Getting Back to Nature: Feralization in Animals and Plants.

Formerly domesticated organisms and artificially selected genes often escape controlled cultivation, but their subsequent evolution is not well studied. In this review, we examine plant and animal feralization through an evolutionary lens, including how natural selection, artificial selection, and gene flow shape feral genomes, traits, and fitness. Available evidence shows that feralization is not a mere reversal of domestication. Instead, it is shaped by the varied and complex histories of feral populations, and by novel selection pressures. To stimulate further insight we outline several future directions. These include testing how 'domestication genes' act in wild settings, studying the brains and behaviors of feral animals, and comparative analyses of feral populations and taxa. This work offers feasible and exciting research opportunities with both theoretical and practical applications.

Domestication Is Not a Dead End

Domesticated animals and plants comprise a rapidly growing proportion of life on our planet [1]. The vast ranges and abundance of these organisms show that domestication (see Glossary) can have remarkable evolutionary payoffs. At the same time, it can induce both plastic and genetic modifications that limit the capacity of an organism to thrive in nature (e.g., 2, 3, 4). Despite this maladaptation, feralization of animals and plants has proven, sometimes to humans’ great frustration, that domestication is not always a one-way process. The flow of domesticated organisms and their genes into noncaptive settings has important conservation implications; it also presents unique opportunities to characterize general and novel evolutionary processes of Anthropocene environments [5]. With these applications in mind, our review summarizes current knowledge regarding the process of feralization and provides a roadmap for further investigation into this tractable, exciting, and understudied research area.

Feralization merits special consideration because its subjects are uniquely distinguished from other animals and plants. Biologists have long appreciated how domestication shapes wild organisms via both deliberate artificial selection by humans and unintended effects of anthropogenic propagation [6]. In recent decades, these effects have been elucidated by intensive studies bridging disparate fields (e.g., anthropology, plant and animal science, and organismal, behavioral, and developmental biology) 7, 8, 9. By contrast, there has been relatively little research into the process of feralization. Here, progress is also hindered by long-held speculations and misconceptions. These include: (i) the idea that formerly domesticated populations are incapable of rapid adaptation, due to their genetic homogeneity or recent establishment [10]; (ii) the idea that captive propagation invariably reduces fitness outside of domesticated settings due to evolutionary tradeoffs and relaxed natural selection (e.g., 2, 11); and (iii) a belief that feralization predictably results in atavism (e.g., [12]). These ideas have received only mixed support from a small but growing body of relevant research. Here, we draw on case studies to: (i) show that routes to feralization are diverse and can facilitate rapid evolution; (ii) synthesize current knowledge concerning feral genotypes and phenotypes; and (iii) outline avenues for future studies.

Pathways to Feralization

Defining Domestication and Feralization

There are many extended discussions of problems surrounding the definition of domestication (e.g., 13, 14, 15). The broadest definitions encompass nonhuman species, such as leaf-cutter ants, that also cultivate mutualists (e.g., [16]). Yet, while these cultivars can feralize [17], such non-anthropogenic processes lie beyond the scope of this review. Others 13, 18 describe domestication as movement along continua of human–animal interactions or, alternatively, as solely the onset of human-facilitated propagation (e.g., [11]). In this review, we expand an operational definition developed for animals [19] to include agricultural and ornamental plants. Except where noted otherwise, we also adopt the inclusion by this definition of both the establishment and subsequent improvement stages of anthropogenic propagation.

Our review also examines how the allele frequencies, traits, and fitness of wild populations can be altered by the introgression of feral alleles from artificially selected sources; thus, it encompasses many wild gene pools that are chiefly derived from undomesticated ancestors 20, 21. Here, we show that even limited introgression from artificially selected sources can have important evolutionary consequences. For clarity, however (except where noted), we use ‘feral’ to describe free-living organisms or populations that are primarily descended from domesticated ancestors.

