Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2009 Jun 1.
Published in final edited form as: Cognition. 2007 Dec 19;107(3):1093–1101. doi: 10.1016/j.cognition.2007.10.005

Acoustic emphasis in four year olds

Elizabeth Wonnacott 1, Duane G Watson 2
PMCID: PMC2491910  NIHMSID: NIHMS50576  PMID: 18070621

Abstract

Acoustic emphasis may convey a range of subtle discourse distinctions, yet little is known about how this complex ability develops in children. This paper presents a first investigation of the factors which influence the production of acoustic prominence in young children’s spontaneous speech. In a production experiment, SVO sentences were elicited from 4 year olds who were asked to describe events in a video. Children were found to place more acoustic prominence both on ‘new’ words and on words which were ‘given’ but had shifted to a more accessible position within the discourse. This effect of accessibility concurs with recent studies of adult speech (Dahan, Tanenhaus, & Chambers, 2002, Watson, Arnold & Tanenhaus 2005). We conclude that, by age four, children show appropriate, adult-like use of acoustic prominence, suggesting sensitivity to a variety of discourse distinctions.

Keywords: Language Development, Language Production, Prosody, Accent, Discourse, Accessibility

In adult speech, certain words may be emphasized over others, using some combination of increased pitch, higher intensity and longer duration in order to signal information about discourse structure. However, little is known about how this ability develops in children. In this paper, we investigate the relationship between acoustic prominence and discourse structure in four-year olds’ spontaneous production.

For adult speech, a standard hypothesis has been that acoustic prominence1 is a function of the givenness of information in the discourse: information that is new is emphasized while information that is given is not (Brown, 1983; Chafe, 1974; Prince, 1981). However, more recently, researchers have argued that the use of emphasis may reflect more fine-grained discourse distinctions. One notion that has played a central role in understanding discourse status is that of the accessibility of information within the discourse, i.e. the degree to which information is activated or salient. Information that has been mentioned frequently, or that is highly topical, is more accessible than information that is new, or information that has been mentioned previously but only peripherally. This property has been shown to affect word form choice and pronoun usage: referents that are highly accessible are typically referred to using pronouns and less accessible referents are typically referred to using full noun phrases (Ariel, 1990; Grosz & Sidner, 1986; Gundel, Hedberg, & Zacharski, 1993).

Dahan, Tanenhaus and Chambers (2002) proposed that the relationship between word form choice and accessibility extends to the use of acoustic prominence such that information that shifts from a low to a high level of accessibility is more likely to be produced with an accent than information that remains highly accessible in a discourse. To test this hypothesis, they examined how listeners interpreted accented information. In an eye-tracking study, participants were given instructions to move objects to new locations on a computer display. In line with the notion of a ‘given-new’ distinction, they found that when the object to be moved was accented, participants were faster at fixating referents that had not been mentioned than those that had been mentioned and had moved previously. However, participants were also faster at fixating an accented referent if the object had been mentioned but had not moved (e.g. it had been referred to as a landmark and so was not highly accessible). These findings suggest that listeners interpret acoustic prominence as signals about accessibility change.

Watson, Arnold and Tanenhaus (2005) found similar patterns in language production. Speakers described the movement of objects on a computer screen to a partner who had to mimic the actions, producing sentences like the ones below in (1):

  • (1) a. Given-NonShift Condition, e.g.

  • Put the bed above the house.

  • Put the bed above the pineapple.

  • b. Given-Shift Condition, e.g.

  • Put the house above the bed.

  • Put the bed above the pineapple.

  • c. New-Condition, e.g.

  • Put the house above the bell.

  • Put the bed above the pineapple.

The critical word in each trial was the direct object of the verb in the second sentence. Speakers used the most acoustic prominence when referring to referents that were new, (as in 1c), and given referents that had shifted accessibility in the discourse (1b). Speakers used less prominence on the critical word when the referent had not shifted its discourse status (1a).

