Introduction
Usually, speakers change their speech style based on their listener by adjusting various acoustic characteristics that are associated with prosody, including mean overall pitch (fundamental frequency (f0) mean) and other pitch-related features (f0 range, variability, contour, etc.) (e.g. Falk, Reference Falk2004; Saint-Georges et al., Reference Saint-Georges, Chetouani, Cassel, Apicella, Mahdhaoui, Muratori, Laznik and Cohen2013). While talking to infants, caregivers tend to use higher overall pitch, wider pitch range, specific pitch contours, and longer utterances (e.g. Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Golinkoff et al., Reference Golinkoff, Can, Soderstrom and Hirsh-Pasek2015). There is ample evidence that the characteristics of infant-directed speech prosody serve multiple functions. These functions include capturing and maintaining the infant’s attention, strengthening the bond between the infant and the caregiver through enhanced positive interactions, facilitating language acquisition, expressing emotions, and conveying information about the speaker’s intentions and identity. As a result, infant-directed speech plays an essential role in the healthy emotional and cognitive development of children (for a review, see Soderstrom, Reference Soderstrom2007). In the past decades, there has been growing interest in more systematic and controlled investigations, which have the potential to reveal more exact functions and related acoustic features in infant-directed speech prosody. In the present study, we focused on the potential functions of two pitch-related characteristics (f0 mean and range) and one utterance length-related feature (call length) of infant-directed speech prosody.
Effect of situation
One approach is to investigate and compare infant-directed speech prosody in and between different situations and contexts. With this method, it has been shown that various pitch characteristics can play distinct functions and roles during tutoring interactions with preverbal infants. More precisely, specific large pitch contours in infant-directed speech (which manifests in a wider f0 range) have the potential to facilitate word segmentation and, thus, language acquisition (e.g. Thiessen et al., Reference Thiessen, Hill and Saffran2005; Trainor & Desjardins, Reference Trainor and Desjardins2002). By contrast, it has been suggested that a higher overall pitch (i.e. f0 mean) not only does not facilitate but actually also impedes word segmentation. Simultaneously, it plays an essential role in capturing and controlling infants’ attention and expressing emotions (e.g. Cooper & Aslin, Reference Cooper and Aslin1994; Trainor & Desjardins, Reference Trainor and Desjardins2002). It has also been suggested that tutoring and playing situations involving objects contain less exaggerated prosody (i.e. lower f0 mean and smaller f0 range) to effectively divide infants’ attention between the object and the speaker (e.g. Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017; Gogate et al., Reference Gogate, Bolzani and Betancourt2006).
The relevancy (i.e. infant directedness) and naturalness (i.e. fixed sentences or text reading versus free speech) of the given situation also affect the pitch characteristics of prosody. When a specific text, such as a story from a book, had to be read to children, speakers used lower f0 mean and smaller f0 range compared to situations where they were allowed to speak freely to the infant (e.g. Shute & Wheldall, Reference Shute and Wheldall1999; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). At the same time, fixed sentences that are pronounced rhythmically and melodically and have typical infant-directed content (e.g. rhymes and playsongs) seem to have distinctive, intense, and consistent acoustic prosody with heightened f0 mean (e.g., Falk & Audibert, Reference Falk and Audibert2021).
Effect of the partners’ needs and capacities
Another feasible approach to studying functions of prosody and related acoustic features is to compare them across different partners with varying emotional needs and cognitive capacities. Using such a comparative method, it has been revealed that people tend to employ strikingly similar acoustics, including higher f0 mean and wider f0 range when talking to infants and pets, which differ significantly from the speech towards unfamiliar adults (e.g. Hirsh-Pasek & Treiman, Reference Hirsh-Pasek and Treiman1982; Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). It has been suggested that one basic function of such exaggerated prosody is to evoke and maintain the attention of partners with limited linguistic competence, whether conspecific or heterospecific (e.g. Hirsh-Pasek & Treiman, Reference Hirsh-Pasek and Treiman1982; Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). Besides the acoustic similarities, there is also evidence that the given context and the naturalness of the situation similarly influence the f0 mean and range of infant- and dog-directed speech (e.g. Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017).
This comparative framework has also revealed a relationship between utterance lengthening (i.e. vowel hyperarticulation) and the linguistic competence of the intended addressee: speakers used the longest vowels towards infants (i.e. future speakers) than towards parrots (i.e. expected future speakers), but not towards dogs or cats (i.e. non-speakers; e.g. Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017; Xu et al., Reference Xu, Burnham, Kitamura and Vollmer-Conna2013). The aforementioned results supported the language tutoring function of utterance lengthening and provided evidence that, similarly to pitch characteristics, speakers adjust these parameters as well to their audience’s expected needs and capacities.