Our discussion of feralization requires a few caveats. First, some feral populations still receive limited, intentional support from humans. For example, feral cats and horses are sometimes provisioned with food, yet remain highly self-reliant compared with their domestic counterparts and do not fulfill an artificially selected utility. Additionally, some taxa have oscillated between feral and domestic states, blurring lines between the two processes (e.g., longhorn cattle that were redomesticated from feral ancestors) [22]. Finally, we acknowledge that feralization need not involve a return to truly ‘wild’ habitats. Instead, it often unfolds within cultivated or disturbed settings (e.g., agricultural fields and cities). Still, its subjects are distinguished from domesticated ancestors by the withdrawal of intentional efforts to support their reproduction. This alters selection regimes in ways that can, both in principle and practice, produce rapid evolutionary changes (Figure 1).

Figure 1

Figure360: An Author Presentation of Figure 1

Evolutionary Forces That Shape Feral Gene Pools and Traits.

The core process of feralization (depicted here with solid black arrows and boxes) is often modified by various forms of gene flow and/or anthropogenic selection (depicted here as dashed gray lines and boxes). For a Figure360 author presentation of Figure 1, see the figure legend at https://doi.org/10.1016/j.tree.2019.07.018

Sources of Feral Populations

To understand how populations evolve, it is usually helpful to examine their sources and genetic structures. Given that feral populations compound demographic and selective effects of domestication with a subsequent ‘re-invasion’, they present unique challenges for DNA-based ancestry reconstructions, as well as for sequence-based tests of adaptation 4, 23. Despite these obstacles, many investigators have succeeded in elucidating pathways to ferality. Gressel [24] delineated two alternative categories, which we illustrate with diverse examples in Table 1. ‘Endoferal’ populations stem from a single domesticated lineage (e.g., a breed or crop), whereas ‘exoferal’ populations are derived via admixture, either among domesticated lineages (e.g., crop varieties) or between domestic taxa and their wild relatives. Current data suggest that both endo- and exoferality are common. Among 23 plants that have feralized into weedy or invasive forms, approximately equal numbers were found to involve endo- versus exoferal origins [25]. Both mechanisms have also produced feral animal populations (Table 2), although their relative roles have not been systematically reviewed.

Table 1

Animal and Plant Domestications That Have Resulted in Feralization, and Their Primary (Artificially Selected) Utilities

Order	Domesticated taxon		Antiquity (years before present)	Food	Companionship	Aid	Security	Ornament	Sport-racing	Warfare	Sport-fighting	Transport or draft	Textiles	Pest control	Pollination
Mammals
Image 1Carnivora	Dog, dingo	Canis lupus	15 000a	Image 2	Image 3	Image 4	Image 5	Image 6	Image 7	Image 8	Image 9	Image 10	Image 11
	House cat	Felis catus	9500a		Image 12			Image 13						Image 14
	American mink	Neovison vison	80b										Image 15
Image 16Lagomorpha	Rabbit	Oryctolagus cuniculus	1300–17 000c	Image 17	Image 18								Image 19
Image 20Perissodactyla	Pig	Sus scrofa	10 300a	Image 21
	Horse	Equus ferus	5500a	Image 22				Image 23	Image 24	Image 25		Image 26
	Ass	Equus africanus	5500a									Image 27
Image 28Artiodactyla	Goat	Capra aegagrus hircus	10 000a	Image 29									Image 30
	Sheep	Ovis aries	10 000a	Image 31									Image 32
	Cow	Bos taurus	10 300a	Image 33								Image 34	Image 35
	Dromedary camel	Camelus dromedarius	3000a	Image 36								Image 37	Image 38
Birds
Image 39Galliformes	Chicken	Gallus gallus	4000a	Image 40				Image 41			Image 42
	Turkey	Meleagris gallopavo	2000d	Image 43
Columbiformes	Street pigeon	Columbus livia	>5000b	Image 44				Image 45				Image 46
Anseriformes	Mallard	Anas platyrhynchos	1000a	Image 47				Image 48
	Muscovy duck	Cairina moschata	Pre-Columbian	Image 49									Image 50
Insects
Hymenoptera	Honeybee	Apis mellifera	9000e	Image 51											Image 52
Lepidoptera	Silkworm	Bombyx mori	7500e										Image 53
Fish
Image 54Salmoniformes, Cyprinodontiformes, Cypriformes, Cichliformes, Anabantiformes	Aquacultural and pet species	e.g., salmon, cichlids, guppies, betas	Variable	Image 55				Image 56			Image 57
Plants
Asterales	Jerusalem artichoke	Helianthus tuberosus		Image 58
Poales	Bread wheat	Triticum aestivum	10 000f	Image 59
	Finger millet	Eleusine coracana	5000f	Image 60
	Grain sorghum	Sorghum bicolor	5000f	Image 61
	Rice	Oryza sativa	7000f	Image 62
	Rye	Secale cereale	5000f	Image 63
Brassicales	Radish	Raphanus raphanistrum	8000g	Image 64
Caryohyllales	Sugarbeet	Beta vulgaris	300f	Image 65