With a few exceptions (see Snow & Balog, 2002), little is currently known about how the ability to use acoustic prominence develops in children, or how children might differ from adults. The question is important since sensitivity to discourse structure may not be fully developed in young children (e.g. Avrutin, 1999; Arnold, Novick, Brown-Schmidt, Eisenband & Trueswell, 2001). However, recent work on preschoolers’ pronoun use suggests at least some sensitivity to discourse accessibility in both comprehension and production. Referents that have occupied subject position previously in a discourse, a position which is associated with high accessibility, are more likely to be referred to with a pronoun by 4-year olds (Hickmann & Hendriks, 1999). Children at ages 2.5 to 3.5 are more likely to respond to a question with an answer that includes a pronoun if the referent was mentioned in the question (Campbell, Brooks, & Tomasello, 2000). 3-year old children are more likely to interpret a pronoun as referring to the subject than to the object (Song & Fisher, 2005). In sum, these findings suggest an ability to track shifts in discourse accessibility. We now ask whether these same discourse distinctions are reflected in their use of acoustic prominence.

In this experiment, we attempted to elicit pairs of SVO sentences from descriptions of events in a video. The agent of the second sentence was the target word and the first sentence established a discourse context. There were three trial types:

  • (2) a. Given–NonShift, e.g.

  • Sentence 1: The lion hit the giraffe.

  • Sentence 2: The lion hit the elephant.

  • b. Given–Shift, e.g.

  • Sentence 1: The giraffe hit the lion.

  • Sentence 2: The lion hit the elephant.

  • c. New, e.g.

  • Sentence 1: The elephant hit the giraffe.

  • Sentence 2: The lion hit the elephant.

The critical manipulation was the accessibility change of the subject in the second sentence. Subjects (and agents in particular) tend to be highly accessible referents in a discourse (Givon, 1990). Thus, in the Given-NonShift trial, the target word ‘lion’ is highly accessible in both the first and second sentences, while in the Given-Shift condition, the target word shifts from a position of low accessibility (theme), to a position of high accessibility (agent) 2. Finally, in the New condition, the target word has not been mentioned before, and thus also “shifts” to a prominent place in the discourse.

If children’s use of acoustic prominence is sensitive to these shifts in discourse structure, we will see differences in the production of the target word across conditions. One possibility is that acoustic prominence will convey less articulated discourse representations in child speech. For example, children might distinguish the simpler, possibly more salient contrast between given and new information, but not more fine-grained accessibility contrasts. If so, we predict more acoustic prominence on the target word in the New condition (2c) than in either of the Given conditions (2a and 2b). On the other hand, if children use acoustic prominence in the same discourse situations as adults, they will also use more acoustic prominence on the target word in the Given-Shift condition (3b) than in the Given-NonShift (3a) condition.

METHODS AND PROCEDURE

Participants

28 children between the ages of 3 years 5 months and 4 years 9 months were tested individually in the presence of their parents. Each child came in to the lab for a 30–40 minute session which began with free play, intended to encourage the child to be comfortable speaking to the experimenter. Children received a small toy at the end of the experiment.

Procedure

In order to elicit the sentences, we showed children short video scenes displaying pairs of hand puppets performing actions. In all of the displays, one puppet was the agent of an action of which the other was the theme. There were three actions (hit, kiss, hug) and five puppets performing the actions. Each trial consisted of the child viewing two scenes (involving the same action) one after the other. After viewing the first scene they were asked to describe it (e.g. “What happened?). After the second scene, they were asked to describe the scene in a manner intended to convey a continuing discourse (e.g. “And then what happened?”). Due to the interactive nature of the experiment, there was some variability in the precise form of the questions. Children also occasionally produced the target sentence before the experimenter had the opportunity to prompt them. Children sometimes needed some coaxing to produce a suitable utterance. Where children did not spontaneously produce a sentence of the correct form, the experimenter would encourage them to produce the required sentence. (The experiment was designed such that the child could push a button to view the video as many times as (s)he wished). If the child did not finally produce the required sentence, the experimenter provided the sentence, using a neutral prosody with neither noun receiving emphasis, and asked the child to copy it. Trials in which this occurred for the target utterance were excluded from the analysis. We also excluded: trials in which the child or experimenter referred to any of the subject, object or verb of the target sentence before the target sentence was uttered; trials in which the child repeated or replaced one of the words in uttering the target sentence (e.g. The lion --- no the bee hit the giraffe); trials in which the target sentence had an unusual form which made it difficult to identify the subject and object nouns for analysis (e.g. ‘it is putting the elephants eye on the giraffes stomach’). The decision as to whether a trial should be included was made by a coder blind to condition, who read a transcript of what occurred in each trial after the context sentence had been uttered. The trials were constructed such that each of the three nouns occurred as the target word once for each condition, making 9 trials in all (3 target nouns * 3 conditions – see Table 1). Trial order was randomized across participants. Participants took a short break between trials.

