Head before heart? Cognitive empathy may emerge before affective empathy in the developing brain

  1. Chiara Bulgarelli
  2. Paola Pinti
  3. Tessel Bazelmans
  4. Antonia Hamilton
  5. Emily J Jones

  • Curated by eLife

    eLife logo

    eLife Assessment

    The authors have presented a study which addresses a recognised gap in the literature, the emergence of the neural correlates of cognitive and affective empathy in children; they introduce a task for measuring both positive and negative empathy in a relatively large group of children aged 3-5. Using functional near infrared spectroscopy they found different brain region activation in cognitive versus affective empathy tasks, as well as age-related changes in the activation of right MPFC and right TPJ. The study represents a valuable contribution to understanding the development of cognitive function with solid evidence for the claims about brain regions and ageing.

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Empathy is crucial for social interactions across all cultures, and is foundational to establishing social cooperation and group ties in human societies. Challenging the current predominant view, we recently proposed that understanding others’ emotions (cognitive empathy) might emerge earlier than actually sharing those emotions (affective empathy) (Bulgarelli & Jones, 2023). Here we test this hypothesis by measuring which empathic component matures first during toddlerhood, a critical period for the development of broader social networks. Addressing this question is critical to understand the mechanisms through which caregivers scaffold empathy development. Traditional approaches are inadequate, as they rely on children’s verbal skills or unfamiliar scenarios that lack ecological validity. In this preregistered study, we employed a novel toddler-appropriate task to dissociate neural and physiological correlates of cognitive and affective empathy in N=90 3-to-6-year-olds using functional near-infrared spectroscopy (fNIRS) and simultaneous heart rate monitoring to identify internal markers of empathy. We found that brain regions supporting affective and cognitive empathy in young children resemble those observed in adults. Importantly, we showed an effect of age on network specialisation with brain activations of cognitive empathy stronger in younger compared to older preschoolers, and brain activations of affective empathy stronger in older compared to younger preschoolers. These results may provide the first evidence that cognitive empathy is engaged early and develops alongside or ahead of affective empathy in preschoolers, challenging existing models and suggesting a new framework for understanding the development of empathy.

Article activity feed

  1. Version published to 10.7554/elife.110899.2 on eLife
  2. Version published to 10.7554/elife.110899 on eLife
  3. eLife

    eLife Assessment

    The authors have presented a study which addresses a recognised gap in the literature, the emergence of the neural correlates of cognitive and affective empathy in children; they introduce a task for measuring both positive and negative empathy in a relatively large group of children aged 3-5. Using functional near infrared spectroscopy they found different brain region activation in cognitive versus affective empathy tasks, as well as age-related changes in the activation of right MPFC and right TPJ. The study represents a valuable contribution to understanding the development of cognitive function with solid evidence for the claims about brain regions and ageing.

  4. eLife

    Reviewer #1 (Public review):

    This paper presents an fNIRS neuroimaging study with a relatively large sample of preschool children (aged 3-5) that measures both positive and negative empathy within a single task. Children watch emotional events and are asked questions about both their own emotions and the emotions of others, allowing the authors to distinguish between affective and cognitive empathy. The authors propose "foundational" models of affective and cognitive empathy and argue that their findings support the idea that cognitive empathy emerges before affective empathy in early childhood.

    Strengths:

    The paper addresses a valuable question by measuring both positive and negative empathy within a single cognitive task. The use of fNIRS with a relatively large preschool sample is commendable, and the pre-registered design strengthens the contribution. The task itself is innovative, well-suited to this age group, and achieves high compliance, which is essential and notably difficult with young children. Overall, the methods are appropriate, and the empirical work is valuable.

    Comments on revised version.

    I thank the authors for their thoughtful and thorough response, and the editor for the opportunity to review the revised manuscript. The current version is substantially improved, with more accurate coverage of the relevant literature and more careful interpretation of findings. The authors have addressed the large majority of the concerns raised in the initial review. Two minor points remain.

    The "foundational" models are still presented as though they are established frameworks, when in fact this paper is their first empirical test. The labelling should reflect that. Additionally, the claim that both empathy components are present together "early in life" needs to specify that this evidence comes from studies of infants in the first year, which is what makes it relevant to the sequencing debate the authors are engaging with.

    (1) I appreciate the authors' clarification that the "foundational" models originate from their prior theoretical paper and that the current study represents the first empirical test. This is helpful context, but it also reinforces the concern. If this is the first empirical test, presenting these as named, italicised models throughout the manuscript still give the impression of established frameworks. I would suggest the authors signal more clearly in the text that these are their own proposed, as-yet-untested organising frameworks rather than concepts with an independent evidence base. I believe doing so would only strengthen the transparency of the contribution and, ultimately, the establishment of the model itself.

    (2) The authors now state in the introduction that "this linear, stage-based account has been challenged by evidence that affective and cognitive dimensions of empathy seem to be present together early in life." However, as written, the claim that both components are present together early on does not in itself challenge a linear stage-based account without specifying the ages at which this evidence was obtained. The claim would be more accurate if the authors noted that both components appear to be present as early as the first year of life (e.g., Roth-Hanania, 2025), which would make it clearer why this evidence is relevant to the sequencing debate.

  5. eLife

    Reviewer #2 (Public review):

    Summary:

    Authors examined neural substrates for cognitive empathy (conceptually understanding others' emotions) versus affective empathy (automatically sharing others' emotions) development in 3-5-year-old toddlers and argued that cognitive empathy emerges earlier than affective empathy, challenging the predominant view that affective empathy develops earlier. The authors developed an empathy test for toddlers while measuring their brain activity with fNIRS (particularly in MPFC, STG, DLPFC, and TPJ) and heart rate. They found different brain region activation in cognitive versus affective empathy tasks, as well as age-related changes in the activation of right MPFC and right TPJ.

    Strengths:

    This work investigated the development of different components of empathy, which is a quite understudied topic. The authors developed an age-appropriate task for toddlers to measure their cognitive empathy and affective empathy, which is likely useful for future research in this field. Their methods are sound, and give the relatively large sample, the results look interesting and relatively solid.

    Weaknesses:

    Although the authors have tempered some of their claims in the revision, parts of the paper still strongly advocate for the earlier developmental emergence of cognitive empathy (e.g., with the foundational affective empathy model and wordings such as "this provide the first direct evidence for the earlier emergence of cognitive empathy", "this groundbreaking empirical evidence in support of a foundational cognitive empathy model") . Given the limitations that the authors currently acknowledge (e.g., longitudinal data needed, possible maturation of other cognitive processes recruited by the task), the evidence does not seem to fully support such strong temporal assertions.

  6. eLife

    Author response:

    The following is the authors’ response to the original reviews.

    Public Reviews:

    Reviewer #1 (Public review):

    This paper presents an fNIRS neuroimaging study with a relatively large sample of preschool children (aged 3-5) that measures both positive and negative empathy within a single task. Children watch emotional events and are asked questions about both their own emotions and the emotions of others, allowing the authors to distinguish between affective and cognitive empathy. The authors propose "foundational" models of affective and cognitive empathy and argue that their findings support the idea that cognitive empathy emerges before affective empathy in early childhood.