Conveying positive emotions, expressing affection, and strengthening attachment are listed among the most important functions of infant-directed prosody, to which heightened and wider-ranged f0 contributes greatly (e.g. Fernald, Reference Fernald, Papoušek, Jürgens and Papoušek1992; Trainor et al., Reference Trainor, Austin and Desjardins2000). Moreover, it has been suggested that the striking acoustic differences between adult- and infant-directed prosody are by-products of speakers’ emotional expressions when interacting with infants and inhibited when talking to adults (Trainor et al., Reference Trainor, Austin and Desjardins2000). Facial expressions accompanied by infant- and adult-directed acoustic prosody seem to support this notion, as more exaggerated facial expressions are displayed towards infants than towards adult partners (e.g. Chong et al., Reference Chong, Werker, Russell and Carroll2003; Gergely et al., Reference Gergely, Koós-Hutás, Filep, Kis and Topál2023). It is important to note, however, that in the aforementioned studies, prosody towards one’s own infant was compared to speech prosody towards a nice but unfamiliar adult partner (i.e. experimenters). As attachment and personal relationships between the interactants greatly impact speakers’ emotions and speech prosody (e.g. Bombar & Littig, Reference Bombar and Littig1996), the feasibility of comparing speech prosody towards unfamiliar adults and own infants has been questioned (Trainor et al., Reference Trainor, Austin and Desjardins2000).
Xu and co-workers (Reference Xu, Burnham, Kitamura and Vollmer-Conna2013) used the same unfamiliar partners (adult, dog, or parrot) with all female speakers in their study and provided evidence that acoustic differences between adult- and pet-directed speech are still evident when familiarity between conditions is equalized (Xu et al., Reference Xu, Burnham, Kitamura and Vollmer-Conna2013). In a recent study, Koós-Hutás and co-workers (Reference Koós-Hutás, Kovács, Topál and Gergely2024) compared facial emotional expressions and emotional states of female and male speakers when interacting with their 6- to 18-month-old infants, their spouses, and their family dog. Contrary to previous findings with unfamiliar adult partners, speakers in this study showed similarly intense emotions and related facial expressions during infant- and adult (i.e. spouse)-directed conditions (Gergely et al., Reference Gergely, Koós-Hutás, Filep, Kis and Topál2023; Koós-Hutás et al., Reference Koós-Hutás, Kovács, Topál and Gergely2024). These results highlight the importance of taking personal relationships into account between the interactants (Trainor et al., Reference Trainor, Austin and Desjardins2000). It is also important to note that, in this study, speakers used less intense and less positive facial expressions with their family dogs than with their infants and spouses suggesting that facial expressions might follow different dynamics and have different functions than pitch characteristics, which speakers use similarly with dogs and infants (Hirsh-Pasek & Treiman, Reference Hirsh-Pasek and Treiman1982; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017; Koós-Hutás et al., Reference Koós-Hutás, Kovács, Topál and Gergely2024).
Effect of the speakers’ sex
According to the current state of the literature, acoustic features as well as utterance length-related properties of infant-directed speech are more similar than different among women and men (for a review, see Ferjan Ramírez, Reference Ferjan Ramírez2022). There is ample evidence that both sexes use higher pitch during infant-directed speech than during adult-directed speech (e.g. Niwano & Sugai, Reference Niwano and Sugai2003; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017; Weirich & Simpson, Reference Weirich and Simpson2019). Pitch range, on the other hand, presents a more variable picture of how sex differences are manifested in infant- and adult-directed conditions. Several studies have reported wider pitch range in female speakers than in male speakers during parent–infant interactions in various contexts and languages, including spontaneous and read speech situations (e.g. Fernald et al., Reference Fernald, Taeschner, Dunn, Papousek, de Boysson-Bardies and Fukui1989; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). However, other studies have found no sex differences in infant-directed pitch range (e.g. Shute & Wheldall, Reference Shute and Wheldall1999; Niwano & Sugai, Reference Niwano and Sugai2003) or have shown that male speakers use a wider range than female speakers (e.g., Warren-Leubecker & Bohannon, Reference Warren-Leubecker and Bohannon1984). When it comes to pet-directed speech, there is evidence that both sexes use similarly heightened pitch and wide pitch range when talking to dogs as opposed to adults, but similar to that of infant-directed speech (Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). Moreover, both sexes hyperarticulate their vowels with infants, but not with dogs and unfamiliar adults (e.g. Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017).
Aims and hypotheses
In the present study, we aimed to investigate the functions of two pitch-related parameters (f0 mean and range) of infant-directed acoustic prosody by comparing them across different situations and partners in both women and men. To achieve this, we analysed speech samples from our recently published comparative study (Koós-Hutás et al., Reference Koós-Hutás, Kovács, Topál and Gergely2024), in which female and male speakers interacted with their own infants (infant-directed condition), own spouses (adult-directed condition), and own family dogs (dog-directed condition) during two free speech situations (attention getting and language tutoring) and one fixed sentences situation with a nursery rhyme (fixed sentences). In addition to f0 mean and range, we also aimed to study one utterance length-related parameter (call length) during the language tutoring situation, to examine whether speakers adjust their uttering in line with the partners’ expected linguistic competence.