From [100].

From [101].

From [102].

From [103].

From [104].

From [105].

From [106].

Table 2

Sources of Feral Animals and Plants

Domestic population crossed with	Definitiona	Examples
Self	Endoferal	Crop rice (Oryza sativa) appears to be particularly prone to feralization, because there is evidence for multiple de-domestication events with varying origins in Asia and North America. Weed rice populations of endoferal origin are present on both continents [98]. Endoferality is common in animals, including serial introductions of rabbits to Australia that have generated genetically distinct endoferal subpopulations [47]
Divergent population (e.g., breed or crop)	Exo–endoferal (intercrop)	In Bhutan, weedy rice is a hybrid of two crop varieties (O.s. japonica × O.s. Indica) [98]. Feral cattle in the New World that were subsequently re-domesticated stemmed from admixture between independently domesticated taurine and indicine aurochs (Bos primigenius), and this admixture may have facilitated adaptation to novel environments outside the native range [22]
Wild conspecific	Exoferal (crop–wild)	SNP diversity of weedy rice is higher in southwest Asia than in the range of wild rice, due to introgression from wild rice and also perhaps from local crop rice landraces [98]. Exoferal (domestic–wild) animals include chickens that hybridize with red junglefowl (Gallus gallus) within the native and introduced ranges of the species 30, 107
Other domesticated species	Exoferal (domestic hybrid)	Feral Jerusalem artichoke (Helianthus tuberosus) and domesticated sunflower (Helianthus annuus) may hybridize in Europe [25]
Other wild species	Exoferal (crop–wild hybrid)	California wild radish is an interspecific hybrid between the crop radish (Raphanus sativus) and the agricultural weed ecotype of native wild radish (Raphanus raphanistrum; [64]). Available evidence suggests that the agricultural weed radish is derived from the native wild radish [106]. Animal examples are rare, but include coyote–dog (Canis latrans × C. lupus) hybrids [58]
Genetically modified organism	Exoferal (transgene hybrid)	Transgenes have been found in several wild plant populations 37, 38, 39, 40. Animal cases are not yet known, partly due to legal, logistical, and technological barriers to the cultivation of transgenic animals

After 24, 25.

Mechanisms of Feralization

Endoferality can occur when individuals from a domestic population escape into local environments in which they can survive and reproduce. This is what most people envision when contemplating feralization. Endoferality can also result from intentional releases of organisms to establish feral descendants. We call this process ‘de-domestication’ (sensu [15]), although the term is used in the plant literature synonymously with atavism (e.g., [26]). Motives for releases of domestic taxa range from ecosystem engineering [27] to providing recreational, nutritional, and/or economic benefits (e.g., hunting and fishing) [28].