    Strengths:

    The paper addresses a valuable question by measuring both positive and negative empathy within a single cognitive task. The use of fNIRS with a relatively large preschool sample is commendable, and the pre-registered design strengthens the contribution. The task itself is innovative, well-suited to this age group, and achieves high compliance, which is essential and notably difficult with young children. Overall, the methods are appropriate, and the empirical work is valuable.

    We thank the reviewer for highlighting the strengths of our work.

    Weaknesses:

    The main concerns relate to the framing of the paper rather than the empirical work itself. The introduction contains several claims that are overstated or inaccurate. The statement that "we know very little about the development of this fundamental social skill during the first years of life" does not reflect the state of the field; empathy in early development has been quite extensively studied (e.g., Davidov et al., Malti et al., Uzefovsky et al., Decety et al., Feldman et al., among others). The view that emotional contagion directly develops into affective empathy is based on early theoretical accounts that have since been challenged by empirical evidence (see Davidov et al., 2025). The claim that cognitive empathy does not require theory of mind is also overstated - it is hard to see how theory of mind, the understanding that others have thoughts, beliefs, and emotions that may differ from our own, would not be required for cognitive empathy. Furthermore, the introduction neglects recent and directly relevant work (e.g., Zach et al., 2025; Uzefovsky et al., 2020; Davidov et al., 2021).

    We are grateful to the reviewer for directing us to this important body of work, and we have adjusted the description of our contribution in the Introduction accordingly, as follows: “While empirical studies on empathy are abundant in adults, we know little about the development of this fundamental social skill during the first years of life. In particular, the developmental ordering of its affective and cognitive components, and the neural changes underlying it, remain debated. However, understanding the mechanisms behind the emergence of the two empathic components is crucial for advancing our understanding of social development and the factors that can support emerging bonds for children in their natural environments” (page 2, Introduction section).

    We also acknowledged the body work cited by the reviewer as follows: “Recently, this linear, stage-based account has been challenged by evidence that affective and cognitive dimensions of empathy seem to be present together early in life, that concern for others does not develop in discrete stages, and that infants seem to show other-oriented responses to both positive and negative emotions, which have been interpreted as reflecting more than emotional contagion alone (Davidov et al., 2021, 2025; Uzefovsky et al., 2019; Zach et al., 2025).” (page 2, Introduction section)

    Regarding the statement on theory of mind, we did not intend to claim that cognitive empathy is completely independent of theory of mind altogether; as we note, cognitive empathy can involve elements of mentalising and recruits overlapping regions in adults. Our point is narrower: that the understanding of others' emotional states (cognitive empathy) does not require the broader capacity to represent another's thoughts and beliefs, which is the aspect of theory of mind that consolidates later in development. We have revised the passage to make this distinction explicit and to avoid overstatement as follows: “Cognitive empathy can involve elements of theory of mind or mentalising (5) and activates similar regions in adults, but specifically refers to understanding others' emotional states rather than the broader ability to represent another person's thoughts (4, 15). Thus, cognitive empathy may not necessarily require the broader belief-representing capacities of theory of mind, as emotional states can be read from external cues.” (page 2, Introduction).

    Most critically, the claim that "no neuroimaging studies have yet investigated brain regions supporting empathy in preschoolers" is inaccurate. Multiple studies have examined brain regions supporting empathy in children within this age range, including work using fNIRS and studies of positive empathy (e.g., Decety et al., 2018; Light et al., 2009; Levy et al., 2019; Bray et al., 2022; Brink et al., 2011). This is also not the first study to measure brain activation in response to positive and negative emotional events in children (e.g., Cheng et al., 2014; Light et al., 2009). These novelty claims need to be corrected.

    We thoroughly revised the literature suggested by the reviewer and we noted that:

    - Light et al. (2009) is an EEG study in 6–10-year-olds during emotional events;

    - Levy, Goldstein & Feldman (2019) is a MEG study of preadolescents (~10–11);

    - Bray et al. (2022) is a structural (voxel-based morphometry) study of empathy in 10-year-olds;

    - Cheng, Chen & Decety (2014) is an EEG/ERP in children aged 3–9 viewing animations of painful situations;

    - Decety, Meidenbauer & Cowell (2018) is an EEG/ERP study of 4-to-5 year-olds viewing painful vs. neutral stimuli.

    Therefore, all these works are either on older children (Light, 2009; Levy, 2019; Bray, 2022), not functional studies (Bray, 2022) used only negatively valenced pain/distress stimuli (but more on physical pain, Cheng, 2014; Decety, 2018), and in the case of EEG/ERP studies indexed temporal neural components rather than the regional activations our task was designed to localise.

    Brink et al. (2011) is indeed an fNIRS study of 4–8-year-olds processing affective and cognitive empathy stories (verbal and cartoon), but they used different stimuli for affective and cognitive empathy, which is what we argue against in the introduction of this work: “Within the foundational affective empathy model, cognitive empathy does not emerge before the age of 4. This has often been based on tasks that rely on children’s verbal abilities (Brink, 2011)” (page 2, introduction section).

    We have therefore adjusted the novelty sentence in the introduction as follows: “While previous studies have investigated neural responses to others' pain or distress as the empathy-eliciting stimulus, which may evoke personal distress in the observing child rather than a shared emotional response, and have largely been restricted to negatively valenced events (Cheng et al., 2014; Decety et al., 2018, Brink et al., 2011), this is the first study to test whether the developing preschool brain responds to emotional scenarios in regions comparable to those documented in adults, using a task that assesses both empathy components with the same stimuli”. (page 3 introduction).

    The use of "explicit" to describe cognitive empathy and "implicit" or "spontaneous" to describe affective empathy is problematic. Affective empathy can be expressed quite explicitly, through facial expressions, verbal statements, and gestures, and framing it as spontaneous overlooks the motivational dimensions of empathy (e.g., Zaki and colleagues). The authors' use of "foundational affective empathy model" and "foundational cognitive empathy model" as though these are established concepts is not well supported by the current evidence base.

    We thank the reviewer for this thoughtful and theoretically important comment. We agree that the terms "explicit" and "implicit/spontaneous" carry connotations that may not map cleanly onto affective and cognitive empathy as constructs. Affective empathy can indeed be expressed and even strategically deployed in highly explicit ways, and we acknowledge that our original framing did not adequately reflect this nuance.

    Regarding the terms foundational affective empathy model and foundational cognitive empathy model: we agree with the reviewer that these should not be presented as established, widely-accepted frameworks. We introduced these terms in a prior theoretical paper (Bulgarelli and Jones, 2023, JCPP Advances) to organise competing hypotheses about the developmental sequence of empathy components, and the present work constitutes the first empirical test of these models. We have referred to this reference through the introduction and discussion when discussing these models.