Table 1.

Sentences for each of the 9 trials (target word in bold).

Context Sentence Target Sentence
New The bee hit the ladybird. The giraffe hit the lion.
The elephant hugged the giraffe. The lion hugged the bee.
The bee kissed the giraffe. The ladybird kissed the elephant.
Given-Shift The elephant hugged the lion. The lion hugged the bee.
The bee kissed the ladybird. The ladybird kissed the elephant.
The ladybird hit the giraffe. The giraffe hit the lion.
Given-NonShift The ladybird kissed the giraffe. The ladybird kissed the elephant.
The giraffe hit the ladybird. The giraffe hit the lion.
The lion hugged the giraffe. The lion hugged the bee.
Open in a new tab

The session was recorded to a computer and analyzed using Praat acoustic analysis software (Boersma & Weenink, 2005)3. A research assistant labeled the word boundaries for subjects and objects in the target utterances and a script was used to automatically extract the acoustic measures. Three types of data were collected for both subjects and objects: (a) the maximum pitch (F0) reached at any point in the word (b) the overall intensity of the entire word and (c) the word’s duration. (Since children did not maintain a constant distance from the microphone, we anticipated that our measure of intensity would be highly variable).

For each child, we calculated the average for each of these measures in each condition. Children with missing cells were excluded. Our criteria excluded all of the data from 14 out of the 28 child participants.

RESULTS

The critical data are the maximum pitch, the intensity and the duration of the target words (i.e. the subject nouns of the target utterances) in each condition. These can be seen in the top portion of Table 2. The maximum pitch is significantly higher in the Given-Shift and New conditions than in the Given-NonShift condition (Given-Shift v Given-NonShift t=1.81, p< 0.05, df=13, New v Given-NonShift t=2.55, p<0.05, df=13). There is no difference between the Given-Shift and New conditions (t=0.77, p>0.05, df=13). Following the same pattern, intensity is significantly greater in the Given-Shift and New conditions than in the Given-NonShift condition (Given-Shift v Given-NonShift t=1.87, p< 0.05, df=13, New v Given-NonShift t=2.49, p<0.05, df=13). Again, there is no difference between the Given-Shift and New conditions (t=0.50, p>0.05, df=13). Recall, that the measurement of intensity was potentially influenced by the fact that children did not maintain a constant distance from the microphone. However, the close correlation with pitch suggests that this factor was not critical. There were no significant differences between any of the conditions for measurements of duration.

Table 2.

Acoustic Data. Data for the target subject nouns (top) with comparison data from object nouns and difference scores. (Standard error in parentheses).

new shift nonshift
SUBJECT (TARGET) NOUN maximum pitch within word (Hz) 312.64 (12.38) 323.93 (17.87) 302.37 (13.72)
intensity across word (dB) 66.51 (1.72) 67.08 (1.69) 65.13 (1.69)
duration of word (secs) 0.63 (0.05) 0.64 (0.04) 0.66 (0.05)
new shift nonshift
OBJECT NOUN maximum pitch within word (Hz) 296.27 (14.99) 279.08 (12.72) 286.63 (9.44)
intensity across word (dB) 62.83 (1.41) 61.81 (1.62) 63.06 (1.59)
duration of word (secs) 0.77 (0.05) 0.8 (0.06) 0.83 (0.03)
new shift nonshift
DIFFERENCE: SUBJ-OBJ NOUN maximum pitch within word (Hz) 16.37 (11.84) 44.85 (13.04) 15.74 (9.51)
intensity across word (dB) 3.68 (0.81) 5.27 (1.36) 2.07 (0.43)
duration of word (secs) −0.14 (0.07) −0.16 (0.06) −0.17 (0.05)
Open in a new tab

