Blue sharks (Prionace glauca) are large-bodied, highly migratory sharks with a global distribution extending throughout temperate and subtropical waters (Nakano and Stevens, 2008; Coelho et al., 2018). Blue shark populations have declined broadly, with high longline bycatch and mortality rates due to extensive overlap with commercial fisheries across much of their global range (Queiroz et al., 2016, 2019). This is true for the North Atlantic (50–79% decline over 30 years; ICCAT, 2015) and Mediterranean (96.5–99.8% decline since the early 19th century; Ferretti et al., 2008) populations as a result of both targeted fisheries (i.e., for fins, meat, squalene) and bycatch (Clarke et al., 2006a, b; Cardeñosa et al., 2020), though the North Pacific population has recently been assessed as not overfished (ISC, 2017). While studies have challenged model-based inferences of shark population declines (Burgess et al., 2005), blue sharks are the major bycatch species in high-seas fisheries regionally (McKinnell and Seki, 1998; Francis et al., 2001) and perhaps globally (Clarke et al., 2006b; Campana et al., 2009). Limited genetic structure has been observed across populations sampled from disparate oceanic regions (King et al., 2015; Taguchi et al., 2015; Veríssimo et al., 2017; Bailleul et al., 2018), though regional populations are managed separately. Mechanisms for high genetic homogeneity across global blue shark populations remain mostly speculative, mainly because studies are lacking in adequate sample sizes and sufficient demographic coverage to allow for robust conclusions to be made (Veríssimo et al., 2017). This limitation spans not only genetic information, but also a robust understanding of other aspects of the species’ life history, particularly movements and migrations in relation to proposed mating and parturition grounds.

In the North Pacific Ocean, data suggest that mature blue sharks migrate to a latitudinal band spanning ∼20–30°N for mating during the early summer months, with pupping typically occurring the following summer after a ∼12-month gestation. Gravid females are believed to move further north to parturition grounds located in sub-arctic waters between 35 and 45°N (Nakano, 1994). Males and females typically segregate spatially prior to mating events, but some overlap between immature and mature individuals of opposite sexes can occur (Maxwell et al., 2019). Broadscale movements in the Pacific have been described with conventional and electronic tagging studies (Musyl et al., 2011; Maxwell et al., 2019), and although some blue sharks exhibit long-distance migrations over thousands of kilometers (Maxwell et al., 2019), most tracking data shows predominately latitudinal movements. However, recent conventional mark-recapture information has shown that some individuals migrate across the northern Pacific Ocean from the western Pacific Ocean (WPO) into the eastern Pacific Ocean (EPO) and vice versa (Sippel et al., 2011), a behavior that also has been observed in the North Atlantic (Howey et al., 2017). The extent of connectivity between North Pacific sub-populations (e.g., WPO and EPO), however, remains unquantified. For example, it remains unclear whether genetic mixing and homogeneity (King et al., 2015) is maintained by trans-Pacific migrations of juveniles or adults, or by some other mechanism such as pupping and recruitment dynamics. Thus, the proportion of blue sharks performing trans-Pacific migrations warrants investigation, as quantifying movement connectivity can clarify regional source-sink dynamics, inform spatial scales of management, and help explain the mechanisms that facilitate genetic homogeneity.

While electronic and conventional tagging approaches have provided useful information on blue shark movement and migration dynamics, studies are limited by high cost (for electronic tagging) and often require high sample sizes and protracted study duration to yield necessary ecological information (Sequeira et al., 2019). Consequently, complementary approaches are required for rapid assessment of blue shark migration and movement connectivity. Intrinsic chemical tracers measured in animal tissues, such as stable isotope (SI) ratios, are useful for reconstructing prior animal migrations (Graham et al., 2010; Trueman et al., 2019; Madigan et al., 2021). Stable isotope analysis (SIA)-based movement studies utilize the distinct isotopic composition of prey baselines (i.e., regional isoscapes) across oceanic sub-regions, driven by local oceanographic and biogeochemical regimes (McMahon et al., 2013; Brault et al., 2018; Espinasse et al., 2020). In the North Pacific Ocean, pelagic prey fields in the EPO and WPO are isotopically distinct, particularly for nitrogen isotope ratios (δ¹⁵N) (Matsubayashi et al., 2020). In the EPO, upwelling of nutrient-rich water in the California Current promotes larger nitrate metabolizing primary producers such as diatoms, which creates a ¹⁵N-enriched isotopic composition of regional prey (Altabet et al., 1999; Montoya, 2007; Madigan et al., 2012a, 2017). Comparatively, WPO waters are nutrient-poor, which promotes dominance of nitrogen-fixing picophytoplankton at the base of the food web and lower (i.e., ¹⁵N depleted) δ¹⁵N composition of regional prey pools (Takai et al., 2007; Fujinami et al., 2018; Ohshimo et al., 2019).