Our first research question was as follows: (1) whether and how different speech situations affect the speakers’ mean pitch and pitch range towards their infants, spouses, and dogs. Heightened f0 mean proved to be crucial for capturing and maintaining the attention of partners with limited linguistic competence (i.e. infants and dogs; e.g. Fernald & Kuhl, Reference Fernald and Kuhl1987; Jeannin et al., Reference Jeannin, Gilbert, Amy and Leboucher2017). However, a heightened f0 mean might impede word segmentation, while a wider f0 range has the potential to facilitate language acquisition (e.g. Trainor & Desjardins, Reference Trainor and Desjardins2002). We hypothesized, therefore, that the attention-getting situation, in which speakers were instructed to get and maintain the focus of their partners on themselves, would evoke higher f0 mean when speaking to infants and dogs compared to adults. Additionally, we predicted that speakers would use a lower f0 mean and wider f0 range when talking to infants compared to dogs during the language tutoring situation. Concerning the fixed sentences situation, in which speakers were instructed to tell three everyday-like sentences along with a nursery rhyme to the partners, we could predict two different outcomes based on the literature. There is evidence that speakers use less exaggerated prosody with their partners during less naturalistic and more restricted situations (e.g., Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017; Jürgens et al., Reference Jürgens, Hammerschmidt and Fischer2011) which suggests lower f0 mean and smaller f0 range in this situation compared to the two free speech situations. On the other hand, it has also been shown that rhythmic and melodic speech and the infant directedness of a speech affect prosody and can evoke intense acoustics from the speakers (e.g. Falk & Audibert, Reference Falk and Audibert2021). Therefore, it is also possible that the fixed sentences situation with a nursery rhyme will evoke similar or even more exaggerated prosody with a higher f0 mean and wider f0 range, irrespective of the type of the partner, compared to the free speech situations.
The second research question of the present study was as follows: (2) whether and how speakers adjust mean pitch, pitch range, and utterance length according to their partners’ expected language competence. If such adjustments occur, we would expect a higher mean f0 and a wider f0 range when addressing partners with developing linguistic skills (i.e. infants) or limited linguistic skills (i.e. dogs) compared to fully competent speakers (c.f. Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). Based on the results of previous studies on hyperarticulation and acoustics (e.g. Trainor & Desjardins, Reference Trainor and Desjardins2002; Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017; Xu et al., Reference Xu, Burnham, Kitamura and Vollmer-Conna2013), we may expect that speakers will use longer utterances (i.e. call length), lower f0 mean, and wider f0 range to facilitate word segmentation for potential speakers (i.e. infant) when uttering a to-be-thought object label (i.e. language tutoring situation). However, shorter utterances (i.e. call length), higher f0 mean, and smaller f0 range are expected when speakers are uttering it to non-speakers (i.e. dogs; Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017; Xu et al., Reference Xu, Burnham, Kitamura and Vollmer-Conna2013). Speakers are expected to use no speech modifications to enhance word segmentation when interacting with equally competent speakers (i.e. their spouses, e.g. Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017).
Alternatively, it is also possible that speakers’ emotions play a more significant role in regulating speech prosody than the audience’s needs and capabilities. Recently, these speakers’ facial expressions and related emotional content were analysed and showed that both female and male speakers in all examined situations used more frequent and intense happy emotions when interacting with their infants and spouses than with their dogs (Koós-Hutás et al., Reference Koós-Hutás, Kovács, Topál and Gergely2024). We can hypothesize that the acoustics of the accompanied speech will follow this emotional pattern of the speakers, and as a “by-product” of happy speech, we can predict heightened and wider-ranged f0 when interacting with the spouses and infants than with the dogs (e.g. Fernald, Reference Fernald, Papoušek, Jürgens and Papoušek1992; Trainor et al., Reference Trainor, Austin and Desjardins2000).
The third research question we aimed to study was as follows: (3) whether and how speakers’ sex affects the two pitch-related and one utterance length-related parameters of their speech. Based on the literature, aforementioned hypotheses, and predictions regarding f0 mean, we expect similar patterns in female and male speakers (e.g. Niwano & Sugai, Reference Niwano and Sugai2003; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). However, a wider f0 range will likely be observed in female speakers compared to male speakers (Fernald et al., Reference Fernald, Taeschner, Dunn, Papousek, de Boysson-Bardies and Fukui1989; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). According to previous results, we also expect female and male speakers to modulate their utterance length similarly (e.g. Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017).
Materials and methods
Ethics statement
This research was approved by the Human Research Ethics Committee (EPKEB) at the Hungarian Academy of Sciences (No. 2022-85). All parents gave their written consent to engage in the research in accordance with ethics approval, and all procedures were carried out in accordance with the relevant rules and regulations of the EPKEB and the applicable laws of Hungary.
