquires 11 /h/-initial words, of which eight are realised with an accurate /h/. This includes the CW form of HARE, which is first produced at 1;2.
Word-final /k/
/k/ is limited word-finally as well as word-initially, and is not attempted in this position until Annalena produces Tag ‘day’ as [tɑːkʰ, ɡɑːkʰ] (first produced at 0;11 as [dada]) at 1;2. Two more /k/-final words are acquired in the same period: weg ‘way, away’ [bakʰ] and the OW ticktack ‘ticktock’ [kəkək]. A few months later, at 1;5, the transition of MUSIC from OW to CW [zɪ:kʰ] takes place.
Cluster /kʁ/
It is unsurprising that the cluster /kʁ/, as well as many other clusters, are limited in Annalena’s output. Elsen (1991) reports that this cluster is not acquired until 1;9, which goes beyond the period analysed here, but even so CROW is not fully acquired until 2;1, when it is produced accurately as [kʁeːə].
This analysis has demonstrated a number of phonological limitations which affect the entirety of Annalena’s output, but which can largely be seen to disappear between the period 1;0 and 1;2. Before these changes take place we can see that Annalena’s OW forms provide, in almost all cases, a better ‘match’ to her phonological abilities than would the CW forms. This demonstrates the presence of a systematic lexical template, in line with the phonological templates discussed by Vihman and Croft (2007). This lexical template makes use of OW forms to facilitate production when the phonological skills necessary to produce the CW are not yet in place. Over time we see a general transition from OW to CW production, as her phonological systems fall into place and she is able to produce the full CW forms. Alongside these transitions we see extensive lexical acquisition, prompted by each of the phonological developments discussed above. In many cases Annalena can be seen to avoid words containing these phonological segments completely, only attempting them once she has the appropriate capacity to do so. Her careful approach to CW production is thus applied throughout her lexicon, but in cases where there is no suitable OW which can be used functionally as a template she must instead wait until her output capacity is sufficiently developed to produce these words.
In some cases these limitations affect Annalena’s OW production as well as her CW production. OW BEE, for example, is produced as a single fricative, but is acquired at 1;2, after she has begun to acquire fricatives. Drawing again upon the example of BEE, after only two months Annalena is able to produce word-medial /n/, and thus she transitions to the CW form, repeatedly moving forwards within the boundaries of her phonological capacity.
It must also be noted that many of the CWs in this analysis contain multiple challenges for Annalena: CAT, DOG, HARE, MUSIC and SHEEP all contain two of the phonological limitations listed above. Annalena does not attempt to tackle both segments simultaneously, but typical to the systematic approach that we observe throughout her production, she adapts CW forms in line with her expanding production capacity, moving gradually towards a more accurate production. The gradual transition to these five CWs is shown in Table 2. We can see that the forms listed under CW1 take on only one of the newly acquired phonological segments, while under CW2 we see a gradual transition towards the second phonological segment, with instability across all forms (and regression in SHEEP) as she does so. By CW3 she is close to the full adult form in all cases.
Transition to CW forms over time.
Furthermore, the acquisition of these forms is not delayed by the extra phonological challenge that they pose; three of the five are acquired as early as 1;2, corresponding in all cases to the timeframe in which these phonological changes start to take hold, and with the transition to CWs across Annalena’s OW forms.
Discussion
The production of OWs (both standard and functional) is prominent in Annalena’s early output and contributes substantially to her early vocabulary. The prominence of OWs does not last, but leads to a period of rapid change in Annalena’s output – from 1;0 to 1;2 – where we can find lexical, phonetic and phonological developments taking place.
Early on, OWs are found to function at a phonetic level, produced with consonants which are not available elsewhere in Annalena’s lexicon. These consonants allow more scope for producing words with wild features, providing an inventory of segments which are appropriate for imitating sounds in her environment (the bilabial trill in CAR, for example, or the labial fricative in DOG). In the early stages of language acquisition, therefore, OWs have been found to facilitate production through the expansion of her lexicon and through the production of a wider variety of consonants. The acquisition of phonemes through OW production enables Annalena to expand her output further, and she is found to utilise these phonemes in the adaptation of new words.
