Abstract
Although domestic dogs were the first domesticated species, the nature of dog domestication remains a topic of ongoing debate. In particular, brain and behavior changes associated with different stages of the domestication process have been difficult to disambiguate. Most modern Western breed dogs possess highly derived physical and behavioral traits because of intense artificial selection for appearance and function within the past 200 years. In contrast, premodern dogs, including primitive/ancient breeds, village dogs, and New Guinea Singing Dogs, have undergone less intensive artificial selection and retain more ancestral characteristics. Consequently, comparisons between modern and premodern dogs can shed light on brain and behavior changes that have occurred recently in the domestication process. Here, we addressed this question using a voxel-based morphometry analysis of structural MRI images from 72 modern breed dogs and 13 premodern dogs (32 females). Modern breed dogs show widespread expansions of neocortex and reductions in the amygdala and other subcortical regions. Furthermore, cortical measurements significantly predicted individual variation in trainability, while amygdala measurements significantly predicted fear scores. These results contrast with the long-standing view that domestication consistently involves reduction in brain size and cognitive capacity. Rather, our results suggest that recent artificial selection has targeted higher-order brain regions in modern breed dogs, perhaps to facilitate behavioral flexibility and close interaction and cooperation with humans.
Significance Statement
This study provides novel insights into the neural changes associated with artificial selection during dog domestication by comparing brain morphology between modern breed dogs and a unique and rare sample of premodern dogs, including ancient breeds, village dogs, and New Guinea singing dogs. Our findings demonstrate that modern breed dogs exhibit significant cortical expansion linked to trainability and premodern dogs show amygdala enlargement associated with heightened fear, suggesting that brain evolution has happened rapidly in a species embedded in the anthropogenic environment.
Introduction
Dogs were the first species to be domesticated, but the nature of dog domestication remains a topic of intense debate (Thalmann and Perri, 2019). The predominant view is that gray wolves, Canis lupus, initiated their own domestication by scavenging around human settlements (Coppinger and Coppinger, 2001; Larson and Fuller, 2014). However, others argue that hunter-gatherers may have intentionally captured and raised wild wolves to keep as pets, hunting companions, or guards (Germonpré et al., 2021; Serpell, 2021). In either case, behavioral shifts which presumably would have become adaptive during the initial wild-to-domesticate transition include reduced flight-or-fight responses, reduced reactive aggression, and potentially increased behavioral flexibility and learning capacity. These behavioral shifts in early dogs are generally understood to have occurred through natural selection, rather than artificial selection (Scott, 1968; Boyko et al., 2009; Lupo, 2019).
Today, most dogs in the world continue to breed freely (Hughes and Macdonald, 2013; Gompper, 2014). Free-breeding dogs who range freely around human settlements and rely on human waste for survival are termed village dogs (Coppinger and Coppinger, 2001). There are also a few smaller populations of free-breeding dogs, including New Guinea singing dogs (NGSDs), that primarily hunt prey rather than rely on anthropogenic sources of food (Boyko and Boyko, 2014; Surbakti et al., 2020).
However, some dog lineages transitioned from an autonomous lifestyle to one controlled by humans. The oldest dog breeds in existence today are called ancient breeds (Parker et al., 2004; Vonholdt et al., 2010). Unlike modern breeds, ancient breeds were traditionally maintained in open populations without pedigree records and were primarily selected for function, not conformation (Boyko and Boyko, 2014; Worboys et al., 2018).
Modern dog breeding practices began in earnest in the United Kingdom during the Victorian period (Worboys et al., 2018). With the advent of dog clubs, shows, and trials, coupled with scientific advances in animal breeding, people started to systematically breed dogs in closed populations for highly specialized morphologies and behavioral tendencies (Pedersen et al., 2013; Worboys et al., 2018). These modern breeds have been subject to intensive artificial selection (Konno et al., 2016; Hansen Wheat et al., 2019), unlike premodern dogs.