Participants
Both parents from 22 families (N=44; 22 women and 22 men; mean age ± standard deviation [SD]: 34.6 ± 4.4 years; urban, heterosexual, and middle-class families) voluntarily participated in this research (Koós-Hutás et al., Reference Koós-Hutás, Kovács, Topál and Gergely2024). Each family had their own infant (6–18 months old; 10 girls and 12 boys; mean age ± SD: 10.2 ± 3.7 months) and a pet dog that is at least 1 year old (13 female and 15 male dogs; mean age ± SD: 6 ± 3.7 years). All the parents were instructed to interact with their baby (infant-directed condition) and their family dog (dog-directed condition). If there were more than one dog in the family, the speakers had the liberty to interact with different dogs, choosing those with whom they felt most comfortable. During the adult-directed condition, they interacted with their spouses. All participants had Hungarian as their first language. Demographic details of the participating interactants are reported in the supplementary material (Table S1).
Procedure
Data collection took place at the participants’ homes in the presence of two experimenters. One of them managed the technical equipment required for the recording, while the other supervised the entire process. Before beginning, each parent signed an informed consent form. After that, each mother and father were recorded individually while interacting with their own infant, dog, and spouse in a within-subject design. Speakers were instructed to occupy seats about 30 centimetres away from the addressee at eye level or lower to avoid data loss of the speaker’s face by gazing down (see Figure 1; Koós-Hutás et al., Reference Koós-Hutás, Kovács, Topál and Gergely2024). Leaning over or touching the addressee in certain circumstances was not strictly forbidden, but the speakers were encouraged to try to maintain their position throughout the interaction. Adult partners (i.e. spouses) were instructed to maintain a sit position during the experiment, and dogs were placed in a sit or down position at the same spot, while infants were sitting in a baby chair or the spouse’s lap or the experimenter’s lap during the interactions (see Figure 1).
Figure 1. Experimental arrangement. (a) Dog-directed condition, (b) adult (spouse)-directed condition, and (c) infant-directed condition.
Speech interactions were recorded in three different situations – attention getting, language tutoring, and fixed sentences – using the same microphone (Zoom F2 recorder with LMF-2 Lavalier microport). Smartphones were also used during the study to record data for a separate analysis, which was reported in another study (Koós-Hutás et al., Reference Koós-Hutás, Kovács, Topál and Gergely2024). Participants were told to engage in natural conversation with the addressees during each recording phase, which consisted of three situations. The order of situations and conditions was counterbalanced across participants.
Attention-getting situation (1 minute)
Participants were told to capture the addressee’s attention and maintain his/her attentional focus (preferably by maintaining eye contact) for one minute. We aimed to observe how the speaker naturally manages to maintain the addressee’s attention, so we did not provide specific instructions to the speakers on how to complete the tasks.
Language tutoring situation (1+1 minutes)
During this situation, speakers were instructed to teach an object–label association to their partners (presentation phase), and then, the partner was asked to select the labelled object (two-way choice task). To do so, the experimenter chose randomly two objects out of five, all of which were novel to the partners (see Figure 2). One object was randomly assigned as a target object and the other one as a non-target object. Then, the experimenter randomly selected one of the predetermined three artificial words (“danidu,” “burida,” and “zibula”) and asked the speaker to label the target object using this word while interacting with his/her baby, dog, or spouse. When creating the words for object labels, we aimed to use novel words without meaning that interactants had never heard before. Note that all labels were required to contain the three syllables necessary to draw vowel triangles (i.e. i, a, u) for future studies aiming to investigate hyperarticulation.
Figure 2. Set of potential target and non-target objects used in the language tutoring situation.
Language tutoring – presentation phase (1 minute)
The speakers’ task was to associate artificial labels with the target object while holding both the target and non-target objects in their hands. Speakers were instructed to use only demonstrative words such as “this,” “that,” “thing,” and “something” when referring to the non-target object. They were told to talk about both the target and non-target objects separately for at least half a minute, using the predetermined label (referring to the target) and the demonstrative words (referring to the non-target) as frequently as it is possible (for a similar method, see Woodward et al., Reference Woodward, Markman and Fitzsimmons1994). The addressee was not allowed to touch the objects during this phase.
Language tutoring – two-way choice task phase (1 minute)
After about a minute, the speaker moved on to the second phase and encouraged the addressee to select the target object with these words: “Which one is the danidu/burida/zibula?”. During this phase, speakers were instructed to hold the two objects still at an equal distance (at arm’s length) from the addressee. If needed, speakers were allowed to encourage the partner verbally to choose without moving the objects. After choosing an object, the addressee was allowed to touch and explore the chosen object, and the speaker was allowed to praise the partner. Then, the speaker kindly asked for the object back from the partner, switched the position of the target and non-target objects in her/his hands, and repeated the whole “choosing” procedure once more.