    To address this, we have revised the manuscript to avoid characterising affective empathy as inherently "spontaneous" or "implicit” and cognitive empathy as “explicit”:

    - “Empathy consists of two components: an affective dimension, which involves sharing others' emotions, and a cognitive dimension, which involves conceptually understanding others' emotions”. (page 2, introduction)

    - “To directly evaluate the explanatory value of the proposed foundational cognitive empathy model against the foundational affective empathy model, we designed a novel task to comparably assess both affective and cognitive empathy in pre-schoolers.” (page 3, introduction)

    - “In fact, if understanding others’ emotions is crucial for sharing them, then anything that supports this understanding becomes crucial for an empathic sharing to arise. Our data suggest that younger preschoolers engage more when explicitly prompted, while older preschoolers may have learned to think about what others are experiencing.” (page 8, discussion)

    - “It is in fact plausible to think that the cognitive dimension of empathy does not arise automatically (page 8, discussion)”

    The conclusions in the discussion go beyond what the data can support. The question of whether cognitive or affective empathy emerges first cannot be adequately addressed with a cross-sectional sample aged 3-5, an age at which affective empathy is likely already well established and cognitive empathy is expected to be developing around the lower end of this range. The cross-sectional design further limits what can be inferred about developmental trajectories during a period of substantial individual variability. Together, these issues make the developmental-precedence conclusions difficult to sustain. The claim that the results demonstrate "the first time that this brain specialisation for stimuli of different emotional valence may be rooted in childhood" is also inaccurate, as there is prior evidence for brain specialisation of emotional valence in early childhood (e.g., Grossmann et al., 2007).

    We have adjusted our conclusions to clarify that we present the cognitive-first (or simultaneous) account as our interpretation of the observed cortical specialisation, rather than as a developmental ordering directly demonstrated by the data. We acknowledge explicitly that a cross-sectional design cannot establish emergence order within a child, and have added this caveat. We note, however, that our findings, and a growing body of work questioning strictly affective-first stage models (e.g., Davidov et al., 2021, 2025), are in our view more consistent with a cognitive-first or simultaneous account, and we have framed our contribution accordingly. Consistent with this we edited the title as follows: “Head before heart? Cognitive empathy may emerge before affective empathy in the developing brain”; the abstract as follows: “These results may provide the first evidence that cognitive empathy is engaged early and develops alongside or ahead of affective empathy in preschoolers, challenging existing models and suggesting a new framework for understanding the development of empathy”, and we have extensively revised the entire discussion to tone down some of our conclusion and make clear that this our interpretation of the results. As the reviewer suggested we also acknowledge the limitation of the cross-sectional design as follows: “We interpret this as challenging years of speculation about the pivotal role of affective empathy in the development of its cognitive dimension, which was nevertheless supported by limited empirical evidence and not robustly investigated. We note, however, that a longitudinal design would be required to build on the current cross-sectional study of 3-to-6-year-olds to precisely establish the order in which these components first emerge within a child.” (page 8, discussion).

    Moreover, we agree with the reviewer that brain specialisation for stimuli of different emotional valence has been shown earlier in development; ours is, to our knowledge, the first study to explore it in the context of an empathy task. We have therefore edited the Discussion as follows: “Our results show that this cortical specialisation for stimuli of different emotional valence, well documented in infancy and early childhood (Grossmann et al., 2007), is also engaged in the context of an empathy task in preschoolers, to our knowledge the first demonstration of valence-specific cortical responses within an empathy paradigm at this age” (page 7, discussion section)

    Appraisal: The empirical contribution, the task design, the fNIRS data, and the analyses are sound and have value for the field. However, in its current form, the paper does not achieve what it sets out to do. The novelty claims are undermined by the omission of a substantial body of relevant prior work, and the developmental conclusions are not adequately supported by the cross-sectional design and age range studied. The abstract similarly overstates the support this study provides for the early emergence of cognitive over affective empathy.

    Impact: With appropriate revision, this work could make a meaningful contribution. The task is well-designed for studying empathy in young children and could be useful to other researchers in the field. The fNIRS data from a large preschool sample are a valuable resource. However, the contribution needs to be framed accurately, both in terms of what is genuinely novel relative to the existing literature and in terms of what conclusions the data can and cannot support.

    We thank the reviewer for highlighting the strengths of this work. We have thoroughly revised the manuscript and, following the reviewer's suggestions throughout, we have adjusted our conclusions and reframed the contribution of this work more precisely.

    Reviewer #2 (Public review):

    Summary:

    Authors examined neural substrates for cognitive empathy (conceptually understanding others' emotions) versus affective empathy (automatically sharing others' emotions) development in 3-5-year-old toddlers, and argued that cognitive empathy emerges earlier than affective empathy, challenging the predominant view that affective empathy develops earlier. The authors developed an empathy test for toddlers while measuring their brain activity with fNIRS (particularly in MPFC, STG, DLPFC, and TPJ) and heart rate. They found different brain region activation in cognitive versus affective empathy tasks, as well as age-related changes in the activation of right MPFC and right TPJ.

    Strengths:

    This work investigated the development of different components of empathy, which is a quite understudied topic. The authors developed an age-appropriate task for toddlers to measure their cognitive empathy and affective empathy, which is likely useful for future research in this field. Their methods are sound, and give a relatively large sample; the results look interesting and relatively solid, except for certain details in the reporting of methods and results.

    We thank the reviewer for highlighting the strengths of our work.

    Weaknesses:

    (1) My major concern is the roles of brain regions hypothesized and found in this paper (MPFC, STG, DLPFC, and TPJ) - the authors seemed to have omitted a large portion of the literature on this topic. Prior works have found that these brain regions may be involved in more than one process, or involved in processes that are common to both cognitive and affective empathy (see Schurz et al., 2021). In particular, MPFC seems to be indicated more often in cognitive empathy, and STG may be involved in both cognitive empathy and intermediate processes, which is contradictory to what the author claimed and hypothesized. Relatedly, when the authors made statements like "these results highlight that regions underpinning affective and cognitive empathy in preschoolers largely resemble those documented in adults" (without proper citations), I found it unconvincing due to the disagreements in the past adult research about brain regions related to empathy, which were not quite discussed in the current paper. It may be helpful if the authors do a more thorough literature review and provide a more comprehensive view of how their results fit in the existing literature.

    We thank the reviewer for this valuable point, and for directing us to Schurz et al. (2021). We have substantially expanded our treatment of the functional heterogeneity of these regions in both the Introduction and Discussion. In particular, two of our regions are well supported across frameworks. DLPFC and TPJ are consistently associated with cognitive empathy and mentalising across lesion, meta-analytic, and review evidence (Shamay-Tsoory, 2011; Fan et al., 2011; Schurz et al., 2021), consistent with our interpretation. We accept the reviewer's point regarding STG, which Schurz et al. (2021) implicate in both cognitive and "combined" processes that engage affective and cognitive functions in parallel. We have revised our interpretation of the STG findings accordingly, no longer treating this region as a clean marker of a single component, and we now discuss it as a region whose engagement is compatible with both component-specific and integrative processing.