Therefore, premodern dogs, including ancient breeds, village dogs, and NGSDs, offer a window on what early dogs may have been like after the wolf-to-dog transition (Hansen Wheat et al., 2019; Thalmann and Perri, 2019). Comparisons between premodern and modern breed dogs are useful for reconstructing the evolution of traits that are not preserved in skeletal remains, such as the brain and its behavioral output. Behavioral shifts in dog evolution have been a topic of ongoing debate (Marshall-Pescini et al., 2017). On the one hand, domestication is generally presumed to reduce brain size across species (Kruska, 1988; Hecht et al., 2023). Domestication is also generally thought to involve reductions in flight-or-fight responses (Coppinger and Coppinger, 2001; Trut et al., 2009; Zeder, 2012). On the other hand, some argue that modern dogs possess social cognitive capacities which surpass those of extant wolves (Hare et al., 2002; Salomons et al., 2021; but see Range and Marshall-Pescini, 2022), and in at least one population of dogs, intraspecies aggression does not appear to be reduced in dogs relative to wolves (Range et al., 2015). Nearly all behavioral, cognitive, and neuroscience research on dogs to date has focused on modern breeds. Because behavior originates in the brain, comparisons between premodern and modern dogs are necessary to understand selection pressures and behavioral shifts resulting from intensified artificial selection during the late stage of the domestication process.
In this study, we collected T2-weighted MRI scans from 85 dogs of modern and premodern ancestry. Using voxel-based morphometry, we find that modern dog breeds show significant, widespread expansion across cortex, while premodern dogs show significantly larger subcortical structures, particularly the amygdala. Cortical and amygdala expansion are significantly predictive of individual variation in trainability and fear scores, respectively, using the Canine Behavioral Assessment and Research Questionnaire (Hsu and Serpell, 2003).
Materials and Methods
Ethics statement
All procedures were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) and Internal Review Board (IRB) of Harvard University.
Subjects
Canine subjects were recruited from the Penn Vet Working Dog Center, America's VetDogs, Fidelco Guide Dogs, and private owners. Rather than target-specific breeds, we included all eligible dogs of single breed ancestry and later categorized their modern breed versus premodern status based on available genetic data (Parker et al., 2004) and historical information derived from the “World Atlas of Dog Breeds” (De Vito et al., 2009). Veterinarians screened all candidate subjects prior to inclusion in the study. All subjects were in good health and had no neurological disorders. Dog owners provided informed consent. A total of 85 dogs (32 females) took part in the study, 13 of which were premodern dogs (Extended Data Table 1). All dogs were working and/or companion dogs.
MRI scan acquisition
Dogs were anesthetized by a licensed veterinarian during their MRI scans to ensure that they did not experience stress or discomfort from being inside the machine. First, dogs were sedated with butorphanol and dexmedetomidine (0.3 mg/kg +3–5 mcg/kg, i.m.). Next, general anesthesia was induced with propofol (to effect, 2–6 mg/kg, i.v.) and maintained with isoflurane O2. A licensed veterinarian and two veterinary technicians continuously monitored dogs under anesthesia using ECG, SpO2, NIBP (oscillometric), and capnography (InVivo Expression MRI-safe unit). T2-weighted structural images (0.667 mm3, 1 average) were collected. The duration of the scanning was ∼1 h per dog. After the scanning completion, dogs were recovered from anesthesia under the observation of a veterinary technician. Dogs were sent home with a bandana and a framed image of their brain.
Image processing and analyses
Data processing was performed with two open-source software packages—FSL (Smith et al., 2004, Jenkinson et al., 2012; https://fsl.fmrib.ox.ac.uk/fsl/fslwiki) and ANTs (Avants et al., 2011; http://picsl.upenn.edu/software/ants/). Brain extraction was accomplished with an in-house pipeline using ANTs (Avants et al., 2011). We generated a study-specific, symmetric template by registering and merging images from 13 premodern dogs with a randomly selected set of 13 modern breed dogs. The images of all subjects were then nonlinearly registered to the template using ANTs (Avants et al., 2011). Bias-field correction and segmentation into gray matter, white matter, and cerebrospinal fluid was accomplished using the FSL FAST tool (Zhang et al., 2001). Total gray matter and white matter volume were added together to determine each subject's total brain volume, which was converted to brain weight in grams.
Each subject's gray matter mask was multiplied by the Jacobian determinant of the ANTS warpfield. This created a Jacobian image containing just gray matter voxels, which is essentially a “heat map” of the spatial transformation that needed to occur to align a subject's brain with the template brain. FSL's randomise tool was then applied to perform Monte Carlo permutation testing of a general linear model comparing whole-brain Jacobian images of modern breed and premodern dogs (Winkler et al., 2014). The threshold used for statistical significance was p < 0.05 after threshold-free cluster enhancement (TFCE) for family-wise error correction (Smith and Nichols, 2009).