Knowledge of regional isotopic baseline variation can be combined with measured predator stable isotope ratios and tissue-specific incorporation rates to identify individuals that have recently migrated from one system into another (Madigan et al., 2021; Shipley et al., 2021). Tissues of recent migrants will reflect an isotopic mix of prey baselines from the prior and current regions, provided that sampling has occurred prior to the consumer reaching isotopic steady-state with prey from their current region (Heady and Moore, 2013; Moore et al., 2016; Madigan et al., 2021). Residents are then defined as those individuals that are at isotopic steady-state with prey baselines in their current region (Madigan et al., 2014). This technique has been applied to characterize trans-Pacific migrations of Pacific bluefin tuna (Thunnus orientalis), using machine learning algorithms to define migrants vs. residents, and thus predicting the extent of mixing between WPO and EPO populations (Madigan et al., 2014). More recently, SIA approaches have been used to determine movement transitions across systems spanning marine, brackish, and freshwater habitats (Moore et al., 2016; Shipley et al., 2021), and have proven to be a robust and insightful approach for clarifying aspects of animal migration.

In the current study, we use regional δ¹³C and δ¹⁵N values from sampled sharks and prey to quantify potential trans-Pacific exchange rates of blue sharks between the WPO and EPO. We use multiple analytical methods with isotopic data to examine the extent of regional residency and foraging connectivity between oceanic sub-regions. This approach provides a tracer-based assessment of habitat use that is complementary to traditional tagging approaches and provides a framework than can be adopted across other study taxa and ecosystems. Inferred movements from isotopic signatures aid in constraining the extent of blue shark movements in the North Pacific basin, clarify the migratory mechanisms that may drive a mixed genetic stock, and can inform appropriate multi-national or regional management strategies for North Pacific blue sharks.

Region-specific stable isotope values of blue sharks and prey allowed for inferences of prior movement patterns and regional connectivity in the North Pacific Ocean, quantifying dynamics that have been observed (e.g., trans-Pacific migrations), but for which frequency and exchange rates are unknown. Our results suggest minimal trans-Pacific movements and indicate potential finer-scale movement and residency dynamics within sub-regions of the EPO and WPO, though key caveats were evident in estimates of finer scale movements. High coherence of prey-estimated blue shark δ¹⁵N values with empirical SI values demonstrated the effectiveness and predictive value of North Pacific isoscapes. Overall, our results demonstrate the efficacy of our multi-analytical stable isotope approach to identify movements of a highly migratory pelagic species in the North Pacific, and the potential for future analyses with consideration of sample treatment and quantification of crucial species-specific isotopic parameters.

Trans-Pacific migrations in North Pacific blue sharks were first demonstrated by conventional tagging efforts in the eastern Pacific (Sippel et al., 2011), and we used SI data to assess the frequency of EPO↔WPO movements. In the WPO, SI datasets revealed no evidence of migrations from the EPO, compared to conventional tagging studies that reported 4 of 205 (∼2%) EPO-tagged (NCC, SCB, and SBaja) blue sharks migrating to the WPO (Sippel et al., 2011). In the EPO, discriminant analysis indicated trans-Pacific migration from the WPO in low proportions (∼5%) of our EPO bootstrapped population estimates, including at least one empirical migrant value (δ¹⁵N = 13.8‰; ETP). However, conventional tagging found no trans-Pacific migration to the EPO from the WPO (n = 207) (Sippel et al., 2011), and more recent satellite tagging in the EPO (n = 47) and WPO (n = 21) showed movements only within those respective ocean regions (Maxwell et al., 2019; Fujinami et al., 2021). While tagging studies have thus provided quantitative metrics of trans-Pacific migrations, tag-inferred movements have limitations, including simplistic movement information and potential non-reporting of tag recovery (conventional tags), limited sample size (electronic tags), and short-term tracks that do not capture long-distance movements (both tag types) (Siskey et al., 2019). Tagging data is also prospective, capturing future rather than prior movements that are potentially biased by tagging location. In contrast, isotopic measurements coupled with isotopic turnover rates and DTDFs can provide retrospective, quantifiable timeframes of prior movements. While SI-inferred movements are limited by isotopic turnover rates of analyzed tissue, prior movements can be characterized due to mobile predators integrating prey isotopic signatures during movements through isotopically distinct regions (Graham et al., 2010; MacKenzie et al., 2012; Carlisle et al., 2015; Madigan, 2015; Trueman and Glew, 2019). As such, isotope- and tag-inferred migration patterns are effective complementary techniques (Carlisle et al., 2012, 2015; Madigan et al., 2015, 2018; Shipley et al., 2021). Here, comparing isotopic estimates to past conventional tagging suggests that blue sharks make both eastward and westward trans-Pacific migrations, though the number of sharks that make these migrations appears to be low (<5%).