    Regarding MPFC, we agree that its functional attribution warrants more careful treatment, but we would respectfully note that the literature does not support a single clean assignment in either direction. While several empathy-specific frameworks associate medial prefrontal regions with cognitive empathy and mentalising (Shamay-Tsoory et al., 2009; Fan et al., 2011; Schurz et al., 2021), a substantial affective-neuroscience literature implicates ventromedial prefrontal cortex in the generation, appraisal and regulation of emotion (Hiser & Koenigs, 2018), with lesion evidence linking damage on its more ventral part to deficits in empathy and emotion recognition. Critically, the spatial resolution of fNIRS does not allow us to distinguish ventral from dorsal medial prefrontal contributions, so we are not in a position to adjudicate between these accounts on the basis of our data. We would note, however, that while we cannot resolve ventral from dorsal contributions, sensitivity analyses in young children indicate that forehead fNIRS configurations register onto medial prefrontal regions, including more ventral medial portions (Bulgarelli et al., 2020, Human Brain Mapping). This supports the interpretability of our MPFC findings, particularly given that the ventral medial portion is the one most strongly implicated in affective processing.

    We have therefore revised the Discussion to acknowledge explicitly that MPFC is implicated in both affective and cognitive aspects of empathy, and we now interpret our MPFC findings as reflecting this region's role as a site of convergence rather than as a marker of a single component. Interestingly, the reviewer’s point is consistent with the results of this work, in which we find MPFC engagement for both affective and cognitive empathy

    We revised the Introduction to reflect the genuine heterogeneity of this region as follows: “At the cortical level, empathy engages a network of regions whose functional roles are partly overlapping (Schurz et al., 2021). Cognitive empathy has been consistently linked to the dorsolateral prefrontal cortex (DLPFC) and the temporoparietal junction (TPJ) (Shamay-Tsoory et al., 2005, 2011; Eres et al, 2015; Marsh et al, 2018). Affective empathy has been associated with activity in medial prefrontal cortex (MPFC) and superior temporal gyrus (STG). Notably, the MPFC has also been associated with cognitive aspects of empathy (Shamay-Tsoory et al., 2009; Hiser & Koenigs, 2018).” (page 2, Introduction section).

    While we already acknowledged the engagement of the MPFC in both affective and cognitive empathy in our results, we revised the discussion as follows to make the point even clearer: “Interestingly, we also found engagement of the medial portion of the PFC for cognitive empathy. This finding has also been documented in some studies on adults (Mitchell, 2009), and it is consistent with the functional overlap between the affective and cognitive sides of empathy (Shamay-Tsoory et al., 2009; Hiser, 2018). This engagement might be interpreted in light of the key role of the MPFC in self-reflection and self-processing (which we recently confirmed in the developing brain too (Bulgarelli, 2019)), a fundamental process for reflecting on others’ mental state and linking them to one’s own (Gusnard, 2001).” (page 7, discussion section).

    (2) Given the disagreement in the past research about the roles of these brain regions, I feel like the authors' hypotheses may be insufficiently justified, and their claim that cognitive empathy develops earlier than affective empathy is a bit overly strong - would it be possible that these brain regions' different rates/patterns of development are irrelevant to specific components of empathy? Given that behavioral data did not show any age difference, and that each brain region can engage in many functions besides empathy (e.g., generic social and emotional processing), I would be more cautious when interpreting these results.

    We thank the reviewer for these cautions. We address each in turn, conceding where our design genuinely constrains our claims and clarifying where we believe our design already accounts for the concern raised.

    Regarding the justification of our hypotheses given disagreement about regional roles, we have substantially revised our treatment of these regions to reflect the functional heterogeneity and ongoing debate in the literature (see our response to comment above), and we now frame our regional assignments as approximate associations rather than one-to-one mappings. We note that this heterogeneity is in fact accommodated by our findings: for example, MPFC was engaged for both affective and cognitive empathy in our data, consistent with its documented role as a region of convergence between the two components. Our hypotheses are now presented in light of this debated literature rather than against a presumed consensus.

    Regarding whether the regions' developmental patterns are "irrelevant to specific empathy components." We agree that the regions examined are multifunctional and engaged in processes beyond empathy, and we have added explicit acknowledgement of this. However, our design specifically guards against a purely generic interpretation. Our findings derive not from overall activation during a single condition, but from within-task contrasts using the same stimuli — emotional versus neutral events, and questions targeting affective versus cognitive empathy, presented to the same children. Generic social, emotional, attentional, and linguistic processing engaged by the scenarios is present across conditions and is therefore largely subtracted out by these contrasts. While we cannot exclude that some portion of the age-related differences reflects maturation of more general processes, a purely generic account is less parsimonious than one involving the empathy-relevant manipulation, since it is specifically the affective-versus-cognitive and emotional-versus-neutral distinctions, not overall task engagement, that varied with age. We now make this explicit in the Discussion as follows: “Importantly, the age-related differences we observed were specific to the empathy-relevant contrasts rather than to overall task engagement. Because our design compared emotional versus neutral events, and questions about emotional states compared to questions requiring cognitive reasoning, using the same stimuli within the same children, generic social, attentional, and linguistic demands were present across conditions and are therefore unlikely to account for the effects. While we cannot fully exclude a contribution from the maturation of more general processes recruited by the task, an account based on these specific empathy-relevant distinctions is more parsimonious than one invoking generic developmental change, since it was these contrasts, not overall activation, that varied with age.” (page 8, discussion section).

    Regarding the absence of behavioural age differences, we would respectfully note that neural and behavioural measures are not expected to track one another in this population, and that a behavioural null does not undermine a neural developmental effect. Indeed, the dissociation between neural and behavioural indices is a central rationale for using fNIRS with preschoolers: behavioural and verbal measures of empathy are known to be unreliable at this age, a point reinforced in our own sample by the substantial proportion of children (~16%) who deliberately responded counter to their evident understanding. Neural measures can therefore reveal developmental change that noisy, verbally-dependent behavioural measures are underpowered to detect. We interpret the neural–behavioural dissociation as consistent with this rationale rather than as evidence against developmental change.

    As mentioned throughout this review, we agree that our central claim was stated too strongly. While our regional findings are interpretable in terms of empathy components, our cross-sectional design and age range cannot establish that cognitive empathy develops earlier than affective empathy as a matter of developmental ordering. We have accordingly softened this claim throughout, reframing our contribution as evidence for the differential cortical engagement of affective and cognitive empathy across the preschool years, and we now present the developmental interpretation explicitly as one account requiring replication (see also our response to Reviewer 1 on this point).

    (3) It would be helpful if the authors report certain parts of their methods and results in more detail.

    (a) During the cognitive/affective empathy tasks, it is not explicitly clear which part of the fNIRS data were included in the analysis.

    We believe the data entering each contrast were already specified through the design matrix regressors and contrast definitions on page 12 of the methods section; however, to make this fully explicit, we have now added the duration of each task phase to the contrast descriptions, as follows: “The betas were then used to compute the following contrasts and test our hypotheses through one sample t-tests:

    - Affective empathy: “(events for the emotional negative scenarios + events for the emotional positive scenarios) > (events for the neutral scenarios), computed over the 8-second event phase;”

    - Cognitive empathy: “(first questions for the emotional negative scenarios + first questions for the emotional positive scenarios) > (questions for the neutral scenarios), computed over the 8-second event phase;”

    - Emotional valence: “(events for the emotional negative scenarios) > (events for the emotional positive scenarios), computed over the 8-second event phase. Fig 1A shows a graphical representation of the contrasts in the block design.”

    We also redirected the reader to Fig. 1A in which is a graphical representation of the contrasts of the block design.

