Old but attuned: the ability to decode babies’ cries does not decline with age

Abstract

In most human societies, grandparents often provide substantial care and support for their grandchildren, including as babies. Given that previous studies have shown that ageing is accompanied by a gradual decline in our ability to identify other people’s emotions, does age also reduce our skill at understanding a baby’s cries? Here, we show that older people with experience of caring for babies remain able to correctly decode the information conveyed by babies’ cries. The results of our psychoacoustic experiments underline that older people were at least as good as younger adults at identifying whether a baby is crying in pain or rather as a result of simple discomfort. These results are consistent with the notion that humans are cooperative breeders where older generations can effectively help younger ones with alloparental care. Favouring intergenerational solidarity is likely to simultaneously benefit the quality of life of young parents and older, knowledgeable carers.

1. Introduction

In humans, it is common for people other than biological parents to look after babies, contributing to care and feeding. This alloparenting probably began with Homo erectus around 2 million years ago [1], and characterizes us as a cooperative breeding species. Alloparents may or may not be genetically related to the children they care for, and their positive impact on infants has been widely demonstrated [2–4]. In particular, the presence of grandparents in the first postpartum months allows support and division of infant care, as well as transmission of knowledge and practice between generations [5,6]. In modern societies, where people are living longer, older adults are likely to be involved in babies’ care [7]. This social investment and connection with people of different ages is one of the keys to successful ageing [8–10], as are physical and intellectually demanding activities [11]. Thus, caring for kin or non-kin children is associated with better cognitive function [12], well-being [13] and reduced social loneliness [14].

Yet, providing appropriate care for babies requires the ability to identify their needs by decoding the visual and vocal signals they emit [15]. Among these signals, crying, which is triggered by hunger, discomfort, isolation or pain, is the most likely to elicit a response from carers [16]. Are older people still able to decode the information in babies cries? Previous studies have shown that ageing makes adults less able to identify emotions [17–21], and recognition of emotional facial expressions and cognitive empathy declines with age [17,22]. It has also been shown that the ability to decode emotions in adult non-verbal vocalizations is affected by ageing [18], and declines sharply after the age of 60 [19]. More specifically, a study involving participants aged between 16 and 68 found that younger adults are better than older ones at identifying the context of infant vocalizations [23]. However, this study did not take into account the participants’ previous experience in caring for babies, a crucial factor now known to modulate the ability to identify the level of pain encoded by cries [24].

Here, we focused on well-characterized baby cries available from our databank [25] and tested whether getting older impairs our ability to identify crying contexts. We tested pain and discomfort cries, which are acoustically different and recognizable by listeners [24–26] as opposed to cries of hunger or isolation [27]. Parents are able to recognize pain cries for their own baby [25], and more broadly, being familiar with a given baby’s cries facilitates pain recognition for this baby, especially for people not currently involved in baby care [24]. Thus, we also investigated whether familiarity with a given baby modulates cries identification in older adults. We predicted that the effect of age would be mitigated in listeners with extensive prior exposure to baby cries and that higher cognitive empathy and familiarity with the cries of a given baby would facilitate the identification of their pain cries, whatever the listener’s age. To test this, we conducted a psychoacoustic experiment with participants aged from 20 to 87 years having different levels of experience with babies: individuals with no experience with babies, individuals who had cared for a baby more than 5 years ago, and individuals who were currently caring for a baby less than 2 years old.

2. Methods

(a). Participants

We tested a sample of 309 participants (mean age = 41.1 ± 17.2 years, min–max = 20–87). Participants were recruited on the online platform Prolific (https://www.prolific.com/), then directed to the online platform Labvanced (https://www.labvanced.com/) where the experiment took place. Once the experiment was completed, participants were paid at the recommended rate of 8.0 GBP per hour. All participants reported having normal hearing. They were distributed as follows (see table 1 for details):

Table 1.