Exoferality, by definition, involves admixture. Sometimes, this gene flow precedes translocation into new environments, as shown by a subset of North American weedy rice that originated from admixture outside of their introduced range [29]. Admixture can also occur at multiple timepoints during and after establishment. For example, archeological, morphological, and genetic evidence suggest that, centuries after Polynesians dispersed red junglefowl (Gallus gallus) into Pacific Oceania, the descendants of these birds hybridized with chickens introduced by Europeans (e.g., 30, 31, 32, 33). These and other exoferal populations (e.g., Table 2) provide tractable systems for studying how gene flow impacts the establishment, fitness, and local adaptation of non-native organisms, a central goal of invasion biology (e.g., 4, 23, 34, 35, 36). In addition, a subset of exoferal gene pools harbor feralized transgenes, an increasingly common phenomenon that raises unique ethical issues and research questions [37]. Transgenes have introgressed into nonagronomic plant populations (e.g., wild cotton and bentgrass 38, 39), into cultivated crops (e.g., canola, soybean, and maize [40]), and into feral plants (e.g., weedy rice and beets 41, 42). Thus, gene flow among domestic, feral, and wild plants comprises an important potential mechanism for transgene establishment and spread.

In the near future, broadening of sampling and analytical tools will likely increase the number of feral populations with known exoferal origins [23]. Ancient DNA can also be used to clarify population ancestries (e.g., 43, 44). Recently, for instance, this approach revealed that modern Przewalski’s horses are in fact feral descendants of horses domesticated by the Botai culture, rather than truly wild [45]. Furthermore, recent introgression from domestic horses has introduced deleterious gene variants to this exoferal gene pool.

The diversity of pathways to feralization (Table 2) raises an interesting issue regarding the modeling of the process. Although endoferal populations provide the clearest insights into how feral selection regimes affect formerly domestic gene pools and traits (i.e., evolution in absentia of admixture), they may also represent a minority of feralization episodes in nature. A parallel conundrum has catalyzed recent revisions of domestication models, since the process involves admixture more often than previously thought, and it can also be difficult to detect [8].

Viewing feralization ‘in light of admixture’ helps to clarify how future gene flow can impact outcomes and consequences of the process. For example, many feral taxa (e.g., weedy rice, dogs, and chickens) appear to exhibit both exo- and endoferal origins across their current ranges. These interpopulation differences result in both genetic and phenotypic variation (e.g., 25, 30, 46, 47), which would likely be affected by further introgression (e.g., admixture between genetically divergent feral populations; e.g., [29]). Admixture from domestic sources can also convert wild populations into exoferal ones [20] and accelerate their responses to new selection pressures [48]. Remarkably, genes from 23 of humanity’s 25 most important domesticated plants have been found in wild populations. The geographical distribution and phenotypic consequences of this crop–wild admixture vary widely by case [49]. The same phenomenon is seen in animals, with examples including wolf × dog, chicken × red junglefowl, and farmed × wild salmonid hybrids. We briefly explore the fitness effects of these exchanges in the following section.

Adaptation in Feral Organisms

Fitness Consequences of Admixture

Several methods are available for assessing how admixture affects fitness in feral populations, including: (i) direct measurements of growth, survival, reproduction, and health in hybrids; (ii) functional analyses of outlier loci detected in genome scans (e.g., 50, 51); and (iii) experimental tests of the effects of these loci in laboratory systems (e.g., [50]). In recipient wild populations of fish, these approaches often find outbreeding depression (e.g., 52, 53). Reductions in hybrid fitness are also seen in weedy plants (e.g., [54]). These patterns can arise through the disruption of coadapted genes, allelic incompatibilities between source populations, and/or when gene variants from one source (e.g., domestic settings) are locally maladaptive in ferals 4, 55, 56. Altogether, this may explain why recipient wild populations often contain a small fraction of genes from domestic sources. Animal examples in which domestic introgression is minor (∼5–10%) include wolves (e.g., 57, 58, 59), wild boar 60, 61, coyotes 58, 62, and partridges [63].