    (b) When the authors did FDR corrections, they should include the q values and adjusted p values. I was also confused about how the FDR correction was conducted - were analyses performed on all 10 ROIs or only the hypothesized regions? I think if the authors have hypotheses about specific regions, they should test their hypotheses first, and then everything else would be exploratory analyses.

    We thank the reviewer for these points. We reported the corrected p-values (q-values) alongside the raw p-values, and the effect sizes in the results section (marked in red in the text).

    Regarding the analysis structure: the one-sample t-tests across all ten ROIs were preregistered (osf.io/crvz2) as exploratory analyses characterising which regions the task recruits, and were FDR-corrected across regions. We then tested for effects of age within the regions functionally defined as engaged by affective and cognitive empathy.

    (c) Additionally, it is unclear what brain template was used and what procedure was followed to map channels of fNIRS data to the template.

    The template used for the co-registration was selected from the Neurodevelopmental MRI Database of the University of South Carolina (http://jerlab.psych.sc.edu/NeurodevelopmentalMRIDatabase/). We have added this in the method section as follows: “Here the channels were coregistered onto an age-appropriate MRI template from the Neurodevelopmental MRI Database of the University of South Carolina (http://jerlab.psych.sc.edu/NeurodevelopmentalMRIDatabase/) (Richards, et al., 2015).”

    The procedure followed to coregister the array was described in our previous work (Bulgarelli et al., 2025): First, reference points, optodes and channel coordinates were estimated for each participant using STORM-Net (https://github.com/yoterel/STORM-Net). This tool estimates the position of the fNIRS optodes on the participant’s scalp, by locating some reference points in a video of the participant wearing the cap against an ideal cap placement onto a head model. Here, we have 3D printed a 5-year-old head model of the MRI template used for the optode coregistration for the offline stage of STORM-NET. Then, a 5-year-old MRI template from the Neurodevelopmental MRI Database of the University of South Carolina (http://jerlab.psych.sc.edu/NeurodevelopmentalMRIDatabase/) was segmented into 5 tissue types (scalp, skull, cerebrospinal fluid, grey matter (GM), white matter) and used to generate a volumetric multilayer mask. Routines from the DOT-HUB toolbox (https://github.com/DOT-HUB) employing iso2mesh were used to create a tetrahedral volumetric mesh, as well as a mesh of the brain (i.e. GM tissue type). For each participant, the MRI mesh model was registered on the participant’s head through an affine transformation using the model’s cranial landmarks (Nasion, Inion, Right and Left Preauricular points, Cz) and the same subject-specific landmarks estimated by STORM-NET. The affine transformation was applied to the GM mesh. The channel coordinates derived from STORM-NET were first co-registered onto the scalp mesh and then projected onto the GM surface mesh. A 1.25 cm radius sphere centred around each channel’s coordinate on the GM mesh was created to evaluate the anatomical regions overlapped by each channel. To this goal the LPBA40 atlas was used, and anatomical labels were assigned to each channel considering a minimum percentage of overlap to that region of 25%. Channels were then assigned to one of the 10 ROIs based on the anatomical label and the spatial location on the brain (see Table 1). Figure 5 provides a graphical representation of the optodes location on the template head and the brain areas covered by the fNIRS array used, where the ROIs are colour-coded.

    References:

    Schurz, M., Radua, J., Tholen, M. G., Maliske, L., Margulies, D. S., Mars, R. B., ... & Kanske, P. (2021). Toward a hierarchical model of social cognition: A neuroimaging meta-analysis and integrative review of empathy and theory of mind. Psychological bulletin, 147(3), 293.

    Recommendations for the authors:

    Reviewing Editor Comments:

    Note that there is a strong recommendation to make the task available to other researchers. Please also consider carefully the task validation (positive, negative, neutral are connoted, not demonstrated). The use of the term 'foundational' model, while understood, was questioned. In addition to the reviewers' comments, I am sure the authors are aware that foundational models have their own meaning in large language models, and thus, this term may be misunderstood. Please take this into account and consider a rephrasing.

    We thank the Editor for these helpful points, which align closely with the reviewers' comments and which we have addressed as follows:

    (i) regarding task availability: We have added an explicit statement that the full task is available upon request to the corresponding author, and we now highlight broader sharing and adaptation of the paradigm as a direction for future work in the Discussion.

    (ii) regarding task validation. We agree that the emotional and neutral conditions are connoted rather than independently demonstrated. We have clarified in the Methods that "neutral" is defined operationally by the non-emotional nature of the scene's central event and the factual focus of its question, while acknowledging that everyday scenarios cannot be entirely free of incidental affective associations. We also now note the converging internal evidence that the conditions functioned as intended: the absence of any significant heart rate change for neutral events (in contrast to both emotional conditions) and the absence of greater neural activation for neutral than emotional scenarios. We further highlight independent validation in more diverse populations as an important future direction.

    (iii) regarding the term "foundational." We appreciate the Editor's and reviewers' point regarding this terminology. We introduced the terms foundational affective empathy model and foundational cognitive empathy model in our prior theoretical paper (Bulgarelli and Jones, 2023, JCPP Advances) where they are defined in detail, and we have retained them here for consistency with that framework and the emerging literature that builds on it. To avoid ambiguity, we now cite our theoretical paper explicitly throughout this work so readers can locate the original definitions. We hope this preserves continuity with our published framework while addressing the concern about potential misreading.

    Reviewer #1 (Recommendations for the authors):

    (1) The opening example, while painting a vivid picture, emphasises affective empathy. It can convey both components for those familiar with the concept, but as described, it may be confusing. A slight adjustment would help make it clear how both understanding and sharing come together in the examples provided.

    We agree with the reviewer’s comment, and we edited the opening to highlight also the cognitive component of empathy as follows:

    “Sharing emotion is universal: a good friend is overjoyed when she gets promoted at work, we recognise why the news matters to her, and we genuinely feel happy for her. Or our sister loses her beloved pet, we understand what the loss means to her, and we feel sad for her loss. This ability to share someone else’s happiness or sadness and understand the reason of their emotion is called empathy.” (page 2 introduction).

    (2) The authors mention a target sample of 75. It would be helpful to know the rationale for this number. Was an a priori power analysis conducted?

    Yes the study was preregistered here https://osf.io/crvz2/overview, in which we wrote:

    - Sample size: Our target sample is 75 3-to-5-year-olds for the experiment, including 10 participants who will be part of the piloting phase. In a preliminary study ran in our lab with the same fNIRS system, we had an attrition rate of 25%, therefore we estimate to have at least 56 toddlers in the final sample.

    - Sample size rationale: Power analysis (G*Power, Faul et al., 2007) estimated that for a repeated measures ANOVA, with empathy component (affective/cognitive) as within-subjects factor, 54 participants are sufficient to detect a small effect (d=0.25), with power=0.97 and alpha=0.05.

    We added this specification also in the method section as follows: “(our target sample was 75 participants, as preregistered here https://osf.io/crvz2/overview).”

    (3) I would recommend reporting corrected p-values rather than raw p-values, and including effect sizes for the t-tests to aid interpretation.

    We reported the corrected p-values (q-values) alongside the raw p-values, and the effect sizes in the results section.