Men and women participants in the psychoacoustic experiment. For each category, the number of parents is indicated in brackets.

age of the participants	20−29 years	30−39 years	40−49 years	50−59 years	60−69 years	70−79 years	>80 years
participants with no previous experience with babies (men (fathers); women (mothers))	20 (0); 18 (0)	6 (0); 5 (0)	7 (0); 5 (0)	6 (0); 5 (0)	6 (0); 5 (0)	3 (0); 4 (0)	0 (0); 0 (0)
participants with past experience with babies (men (fathers); women (mothers))	5 (0); 6 (0)	5 (3); 7 (7)	12 (10); 6 (5)	6 (4); 5 (5)	5 (5); 6 (6)	6 (6); 6 (6)	1 (1); 5 (5)
participants with current experience with babies (men (fathers); women (mothers))	16 (9); 38 (3)	18 (16); 29 (21)	6 (6); 7 (6)	5 (5); 5 (3)	6 (6); 4 (4)	1 (1); 3 (3)	0 (0); 0 (0)

• —In total, 90 participants had no experience with babies (‘no experience’ group; 48 men, 42 women; none of them were parents; mean age = 40.0 ± 17.8 years, min–max = 20–76).
• —About 81 participants had cared for one or more babies in the past, more than 5 years ago (‘past experience’ group; 40 men, 41 women; 63 of them were parents; mean age = 51.1 ± 18.6 years, min–max = 20–87).
• —In total, 138 participants were currently looking after at least one baby under 2 years old (‘current experience’ group; 52 men, 86 women; 83 of them were parents; mean age = 36.0 ± 13.2 years, min–max = 20–72).

Experience with babies was considered when participants reported taking care of a baby, either their own, a relative (sibling, nephew, grandchild, etc.) or unrelated through babysitting or their professional activity.

(b). Sound stimuli

The recordings used to produce stimuli for the psychoacoustic experiment were cries extracted from the database of the ENES Bioacoustics Research Laboratory [25]. Pain cries were recorded at the paediatrician’s office during a vaccination session (n = 22 babies, 10 boys and 12 girls; age = 60.3 ± 3.4 days on the day of vaccination). Discomfort cries were recorded at the baby’s home during bathing, undressing or dressing by the parents (time between recording sessions for discomfort cries and pain cries for a given baby = 6.9 ± 3 days; for details concerning recording conditions see [25]).

These raw recordings were segmented into 284 cry sequences using Audacity software (https://www.audacityteam.org/). We selected only those sequences with an excellent signal-to-noise ratio (mean sequence duration = 6.3 ± 1.1 s; four sequences of pain cries per baby and 2−19 sequences of discomfort cries per baby). Furthermore, in order to limit pseudo-replication of stimuli in the playback experiment, we retained only those babies for whom we had at least 10 sequences of discomfort cries and four sequences of pain cries (n = 9; three boys, six girls; total = 150 sequences). The sound intensity of the stimuli was normalized to 100% of the maximal amplitude using the ‘normalize’ function in tuneR package (cran.r-project.org/web/packages/tuneR/index.html).

(c). Experimental procedure

The experimental procedure consisted of two successive phases: a training phase in which participants listened to the discomfort cries of an assigned baby, thus creating familiarity with this assigned baby, and a testing phase in which participants were asked to rate as ‘discomfort cry’ or ‘pain cry’ the cries of their assigned baby and those of an unknown baby (figure 1a; same procedure as in [24]). Each participant was trained and then tested with different cries, thus preventing idiosyncratic acoustic characteristics of individual babies from biasing the results. Randomization of cry stimuli was achieved using the ‘sample’ function in R statistical software (https://www.r-project.org/).

Figure 1.

Ageing does not impair the ability to identify whether a baby cry expresses pain or simple discomfort. (a) The experimental protocol consisted of two successive phases: a training phase in which participants listened to discomfort cries from an assigned baby, followed by a testing phase (playback tests) in which the participants had to identify cries from their assigned baby and cries from an unknown baby as ‘discomfort cry’ or ‘pain cry’. (b) Results of the psychoacoustic tests. Solid lines represent medians of posterior distributions and light areas 95% credible intervals (CIs). Dotted lines indicate the 50% chance level.