However, in some situations, exoferal hybrids can have higher fitness than source populations. In greenhouse common gardens, functional traits of California wild radish were either phenotypically intermediate between the source populations of these hybrids or ‘domestic-like’ [64]. In exception to this pattern, California wild radish fruits were heavier than either parental taxon [64], were better protected against house finch damage [65], and had higher fitness in three common gardens within the invasive range of the hybrid [66]. This apparent hybrid vigor may help explain the capacity of the exoferal hybrid to thrive in noncultivated habitats and displace both domestic and wild progenitors.

Alleles involved in domestication and improvement can also facilitate adaptation in animals. For example, admixture between independently domesticated cattle likely facilitated the adaptation of the longhorn to feral conditions within the New World [22]. In general, we suspect that alleles that were artificially selected to enhance production (e.g., accelerating growth or fecundity) may often prove beneficial in nature, particularly during the establishment and expansion of feral populations (e.g., 50, 67). Still, more work is needed that examines the genetic basis of fitness-related phenotypes in feral settings. These studies should also compare genotype–phenotype relationships across populations and/or conditions, because hybrid fitness can vary sharply between environments (e.g., in carrots, radish, and salmon 25, 68, 69), and because plasticity can be important in colonizing novel ones 34, 70. Thus, accounting for gene × environment interactions will be essential for forecasting future feralization trajectories in the variable and changing environments of the Anthropocene 5, 71.

Effects of Domestication and Improvement Alleles

Domestication has produced consistent, correlated changes in a variety of species, such that domestication syndromes are commonly observed in both animals and plants [72]. The genetic mechanisms that produce these shared phenotypes within evolutionarily distant taxa is an area of intensive current research. In animals, one hypothesis proposes that syndromes arise through correlated effects of tameness selection on the development of neural crest-derived cells 73, 74. This idea is supported by emerging findings of parallel evolution in pathways that control neural crest cell fates in distantly related taxa (e.g., 59, 75). Plant domestication syndromes involve an array of traits, including attenuated seed dormancy and dispersal, vertical growth forms, increased seed size, accelerated growth, and palatability [11]. As in mammals, many of these traits involve complex gene networks and biochemical pathways that are evolutionarily conserved in distantly related taxa. At present, the extent to which domestication modifies homologous genomic loci to produce animal and plant domestication syndromes is not clear (e.g., [76]). Fortunately, emerging discoveries within this area (e.g., 73, 77) will soon enable us to determine whether (and how) domestication syndromes evolve under feralization.

In addition to exhibiting similarities in the form of syndromes, domesticated taxa are also differentiated from one another by their unique ancestries, cultivation or husbandry methods, and artificially selected utilities. Central goals in domestication research are to determine: (i) which genetic changes were directly selected by humans; and (ii) which variants and traits were crucial for the onset of domestication [11]. However, to understand feralization, it is important to examine the frequencies and functions of both domestication and improvement alleles. Together, these features distinguish contemporary domestics from their wild relatives, and we suspect that they can both contribute to the local adaptation or maladaptation of feral populations.

Table 3 provides diverse examples of loci with major effects on domestic phenotypes. In domesticated settings, functional impacts of these genes are sometimes known. By contrast, their allele frequencies and phenotypic effects are largely unstudied in feral populations. This offers compelling directions for future research, including determining the significance of: (i) mutations and structural variants arising de novo within domestic populations (versus ancestral variants recruited by soft sweeps or drift); (ii) gene variants affecting protein structures and gene expression; and (iii) fitness consequences of domestication versus improvement alleles. Expanding this work to include polygenic traits will be similarly important for understanding feralization, because many domestication-related phenotypes are only partly attributable to loci of major effect 78, 79. However, these are more technically challenging to characterize, and further work is first needed to elucidate their modification by domestication.