Fixed sentences situation (1 minute)
Participants were instructed to recite a nursery rhyme and three previously specified sentences to the addressee. The fixed three sentences were as follows: (#1) Nézd csak, milyen szép idő van odakint! (in English: Just look! What nice weather!), (#2) Akarsz sétalni egyet? (in English: Do you want to go for a walk?), and (#3) Úgy látom, unatkozol. Nem csinálunk valami mást? (in English: You seem really bored. Shouldn’t we do something else?).
Apart from the three fixed sentences, speakers were also asked to recite the following well-known Hungarian nursery rhyme: Cini-cini muzsika; táncol a kis Zsuzsika; jobbra dől, balra dől; tücsök koma hegedül (in English: “Cini-cini music plays; little Susan dances away; leaning to the right, leaning to the left; the cricket buddy plays the fiddle”).
Data analysis
Acoustic analysis
We used acoustic data (i.e. the audio file recorded by the microport) from our recent study in which only the facial prosodic features of the speakers were analysed (Koós-Hutás et al., Reference Koós-Hutás, Kovács, Topál and Gergely2024). The analysis of the acoustic recordings from all three situations was done in line with Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017, with the help of the Praat software (version 6.0.05; Boersma & Weenink, Reference Boersma and Weenink2021). It is important to note that for the analysis we used only recordings of the Zoom microport and not the smartphones. At first, we used a semi-automatic script to annotate the recordings, defining and labelling pauses and calls and excluding background sounds. We applied a call-based approach for our analyses similar to Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017. One call, in terms of bioacoustics, can be considered as a functional unit in the speech stream intonation contour which usually contains one voiced sound. Calls are separated by pauses, breathtaking, and unvoiced sounds, similarly to utterance units. The baseline search range was defined between 75 Hz and 500 Hz, and before the pitch extraction, the coder checked visually the detection of the pitch contour for halving and doubling errors and modified the range if it was necessary. This way we could ensure the minimal level of artefacts in the measurements and we could also exclude intermittent vocalizations as well as remaining background noises from the sample. Then, we exported the following acoustic characteristics of each call from the programme:
f0 mean: It refers to the mean of the fundamental frequency (f0, perceived as pitch) of each call (40148 calls in total, 13620 in adult-directed, 13274 in dog-directed, and 13254 in infant-directed conditions). The analysis was performed using Praat’s built-in cross-correlation-based pitch extraction method.
f0 range: The Praat software’s built-in function was used to calculate each call’s f0 range by subtracting f0 minimum from f0 maximum.
Call length: The Praat software’s built-in function was used to analyse the call length of the object labels. This analysis aimed to investigate whether speakers uttered the label differently when talking to infants, dogs, and adults. When labels were not isolated, we manually separated them in Praat software by using tiers, ensuring that all labels (i.e. danidu/burida/zibula) were analysed as a single continuous call (2112 calls/labels in total).
Statistical analysis
RStudio (https://www.rstudio.com/) was used for the statistical analysis (R version 4.2.3 using RStudio 2023.06.0+421, R Core Team 2023). To analyse f0 mean and range, we used generalized linear mixed models (nlme and lme4 package and glmer and lme functions; Bates et al., Reference Bates, Mächler, Bolker and Walker2015; Pinheiro & Bates, Reference Pinheiro and Bates2000) with the Akaike information criterion (AIC)-based backwards elimination (MASS package and drop1 function; Venables & Ripley, Reference Venables and Ripley2002) to find parsimonious models. Due to the anatomy-based difference in f0 mean of women and men (Titze, Reference Titze1989), f0 mean of female and male speakers was analysed with separate models for the whole dataset and for the object label analysis. As the data distribution was skewed towards low values, we normalized them with log transformation. Also, as fixed sentences situation had lower variance, we controlled for heteroscedasticity in these models by adding situation-dependent weights to the model. In f0 mean models, for the whole dataset, condition (infant-, adult-, and dog-directed), situation (attention getting, language tutoring, and fixed sentences), and their interaction were included as fixed effects. For f0 range and call length analysis, female and male speakers were included in the same model; therefore, the effect of sex (female and male) and all two- and three-way interactions with condition (f0 range and call length) and situation (f0 range) were included. In object label models (f0 mean, f0 range, and call length variables), condition, sex, and their interaction were included. First, we included speaker identity number (ID) and family ID to the models as random intercepts (speaker nested in family) to control for dependence and repeated measurements. After comparing model performance (compare performance function) and checking the explained variance, family ID was dropped out as it explained no variance, and only speaker ID was included as a random intercept in all final models. For post hoc pairwise comparisons, we used the Tukey method (emmeans package; Lenth, Reference Lenth2023).
Results
First, we will present the significant interactions and main effects (i.e. situation, condition, and speakers’ sex) for all analysed prosodic features (i.e. f0 mean and range, call length). Then, we will present the post hoc analysis and pairwise comparisons according to the research questions (for summary, see Table 1).