The low trans-Pacific exchange inferred here improve understanding of blue shark population dynamics in the North Pacific. Genetic analyses have shown global panmixia across regional blue shark populations, with minimal evidence of regional population structure (Taguchi et al., 2015; Veríssimo et al., 2017; Bailleul et al., 2018). This lack of observed population structure requires some mechanism of regional population mixing, and while maintenance of population genetic homogeneity does not necessarily require high mixing (Bremer et al., 2005; Waples and Gaggiotti, 2006), the minimal trans-Pacific exchange we observed here is unlikely to be the primary driver. Current understanding of blue shark life history provides several alternative scenarios for mixing of EPO and WPO populations. If WPO and EPO sharks mate (central-southern waters; ∼20–30°N) and pup (northern waters; ∼35–45°N) (Nakano, 1994; Nakano and Stevens, 2008) in their respective ocean basins, EPO- and WPO-origin young-of-the-year (YOYs) could subsequently recruit to either the WPO or EPO. In addition, the Central Pacific Ocean (CPO) (i.e., waters around Hawaii) could serve as a mixing region for WPO and EPO sharks, as some exchange of blue sharks in this region has been observed previously (Sippel et al., 2011). Consequently, dispersive YOY recruitment and/or partial, temporary mixing of adults in the CPO (if for mating or parturition) could drive genetic mixing while maintaining regional WPO and EPO isotopic signatures that indicate regional residency of juveniles and adults. Alternatively, genetic analysis has demonstrated the possibility of “genetic time-lag” effects in blue sharks (Bailleul et al., 2018), with the possibility of discrete sub-populations despite genetic homogeneity; thus, WPO and EPO separation observed here could indicate discrete, minimally mixed sub-populations. As such, the degree to which YOY recruitment and adult mixing contribute to North Pacific blue shark population dynamics warrants further study.

Comparison of blue shark and prey values between the WPO and EPO demonstrated that δ¹⁵N, and not δ¹³C, serves as a regional diagnostic tracer in the North Pacific Ocean as has been previously observed in Pacific bluefin tuna (Madigan et al., 2017). We observed high overlap of WPO and EPO prey δ¹³C values, but almost no overlap of prey δ¹⁵N values (Figure 2). Mechanisms for this have been demonstrated, with photosynthetic pathways in pelagic primary producers (C₃ photosynthesis) varying more with latitude (i.e., due to variable productivity, temperature, and seawater pCO₂ regimes), rather than longitude (Bowen, 2010; Magozzi et al., 2017; Brault et al., 2018; Ohshimo et al., 2019). Coherence of prey-based estimated δ¹⁵N blue shark values with shark-derived values (Figure 3) further supports both the robustness of the δ¹⁵N isoscape approach and the marked separation between δ¹⁵N values of WPO and EPO sharks. Similar patterns of distinct EPO and WPO isotopic signatures have been found in Pacific bluefin tuna, a large-bodied pelagic teleost in the WPO and EPO that makes seasonal migrations through the same WPO and EPO regions (Boustany et al., 2010; Madigan et al., 2017; Tawa et al., 2017). In Pacific bluefin tuna, δ¹⁵N-based estimates of trans-Pacific migration were demonstrably effective based on coherence with other chemical tracers, while non-differentiation of tuna δ¹³C made it an ineffective migration tracer (Madigan et al., 2017). This differs from more typical isotopic applications to ecology, which generally use δ¹⁵N to estimate trophic dynamics and δ¹³C to trace energy source and foraging location(s), due to lower trophic fractionation of δ¹³C than of δ¹⁵N (Post, 2002). However, more recent studies have demonstrated that in some systems and for certain predators, baseline δ¹⁵N values across ecoregions can result in predators acquiring regional δ¹⁵N signatures that outweigh trophic effects (Graham et al., 2010; Hobson et al., 2010; Madigan et al., 2017; Shipley et al., 2021). Isoscapes will be basin- and ecosystem-dependent. In the North Pacific, the δ¹⁵N gradient is caused by upwelling-driven enrichment of ¹⁵N in the EPO and low δ¹⁵N values due to oligotrophic N-fixation in the WPO (Takai et al., 2007; Madigan et al., 2012a, 2017; Fujinami et al., 2018; Ohshimo et al., 2019). The finer scale structure of δ¹³C and δ¹⁵N values in the WPO and EPO shown here demonstrate the utility of SIA in these ecosystems to track inter- and intra-basin predator movements.