At the beginning of the training session, participants received the following instruction: ‘You are going to hear several cries of one baby, who will be named YOUR baby for the rest of this study. YOUR baby is in the bath, it is uncomfortable, YOUR baby is crying.’ Then, they listened to a succession of four discomfort cries from their assigned baby. This training was repeated a day later, using four more discomfort cries from the same assigned baby. Participants then performed their test session a few hours after the end of the second training session (mean = 6.7 ± 5.6 h, range [3–29.3 h]). During the test, they listened to two successive sets of four cries: two pain cries and two discomfort cries presented in a random order. One of the two sets of cries came from their assigned, ‘familiar’ baby. The other set of cries came from an unknown baby. The order of presentation of the two sets, familiar and unknown, was balanced among the participants. Before the start of each set of cries, participants were informed whether they would hear the cries of their assigned baby or an unknown baby. After each cry, participants were asked to estimate in a two-alternative forced choice whether the cry they heard was due to discomfort or pain. Participants were not told how many cries of discomfort or pain they would listen to.

(d). Data analysis

Statistical analyses were performed using Bayesian mixed models fitted with the brms R package [28]. Response scores were modelled using a logistic function (Bernoulli family). The outcome variable was success in recognizing the context of crying (discomfort or pain). The participant’s age, experience with baby care, as well as the baby’s status (familiar or unknown), were included in the model as fixed factors. The participant’s identity and that of the baby were included as random factors. The model structure in brms syntax was as follows:

$${\displaystyle \mathrm{s}\mathrm{u}\mathrm{c}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}\sim \mathrm{a}\mathrm{g}\mathrm{e}\times \mathrm{l}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{l}\mathrm{e}\mathrm{x}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}\times \mathrm{b}\mathrm{a}\mathrm{b}\mathrm{y}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{u}\mathrm{s}+(1|\mathrm{s}\mathrm{u}\mathrm{b}\mathrm{j}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{I}\mathrm{D}+\mathrm{b}\mathrm{a}\mathrm{b}\mathrm{y}\mathrm{I}\mathrm{D}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{m})}$$

To assess the effect of the listener’s sex, we built another model by adding sex as a fixed factor in the previous model.

For each model, 3000 iterations were run over two Markov chain Monte Carlo (MCMC) chains, with the first 1000 iterations of each chain used to tune the algorithm. To improve convergence when modelling response scores, we specified mildly informative conservative priors. Results are summarized as medians of the posterior distributions and 95% credible intervals (CIs). The chance level was set at 50% as it was a two-alternative forced choice (discomfort/pain) with a balanced number of the two conditions. CIs for estimates above the chance level indicate a credible effect, given the observed data and model structure [29]. Similarly, when contrasting two conditions, CIs excluding the null value can be inferred to indicate a credible difference between the conditions. In the same way, the distribution of posterior values over a given threshold can be quantified and indicate a credible difference when more than 97.5% of the posteriors are over the threshold value and a tendency when 95–97.5% of the posteriors are over the threshold.

3. Results

The participants, on average, successfully identified the ‘discomfort’ or ‘pain’ crying contexts (mean correct identification of the crying context = 61.4%, 95% CI [53.1, 69.0]; chance level = 50%), and we observed no significant decline in this ability with the age of the participants. The recognition accuracy even tended to increase with age since the crying context was correctly identified in 57.3% of trials at the age of 20 (95% CI [48.4, 65.7]), while this rate of success increased to 64.4% of trials over the age of 70 ([55.1, 72.7]; 7.0% better [−0.6, 14.7], see electronic supplementary material, table S1). Regardless of age, participants were more successful in identifying the crying context when they heard a cry from their assigned (familiar) baby than a cry from an unknown baby (6.4% better [1.5, 11.3]).