    (4) In Figure 5, there appears to be a child older than 6, though the stated age range is 3-5. Could the authors clarify this?

    We thank the reviewer for spotting this. Our initial plan was to recruit children aged 3 to 5 years (before they turned 6); however, one participant had to reschedule and had just turned 6 at the time of testing. We have corrected the age range throughout the manuscript to reflect this.

    (5) The authors state: "we did not observe any greater brain activations for neutral events compared to emotional events... suggesting that our task is well-suited to elicit empathic reactions in young children" (p. 6). Could the authors explain why the absence of greater neutral activation supports this conclusion?

    Our intended point was: had neutral events or non-emotional reasoning elicited greater activation in any region, this might indicate that condition differences reflected general task demands (e.g., difficulty, novelty, attentional capture) rather than emotional/empathic processing. The absence of any such reverse effect, combined with the significant emotional > neutral activations in empathy-related regions, supports the interpretation that our task selectively engaged empathy-related processing. We have rewritten the passage to make this reasoning explicit rather than implying that the null result alone is sufficient as follows: "Interestingly, we did not observe any greater brain activations for neutral events compared to emotional events, or for questions about neutral states compared to questions about emotional states. Together with the significant emotional > neutral activations reported above, this indicates that condition differences were specific to emotional processing rather than reflecting general differences in task difficulty or attentional demands." (page 7, discussion)

    (6) In the results, "all p <0.2" should read "all p > 0.2" to indicate non-significant differences.

    We apologise for the oversight and we corrected as suggested.

    (7) Figure 4A (left plot) shows an editing menu that needs to be removed.

    We apologise for the oversight, the editing menu has now been removed from the figure.

    (8) "Children with low-empathy traits lack affective empathy, while their cognitive empathy remains intact" is confusing and should be revised to "children with high callous-unemotional or psychopathic traits" for accuracy.

    Following the reviewer’s suggestion, we have replaced “children with low-empathy traits” with “children with high callous-unemotional traits” in the discussion section (page 9) as follows:

    - “On the other hand, it has been proposed that children with high callous-unemotional traits, characterised by reduced affective empathy alongside relatively intact cognitive empathy, may be able to understand others' emotional states without sharing them (64, 65), which may enable them to manipulate others without remorse.”

    - “However, more recent evidence suggests that both children with high callous-unemotional traits and autistic children may be deficient in both dimensions of empathy (67, 68).”

    Consistent with several scholars (Frick and White, 2008; Viding & McCrory, 2012; Frick, 2014), we refrain from using the term "psychopathy" in reference to children, as it is considered more appropriate for adults.

    Reviewer #2 (Recommendations for the authors):

    (1) It could strengthen the paper if the authors validate their task/stimulus with another population. For example, I wonder how the authors decided whether a scenario is neutral, since the example in Figure 1 (where kids wear their wellies and jump in the puddles) seems quite positive to me.

    We thank the reviewer for both suggestions. Regarding validation in another population, we agree this would strengthen confidence in the generalisability of the task, and we now highlight it as an important direction for future work in the Discussion, as follows: “Furthermore, the remarkably high percentage of participants who answered all questions correctly or as expected confirms that the scenarios were well-suited for this age group. Building on this, future work should validate the scenarios in independent and more diverse populations, including children from different cultural and linguistic backgrounds and across a broader age range, to confirm that the stimuli are interpreted as intended and to establish the generalisability of the paradigm.” (page 9, discussion section).

    Regarding the definition of "neutral," we appreciate the opportunity to clarify, as our use of the term is operational. In our task, what defines a scenario as neutral is the event itself: a non-emotional occurrence such as the teacher bringing wellies, rather than an emotional event. We acknowledge that it is difficult to design scenarios that are entirely free of any emotional association, but the defining event in the neutral scenarios was itself neutral, and the associated question probed a fact (e.g., what a character was doing or wearing) rather than an emotional state. We have revised the Methods to make this operational definition explicit, as follows: “For the neutral scenarios, we acknowledge that scenarios depicting everyday activities cannot be entirely free of incidental affective associations; what defined a scenario as neutral was the non-emotional nature of its central event and the factual focus of its question” (page 11, stimuli and procedure section).

    Importantly, our physiological data support the view that the neutral scenarios functioned as a non-emotional baseline: unlike the emotional conditions, the neutral condition produced no significant change in heart rate between the set-up and event phases of the scenario (t(90) = 0.65, p = 0.513), whereas both the emotional negative and emotional positive conditions did. The neutral scenarios also did not elicit greater neural activation than the emotional scenarios in any region. Together, these results indicate that the neutral scenarios served as an emotionally neutral baseline as intended, regardless of whether an individual scene depicts a mildly pleasant activity.

    (2) It would be helpful for future research if the authors made their task stimuli publicly available.

    We would be very happy to share the task with anyone who is willing to use it upon reasonable request. We have added this in the code and data availability section and in the discussion section as follows: ”This task can be shared with the scientific community upon request to the corresponding author and potentially used with children of different ages and backgrounds” (page 9, discussion section).

    (3) I found it weird that the authors corrected for motion artifacts using a wavelet filter but not a spline, with the reason being "fNIRS data quality was overall good". Do the results change using spline interpolation?

    Our decision not to apply spline interpolation in addition to wavelet filtering was deliberate and methodologically motivated, rather than based solely on overall data quality. Spline interpolation (Scholkmann et al., 2010) is designed to correct large, discrete motion artifacts (abrupt shifts and spikes) by modelling and subtracting them from identified segments. In our paradigm, preschoolers were seated and watched a screen without any movement requirement, so the data were characterised by low-amplitude, diffuse motion rather than large discrete artifacts. Wavelet filtering (Molavi & Dumont, 2012) is well-suited to this profile, as it attenuates motion-related components across the signal without requiring discrete artifact detection. Because each successive correction step alters the underlying signal, we chose the most parsimonious pipeline appropriate to our data, deviating from the preregistration in the conservative direction of minimal manipulation. Consistent with this, the recovered haemodynamic response functions shown in Figures 2–3 are smooth and free of residual spikes or motion-related artifacts, indicating that the wavelet-based correction adequately recovered clean haemodynamic responses.

    (4) The number of participants involved was confusing throughout the paper.

    (a) The degrees of freedom kept changing throughout the results section (there were 79, 85, 86, 87, 90, etc.). If the authors excluded participants in certain analyses, they should say so and also state the reason/criteria.

    The varying degrees of freedom reflect the fact that the number of participants contributing to each analysis differed depending on data quality at the relevant channels: when an ROI had no channels surviving our signal-quality and motion-correction criteria for a given participant, that participant could not be included in analyses involving that ROI. As ROIs differed in how many participants had at least one usable channel, the number of participants, and therefore the degrees of freedom, varies across regions and analyses. We have clarified this in the Methods as follows: “When an ROI had no channels surviving our quality criteria for a given participant, that participant was excluded from analyses involving that ROI. Therefore, degrees of freedom therefore may vary across regions.” (page 12, fNIRS data processing section).