We used niche overlap of blue shark δ¹³C and δ¹⁵N values to quantify finer scale connectivity dynamics within EPO and WPO sub-regions. In the EPO, isotopic niche overlap suggested variable mixing between the NCC, SCB, SBaja, and ETP (Figure 5). High exchange was suggested between the NCC and SCB; this is supported by conventional (Sippel et al., 2011) and electronic tag studies, which also found sexual segregation between these two regions (Maxwell et al., 2019). Lower overlap between SBaja and the NCC/SCB suggests that sharks may be more resident to this region. Regional proximity likely plays a role in exchange dynamics, as in general, more proximate regions showed greater migratory exchange (Figure 5). Collectively, estimated connectivity in the ETP and SBaja supports the premise that coastal and pelagic waters off SBaja may serve as both an overwintering ground for juveniles and a potential reproductive function for adults (Vögler et al., 2012), with isotopic overlap suggesting migration from SBaja to the ETP (Figure 5). Importantly, movement dynamics related to size and sex structure have been observed across these regions (Nakano and Stevens, 2008; Vögler et al., 2012; Maxwell et al., 2019), which we could not evaluate here due to a lack of size- and sex-specific information to match with blue shark isotopic values. Improved insight into size- and sex-specific movement dynamics could be accomplished with robust sampling of blue sharks for empirical isotopic measurements, coupled with size and sex metadata across the defined study regions. This could easily be achieved through sampling of fisheries [by]catch, potentially incorporating tissues of different turnover rates (i.e., fast [plasma, liver]; slow [muscle]; Thomas and Crowther, 2015; Vander Zanden et al., 2015).

While mixing models provided exploratory and potentially informative estimates of regional connectivity, results of finer scale sub-regional movements appeared biased and thus unreliable. Although the δ¹⁵N values of sharks and DTDF-corrected prey highly overlapped, shark δ¹³C values were low relative to prey in both the EPO and WPO (Supplementary Figures 2, 3). Furthermore, prey mixing polygons (Smith et al., 2013) showed up to 32% and 55% of blue shark values falling outside of the simulated prey mixing space, due to these low shark δ¹³C values (Supplementary Figure 1 and Supplementary Tables 2, 3). This likely biased mixing model estimates toward the lowest δ¹³C prey input(s); this confounding factor was observed in the two-isotope mixing models in both the WPO (high input of Kuroshio-Oyashio) and EPO (high input of SCB) (Supplementary Figures 2, 3). A similar effect of low blue shark δ¹³C values was observed in a previously published estimate of diet based on isotope mixing models in the SBaja sub-region, in which low blue shark δ¹³C values relative to prey resulted in the two lowest δ¹³C prey items (pelagic octopus Argonauta spp. and pelagic red crab P. planipes) dominating diet estimates (Hernández-Aguilar et al., 2015). Consequently, mixing model results here demonstrate that biased results will likely occur when predator SI values are substantially offset from prey inputs. While this is unsurprising and has previously been identified in the literature, studies continue to adopt this approach and report findings without identifying potential bias. It is imperative that investigators use appropriate diagnostic tools (i.e., quantitative assessment of prey vs. predator data following Smith et al., 2013) to ensure results are accurate or that potential bias is reported. For blue sharks, low δ¹³C values that are irreconcilable with local prey is consistent across studies and requires clarification.