When considering Annalena’s early consonant inventory (Figure 4a), one could assume that she would be able to produce CWs COW, BEE, CROW, DUCK, DOG and HORSE, as all of these forms contain segments that have already occurred in her output. However, while her phonetic capacity is relatively developed even by 1;0 – indeed, Annalena is already able to produce over 70 word forms at this point – phonologically speaking, her output is heavily restricted. It is here that we see the implementation of OWs on a phonological level, offering an alternative form which can be used to communicate referential meaning. In most cases OWs match her phonological ability – OW BEE, for example, appears at 1;2, once Annalena is able to produce fricatives, while others have simple CV or consonant harmony structures. We do see examples of OWs which override these limitations – fricatives are found in OW DOG and CAR, and medial /n/ in CAT – but these apparent exceptions to Annalena’s rigid phonological rules should be considered in terms of their initial appearance in her output. All three of these forms were first produced at 0;11, before Annalena had 40 words in her lexicon; DOG and CAR were acquired before the 25-word point. Annalena thus has ample experience of these forms, and is able to continue producing segments within these forms which are otherwise avoided in her expanding lexicon. This highlights regression in Annalena’s output, which is typically found when infants begin implementing systematic templates (Vihman & Croft, 2007): early forms are produced with a higher level of accuracy before an increase in word acquisition, and thus the implementation of templates, prompts regression as words are adapted to fit these templates. Such regression can also be identified on a phonological level in Annalena’s output: for example Teddy is produced as [tεdə] at 0;10 before being adapted to a consonant harmony template as [bebɪ] at 1;0 (this template accounts for 66% of her word forms at 1;1), and Zeh ‘toe’, produced as [teː] at 0;11 is adapted as [deɪç] to fit a /ç/-final template at 1;0.
Elsen considers the principle of avoidance in a paper exploring the lexical and phonological constraints in Annalena’s output (1994). Here she describes how Annalena produced ‘simple forms … which enabled [her] to speak of corresponding referents successfully’ (p. 313) and while to some extent her argument supports observations made in this analysis, it seems that Annalena’s approach is more than simply the ad hoc avoidance or substitution of specific words and phonemes. Indeed, her approach is highly methodical, whereby a consistent template is implemented in order to facilitate production. The systematicity of Annalena’s OW template is striking, with a pattern force which brings about the creation of new OWs even in the absence of an established OW form in the adult language. A similarly systematic approach can be seen in later word production, as she lets go of these templates across the board in order to bring together her phonological and lexical abilities.
‘Wild’ features have been identified as characteristic of Annalena’s production of OWs, allowing her to create novel forms to serve in the implementation of the OW template. From Elsen’s own accounts (Elsen, 1991, 1994) we have a wealth of evidence showing that many OWs – BIRD, CROW, DUCK, HARE, HORSE and PIG; or 38% of her functional OWs – are not based on regularised OWs from the input, and must thus be adapted by Annalena to fit her OW template. This brings us to the discussion of the role of iconicity in onomatopoeia (Jespersen, 1950), as wild features provide what could be considered here as ‘added’ iconicity; this could be proposed to facilitate Annalena’s lexical development, especially during a period of marked change in her output.
However, this analysis suggests that it is the phonetic and phonological features of OWs which appear to have the most prominent role. It has been proposed that onomatopoeia may be easier to produce in early development (Kunnari, 2002), and indeed the OWs observed here have been found to exhibit features that do not necessarily match those of the language being acquired, or the expected output constraints that are implemented elsewhere in the lexicon. In most cases the ‘articulatory shape’ of these OWs is variable, owing to the marginal nature of these words; their fuzzy phonological boundaries appear to facilitate production by providing an impressionistic template which incorporates lexical meaning while also allowing a wide margin of phonological error. Indeed, many common characteristics of OWs – prosodic and extra-phonetic modifications, reduplication, vowel lengthening – are likely to require less planning, and demand less precise motor skills in production; it is therefore unsurprising that they feature so prominently in Annalena’s early output. In addition, these common features appear to enable the development of phonological precision through the production of word forms which are motorically uncoordinated and variable, but which can still be understood and are thus productive in communication.
The OW template is implemented during a period of marked change across Annalena’s output: we see the development of the consonant inventory as well as a steep increase in her lexical and phonological capabilities. The output is very much in transition during this period (1;0 to 1;2), but the OW template remains stable until these changes are established, after which point a general move towards the production of CW forms takes place. Again this demonstrates the systematic implementation of this template, and Annalena’s reliance on OW forms in maintaining stability in an increasingly unstable output.
Conclusion
OWs have been shown to provide a lexical template which allows Annalena to bridge a gap in her production capacity, as in all cases CW forms present phonological challenges beyond Annalena’s ability. This serves as a bootstrapping mechanism for word learning, as OWs provide a second option to some of the rigidly structured adult forms which Annalena eventually moves towards. Selection on both a phonological and a lexical level is apparent as she matches the words she produces with her phonological capacity.
From this case study it can be posited that OWs are more than just an irrelevant feature of a temporary babytalk lexicon. OWs were found to be abundant in the output, working alongside the developing lexicon rather than existing as a subsidiary feature to CWs. In Annalena’s case at least, onomatopoeia has been found to be lexically meaningful while also serving productively at the front line of phonological development.
Footnotes
Acknowledgements
The author would like to thank Marilyn Vihman, Tamar Keren-Portnoy, Chloë Marshall and the two anonymous reviewers for their invaluable feedback and support in the writing of this article. Thanks also go to Hilke Elsen for the fascinating and detailed account of her daughter’s language development.
Funding
The author’s PhD is being supported by the ESRC (grant number ES/J500215/1).