Table 3

Examples of Loci Involved in the Domestication or Subsequent Improvement of Plant and Animal Morphology and Physiology, and Their Significance to Feralization

Trait	Gene(s)	Domestic phenotype	Domesticated variant present in ferals	Fitness effects in the wild	Refs
Animals
Morphology	TYRP1	Melanic coat color in sheep	+	Artificially selected ‘light color’ phenotype was positively selected in feral Soay sheep	[108]
	CBD103	Melanic coat color in wolves	+	A continent-wide selective sweep in wolf × dog hybrids may result from the domesticated variant enhancing survival	[109]
	MC1R	Coat color in pigs	+	Domestic phenotypes involving this locus are common in Pacific feral pigs, perhaps indicating relaxed or positive selection	[61]
	RXFP2	Horn type in sheep (normal or scurred)	+	In feral Soay sheep, male heterozygotes have high fitness due to a balance of sexual costs and longevity gains of an artificially selected allele producing smaller horns. RXFP2 genotypes were not found to affect female survival or fitness	[110]
Growth and physiology	?	Increased fecundity in pigs	+	Domesticated gene variants may increase fecundity in admixed wild populations near farms. This example highlights the many cases where causal genes are not yet known	[60]
	IGF1, GHR, IGFII, THR	Increased growth in Salmon	?	Effects of alleles from wild-type, domestic, and/or transgenic origin can vary across environments. Domesticated alleles are often deleterious	[111]
Plants
Growth and physiology	SH4, qSH1	Delayed seed shattering in rice	+	Domesticated phenotype is absent in weedy derivatives of domestic rice, although they do carry the domesticated allele at sh4. Compensatory mutations may have been positively selected to facilitate weediness	78, 112
	CBF	Stress tolerance in barley	?	Unknown, but may affect abiotic stress tolerance. HvCBF4 is important for salt tolerance in wild Tibetan barley, the source of domesticated barley	[113]
	FRI	Flowering time in rapeseed	?	Unknown, although multiple orthologs are important for flowering time in rapeseed (Brassica napus)	[114]
	?	Life history and morphology	+	A mixture of crop and wild traits were positively selected in outplanted hybrid sunflowers	[115]

Another novel and potentially transformative goal for future studies is to characterize structural and functional properties of feral microbiomes, which affect an array of fitness-related traits and can evolve rapidly during feralization [80]. For example, even after many generations outside of captivity, feral chickens retain legacies of captive husbandry within their digestive microbiota (e.g., a somewhat attenuated resistance to agroindustrial antibiotics). Nonetheless, these feral microbiomes are also both divergent from, and more variable than, those of farmed poultry reared on a variety of diets [80]. The causes and consequences of microbiome divergence have broad basic and applied significance, and merit further (e.g., comparative) analyses.

Direct Observations of Selection in Feral Populations

One of the most powerful tools for identifying adaptive changes during feralization is to analyze long-term pedigrees; an island population of Soay sheep studied since the 1960s offers one example [81]. In this case, pedigrees were used to infer the selection pressures on several phenotypes with domestic origins. Here, a genetic polymorphism affecting coat color is known, with the heritable black phenotype having a large body size and higher fitness [82]. However, due to the linkage between a major gene for black coloration and a quantitative trait locus (QTL) with antagonistic effects on size and fitness, black coloration is declining in this population.

Plasticity and Reversion of Feral Traits

Feral Brains and Behaviors

Phenotypic plasticity can be crucial in the colonization of novel environments [70]. Animal brains are of central importance for behavioral plasticity, and many domestic animals have diminished brain volumes [83]. This pattern is attributed to the relative simplicity of domestic environments [84], to artificial selection for docility and tameness, and to correlational selection on other traits [85]. Thus, feralization offers unique opportunities to study how brains and behavioral traits evolve when domestic animals transition into highly heterogenous and unpredictable environments.

