Diverse yet limited resources are a defining force in the evolution of biodiversity (Ford et al., 1973; Janz et al., 2001; Gavrilets et al., 2007; Lamichhaney et al., 2015; Anderson and Weir, 2022), but long time scales and shifting conditions can obscure potential links between ecology and speciation (Hendry, 2009; Sadedin et al., 2009). Adaptive divergence through resource specialization is well-studied in island systems (Gavrilets and Vose, 2007; Sadedin et al., 2009; Lamichhaney et al., 2015), and may also be important in the complex, interconnected landscapes that cover most of the globe (Hendry, 2009; Foote, 2012). Research focused on several wide-ranging vertebrates has revealed the potential for socially learned behaviors to mediate adaptive radiation, leading to rapid evolution of specialized forms, or ecotypes (Benkman, 1993; Sewall et al., 2004; Pilot et al., 2010; Foote et al., 2016). Red Crossbills (Frigillidae: Loxia curvirostra), finches specialized to forage on conifer seeds by prying open cones to extract seeds, form cultural groups with discrete flight call characteristics and show varying ecological and morphological distinctness (Benkman, 1993; Groth, 1993a; Groth, 1993b; Benkman et al., 2009). Flight call characteristics are thought to be entirely learned during the first autumn after hatching, and are not known to be dictated by morphology (Groth, 1993b). These ecotypes (hereafter “call types”) are broadly sympatric and engage in dynamic movements across overlapping ranges as they seek seed-producing conifers. Call types are thought to represent cohesive evolutionary lineages, as their flight calls are distinct and consistent, intermediate flight calls are rare, and types appear to largely mate assortatively (Groth, 1993a; Benkman et al., 2009). Call Types also differ in average bill size, which may facilitate their specialization on different conifer species (Benkman, 1993; Groth, 1993b). Despite the description of 12 North American call types (Young and Spahr, 2017; Young et al.), their distribution and ecology remains incompletely known, and multiple hypotheses for the causes of their adaptive radiation have been proposed (Benkman, 1993; Kelsey and Hahn, 2008; Martin et al., 2020). The existence of call types potentially specialized to forage on particular coniferous species provides interesting potential to illuminate the role of ecology in crossbill speciation.

The irruptive or nomadic behavior of Red Crossbills has evolved in response to spatially and temporally fluctuating availability of food resources as the mast cycles of conifers produce a shifting mosaic of available cone crops (Groth, 1993b; Benkman, 1999; Newton, 2006; Kelsey et al., 2008; Young, 2011; Cornelius et al., 2013; Teitelbaum and Mueller, 2019). Locating and exploiting cone crops requires movements across large geographic areas, especially if call types are specialized to particular conifer species or co-occurring conifer species with similar cone structures are geographically asynchronous in their mast cycles (Benkman, 1999; Kelsey et al., 2008; Cornelius et al., 2013). Crossbills breed in the late summer and facultatively in late winter, taking advantage of seasonal variation in food availability (Cornelius et al., 2013). Crossbill movements are known to vary in scale from local to continental (Groth, 1993b; Newton, 2006; Young, 2011). Each call type’s unique vocalizations have allowed researchers to track type-specific movements across vast areas of North America and Eurasia (Groth, 1993a; Groth, 1993b; Benkman, 2003; Martin et al., 2019), despite extreme similarity in morphological appearance. By recording flight calls, our understanding of spatial and temporal patterns of movement along with information on choice of coniferous tree species for foraging is quickly improving.