There are three possible explanations for the consistently low δ¹³C values observed in blue sharks: (i) feeding in a region, within timeframes, or on prey not represented in regional studies; (ii) estimated Δ¹³C values for blue sharks are too high, and fractionation between prey and blue shark muscle is lower than our estimated DTDF values; or (iii) shark sample preparation for SIA resulted in artificially low δ¹³C values across studies. Missing or isotopically misrepresented prey sources (scenario [i]) is a common issue across studies (Smith et al., 2013); here, sampling too close to coastlines could result in unrealistically high prey δ¹³C and/or δ¹⁵N values for a pelagic shark. However, most studies used in these analyses explicitly sampled in offshore epi- and mesopelagic zones. In addition, prey data are available for two offshore regions that were not included here, the TZCF and CPO, and prey δ¹³C values from those regions (−18.2 ± 0.8‰ and −17.7 ± 0.8‰, respectively) (Gould et al., 1997; Choy et al., 2015) are also not low enough to explain the observed shark δ¹³C values. Temporal variability in isotopic values, while possibly influential, is also an unlikely explanation here, as studies spanned multiple seasons and years and different EPO and WPO values are driven by coarsely consistent oceanographic conditions (Madigan et al., 2017). It is possible that our DTDF estimates are imprecise (scenario [ii]), as they were calculated as diet-based DTDFs following Caut et al. (2009) and Hussey et al. (2014), rather than empirically derived from laboratory experiments. Differential amino acid composition and subsequent ¹³C fractionation in blue sharks could drive atypical DTDFs (McMahon et al., 2010). Laboratory-derived Δ¹³C DTDFs available for other elasmobranch species (Hussey et al., 2010; Logan and Lutcavage, 2010; Kim et al., 2012) are rarely <1.0‰ (the mean DTDF applied here), and a DTDF of ≤0‰ would be necessary for blue shark values to be highly coherent with prey. Negative Δ¹³C values for sharks are rare in available studies, though one negative value (−0.5‰) has been estimated for one prey type (of 8 total diet items in natural diet) in the catshark Scyliorhinus canicula, so this DTDF in wild blue sharks cannot be ruled out (Caut et al., 2013). Similar observations in other ocean basins of low blue shark δ¹³C values relative to prey and/or other sharks (Rabehagasoa et al., 2012; Li et al., 2014; Kiszka et al., 2015) support the possibility that a unique aspect of blue shark muscle composition or physiology could drive atypically low Δ¹³C, but this likely can only be validated with captive studies. Finally, it is possible that sample preparation partially contributed to low δ¹³C values in at least some studies (scenario [iii]). Studies included here performed lipid and/or urea extraction, but not always both (Madigan et al., 2012a; Li et al., 2014; Hernández-Aguilar et al., 2015; Fujinami et al., 2018), or mathematically corrected for lipids (Miller, 2006). Many of these studies preceded thorough published analyses demonstrating the importance of both lipid and urea extraction in elasmobranch tissues, which prevents artificially low δ¹³C and/or δ¹⁵N values (Carlisle et al., 2016; Li et al., 2016; Arostegui et al., 2019); recent work has also shown that water rinses for urea extraction may actually increase δ¹³C values (Carlisle et al., 2016). Our analyses, which draw upon published studies, underscore the need for standardized, consistent sample treatment in ongoing and future isotopic studies (Wolf et al., 2009; Shipley and Matich, 2020), while also noting the value of large archival datasets to address questions at an ocean basin scale. While laboratory-derived DTDFs will be difficult to obtain directly for blue sharks in captivity, improved understanding of ¹³C dynamics in this species will be necessary to refine the accuracy of mixing model results, an issue that may also be applicable across other elasmobranch species.