Brain–body allometry analyses
To determine if premodern and modern breed dogs show different brain–body allometric scaling patterns, a least-squares phylogenetic analysis of covariance (pANCOVA) was performed (Smaers and Rohlf, 2016) using phylogenetic data from Parker et al. (2017). This analysis compared three modern breeds—Labrador retriever, border collie, and German shepherd—with four premodern breeds—saluki, samoyed, Shiba inu, and Siberian husky. Two additional nonphylogenetic versions of this analysis were also performed. The first version consisted of a pANCOVA with a nulled phylogeny, which included all dog types in the dataset. The second version involved two linear regression models—one model with two intercepts and one slope and another with one intercept. These models were then compared using an ANOVA.
We additionally performed two more voxel-based morphometry analyses to further interpret the findings of our pANCOVA analysis. One voxel-based morphometry analysis investigated the relationship between brain size scaling and regional gray matter expansions and contractions. The other voxel-based morphometry analysis investigated differences in gray matter volume between modern breed and premodern dogs, while controlling for brain size by including it as a covariate.
Canine Behavioral Assessment and Research Questionnaire
The owners of 61 canine subjects (51 modern breed) completed the Canine Behavioral Assessment and Research Questionnaire (C-BARQ; Hsu and Serpell, 2003). Questions 1 through 8 from the trainability subscale were averaged to create a trainability score for each subject (reverse coded items were converted). The same approach was used to generate scores for stranger-directed fear (Questions 36, 37, 39, 40), dog-directed fear (Questions 45, 46, 53, and 54), and nonsocial fear (Questions 38, 41, 42, 44, 47, 48). We also generated an “overall fear” score by averaging questions from all three fear subscales together (Questions 36–54). We generated five separate linear regression models to determine whether each score was significantly predicted by expansions in brain anatomy (Table 1).
Summary of linear regression models generated to test the relationship between C-BARQ behaviors and measure of amygdala and cortical expansion
Results
Regional gray matter volume
First, we investigated whether regional gray matter volume across the whole brain varied significantly between 72 modern breed and 13 premodern dogs. Our voxel-based morphometric analysis of T2-weighted images found that modern breed dogs showed expansion throughout the cortex. Specifically, expansions were observed in the frontal gyrus, gyrus rectus, prorean gyrus, composite gyrus, coronal gyrus, precruciate gyrus, postcruciate gyrus, sylvian gyrus, ectosylvian gyrus, suprasylvian gyrus, and marginal gyrus (Fig. 1, Extended Data Fig. 1, Extended Data Table 2). Additionally, modern breed dogs showed volumetric reductions in subcortical regions, including the amygdala, right nucleus accumbens, hippocampus, and cerebellum (Fig. 2, Extended Data Fig. 1, Extended Data Table 3). An additional area of reduction was detected abridging the dorsal aspect of the hypothalamus and ventral aspect of the thalamus surrounding the third ventricle. While the resolution of structural MRI does not allow for precise identification of the borders of hypothalamic and thalamic nuclei, the anatomical location of this cluster is consistent with the nucleus reuniens, paraventricular thalamic nucleus, centromedial thalamic nucleus, posteromedial ventral thalamic nucleus, dorsomedial hypothalamic nucleus, and periventricular hypothalamic nucleus. This cluster extends posteriorly along the gray matter surrounding the third ventricle, consistent with the location of the posterior hypothalamic nucleus and posterior paraventricular hypothalamic nucleus.
Significantly enlarged regions of gray matter volume in modern breed dogs compared with premodern dogs. Colored regions are all p < 0.05 after correction for multiple comparisons; t statistic values are illustrated. Visualized on a canine brain atlas created by Johnson et al. (2020).
Significantly enlarged regions of gray matter volume in premodern dogs compared with modern breed dogs. Colored regions are all p < 0.05 after correction for multiple comparisons; t statistic values are illustrated. Visualized on a canine brain atlas created by Johnson et al. (2020).
Brain–body allometry
Next, we examined whether premodern and modern breed dogs show different allometric scaling patterns by performing a pANCOVA that incorporated phylogenetic data from Parker et al. (2017). We found no significant difference in the slope or intercept for modern breed versus premodern dogs (F = 0.49, p = 0.65). Two follow-up analyses using nonphylogenetic approaches found comparable results (both F = 1.42, p = 0.26).