How and why has such morphological and vocal divergence occurred within Red Crossbills? Treatments of ecology and evolution in the Red Crossbill complex suggest two primary mechanisms that differ in the relative importance they ascribe to geographic isolation and resource specialization (Benkman, 1993; Kelsey and Hahn, 2008; Kelsey et al., 2008; Martin et al., 2020). In western North America, for example, studies of foraging performance and preference have suggested that each call type is adaptively specialized on a single conifer species, a postulate known as the “key conifer hypothesis” (Benkman, 1993; Benkman, 2003; Irwin, 2010). This hypothesis predicts that each call type forages most efficiently on a single tree species, causing close specialization on this species particularly during periods of resource scarcity (Benkman, 2003). Conifer seeds ripen in late summer and their availability diminishes through the following spring, so call types would be expected to forage mostly on their key conifers from late winter to early summer when seeds are scarce and the call type best adapted to each key conifer can efficiently extract its limited resources (Benkman, 1993; Cornelius and Hahn, 2012). The foraging efficiency data which underscore the key conifer hypothesis are compelling, but the geographic scope is limited and they omit some ecologically important conifer species (Benkman, 2003; Kelsey et al., 2008). Moreover, recent studies in western North America and Eurasia have proposed that call types are closely connected to geographic regions, or “core zones of occurrence”, to which they cyclically return, potentially allowing slow differentiation in parapatry (Kelsey and Hahn, 2008; Young et al., 2011; Martin et al., 2020). Call types are undoubtedly distinct in ecological preferences, even in sympatry (Benkman, 2003; Benkman, 2007; Young, 2011) but the small body of field data on habitat preferences has shown mixed results for conifer specialization (Kelsey and Hahn, 2008; Kelsey et al., 2008).

We evaluated associations of Red Crossbills with their hypothesized key conifer species in the Pacific Northwest of North America (Oregon, USA). No focused publications have treated call type distribution and ecology in the Pacific Northwest (but see (Groth, 1993b; Kelsey et al., 2008)), yet the region contains five proposed key conifers (henceforth PKCs), at least two other conifer species which crossbills are known to forage on, and five regularly occurring flight call types (Benkman, 2003; Meyers et al., 2015). By recording crossbills and quantifying the availability of cones at hundreds of sites across the Pacific Northwest over two years, we evaluated the prediction of the key conifer hypothesis that each call type should associate with its PKC, especially during the seasons of relative resource scarcity.

We conducted 923 surveys between June 2020 and October 2022, 633 of which encountered Red Crossbills ( Figure 1 ). Three call types (2, 3, and 4) occurred in all five survey regions, while two types (5 and 10) were restricted to one or two adjacent ecoregions. We encountered Type 2s on 118 surveys ( Figure 3A ), with the most observations in the Leeward Cascades (72) and Coast Range (26). We found Type 3s on 365 surveys spread widely throughout all 5 regions ( Figure 3B ). Type 4s were found on 58 surveys, with most encounters in the Coast Range (31) and Leeward Cascades (21) ( Figure 3C ). In contrast, type 10s were only found on the Outer Coast (78 surveys) and Type 5s were sparsely distributed only on the Cascade Crest (5 surveys) ( Figure 3D ).

The four frequently detected call types (2, 3, 4, 10) varied considerably in their movements across ecoregions dominated by hard-coned pines (Leeward Cascades), regions dominated by soft-coned conifers (Coast Range and Windward Cascades) and areas where these conifer types significantly overlap (Outer Coast, Cascade Crest). Type 2s were found consistently in the range of their PKC, ponderosa pine, but also irrupted into the Coast Range in the autumn of 2020 and were detected twice near the Outer Coast in the summer of 2022 ( Figure 3 ). Type 3s and Type 4s were found throughout the range of their PKCs (western hemlock and Douglas-fir, respectively), but their presence was not consistent in any habitat or region. Both types were present in the Coast Range in the summer, fall, and winter of 2020-2021. In the summer of 2021, Type 3s were encountered in smaller numbers throughout the range of their PKC, but were found more abundantly breeding elsewhere on the Cascade Crest. In the winter and spring of 2022, Type 3s and Type 4s were located on surveys in the low-elevation ponderosa pine forests of the Leeward Cascades. Type 3s were found in small numbers on the Outer Coast throughout the survey period. Type 4s were generally scarce, but were more concentrated during their 2020 irruption into the Coast Range. Type 10s were much less irruptive and were found consistently and exclusively in the range of Sitka spruce (their PKC) ( Figure 3D ).