As illustrated in figure 1b, experience with babies improved participants’ ability to identify the crying context in familiar babies. Participants with no experience with babies correctly identified 60.2% [50.1, 69.7] of cries, while those with past experience recognized 63.1% [53.1, 71.9] of cries and those currently caring for a baby 70.4% [60.0, 78.7] of cries (respectively 2.8% [−5.8, 11.2] and 10.0% [0.5, 18.9] higher). At the age of 20, participants with no previous experience with babies were unable to identify the crying context of their familiar baby (53.1% [41.9, 64.0]). However, these participants with no experience performed better at the age of 60 (61.8% [50.6, 72.2]), and this progression was even more marked in the over-70s (65.7% [51.0, 78.2]). Participants aged 20 with past experience with babies also failed to reliably recognize the crying context of familiar baby cries (59.2% [46.2, 71.2]). However, these participants with past experience performed better from the age of 35 (61.0% [50.3, 70.8]), and even better after the age of 70 (65.9% [53.4, 76.7]). Finally, participants currently caring for a baby under 2 years old were able to identify whether cries of a familiar baby had been emitted in a discomfort versus a pain context from the age of 20 (66.2% [56.4, 74.7]). This ability was even improved in the over-70s (73.5% [58.1, 84.8]; see electronic supplementary material, table S1).

Participants with no experience of babies were not able to reliably identify the crying context of unknown babies (58.2% [47.4, 67.9]). We observed a similar result in participants with past experience with babies at the age of 20 (50.5% [37.0, 64.0]). However, these people with previous experience with babies showed better performance after the age of 70 (63.3% [49.8, 74.6]). Finally, the age of the most experienced participants (with babies under 2 years of age) did not seem to influence their ability to identify the crying context of an unknown baby (electronic supplementary material, table S1).

Empathy scores were positively associated with age, but there was no statistical association between empathy scores and the ability to identify crying context (see supplementary methods and results, electronic supplementary material, tables S2 and S3). Furthermore, the model including the sex of the participants did not explain the data any better (WAIC of 3254.2 versus 3240.3 without) and suggested no difference between men and women regarding their ability to identify crying contexts (−0.1% [−5.6, 5.0]).

4. Discussion

In this study, we showed that as adults age, they retain the ability to identify whether a baby cry is emitted in a context of simple discomfort or real pain. Our results further demonstrate that experience in caring for babies improves this ability [24].

Babies' cries are signals that carry information about the baby’s identity and age (‘static’ information) and immediate needs (‘dynamic’ information) [27,30]. Knowing how to identify this information in the cry signal—in other words, being able to decode the information conveyed by the cry—is essential for carers to provide appropriate care. Previous studies have shown that former experience with babies considerably improves the ability to decode cries, both for static information [31] and for dynamic information [24]. For example, mothers who have already had several children learn to identify their newborn more quickly than primiparous mothers [31]. Parents of young children and healthcare professionals working with babies are much better at identifying pain encoded in babies' cries than inexperienced people [24]. The results of the present study confirm the importance of experience and highlight its role throughout life.

In our study, regardless of age, participants who were currently caring for a baby were the most able to determine whether a cry signals pain or simple discomfort, and therefore whether the baby needs immediate help or not. This ability was not lost in the elderly, older participants reporting caring for a baby under 2 years old (usually a grandchild) remaining able to identify the crying context. Regarding adults who had never cared for babies, while young ones identified the context of a familiar baby’s cry only marginally better than chance, older people did significantly better than chance. A comparable improvement in performance was also found in participants who had cared for a baby in the past. This beneficial effect of age on the ability to decode babies' cries could be explained by the fact that human adults are probably indirectly exposed to babies' cries throughout their lives via their social network (family and friends), the media or on various occasions such as public transport journeys. However, it should be noted that, in our sample, there was a higher proportion of parents among older participants. Indeed, those with no experience were necessarily non-parents, as were younger participants with more than 5 years' experience, mainly acquired through babysitting or looking after younger siblings. In contrast, older participants with past or current experience were almost all parents. It is possible that this favoured the older people’s ability to decode babies' cries.

Our results run counter to previous findings, which have shown an age-related decline in emotion recognition in faces, bodies, spoken sentences [20], as well as in non-linguistic [21] and non-verbal vocalizations [19]. One explanation could be that the two-alternative forced choice task used in our study was simpler than in previous works, where participants may be asked to discriminate up to 18 different emotions [21]. Indeed, more difficult tasks, such as combining modalities, are likely to create particular difficulties for older people [20]. On the other hand, our results may be specific to the pain modality. In fact, to our knowledge, pain is almost never included in studies on emotion recognition. In a study where young and old participants were asked to rate pain expressed by facial expressions, they showed that older observers’ performance was only slightly reduced, being sufficient to differentiate pain levels [32]. Thus, the preservation of the ability to identify pain despite ageing may be due to the motivational relevance and survival value of pain cues, their identification being essential to provide adapted care.