Although brain volume did not differ significantly between the two groups, we questioned whether variation in brain size might still be significantly associated with variation in regional gray matter morphometry. Previously, we had observed that total brain size largely accounts for variation in cortex volume, which is linked to breed-average trainability (Hecht et al., 2021a). In order to assess the extent to which variation in brain size across modern and premodern dogs accounts for the results observed here, we then performed an additional voxel-based morphometry analysis investigating the relationship between brain size and regional gray matter volume. We found that increased brain size is associated with widespread expansion across the cortex, in regions very similar to those that are expanded in modern breed dogs compared with premodern dogs (Fig. 3, Extended Data Fig. 2, Extended Data Table 4). Furthermore, reduced brain size is associated with volumetric reduction in subcortical regions, again very similar to those that are reduced in modern breeds compared with premodern dogs (Fig. 3, Extended Data Fig. 2, Extended Data Table 5).
Regions of gray matter volume that are significantly enlarged with increasing brain size (light blue) and decreasing brain size (orange). Colored regions are all p < 0.05 after correction for multiple comparisons; t statistic values are illustrated. Visualized on a canine brain atlas created by Johnson et al. (2020).
Finally, to assess whether there is any variation in gray matter morphology that is not accounted for by variation in total brain size, we compared regional gray matter differences between premodern and modern breed dogs while controlling for total brain size. We found remaining regions of significant cortical expansion in modern breeds located in parts of the left prorean gyrus, anterior suprasylvian gyrus, and postcruciate gyrus (Fig. 4, Extended Data Fig. 3, Extended Data Table 6).
Significantly expanded regions of gray matter volume in modern breed dogs compared with premodern dogs while controlling for brain size. Colored regions are all p < 0.05 after correction; t statistic values are illustrated. Visualized on canine a canine brain atlas created by Johnson et al. (2020).
C-BARQ trainability and fear scores
Owners of 61 of the dogs in this dataset completed the C-BARQ (Hsu and Serpell, 2003; Duffy and Serpell, 2012). We generated five linear regression models to determine if amygdala/cortical expansion significantly predicts individual C-BARQ trainability, general fear, stranger-directed fear, dog-directed fear, and nonsocial fear scores (Table 1). We found that cortical expansion significantly predicts increased trainability (p = 0.00019, t value = 4.0), while expansion of the amygdala significantly predicts increased overall fear (computed as the average of all C-BARQ fear scales; p = 0.0016, t value = 3.3), stranger-directed fear (p = 1.25 × 10−7, t value = 6.0), dog-directed fear (p = 0.024, t value = 2.3), and nonsocial fear (p = 3.66 × 10−5, t value = 4.5; Fig. 5, Table 1). Furthermore, in five additional linear regression models, we found that modern breed versus premodern status significantly predicts individual C-BARQ trainability (p = 3.8 × 10−05, t value = −4.46), overall fear (p = 0.023, t value = 2.34), stranger-directed fear (p = 0.040, t value = 2.11), dog-directed fear (p = 0.041, t value = 2.1), and nonsocial fear scores (p = 0.0166, t value = 2.5; Fig. 6, Table 2).
Relationship between regional expansion in gray matter volume and behavior scores in modern breed and premodern dogs. Regional expansion is shown in the Jacobian warp values for each individual dog; positive numbers indicate expansion relative to the average brain template whereas negative numbers indicate contraction. Each panel corresponds to the relationship between brain expansion and the following C-BARQ behaviors—trainability (A), overall fear (B), stranger-directed fear (C), dog-directed fear (D), and nonsocial fear (E). Numbered points represent individuals with the highest behavior score and/or most regional brain expansion: (1) Joseph, a companion Labrador retriever; (2) Otis, a Labrador retriever bred and trained for service work; (3) Jade, a companion Korean village dog; (4,5, and 7) Kopi, a companion New Guinea singing dog; (6) Sugar, a companion Indian village dog. p values derived from linear logistic regression models (Table 1).
Comparison between modern breed and premodern dogs for C-BARQ scores for trainability (A), overall fear (B), stranger-directed fear (C), dog-directed fear (D), and nonsocial fear (E). p values derived from linear regression models (Table 2).
Summary of linear regression models generated to test the relationship between C-BARQ behaviors and ancestry (modern breed vs premodern status)
Discussion
The present study investigated differences in relative brain size and regional gray matter volume between modern breed dogs, which have undergone a high degree of artificial selection, and premodern dogs, which have been subject to less intensive artificial selection (Boyko and Boyko, 2014). We performed whole-brain voxel-based morphometry on T2-weighted MRI scans of 85 working and companion dogs. Additionally, owners of 61 of the canine subjects completed the C-BARQ (Hsu and Serpell, 2003), a widely validated assessment of dog behavior. Using these two sources of information, we find evidence that dogs’ brains and behaviors have continued to evolve after the initial wolf-to-dog transition.