In our study area, each of four Red Crossbill call types that we frequently encountered showed unique distributional dynamics and conifer associations. Our findings reveal a variety of levels of correspondence between geographic regions, primary key conifer species and crossbill call types. Some types, e.g., types 5 and 10, appeared to be strongly affiliated with a particular ecoregion and conifer species, while the other three types (2, 3 and 5) wandered more widely and occurred frequently in areas lacking their presumed PKC. Our observations that call types 2, 3 and 4 are widespread and associated with both their PKCs and additional tree species suggests potential multi-conifer associations, not reliance on single PKCs. In contrast, type 10 was apparently resident in a narrow band in the Outer Coast ecoregion and strongly associated with Sitka spruce. The patterns we observed suggest that the hypothesis of primary key conifer associations leading to evolutionary specialization of crossbill morphology may be reasonable for some call types but not all of them. Instead, three of the call types we studied appear to associate instead with several ecologically associated conifer species, which suggests regional specialization on sites with multiple food sources is more important than dependence on single conifer species.

Crossbill types with smaller bill morphologies, such as type 3 (the smallest) and type 4 have been thought to specialize on soft-coned conifers due to low foraging efficiency on hard-coned pines (Benkman, 1993; Benkman, 2003). However, our observations indicate that these call types may have broader foraging preferences than we expected. During the two years of our study, all PKCs and two other potentially important conifers (Engelmann spruce and mountain hemlock) in the region showed marked periods of cone abundance and scarcity, which appeared to drive some of the movements of call types across our study area. However, hard-coned lodgepole and ponderosa pines and coastal Sitka spruces showed less pronounced variation in cone abundance. The consistency of cones on these tree species may help explain the apparent residency of type 10s in association with Sitka spruce and the consistent occurrence of type 2s in regions dominated by hard-cones pines. In contrast, Types 3 and 4 were associated with a variety of soft- and hard-coned conifers and were not consistently present in any ecoregion, apparently following abundant cone crops on a variety of conifers. Type 4s were generally found either in areas with large cone crops on their PKC, Douglas-fir, or in association with lodgepole and ponderosa pines, while type 3s showed the most diverse preferences, associating with significant cone crops on western hemlock (PKC), mountain hemlock, Douglas-fir, Sitka spruce, and Engelmann spruce. In early 2022, types 3 and 4 were notably observed in high numbers in forests of lodgepole and ponderosa pines. During this period, we found Type 3s foraging on ponderosa pine seeds, an observation which was independently confirmed in the same season (Cornelius, 2023). These observations contrast with the expectation that types 3 and 4 are specialists which would struggle to forage on hard-coned pines. However, dramatic shifts in conifer species selection are consistent with the concept of Red Crossbills as rich patch exploiters which rely on biogeographic networks of resources to maximize the reliability of foraging opportunities (Cornelius et al., 2013).