Table 1. Summary of the acoustic features of object labels during the language tutoring situation in all three conditions
Significant interactions and main effects
According to the f0 mean (all calls), model selection showed a significant interaction effect of condition × situation in both female (LRT: χ2 4=133.93, p<0.001) and male (LRT: χ2 4=98.885, p<0.001) speakers. According to the f0 range (all calls), model selection showed a significant interaction effect of condition × situation (LRT: χ2 4=16.04, p<0.001), speakers’ sex × condition (LRT: χ2 2=8.03, p=0.018), and speakers’ sex × situation (LRT: χ2 2=10.28, p=0.006). When it comes to the object labels, the model selection of f0 mean (labels) showed a significant main effect of condition in both female (LRT: χ2 2=98.83, p<0.001) and male (LRT: χ2 2=36.71, p<0.001) speakers. In object labels, the model selection also showed a significant interaction effect of speakers’ sex × condition both for f0 range (labels, LRT: χ2 2=6.72, p=0.035) and for call length (labels, LRT: χ2 2=20.86, p=0.035).
Effect of speech situation
Speakers used similarly high pitch during fixed sentences and attention-getting situations when interacting with their infants (p>0.05), but a lower f0 mean was observed during the language tutoring situation in both male and female speakers (all p<0.05; see Figure 3 for summary, and see Table S2 and Figure S1 for detailed statistics). Speakers used the highest f0 mean during the fixed sentences situation (all p<0.05) and a similarly lower one in attention-getting and language tutoring situations when talking to their dogs (p>0.05; see Figure 3 for summary, and see Table S2 and Figure S1 for detailed statistics). When interacting with their spouses, both sexes used the highest f0 mean during fixed sentences, followed by language tutoring and finally during attention-getting situations (all p<0.05; see Figure 3 for summary, and see Table S2 and Figure S1 for detailed statistics).
Figure 3. Fundamental frequency (f0) mean (Hz) of female and male speakers during situations across all three conditions.
Pairwise comparisons revealed general patterns of speech situation on speakers’ f0 range. The widest range was observed during the fixed sentences situation, followed by language tutoring and finally in attention-getting situation in both sexes across all three conditions (all p<0.05; see Figure 4 for summary, and see Table S3 and Figure S2 for detailed statistics).
Figure 4. Fundamental frequency (f0) range (Hz) of female and male speakers during situations across all conditions.
Effect of the partners’ linguistic competence
Pairwise comparisons of f0 mean showed that both female and male speakers used a higher f0 mean towards their infants and dogs than towards their spouses in all three situations (all p≤0.001; see Figure 3 for summary, and see Table S2 and Figure S1 for detailed statistics). F0 mean was similar towards infants and dogs in female speakers during the language tutoring situation and in male speakers during the fixed sentences situation (both p>0.05; see Figure 3 for summary, and see Table S2 and Figure S1 for detailed statistics). However, the pattern of f0 mean towards infants compared to dogs exhibited greater diversity. In the attention-getting situation, speakers from both sexes employed a higher f0 mean towards infants than towards dogs (all p<0.05; see Figure 3 for summary, and see Table S2 and Figure S1 for detailed statistics). In the language tutoring situation, male speakers used an even higher f0 mean towards dogs than towards infants, while female speakers maintained a similar f0 mean towards dog and infant partners during this situation (Figure 3; see Figure 3 for summary, and see Table S2 and Figure S1 for detailed statistics). During the fixed sentences situation, female speakers used a higher f0 mean with infants than with dogs, while male speakers maintained a similar f0 mean across infant-directed and dog-directed interactions in this situation (Figure 3; see Figure 3 for summary, and see Table S2 and Figure S1 for detailed statistics).
In both female and male speakers, the widest f0 range was observed towards infants, then towards dogs, and finally towards adults in almost all situations. The only exception was detected in the fixed sentences situation, during which infant- and dog-directed speech contained a similar f0 range (see Figure 4 for summary, and see Table S3 and Figure S2 for detailed statistics).
Pairwise comparisons showed that male speakers used the highest f0 mean when uttering object labels towards their dogs, followed by their infants and finally towards their spouses (all p<0.05; see Table 1 for summary, and see Table S4 and Figure S3 for detailed statistics). At the same time, female speakers used similarly high f0 mean when conveying object labels to their dogs and infants, while they also used a lower f0 mean when conveying the object labels to their spouses (see Table 1 for summary, and see Table S4 and Figure S3 for detailed statistics).
Pairwise comparisons also showed that female speakers used a wider f0 range of object labels when speaking to infants compared to dogs or adults (all p<0.05). However, they used a similar range when addressing dogs and adults (all p>0.05; see Table 1 for summary, and see Table S4 and Figure S4 for detailed statistics). Additionally, male speakers used a similar range when conveying object labels to infants, dogs, and adults (all p>0.05, Figure 4; see Table 1 for summary and Table S4 for detailed statistics). Consistent with the f0 range model results on the whole dataset, female speakers generally exhibited a wider f0 range than males across all conditions (all p>0.05, Figure 4; see Table 1 for summary and Table S4 for detailed statistics).