    (b) When the authors used percentages of participants to describe the results, it is unclear to me what the population was (was it all 95 participants, or the 87 participants with valid fNIRS + heart data)? It was confusing when it says "10% did not complete all trials", and 10% of either 95 or 87 is a decimal number and not a whole number. Maybe the authors should just include the exact numbers in addition to the percentages. It would also be clarifying if the percentages were rounded to the same decimal place (i.e., 73.3% and 57.0%, rather than 73.3% and 57%; the same thing also applies to the p-values and other numbers).

    We thank the reviewer for asking us to double-check the numbers of the verbal responses to the task questions. These were originally calculated on the N = 90 participants with good fNIRS data. On reflection, we consider it more appropriate to calculate them on all N = 95 participants who observed the stimuli, regardless of fNIRS data quality, since verbal responses are independent of fNIRS signal quality. We have revised these figures accordingly and now report the exact number of participants alongside each percentage (to two decimal places throughout for consistency). Here the revised version of the results section (page 4):

    “Verbal responses to the task questions

    First, we assessed compliance with our task by evaluating how many of the 16 proposed trials (8 emotional and 8 neutral) participants attended (out of the total N=95 of participants who observed the stimuli). 88.42% (N= 84) of participants completed all 16 trials (mean number of emotional scenarios attended = 7.83/8, mean number of neutral scenarios attended = 7.79/8). Among those who did not complete all trials, all attended more than 75% of the emotional scenarios, while 63.64% (N=7) attended more than 75% of the neutral scenarios. There was no significant effect of age on the number of trials included (all p > 0.1). This suggests that our novel task is highly suitable for children aged 3 to 5 years.

    We then explored verbal responses to the task questions, both in the emotional and neutral scenarios. 86.32% (N=82) of the participants answered congruently with the character’s emotions at the first emotional question (‘how does s/he feel?’), 72.63% (N=69) of the participants answered congruently (i.e. what we would expect the child to feel) with the character’s emotions at the second emotional question (‘how does this make you feel?’), and 55.79% (N=53) of the participants made no mistakes, and 26.32% (N=25) made one mistake on the question in the neutral scenarios. Older age was associated with a lower percentage of mistakes when asked how the other child felt in the emotional scenarios (r(93) = -0.343, p = 0.001) and about events within the neutral scenarios (r(93) = -0.357, p = 0.001). Age did not associate with accuracy when asked ‘how does this make you feel?’ (r(93) = -0.147, p = 0.157).”

    (5) Given that the results come before methods in this paper, it would be helpful for readers if the authors included some very brief description of the numbers being reported in the results section (e.g., the beta values) as well as the experimental procedure.

    We have added a brief description of the task at the start of the Results, as follows: “N = 95 preschoolers watched 16 scenarios (8 emotional, with 4 positive and 4 negative, and 8 neutral) featuring child characters experiencing emotional or neutral events, while neural activations and heart rate were recorded. In emotional scenarios, the emotional event served to elicit affective empathy, and was followed by a question about the character's emotional state ('how does s/he feel?', cognitive empathy), a question about the child's own feelings ('how does this make you feel?'). Neutral scenarios followed the same structure with a non-emotional event and a factual question. Full details are provided in the Methods (see also Figure 1A).” (page 4, results section).

    We have also added and a short explanation of the beta values and their interpretation where neural results are first reported as follows: “Beta values index the strength of the haemodynamic response for each condition derived from a general linear model; a positive value reflects greater activation for the condition of interest relative to the comparison condition (see Methods).” (page 4, results section).

    (6) The paper mentioned "see Table 1 and Figure 4" when describing fNIRS data, but I didn't see a Table 1, and Figure 4 does not contain fNIRS data (maybe it is supposed to be Figure 3?).

    We refer to Table 1 and Figure 4 of our other work C. Bulgarelli, P. Pinti, N. Aburumman, E. J. H. Jones, Combining wearable fNIRS and immersive virtual reality to study preschoolers’ social development: a proof-of-principle study on preschoolers’ social preference. Oxford Open Neuroscience 2, 1–15 (2023). in which we described the coregistration and the process to create the ROIs.

    (7) I found this statement confusing: "There was no significant effect of age on the number of trials included (*all p < 0.25*)". Were p-values < 0.05 also true or not? Or were they between 0.05 and 0.25? There was another sentence that says p < 0.2 (was p < 0.05 also true or not?)

    We apologise for the oversight. The correct sign of the p-value was >, not <. We have corrected the statement as follows: “There was no significant effect of age on the number of trials included (all p > 0.25)”. (page 4 results section).

    (8) For a question like "how does this make you feel?", how is accuracy defined? Is it just whether or not the answer is congruent with the emotion shown in the picture?

    Yes, accuracy is defined by whether the participant's answer is congruent with the emotional event (i.e., what we would expect the child to feel). We specified this in the text as follows: 73.3% of the participants answered “congruently (i.e. what we would expect the child to feel) with the character’s emotions at the second emotional question (‘how does this make you feel?’)” (page 4 results section).

    (9) Relatedly, when discussing the results between age and accuracy of 'how does this make you feel?' (r(90) = -0.180, p = 0.089), I think concluding that "any age-related changes in general performance were not related to empathy" is overly strong. Since p < 0.10, there may be very weak evidence of an association, and I would be more cautious when interpreting this result.

    We thank the reviewer for this point. Following our re-analysis of the verbal responses on the full sample (N = 95; see our response to comment 4b), this correlation is now r(93) = −0.147, p = 0.157, i.e. non-significant. We agree that our original interpretation was too strong, and we have revised the sentence accordingly. We no longer conclude that age-related performance changes were unrelated to empathy; instead, we report simply that accuracy on the 'how does this make you feel?' question was not significantly associated with age.

  7. Version published to 10.1101/2025.03.24.645025v2 on bioRxiv
  8. Version published to 10.7554/elife.110899.1 on eLife
  9. eLife

    eLife Assessment

    The authors have presented a study which addresses a recognised gap in the literature, the emergence of the neural correlates of cognitive and affective empathy in children; they introduce a task for measuring both positive and negative empathy in a relatively large group of children aged 3-5. The task was combined with functional near-infrared spectroscopy to examine brain regions involved in the task. The findings are interpreted as providing evidence for the earlier emergence of cognitive than affective empathy. The study represents a valuable contribution to understanding the development of cognitive function, but in its current form, the strength of support for the conclusion is incomplete due to limited support for the comparison to the adult literature and a need to more clearly justify the pre-selected brain regions, their links to empathy and the justification of the hypotheses.

  10. eLife

    Reviewer #1 (Public review):

    Public Review

    This paper presents an fNIRS neuroimaging study with a relatively large sample of preschool children (aged 3-5) that measures both positive and negative empathy within a single task. Children watch emotional events and are asked questions about both their own emotions and the emotions of others, allowing the authors to distinguish between affective and cognitive empathy. The authors propose "foundational" models of affective and cognitive empathy and argue that their findings support the idea that cognitive empathy emerges before affective empathy in early childhood.

    Strengths:

    The paper addresses a valuable question by measuring both positive and negative empathy within a single cognitive task. The use of fNIRS with a relatively large preschool sample is commendable, and the pre-registered design strengthens the contribution. The task itself is innovative, well-suited to this age group, and achieves high compliance, which is essential and notably difficult with young children. Overall, the methods are appropriate, and the empirical work is valuable.