Past treatments of call type ecology have often focused on the importance of resource scarcity in the evolution of specialization (Benkman, 1993; Benkman, 2003; Kelsey et al., 2008), but we did not find greater specialization in expected seasons of low seed abundance. Assuming that late winter and early spring represents a period of resource scarcity due to dwindling conifer seeds and low temperatures, two irruptions we observed showed especially clear departures from this prediction. In the winter and spring of 2021, Type 2s irrupted into the Coast Range, where PKC ponderosa pine is rare on the landscape and seldom produces significant cones. Later, in the winter and spring of 2022, Type 3s and 4s irrupted extensively into the low-elevation East Cascades, where their PKCs and all soft-coned conifers are rare. Moreover, types 2, 3, and 4 are known to irrupt into eastern North America in fall, winter, and spring, where their PKCs are absent (Young, 2011), suggesting that greater specialization is unlikely during this period. These patterns raise questions about whether the winter and early spring, when breeding often occurs, is a true season of scarcity. The hypothesis that crossbills are rich patch opportunists suggests that they generally have access to abundant resources throughout the annual cycle and can limit breeding behavior under resource scarcity and high physiological demands, exemplified by an autumn pause in breeding even when resources are abundant (Cornelius et al., 2013). Red Crossbill fat deposits peak in May and June preceding regular migratory movements (Cornelius and Hahn, 2012), highlighting late spring and early summer as a potentially highly demanding period, but this may coincide with higher insect consumption (Kelsey et al., 2008). Generally, erratic and opportunistic movements of Type 3s and Type 4s throughout the annual cycle in our study did not show more consistent associations with key conifers in any of the periods between late winter and early summer.

Our data suggest that foraging preferences affect call type distributions, with some types more strongly affected by single versus multi-species conifer associations. The tight affinity between Type 10s and Sitka spruce despite the presence of other abundant conifers on the Oregon Coast shows strong evidence that the key conifer hypothesis can explain the distribution of resident or near-resident Crossbill populations. Similar patterns may be present in Cassia Crossbills (Loxia sinescurius) in Idaho and Type 8 Red Crossbills in Newfoundland, both associated with semi-serotinous tree species that bear reliable seed crops (Benkman, 1989; Benkman et al., 2009). Similarly, Type 4 seems to show a geographic affinity for the east slope of the Oregon Cascades, even though it had no overriding association with a single conifer. Type 3s continually shifted both their geographic and conifer associations, with high concentrations found from the Sitka spruce forests of the Outer Coast to the ponderosa pine forests of the East Cascades. This apparent generalist strategy highlights the potential importance of diverse networks of conifer ecosystems in the adaptive radiation of call types.

Ultimately, a multitude of selective forces act on Red Crossbills, and it is possible (or perhaps probable) that geography and resource specialization have worked in tandem to shape their evolution (Benkman, 2003; Kelsey et al., 2008; Cornelius and Hahn, 2012). Our results add to a growing body of evidence that Red Crossbill call types are each associated with a unique network of ecological resources which provides food security over broad geographic scales (Benkman, 1999; Cornelius et al., 2013; Martin et al., 2019). Call types in our data were distinguishable based on the set of conifers with which they were associated but varied in the number of tree species they favored. These discrete and consistent foraging preferences suggest ecological drivers for call type divergence and are consistent with previous findings that crossbills can specialize both on single, reliable conifer species or on regions of conifer diversity Benkman (1987). While our study is limited by its scope in space and time, the conifer associations observed may have geographic implications in a broader context. Type 3 was the only type to associate with western hemlock, a small-coned conifer restricted to moist, coastally influenced forests in the Pacific Northwest (Meyers et al., 2015). In contrast, Types 2 and 4 were only associated with conifers found throughout the interior west and into the Rocky Mountains. These patterns hint at potential core geographic areas proposed by Kelsey (Kelsey et al., 2008). Further study across broader geographic regions are needed to determine if our conclusions apply more widely across the ranges of the key conifer species. Crossbill call types rely on diverse resource networks and have the potential to evolve rapidly. Thus, they may be interesting indicator species that reveal biotic responses to structural and functional changes in the environment. Rangewide, multi-year studies of foraging, assortative mating, and genetics are needed to further elucidate the evolutionary origins of Red Crossbill call types.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethical approval was not required for the study involving animals in accordance with the local legislation and institutional requirements because This was a purely observational study.

CC: Conceptualization, Formal Analysis, Investigation, Methodology, Visualization, Writing – original draft, Writing – review & editing. WR: Conceptualization, Formal Analysis, Methodology, Visualization, Writing – review & editing. MY: Conceptualization, Methodology, Writing – review & editing.