We found that both sexes uttered the object label longer to their infants and their spouses than towards their dogs, while they used a similar call length towards their infants and their spouses (Figure 5; see Table 1 for summary and Table S4 for detailed statistics). Call length was similar between sexes in all conditions (all p>0.05, Figure 5; see Table 1 for summary and Table S4 for detailed statistics).
Figure 5. Call length of object labels of female and male speakers in all three conditions. Within the boxplots, the horizontal line represents the median, the box indicates the quartiles, the whiskers represent the range, and the dots represent the individual data points.
Effect of speakers’ sex
In line with our hypothesis, pairwise comparisons revealed general patterns of the sex on speakers’ f0 range. Female speakers used a wider f0 range than male speakers during all situations and across all conditions (all p<0.05; see Figure 4 for summary, and see Table S3 and Figure S2 for detailed statistics).
Discussion
In the present study, we investigated and compared two pitch-related parameters (f0 mean and range) as well as one utterance length-related parameter (call length) of female and male speakers’ speech during interactions with their own infants (infant-directed speech), their own family dogs (dog-directed speech), and their spouses (adult-directed speech). These interactions were observed in two free speech situations (attention getting and language tutoring) and one fixed sentences situation with a nursery rhyme (fixed sentences). Our aim was to study whether and how the different situations, the partners’ expected linguistic competence, the speakers’ emotions, and sex affect these prosodic features.
Effect of situation
Towards infants, f0 mean and range followed the hypothesized pattern, with f0 mean being higher during the attention getting and f0 range being wider during the language tutoring situation. This supports the notion that f0 mean plays a crucial role in controlling and directing infants’ attention towards the speaker, while f0 range contributes significantly to language acquisition (e.g. Trainor & Desjardins, Reference Trainor and Desjardins2002). Conversely, in adult-directed speech, we observed an opposite trend, with speakers using a lower f0 mean during attention getting compared to the language tutoring situation. This suggests that, with other adults, speakers could use engaging linguistic content rather than relying solely on intense acoustic prosody to capture and maintain their spouses’ attention. Interestingly, however, dog-directed f0 mean showed no difference between the two free speech situations (i.e. attention getting vs. language tutoring). This suggests that speakers did not expect their dogs to form quick object–label associations easily (e.g. Fugazza et al., Reference Fugazza, Dror, Sommese, Temesi and Miklósi2021) and therefore maintained their high pitch to facilitate their canine partner’s attention during the language tutoring situation (e.g. Jeannin et al., Reference Jeannin, Gilbert, Amy and Leboucher2017).
The analysis of the fixed sentences, which contained a nursery rhyme, revealed that speakers utilized the most exaggerated prosody, characterized by higher f0 mean and wider f0 range, across all partners (i.e. infants, spouses, and dogs). This finding aligns with our second hypothesis and suggests that the infant-directed nature of the nursery rhyme strongly influenced speech prosody, resulting in a typical rhythmic and melodic speech style with exaggerated acoustics regardless of the partner (e.g. Falk & Audibert, Reference Falk and Audibert2021). These results underscore the significance of speech content and its relevance as a factor in the infant-directed nature of a given situation for future comparative prosody research.
Effect of the partners’ linguistic competence
In line with our hypotheses, speakers adjusted their speech prosody to their partner’s needs and capacities. Specifically, they used a higher and wider ranged f0 in general, when talking to their infants and dogs compared to when speaking to their spouses. When speakers were attempting to form object–label associations with their infants, they utilized longer utterances (i.e. call length), and female speakers also employed a wider pitch range (i.e. f0 range). Contrary to our predictions, speakers also used a higher overall pitch (i.e. f0 mean) when addressing infants while uttering the object label. High pitch might impede word segmentation while also having the potential to capture and maintain infants’ attention (Trainor & Desjardins, Reference Trainor and Desjardins2002). It is possible that speakers had to employ more attention-getting cues when uttering the label because infants focused less on the target object, particularly when a non-target object was presented simultaneously. Further analysis of the partner’s looking behaviour and attentional states is needed to explore this possibility. When uttering the object label to adults (i.e. their spouses), as expected, speakers used lower mean pitch and smaller pitch range; however, they also employed longer utterances. Previous studies have shown that hyperarticulated vowels and longer utterances are also used towards adults if they are linguistic foreigners (e.g. Uther et al., Reference Uther, Knoll and Burnham2007). Object labels in the present study were artificial words that might resemble foreign phrases, potentially prompting longer utterances from the speakers. Lastly, and in line with our hypotheses, speakers used higher pitch, narrower pitch range, and shorter utterances when uttering the object label to their dogs. These results further support the notion that people tend to adopt a speech style with their dogs aimed at maintaining canine attention, but without the use of language learning aids and likely without word tutoring intentions (e.g. Burnham et al., Reference Burnham, Kitamura and Vollmer-Conna2002; Xu et al., Reference Xu, Burnham, Kitamura and Vollmer-Conna2013; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017).