Do our results mean that older people are systematically able to decode babies' cries? Probably not. Indeed, advancing age is typically accompanied by a decline in sensory and cognitive abilities, leading to increasing difficulties in perceiving and analysing sensory stimuli from the environment [11,33–35]. Although our study is limited by the fact that it does not include auditory and cognitive assessment, our sample likely included people who had retained good sensory and cognitive abilities despite their age. First, all participants reported normal hearing. While the oldest participants almost certainly had age-related hearing loss, this usually affects high frequencies, starting at around 2 kHz and becoming increasingly significant as sounds approach 8 kHz [36]. As babies' cries are broadband signals extending over a wide frequency spectrum into low frequencies [25], it is unlikely that our participants had issues hearing them. Second, the perception of roughness, which is the main acoustic property characterizing pain in cries [25], is little affected by age-related hearing loss [37]. With regard to the potential impact of cognitive decline with age, participants were enrolled on Prolific and had to be able to complete the experiment online, with three successive sessions. This involved regularly accessing their mailbox, following a link to the next session and following instructions. Being able to perform these tasks is highly likely to have selected people with conserved cognitive abilities. Furthermore, neither age-related auditory loss [38] nor general cognitive abilities [18] have been linked to a decline in emotion recognition. Moreover, our participants showed no decline in cognitive empathy with age, all of which is in favour of having selected elderly people who aged well [10]. Finally, our study does not allow us to infer potential effects of age on the behavioural response to crying babies. This impact could be positive—an older, more experienced person could possibly make better decisions than a younger one—or negative—age could increase the stress associated with listening to a pain cry or make the person less able to act promptly and appropriately. These results may be specific to crying understanding since other emotions were not evaluated. Future experiments, including a larger range of conditions, such as laughter or fear, are required to determine whether the present findings may be extended to other infant vocalizations.

In conclusion, our work underlines that experience in caring for babies enhances the ability to decode the information conveyed in their cries and that this expertise can be preserved at advanced ages. Our results show that older people, likely those who have aged well, are still capable of efficiently decoding babies' cries. By highlighting that older persons retain the ability to assess babies’ needs by monitoring the babies’ vocal solicitations, this study highlights how older carers (grandparents or else) may provide young parents with help in baby care [3] by sharing their expertise through advising and direct caregiving. This observation reinforces the idea that favouring the intergenerational solidarity and cohabitation of generations [7,10,39], as in, e.g. multi-generational housing projects are likely to simultaneously benefit the quality of life of young parents and older [14], knowledgeable carers.

Ethics

The experiment was approved by the local ethics committee (October 2019 Comite Ethique du CHUde Saint-Etienne, Institutional Review Board: IORG0007394), and informed consent was obtained from all participants.

Data accessibility

The dataset and the R code for analysing data can be downloaded from https://osf.io/c27fp/.

Supplementary material is available online [40].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

S.C.: conceptualization, data curation, formal analysis, investigation, methodology, writing—original draft; C.C.: methodology, writing—review and editing; R.P.: funding acquisition, writing—review and editing; D.R.: funding acquisition, writing— review and editing; C.F.: conceptualization, funding acquisition, methodology, writing—review and editing; N.M.: conceptualization, funding acquisition, methodology, project administration, writing—original draft.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

The research has been funded by the ANR (BABYCRY project; SCREAM project), the AXA Foundation (NM), the Lyon IDEX Fellowship (DR), the Labex CeLyA (France), the University of Saint-Etienne (France), the Centre National de la Recherche Scientifique (CNRS, France), the Institut National de la Santé et de la Recherche Médicale (Inserm, France) and the Institut universitaire de France (N.M. and D.R.). S.C. was a recipient of a PhD grant from the Fondation pour la Recherche Médicale (FRM; code ECO202006011648). C.F. was a recipient of a research grant from the FRM (ARF202110014013). This publication was completed while N.M. was Visiting Miller Professor at the University of California, Berkeley (Miller Institute for Basic Research in Science).