    Weaknesses:

    The main concerns relate to the framing of the paper rather than the empirical work itself.

    The introduction contains several claims that are overstated or inaccurate. The statement that "we know very little about the development of this fundamental social skill during the first years of life" does not reflect the state of the field; empathy in early development has been quite extensively studied (e.g., Davidov et al., Malti et al., Uzefovsky et al., Decety et al., Feldman et al., among others). The view that emotional contagion directly develops into affective empathy is based on early theoretical accounts that have since been challenged by empirical evidence (see Davidov et al., 2025). The claim that cognitive empathy does not require theory of mind is also overstated - it is hard to see how theory of mind, the understanding that others have thoughts, beliefs, and emotions that may differ from our own, would not be required for cognitive empathy. Furthermore, the introduction neglects recent and directly relevant work (e.g., Zach et al., 2025; Uzefovsky et al., 2020; Davidov et al., 2021).

    Most critically, the claim that "no neuroimaging studies have yet investigated brain regions supporting empathy in preschoolers" is inaccurate. Multiple studies have examined brain regions supporting empathy in children within this age range, including work using fNIRS and studies of positive empathy (e.g., Decety et al., 2018; Light et al., 2009; Levy et al., 2019; Bray et al., 2022; Brink et al., 2011). This is also not the first study to measure brain activation in response to positive and negative emotional events in children (e.g., Cheng et al., 2014; Light et al., 2009). These novelty claims need to be corrected.

    The use of "explicit" to describe cognitive empathy and "implicit" or "spontaneous" to describe affective empathy is problematic. Affective empathy can be expressed quite explicitly, through facial expressions, verbal statements, and gestures, and framing it as spontaneous overlooks the motivational dimensions of empathy (e.g., Zaki and colleagues). The authors' use of "foundational affective empathy model" and "foundational cognitive empathy model" as though these are established concepts is not well supported by the current evidence base.

    The conclusions in the discussion go beyond what the data can support. The question of whether cognitive or affective empathy emerges first cannot be adequately addressed with a cross-sectional sample aged 3-5, an age at which affective empathy is likely already well established and cognitive empathy is expected to be developing around the lower end of this range. The cross-sectional design further limits what can be inferred about developmental trajectories during a period of substantial individual variability. Together, these issues make the developmental-precedence conclusions difficult to sustain. The claim that the results demonstrate "the first time that this brain specialisation for stimuli of different emotional valence may be rooted in childhood" is also inaccurate, as there is prior evidence for brain specialisation of emotional valence in early childhood (e.g., Grossmann et al., 2007).

    Appraisal:

    The empirical contribution, the task design, the fNIRS data, and the analyses are sound and have value for the field. However, in its current form, the paper does not achieve what it sets out to do. The novelty claims are undermined by the omission of a substantial body of relevant prior work, and the developmental conclusions are not adequately supported by the cross-sectional design and age range studied. The abstract similarly overstates the support this study provides for the early emergence of cognitive over affective empathy.

    Impact:

    With appropriate revision, this work could make a meaningful contribution. The task is well-designed for studying empathy in young children and could be useful to other researchers in the field. The fNIRS data from a large preschool sample are a valuable resource. However, the contribution needs to be framed accurately, both in terms of what is genuinely novel relative to the existing literature and in terms of what conclusions the data can and cannot support.

  11. eLife

    Reviewer #2 (Public review):

    Summary:

    Authors examined neural substrates for cognitive empathy (conceptually understanding others' emotions) versus affective empathy (automatically sharing others' emotions) development in 3-5-year-old toddlers, and argued that cognitive empathy emerges earlier than affective empathy, challenging the predominant view that affective empathy develops earlier. The authors developed an empathy test for toddlers while measuring their brain activity with fNIRS (particularly in MPFC, STG, DLPFC, and TPJ) and heart rate. They found different brain region activation in cognitive versus affective empathy tasks, as well as age-related changes in the activation of right MPFC and right TPJ.

    Strengths:

    This work investigated the development of different components of empathy, which is a quite understudied topic. The authors developed an age-appropriate task for toddlers to measure their cognitive empathy and affective empathy, which is likely useful for future research in this field. Their methods are sound, and give a relatively large sample; the results look interesting and relatively solid, except for certain details in the reporting of methods and results.

    Weaknesses:

    (1) My major concern is the roles of brain regions hypothesized and found in this paper (MPFC, STG, DLPFC, and TPJ) - the authors seemed to have omitted a large portion of the literature on this topic. Prior works have found that these brain regions may be involved in more than one process, or involved in processes that are common to both cognitive and affective empathy (see Schurz et al., 2021). In particular, MPFC seems to be indicated more often in cognitive empathy, and STG may be involved in both cognitive empathy and intermediate processes, which is contradictory to what the author claimed and hypothesized. Relatedly, when the authors made statements like "these results highlight that regions underpinning affective and cognitive empathy in preschoolers largely resemble those documented in adults" (without proper citations), I found it unconvincing due to the disagreements in the past adult research about brain regions related to empathy, which were not quite discussed in the current paper. It may be helpful if the authors do a more thorough literature review and provide a more comprehensive view of how their results fit in the existing literature.

    (2) Given the disagreement in the past research about the roles of these brain regions, I feel like the authors' hypotheses may be insufficiently justified, and their claim that cognitive empathy develops earlier than affective empathy is a bit overly strong - would it be possible that these brain regions' different rates/patterns of development are irrelevant to specific components of empathy? Given that behavioral data did not show any age difference, and that each brain region can engage in many functions besides empathy (e.g., generic social and emotional processing), I would be more cautious when interpreting these results.

    (3) It would be helpful if the authors report certain parts of their methods and results in more detail.
    a) During the cognitive/affective empathy tasks, it is not explicitly clear which part of the fNIRS data were included in the analysis.
    b) When the authors did FDR corrections, they should include the q values and adjusted p values. I was also confused about how the FDR correction was conducted - were analyses performed on all 10 ROIs or only the hypothesized regions? I think if the authors have hypotheses about specific regions, they should test their hypotheses first, and then everything else would be exploratory analyses.
    c) Additionally, it is unclear what brain template was used and what procedure was followed to map channels of fNIRS data to the template.

    References:

    Schurz, M., Radua, J., Tholen, M. G., Maliske, L., Margulies, D. S., Mars, R. B., ... & Kanske, P. (2021). Toward a hierarchical model of social cognition: A neuroimaging meta-analysis and integrative review of empathy and theory of mind. Psychological bulletin, 147(3), 293.

  12. Mentioned by @biorxivpreprint.bsky.social on Bluesky

    Head before heart: cognitive empathy emerges before affective empathy in the developing brain https://www.biorxiv.org/content/10.1101/2025.03.24.645025v1

  13. Mentioned by @biorxiv-neursci.bsky.social on Bluesky

    Head before heart: cognitive empathy emerges before affective empathy in the developing brain https://www.biorxiv.org/content/10.1101/2025.03.24.645025v1

  14. Version published to 10.1101/2025.03.24.645025v1 on bioRxiv