In our voxel-based morphometry analysis of regional gray matter volume, we found significant and widespread expansion of the cortex in modern breed dogs compared with premodern dogs. This includes large extents of frontal, temporal, parietal, occipital, and cingulate cortex. The cortex is broadly implicated in higher-order sensory processing, action planning, cognition, and flexible/learned behavior. Given previous findings that modern breed dogs are more trainable than ancient breed dogs (Turcsán et al., 2011 ; Smith et al., 2017), we assessed (1) whether trainability significantly differs between modern breed and premodern dogs and (2) if cortical expansion significantly predicts a dog's trainability score. As expected, modern breed dogs were rated as significantly more trainable than premodern dogs (Fig. 3). Furthermore, cortical enlargement positively predicted trainability (Fig. 4). These results build upon a previous finding by Hecht and colleagues that there is a significant association between breed-average trainability scores and cortical expansion (2021), although this is the first direct assessment of brain anatomy and trainability ratings within the same individuals. Together, these results suggest that artificial selection in modern breed dogs has acted to increase trainability and cortical volume. Why might this have occurred? In their most common role(s) as household pets and/or working dogs, modern breed dogs interact closely with humans. This niche may have favored dogs with larger, more plastic cortices that were better able to learn novel behaviors and skills from humans. Consistent with this interpretation, the dogs with the highest trainability scores and greatest cortical expansion were purpose-bred working dogs (Fig. 4). Brain tissue is metabolically expensive (Aiello and Wheeler, 1995), which generally limits brain enlargement across species (Isler and van Schaik, 2014), but since household pets and working dogs are well provisioned and have a stable diet, their brains may have been released from this constraint (Hecht et al., 2023).
Compared with premodern dogs, modern breeds also show gray matter reduction in several subcortical structures, most markedly including the amygdala (Fig. 2). The amygdala processes both positive and negative environmental stimuli but is especially well-known for its role in fear learning (Janak and Tye, 2015). In a post hoc analysis, we further discovered that expansion of the amygdala significantly predicts owner-reported overall fear (computed as an average of all C-BARQ fear scores), stranger-directed fear, dog-directed fear, and nonsocial fear (Fig. 5, Table 1). Our findings are therefore consistent with the idea that the flight-or-fight response may have been targeted by selection over the course of dog domestication (Price, 1984), not just in the initial transition of early modern dogs from a now-extinct wolf ancestor, but also in more recent, ongoing selection after domestication had occurred. Premodern dogs may have historically benefited from having elevated fear levels as they lived more independently from humans and may have needed to evaluate and avoid threats in the environment by themselves. On the other hand, modern breed dogs have evolved to be highly integrated into human society, and survival in this environment might require reduced fearfulness. Humans may also directly select against behaviors resulting from flight-or-fight processing, particularly in the case of working dogs, where these behaviors can conflict with work goals. Notably, portions of the hypothalamus, which is also centrally involved in flight-or-fight processing, were also reduced in modern breed dogs (Fig. 2).
To evaluate whether relative brain size is different between modern breed and premodern breed dogs, we conducted a pANCOVA (Smaers and Rohlf, 2016). This analysis did not show a significant difference in relative brain size between modern breeds and premodern dogs. Additional nonphylogenetic analyses produced the same result. However, when we subsequently performed a whole-brain voxel-based morphometry analysis investigating the relationship between total brain size and regional gray matter volume, we found widespread cortical expansion and contraction of the amygdala with increasing brain size, closely mirroring modern/premodern differences (Extended Data Fig. 3). Therefore, while relative brain size differences between premodern and modern dogs did not reach significance in this sample, variation in absolute brain size between these two groups accounts for nearly all of the variation in regional brain anatomy. This is consistent with the “late equals large” hypothesis by Finlay and colleagues, which proposes that as brains scale larger, late-developing cortex disproportionately expands, while earlier developing regions, including the amygdala, disproportionately shrink (Finlay et al., 2001).