Recently, it has been demonstrated that speakers of the present study express similarly intense happy emotions and emotional valence when interacting with their infants and spouses, while exhibiting less intense and less positive emotions when communicating with their dogs (Koós-Hutás et al., Reference Koós-Hutás, Kovács, Topál and Gergely2024). If the pitch-related features of their speech were to follow this pattern, one could conclude that acoustics are “by-product” of their happy emotions, as previously suggested (Trainor et al., Reference Trainor, Austin and Desjardins2000). Our results, however, did not support this notion. Instead, we found that speakers used a higher and more variable pitch when addressing their dogs (and infants) compared to their spouses. This suggests that at least in dog- and adult-directed prosody, the facial and acoustic modalities of prosody exhibit different patterns. These results also suggest that pitch characteristics are not only “by-products” of a more emotional speech style, but also they are functional modifications and are probably adjusted to the partners’ emotional needs and cognitive capacities (Trainor et al., Reference Trainor, Austin and Desjardins2000; Koós-Hutás et al., Reference Koós-Hutás, Kovács, Topál and Gergely2024).
Effect of the speakers’ sex
In line with previous studies, we found more similarities than differences in the acoustic prosody of female and male speakers towards their infants, spouses, and dogs (e.g. Niwano & Sugai, Reference Niwano and Sugai2003; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). Across situations, both sexes used their f0 mean and range similarly when speaking to the same type of partner (i.e. infant, spouse, or dog). Moreover, there were no discernible differences between the sexes in the analysis of object labels. In line with prior studies and our hypothesis, the only consistent difference between the two sexes was found in their pitch range: female speakers generally employed a wider f0 range than male speakers across all partners and situations (e.g. Fernald et al., Reference Fernald, Taeschner, Dunn, Papousek, de Boysson-Bardies and Fukui1989; Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). We also identified minor differences in the f0 mean of female and male speakers, contrary to our prior expectations. Male speakers, for instance, exhibited a higher f0 mean when addressing their dogs compared to their infants during the language tutoring situation, while female speakers did not differentiate between partners in terms of f0 mean. Prior research has demonstrated that during tasks involving easy problem-solving, which includes praise, speakers tend to use higher pitch when talking to dogs than to infants (Gergely et al., Reference Gergely, Faragó, Galambos and Topál2017). It is possible that male speakers praised their dogs more than their infants during the object–label association task or that they required more attention-getting cues to maintain the dog’s focus in this setting. Future investigations are needed to test these hypotheses. Moreover, during the fixed sentences situation, female speakers employed a higher mean pitch in their infant-directed speech compared to their dog-directed speech, while male speakers maintained a similar mean pitch when addressing dogs and infants in this scenario. There is evidence that women engage in more frequent singing and rhyming activities with their infants than men, potentially contributing to this discrepancy in the results (e.g. Yan et al., Reference Yan, Jessani, Spelke, De Villiers, De Villiers and Mehr2021).
Conclusions
The present study supports the well-known phenomenon of more intense acoustic prosodic speech when talking to infants and dogs is still observable when compared to spouse-directed speech. In a comparative framework, we provided further evidence that mean pitch has an important attention-getting function, while pitch range might facilitate language acquisition. Our results suggest that infant-, spouse-, and dog-directed speech prosody conveys more than just positive emotional attitudes; it has the potential to serve specific functions such as capturing attention and aiding language acquisition according to the partners’ needs and capacities. Heightened and more variable pitch was found when speakers were reciting a nursery rhyme to both the infant and the dog as well as to their spouses. This finding may indicate that the infant-directed content and context of the speech could have a greater influence on the acoustic prosody than the type of partner. We also found that major patterns of pitch and utterance length modifications are presented similarly in female and male speakers, but female speakers tend to use a wider pitch range in general. In summary, these results highlight the importance of studying the context, content, and addressee-specific features of prosody in a comparative framework to better understand its exact functions and roles.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.17632/z868c5v5yy.1.
Acknowledgements
This study was supported by the Hungarian Scientific Research Fund (NKFIH grant no. FK142968), Hungarian Brain Research Program (HBRP) 3.0 NAP, János Bolyai Research Scholarship (BO/751/20 and BO/00361/24) of the Hungarian Academy of Sciences, and European Research Council (ERC) under the European Union’s Horizon 2020 Research and Innovation Programme (950159). We are grateful to the participating families and to Anna Dallos and Mandula Koós-Hutás for their help in data acquisition.
Competing interest
The authors declare that there are no competing interests.
Open access