Table 4 lists several known features of the brains and behaviors of domestic and feral animals. Somewhat surprisingly, many studies have found no effect of feralization on brain volumes 86, 87. Here, evolution may be hindered by a lack of essential genetic variation or insufficient time. The latter hypothesis is consistent with findings from dingoes, which are likely among the oldest feral populations (since ∼3000–8600 y before present). Dingo brains are larger and more encephalized than those of domestic dogs of similar body size, although variation among dog breeds complicates these comparisons [88]. Feralization may also drive subtler changes in brain structure and function. For instance, pigs were released on the Galapagos Islands ∼100 years ago to serve as meat reserves. Over the decades that followed, proportional sizes of differently-specialized brain regions diverged from those of domestic pigs [89]. Effects of domestication and feralization on brain function are also evident in molecular data, including: (i) comparative studies of domestic mammals revealing divergence in brain-specific miRNAs [90]; and (ii) evidence of selective sweeps at loci controlling neuronal development in feral chickens [50].

Table 4

Effects of Domestication and Feralization on Behavior-Related Phenotypesa

Behavioral trait	Δ Domestic (versus wild) phenotype		Δ Feral (versus domestic) phenotype
Brain volume	↓	Diverse mammals, birds, fish [86]	=	Diverse mammals [86], with exception of dingo [88]
Proportional size of brain regions	↕	Altered allometry of motor, limbic, and sensory regions in diverse taxa. Most pronounced regressions affect limbic regions [86]	↕	In exception to many examples of stasis [86], dingoes and pigs show partial ‘wild-type’ reversions 88, 89
Gene expression in brain	↕	Dogs [116], cows, horses, pigs, rabbits [90]	?
Aggression toward conspecifics	↕	Reduced agonism in many taxa, including fish and dogs. Increased agonism in some fighting breeds (e.g., bulls and cockerels [12])	↑	Roosters [92]
Predator avoidance	↓	Chickens, pheasants, rodents, fish 19, 86, 93, 117	↑	Chickens [92], guppies [93]
Habitat selectivity	↓	Deer mice [12]	?
Neophobia	↓	Mice, rats [19]	↑	Chickens [92]
Stress response	↓	Guinea pigs, foxes, mice [116]	?
Reproductive seasonality	↓	Foxes [116], chickens [118], dogs [19]	?
Diet selectivity	↓	Cats [86]	↓	Salmon parr [12]
Vocalization	↕	Higher rates in dogs, birds, guinea pigs [12], reduced diversity in birds [19]. Rates are also variable among breeds [117]	?

↑trait magnitude is higher; ↓trait magnitude is lower; ↕trait change varies by case (e.g. among previously-studied taxa, contexts, or populations).

At the level of behavior, domestication has often reduced fearfulness, agonism, and overall behavioral responsivity 19, 91; these effects can also be modified in ferals. For example, feral roosters, quails, and guppies were found to be more fearful, agonistic, and alert to potential predators compared with domestics 92, 93 (C.R. Nichols, PhD thesis, University of British Columbia, 1991). There are many other known differences between the social behavior and communication of feral animals and domestics (e.g., 30, 94) (Table 4). Both plasticity (e.g., learning) and genetic evolution can impact these traits [19] and their relative roles have not been systematically examined. Furthermore, fitness consequences of behavioral variation in feral populations remains poorly studied.

Other Feral Traits

While we have emphasized behavioral traits in the preceding section, animal and plant morphology and physiology have, likewise, been profoundly altered by domestication. By way of example, domestication has altered plant chemical defenses mediating herbivory in cultivated and wild settings [3]. These changes, and possibly subsequent ones, likely impact fitness in feral plants, although this has not yet been studied. Alongside many other examples of morphological and/or physiological trait change (e.g., Table 3), this shows how feralization research could both deepen, and expand upon, ecologically enlightened views of the fitness consequences of domestication [3].