Interestingly, though, even after controlling for total brain size, some regions of cortex are still significantly larger in modern breed dogs. This is suggestive of focal, mosaic adaptation which occurs on top of the backdrop of global, concerted, allometric change (Striedter and Northcutt, 2019). These regions include portions of the postcruciate gyrus (motor cortex), anterior suprasylvian gyrus (temporal cortex), and prorean gyrus (prefrontal cortex). Notably, the prorean gyrus is implicated in the emergence of cooperative pack hunting in the carnivore lineage (Radinsky, 1969) and has been affected by selection on social behavior in the Russian farm-fox model (Hecht et al., 2021b). The anterior suprasylvian gyrus activates in response to spoken human language in awake dog fMRI studies (Andics et al., 2016; Cuaya et al., 2022; Bálint et al., 2023). Together, this suggests that cortical areas supporting social cognition and social communication have been under specific positive selection in modern breeds.
An additional possibility is that experiential differences between modern and premodern dogs in this dataset may account for a portion of the observed neuroanatomical variation, as some of the modern breed dogs were enrolled in working dog training programs. However, our results cannot be attributed entirely to plasticity effects for several reasons. First, we previously observed a link between cortical expansion and trainability scores in a study that assessed neural correlates of breed-average C-BARQ scores (Hecht et al., 2021a); this prior analysis was blind to individual variation in experience and driven entirely by heritable, breed-level traits. Second, the brain–behavior associations we report here occurred in both working and nonworking individuals. And third, the anatomical extent and magnitude of the current results far exceed any prior report of experience-dependent plasticity and are more in line with innate, heritable traits. The effect of specific training experiences on brain organization is an important question for future research.
There are several limitations of our study that should be noted. First, the ancestral population of early domestic dogs is extinct and their brains and behavior are not available for study. Today's premodern dogs are not “living fossils,” although they presumably possess more ancestral traits than highly derived, artificially selected modern breeds. Second, there are additional lineages of premodern and modern dogs which are not represented in this sample and should be examined in future research, along with variation among breeds/types within each group. Additional important topics for future research include examining other aspects of brain organization, such as connectivity patterns and cell type distributions, as well as underlying genetic architecture.
Together, our findings suggest that the brain–behavior phenotype of recent dog domestication involves reduction in flight-or-fight processing, as mediated by the amygdala, as well as increases in learning and behavioral flexibility, as mediated by distributed cortical regions. Interestingly, these neural changes are paralleled in rabbit domestication (Brusini et al., 2018). Our results are consistent with a dual-phase model of dog domestication which posits early attenuation to the human socio-ecological niche followed by additional adaptative changes supporting behavioral specialization for particular contexts within that niche (Tancredi and Cardinali, 2023). Our findings challenge the long-standing idea that domestication consistently involves reductions in brain size (Kruska, 1988). Furthermore, our results are consistent with the emerging pattern that brain evolution in domestic dogs has been shaped by both concerted developmental patterning in brain scaling (Finlay et al., 2001; Hecht et al., 2021a) and by focal, mosaic change (Hecht et al., 2019). While dogs were the first species domesticated by humans, they are far from the only ones. Among mammals, the biomass of domesticates now far outweighs that of wild species by 2 orders of magnitude (Bar-On et al., 2018), and increasing anthropogenic influence is arguably now a significant selective force for all animals on the planet. Thus, our results raise the possibility that comparable neuro-evolutionary changes may also be occurring in other animals and further research is warranted in additional species.
Footnotes
This work would not have been possible without the contributions of the participating dogs and their humans. We thank the contributions of the staff and leadership of the participating working dog organizations, including the Penn Vet Working Dog Center: Cindy Otto and Clara Wilson; America's VetDogs: Grete Eide, Valerie Cramer, and Paula Giardinella; Fidelco Guide Dogs: Tommy Mourad and Deb Leamy; the support staff of the Canine Brains Project: Minerva Abdulla and Katie Dabney; Canine Brains Project student research assistants: Sonya Ganeshram and Katie Sierra; the veterinary team of the Canine Brains Project: Lauren Duffee, Erika Militana, Suzanne Sutton, Sara Nath, Lauren Baker Lasker, Emily Finn, Jana Mazor-Thomas, Christina Campbell, Julia Campellone, Miriam Applin, Sara Gomez, Eryn Moitoza, Madysen Delosh, Danielle Husley, and Laurie Rassbach; and the support staff of the Center for Brain Science Neuroimaging Core at Harvard University: Larry White, Ross Mair, Jenn Segawa, Tim O’Keefe, and Caroline West. This work was supported by the National Science Foundation (2238071) and Alfred P. Sloan Foundation.
The authors declare no competing financial interests.
This paper contains supplemental material available at: https://doi.org/10.1523/JNEUROSCI.2032-24.2025
- Correspondence should be addressed to Sophie A. Barton at sbarton{at}g.harvard.edu.