The target noun data thus demonstrate that this word is produced with greater pitch and intensity in the Given-Shift and New conditions than in the Given-NonShift condition. This result is in line with the adult data reported by Watson et al. (2005) and suggests that children emphasize words which have shifted in discourse accessibility. However a potential objection is that the properties of increased pitch and intensity could theoretically apply across a child’s entire utterance, rather than specifically to the target noun. To explore this possibility, we also performed equivalent analyses of acoustic measures of the object noun in each sentence. These data are also shown in Table 2. It can be seen that these data do not pattern with the data collected from subject nouns. (In particular, note that object nouns were actually produced with lower pitch and less intensity in the Given-Shift than in the Given-NonShift condition). There were no significant differences between any of the conditions for any of the three measures for these data. The hypothesis that entire utterances are emphasized in the Given-Shift and New conditions is thus unsupported.

To further test the extent to which subject nouns receive greater comparative emphasis in New and Given-Shift conditions, we also calculated the difference in the pitch and intensity between subject and object nouns in each utterance. This provides a measure of the degree to which the target word “stood out” from other words in the sentence. The average for each condition is also shown in Table 2. Although subject nouns were produced with overall greater pitch and intensity than object nouns across conditions, the difference in intensity was found to be significantly greater in the Given-Shift and New conditions than in the Given-NonShift condition (Given-Shift v Given-NonShift t=1.95, p< 0.05, df=13, New v Given-NonShift t=2.11, p<0.05, df=13), and the difference in pitch was found to be significantly greater in the Given-Shift than in the Given-NonShift condition (Given-Shift v Given-NonShift t=2.43, p<0.05, df=13). There was no reliable difference in the increase in pitch for the New condition than the Given-Shift condition (t=0.05, p>0.05, df=13). While the reasons for this lack of difference are currently unclear, overall these data support the hypothesis that our child participants use increased pitch and intensity specifically on the target nouns in the Given-Shift and New conditions.

DISCUSSION

The current work constitutes an important first step in investigating accent placement by young children. Four-year olds were found to use greater pitch and intensity on subject nouns that had shifted in discourse accessibility and less pitch and intensity on subject nouns that had not shifted.

One unexpected aspect of these data was that although emphasis was realized in differences in both pitch and intensity, there were no corresponding changes in duration. This is in contrast to Watson et al. (2005) who report increases in all three measures of acoustic prominence. However, even in adults, the extent to which the contribution of pitch, intensity and duration contribute to acoustic prominence is not well understood. Recent work suggests that these factors may be tied to differing functions in the discourse. For example, Watson et al. (in press) found that pitch and duration were tied to the predictability of information while intensity was tied to importance of information. Although the reason why children do not use duration to signal acoustic prominence in the current study is unclear, for our current purposes, the critical result is that children's use of pitch and intensity correlates with differences in accessibility status.