Reconsidering Reversions

Many early naturalists reported that feral organisms invariably revert to the ‘wild-type’ traits of their ancestors. While Darwin took interest in the atavism of feral domestics, he also questioned its ubiquity [6]. Today, genomic studies are proving, intriguingly, that even when feralization restores ancestral phenotypes, this reversion can involve novel genetic mechanisms. For example, grain crops have been selected by humans to retain seeds until their harvesting. Given that seed dispersal is a crucial adaptation for most wild plants, reversion to dispersive phenotypes should be common in feralized grain crops. Seed dispersal in rice is called shattering, and this trait has been well studied in weedy rice. A key gene in the decreased shattering of domesticated rice is sh4 [95], but reversions to a shattering phenotype in US weedy rice are not caused by changes at this same locus [96]. Rather, they are controlled by different genomic regions in each of the two weedy rice groups, suggesting independent restorations of a ‘wild-type’ trait [97]. By contrast, in Southeast Asia, shattering in weedy rice is caused at least in part by adaptive introgression of wild alleles at sh4 [98]. Finally, in feral chickens and sheep 50, 51, genome scans found only limited overlap between outlier loci (i.e., candidate ‘feralization loci’), and genome regions that are known to have evolved under domestication. Altogether, these examples show that, at the genetic level, domestication-related changes are not predictably reversed by feralization. In systems where phenotypic reversion has occurred despite this (e.g., in weedy rice), we can now begin to disentangle how stochastic factors, the reversibility or irreversibility of evolution, and/or differences between ancestral and feral environments (e.g., emergent competition with domesticated counterparts [99]) steer the process of feralization.

Concluding Remarks and Future Directions

There is ample evidence that the evolution of feral populations is shaped by their unusual environments and histories. However, a robust understanding of feralization necessitates more studies that elucidate causal roles of selection pressures and genetic variation in the evolution of feral traits and fitness. A search for convergent ‘feralization syndromes’ could help illuminate proximate and/or ultimate mechanisms that drive feralization. At the same time, the process of feralization itself will continue to evolve. For example, genome editing is poised to alter domestication processes, and may generate novel feral populations as a byproduct [4].

In addition to providing Outstanding Questions, we close with some limitations of prior studies. First, many researchers have compared feral taxa to domestic relatives that are not their original source population(s). Therefore, differences in phenotypes and genotypes cannot be conclusively attributed to feralization. Furthermore, few studies have explicitly accounted for effects of differing methods and objectives of artificial selection (e.g., Table 1) on descendent feral populations. Lastly, the literature contains few comparative studies across feral populations or species. Nonetheless, the fact that feralization has often occurred to the same domesticated species in separate parts of the world offers opportunities to identify the constraints and pressures, be they environmental or genetic, that shape the course of feralization. After decades of intensive study, domestication research continues to provide stunning and practical evolutionary insights. Clearly, the open frontiers of feralization research hold equally exciting prospects for investigators bold enough to venture beyond the farm (see Outstanding Questions).

How predictable is feralization? We will need more comparisons of gene pools and traits (both among populations and between species) to answer this question, which has both applied and conceptual significance.

What genetic mechanisms drive feralization processes? Selection scans find evidence of rapid evolution in feral genomes, but have limited ability to detect many kinds of change that could impact feralization (e.g., structural rearrangement, epistasis, soft sweeps, balanced polymorphism, and heritable epigenetic change).

How do social and natural selection act in feral settings? Findings of ‘feralization genes’ suggest compelling hypotheses about fitness effects in nature, but these effects remain largely untested in natural settings (especially in nonplant and -fish models).

How do gene × environment relationships influence feralization? Both theory and empirical studies are needed to better understand how heterogenous and/or changing environments affect formerly domesticated populations. One important variable is the microbiome, which can evolve more quickly than (feral) host genomes and impacts a multitude of functional traits.

How does admixture affect feral population fitness and persistence? Recent work shows that admixture (e.g., between domesticated lineages and wild relatives) can facilitate rapid evolution. Over longer timeframes, we also see significant variability in the dynamics and persistence of feral populations. We do not yet know if and/or how admixture contributes to this variability in feralization outcomes.

Can feralization syndromes be identified? If feral animals exhibit parallel distinctions from domestic and/or nondomesticated counterparts, we can test whether these involve homologous mechanisms (e.g., genes or pathways). This is among the most feasible, and potentially transformative questions on the near horizon of feralization studies.

Alt-text: Outstanding Questions