While the data used here came from known high-use regions for blue sharks in the North Pacific, there are other relevant regions that could not be included in our analyses. In particular, the North Pacific Transition Zone (NPTZ) (also referred to as the Transition Zone Chlorophyll Front) has been shown as a region of high blue shark abundance (Pearcy, 1991; Polovina et al., 2001; Kubodera et al., 2007; Vögler et al., 2012). Recently, the NPTZ has been demonstrated as a migratory corridor for pregnant females that were satellite tagged in the WPO (Fujinami et al., 2021), has been suggested as a nursery ground for young sharks (Nakano and Stevens, 2008), and may be a migratory corridor for other blue shark life stages/sexes as well as tunas, swordfish, and turtles (Block et al., 2011). While blue shark isotope data were not available from the NPTZ, prey δ¹³C and δ¹⁵N appear to be similar to the WPO (Gould et al., 1997). As such, blue sharks with long-term residency in the NPTZ may be isotopically indistinguishable from WPO residents. It is currently unknown whether blue sharks in the NPTZ are residential to the frontal region for adequate timeframes (i.e., months to >1 yr) to acquire the NPTZ isotopic signal or whether they use the region temporarily during inter-region migrations; the WPO migrants we observed here in the NCC could reasonably be entering the California Current via the NPTZ. Isotopic sampling of NPTZ sharks, likely accessible from bycatch in offshore fisheries, could ascertain migration dynamics in NPTZ sharks. Similarly, blue sharks are common in the Central Pacific (i.e., waters around Hawaii; CPO), where conventional tagging suggests mixing with both the EPO and CPO (Sippel et al., 2011). With no isotopic characterizations of blue sharks in the CPO, we could not include this region here; however, isotopic analysis of CPO sharks could reveal the extent to which these sharks are residential to the region, as local prey seem to be distinctive from other regions, particularly prey δ¹⁵N values (Choy et al., 2015). Collecting samples across the spatial range of North Pacific fisheries is tractable, especially since a small biopsy sample from a subsequently released, live shark is sufficient for SIA.

This study was limited to the use of bootstrap-estimated blue shark δ¹³C and δ¹⁵N values rather than using direct empirical measurements. While our approach allowed for population-wide estimates based on empirically derived data distributions, some regional blue shark studies had relatively low sample sizes (e.g., n = 9 in SCB, n = 10 in NCC), and in this context, the “tails” of bootstrapped distributions likely lead to unrealistically high and low SI values (Madigan et al., 2017). Bootstrapped estimates can also lead to de-coupled shark δ¹³C and δ¹⁵N values (see shark data ellipses in Figures 2, 4 and Supplementary Figure 2), when empirical δ¹³C and δ¹⁵N values are often positively correlated. This could affect mixing model estimates of prey/regional contributions to blue shark diet. The lack of size and sex data also limited our ability to undertake more detailed reconstruction of blue shark residency and movement dynamics across life history. Finally, isotopic identification of migrants is dictated by tissue turnover to steady-state conditions (here, likely ∼0.5 to 1.5 year before sampling), precluding identification of long-distance migrations that occurred prior to these timeframes. As such, results here should be taken in the context of these limitations and be viewed as a preliminary framework for regionally focused, empirical investigation.

Our results, drawing upon published δ¹³C and δ¹⁵N data for blue sharks and prey sampled at multiple locations in the EPO and WPO, provide a new and replicable means to assess blue shark residency and migration dynamics in the North Pacific. The analyzed data provide strong evidence for limited direct migrations between the WPO and EPO and reiterate the utility of δ¹⁵N isoscapes for the reconstruction of migratory predator movements in the North Pacific Ocean. Limited trans-Pacific migrations suggest that other mechanisms maintain genetic homogeneity of the North Pacific blue shark population, including YOY movements and/or partial mixing of adults in the Central Pacific. Regional structure in δ¹³C and δ¹⁵N data have promise for further quantification of finer-scale blue shark movements, increasing the resolutions of movement patterns suggested here, but consideration of isotopic parameters (e.g., accurate species-specific DTDFs), appropriate sample preparation of shark tissues, and length/sex metadata of sampled sharks are necessary. With emerging research showing varying residency and trans-regional movements in migratory predators, isoscapes can employ high sample sizes across a breadth of animal life stages, regions, and timeframes to reconstruct habitat use of highly mobile marine animals. Through these isotopic approaches, population-level estimates of movement dynamics are feasible on scales that may not be readily available from conventional tagging or telemetry studies.

The data analyzed in this study is subject to the following licenses/restrictions: Data came from published manuscripts, which are only publicly available through purchase of publications. Requests to access these datasets should be directed to daniel.madigan@stonybrook.edu.

DJM, ONS, NEH, and ABC conceived the study. DJM, ONS, and ABC performed the analyses. DJM and ONS wrote the manuscript, with input from all authors. All authors interpreted results and refined analyses.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.