We suggest that the different treatment of target nouns across the conditions reflects children's ability to track shifts in discourse accessibility. Note that the effects cannot be attributed to priming associated with simple lexical repetition (a phenomenon well documented in the literature, see Bard & Aylett, 1999). That is, the lower pitch and intensity in the Given-NonShift condition cannot merely be a result of previous mention since the critical word in the Given-Shift condition was also previously mentioned, (and in fact was mentioned more recently - a factor which would actually be expected to boost lexical priming). Whilst we acknowledge that it is in theory possible that lexical priming could be boosted by the particular conditions seen in the Given-NonShift condition, specifically the fact that when the lexical item is repeated it occurs in the same syntactic position and with the same thematic role, we point out that such an account, which could equally apply to the equivalent adult data, could not explain non-prosodic discourse phenomena, such as appropriate pronoun use. Since previous research has found that children are sensitive to discourse accessibility in pronoun production, we consider that the most plausible account of the current findings is that children’s use of acoustic prominence reflects shifts in discourse accessibility. However, we acknowledge that our findings, and the adult data in Watson et al. (2005), are consistent both with an account in which the speaker explicitly represents the discourse, and with an account in which accessibility reflects a more passive aspect of processing. The extent to which prominence marking is listener or speaker centered is an important issue for future research.

Further research is also necessary in order to understand exactly how fine-grained the relationships between discourse shifts and acoustic prominence are. Watson et al. (2005) found what looked like a gradient relationship between prominence and accessibility shift: the greater the shift in accessibility, the greater the degree of acoustic prominence. Whether this also holds for child speech is an open question. A second question is whether children can understand accented information to refer to information that is not accessible in the discourse. The work here represents a first step in answering these questions.

Acknowledgements

This work project was supported by NIH grants 5T32 MH 19942-08 and NIH 2T32 DC 0000 35-11. We are very grateful to Elissa Newport for her helpful discussion on this topic, and for allowing us the use of her laboratory and resources. We would also like to thank Katie Daniels and Kathleen Eberhardt for their help in data collection and analysis.

Footnotes

1

We are focusing on what have traditionally been called ‘presentational accents’ (or H* in ToBI notation). However accenting is often discussed as a categorical phenomenon, and some researchers have argued that distinctions in the salience of a word may be continuous (e.g. see Ladd, 1996 for a discussion). We therefore use the more neutral term ‘acoustic prominence’ when discussing differences in salience.

2

An alternative possibility is that it is the change in accessibility, rather than the increase in accessibility, which leads to prominence (see Terken & Hirschberg, 1994). The current work does not differentiate between these two possibilities.

3

We also conducted a subjective analysis of prosody. This was carried out by two coders, blind to experimental condition, who coded each target utterance as ‘subject-accented’ or ‘subject-de-accented’ using a variant of the ToBI coding system. The vast majority of utterances were coded as accented (New 90%; Given-Shift 96%; Given-NonShift 90%) and there were no significant differences across condition. This study thus adds to a growing body of research which finds differences in acoustic prominence across different discourse conditions without concurrent differences in the rate of accenting (Watson, Arnold, & Tanenhaus, 2005; Bard & Aylett, 1999), and which challenges the notion that acoustic emphasis can be described in terms of a simple categorical distinction between accented and de-accented words. The relationship between acoustic prominence and the subjective perception of accenting will be an important topic for future research, but is beyond the scope of the current study.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Contributor Information

Elizabeth Wonnacott, Department of Experimental Psychology, University of Oxford.

Duane G. Watson, Department of Psychology, University of Illinois – Urbana Champaign

References

  1. Akhtar N, Carpenter M, Tomasello M. The role of discourse novelty in early word learning. Child Development. 1996;67(2):635–645. [Google Scholar]
  2. Arnold JE, Novick JM, Brown-Schmidt S, Eisenband JG, Trueswell JC. Knowing the difference between girls and boys: The use of gender during on-line pronoun comprehension in young children. In: Do AH-J, Domínguez` L, Johansen A, editors. Proceedings of the 25th annual Boston University Conference on Language Development; Cascadilla Press; 2001. pp. 59–69. [Google Scholar]
  3. Avrutin S. Development of the syntax–discourse interface. Kluwer: Dordrecht; 1999. [Google Scholar]
  4. Ariel M. Accessing noun-phrase antecedents. London: New York: Routledge; 1990. [Google Scholar]
  5. Bard EG, Aylett MP. ICPHS-99. San Francisco, CA: 1999. The dissociation of deaccenting, givenness, and syntactic role in spontaneous speech. [Google Scholar]
  6. Boersma P, Weenink D. [Retrieved December 11, 2005];Praat: doing phonetics by computer (Version 4.3.36) [Computer program] 2005 from http://www.praat.org/
  7. Bolinger DLM. Intonation and its parts: Melody in spoken English. Stanford, Calif.: Stanford University Press; 1986. [Google Scholar]
  8. Brown G. Prosodic structure and the given/new distinction. In: Cutler A, Ladd DR, editors. Prosody: Models and measurement. Berlin; New York: Springer-Verlag; 1983. pp. 67–77. [Google Scholar]
  9. Brown RW. A first language: The early stages. Cambridge, Mass.: Harvard University Press; 1973. [Google Scholar]
  10. Campbell AL, Brooks P, Tomasello M. Factors affecting young children's use of pronouns as referring expressions. Journal of Speech, Language, and Hearing Research : JSLHR. 2000;43(6):1337–1349. doi: 10.1044/jslhr.4306.1337. [DOI] [PubMed] [Google Scholar]
  11. Chafe WL. Language and consciousness. Language. 1974;50(1):111–133. [Google Scholar]
  12. Cutler A, Dahan D, van Donselaar W. Prosody in the comprehension of spoken language: A literature review. Language and Speech. 1997;40(Pt 2):141–201. doi: 10.1177/002383099704000203. [DOI] [PubMed] [Google Scholar]
  13. Dahan D, Tanenhaus MK, Chambers CG. Accent and reference resolution in spoken-language comprehension. Journal of Memory and Language. 2002;47(2):292–314. [Google Scholar]
  14. Givon T. Syntax: A Functional Typological Introduction. Amsterdam: John Benjamins; 1990. [Google Scholar]
  15. Grosz BJ, Sidner CL. Attention, intentions, and the structure of discourse. Comput.Linguist. 1986;12(3):175–204. [Google Scholar]
  16. Gundel JK, Hedberg N, Zacharski R. Cognitive status and the form of referring expressions in discourse. Language. 1993;69(2):274–307. [Google Scholar]
  17. Hickmann M, Hendriks H. Cohesion and anaphora in children's narratives: A comparison of English, French German, and Mandarin Chinese. Journal of Child Language. 1999;26(2):419–452. doi: 10.1017/s0305000999003785. [DOI] [PubMed] [Google Scholar]
  18. Ladd DR. Intonational Phonology. Cambridge: Cambridge University Press; 1996. [Google Scholar]
  19. Perner J, Leekam SR. Belief and quantity: Three-year olds' adaptation to listener's knowledge. Journal of Child Language. 1986;13(2):305–315. doi: 10.1017/s0305000900008072. [DOI] [PubMed] [Google Scholar]
  20. Prince E. Toward a taxonomy of given-new information. In: Cole P, editor. Radical pragmatics. New York, NY: Academic Press; 1981. pp. 223–255. [Google Scholar]
  21. Snow D, Balog HL. Do children produce the melody before the words? A review of developmental intonation research. Lingua. 2002;112(12):1025–1058. [Google Scholar]
  22. Song H, Fisher C. Who's "she"? Discourse prominence influences preschoolers' comprehension of pronouns. Journal of Memory and Language. 2005;52(1):29–57. [Google Scholar]
  23. Tomasello M, Call J, Hare B. Chimpanzees understand psychological states - the question is which ones and to what extent. Trends in Cognitive Sciences. 2003;7(4):153–156. doi: 10.1016/s1364-6613(03)00035-4. [DOI] [PubMed] [Google Scholar]
  24. Watson D, Arnold JE, Tanenhaus MK. Not just given and new: The effects of discourse and task based constraints on acoustic prominence; Poster presented at the 2005 CUNY human sentence processing conference; Tucson, AZ. 2005. [Google Scholar]
  25. Watson D, Arnold JE, Tanenhaus MK. Tic Tac TOE: Effects of predictability and importance on acoustic prominence in language production. Cognition. doi: 10.1016/j.cognition.2007.06.009. (in press) [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES