The obvious physical differences between air and water, and the consequences thereof for organisms inhabiting both media, have led to a separation of marine and terrestrial ecology into largely discrete disciplines (Dawson and Hamner, 2008). Yet, careful comparisons across realms can be a valuable tool to understand and tease out universal environmental rules that determine ecological processes (Webb, 2012). Giving consideration to conceptual cross-discipline similarities—even if they are not unequivocal—may provide impulses for new research avenues. In the context of dispersal, the main characteristic distinguishing most marine from terrestrial organisms is a fundamental difference in life history strategies. Most terrestrial species have relatively simple life histories and dispersal is often undertaken by well-developed individuals (Carr et al., 2003). Marine species, in contrast, are often characterized by external fertilization and the production of large numbers of small, pelagic and potentially dispersive eggs and larvae (Sale et al., 2006). Due to the strong influence of transport mediated by convective forces, the spatial scales of dispersal in the sea are hence generally thought to exceed those on land (Carr et al., 2003; Kinlan and Gaines, 2003). The general shape of the dispersal kernel (a probability function of dispersal distance), however, was recently shown to essentially resemble that of many terrestrial species in several marine organisms: a leptokurtic probability distribution with the mode around the origin and a “fat tail” of variable dispersal distances (Buston et al., 2011; Almany et al., 2013; D'Aloia et al., 2013; Lowe and McPeek, 2014). The selective advantage of staying near the natal site thus appears to be high in both realms (see Section Costs of dispersal). The underlying processes leading to the observed patterns, on the other hand, may differ profoundly.

Much of the theoretical progress in dispersal research in terrestrial systems has been achieved by the mechanistic partitioning of dispersal into its three successive stages: emigration, transfer, and settlement (Lidicker and Stenseth, 1992; Baguette and Van Dyck, 2007; Gibbs et al., 2010). In a terrestrial metapopulation context, emigration (or philopatry) is often treated as a decision by the dispersing individual and transfer involves the active movement of organisms through unsuitable habitat. It is hence questionable whether this conceptual approach is suitable for marine systems, where dispersal is driven to large extents by external, stochastic forces. Nonetheless, we argue that various dispersal concepts from terrestrial ecology can be adopted—or should at least be examined—in marine systems to advance our understanding of the processes that drive variation in larval dispersal. It has long been established that larvae of a wide range of marine taxa are far from passive particles. Instead they possess well-developed behavioral and navigational capacities that interact with physical transport processes to alter dispersal outcomes (e.g., fishes: Stobutzki and Bellwood, 1997; Atema et al., 2002; crustaceans: Jeffs et al., 2003, 2005; Radford et al., 2007; corals: Raimondi and Morse, 2000; Vermeij et al., 2010; sponges: Abdul Wahab et al., 2011). While thus there is little doubt that dispersal is not entirely neutral, we still lack a thorough understanding of the relative importance of larval behavior over extrinsic forces on dispersal trajectories.

At the most basic level, dispersal variability among individuals is driven by extrinsic (stochastic) and intrinsic (quality- or trait-based) factors (Lowe and McPeek, 2014). The combination of small offspring size with strong physical forcing in marine systems has led to a focus on neutral processes to explain larval dispersal and gene flow in marine systems, i.e., oceanography and environmental factors (e.g., Cowen et al., 2006; White et al., 2010; Nanninga et al., 2014). The obvious effects of stochastic forces on minute, undeveloped individuals clearly generate strong neutral variation in larval dispersal. Due to spatially stratified currents, for instance, identical larvae may end up in vastly different places if they are released even small distances from one another (Vikebø et al., 2005). While there is little doubt about the paramount importance of extrinsic drivers of larval dispersal variation, we also need a more thorough focus on intrinsic drivers and a better understanding of the relative importance of neutral vs. non-neutral factors in regulating individual dispersal trajectories (Lowe and McPeek, 2014).

Polymorphism in dispersal related biological traits may be adaptive or non-adaptive. Variability in offspring size around a population mean, for instance, may be non-adaptive if stochasticity in resource availability or external stressors through time affect maternal allocation to progeny (Fox and Czesak, 2000; see Section Parental effects). However, phenotypic variation is commonly non-random within a population (Ronce, 2007). A single phenotype is unlikely to perform well in all environmental conditions. Metapopulation theory thus predicts that there should be a selective advantage in plastic dispersal strategies in variable environments or habitats near their carrying capacity (Clobert et al., 2001; Bonte et al., 2010). If the future environment of offspring is unpredictable, it may be beneficial for parents to hedge their bets by producing variable phenotypes that differ from one another in dispersal related traits (Crean and Marshall, 2009; Knott and McHugh, 2012) (Figure 1).

Another source of variation in larval dispersal may be short-term environmental stochasticity affecting individual condition. Past experience and adaptive responses to a variable environment may thus cause variability in the internal state of individuals (e.g., gut fullness, energy reserves), which in turn affects behavioral strategies during dispersal (Fiksen and Jørgensen, 2011; Jørgensen et al., 2013) (Figure 1). Behavioral responses to external cues (e.g., predators, food) are tightly linked to physical oceanography and small variations in vertical migration, for instance, may have vast consequences for dispersal trajectories (Fiksen et al., 2007; Vikebø et al., 2007).

In either case—evolved dispersal strategies or variable behavioral trade-offs—individual-level adaptations have strong effects on emergent population-level properties such as gene flow, demographic connectivity, species distribution, persistence, and resilience (Jopp and Reuter, 2005) (Figure 1). For instance, dispersal may enhance phenotypic plasticity within a species or population, when phenotypic variation is maintained by dispersal among habitats with variable environmental conditions (Zhang, 2006; Stamps, 2007). Long distance dispersal may thus favor individuals with high levels of plasticity, or alternatively, plasticity may evolve as a function of dispersal scale (as habitat differentiation tends to increase with spatial distance) (Sotka, 2012) (feedback loop 1, Figure 1). On evolutionary scales, phenotypic variability among individuals may affect dispersal strategies, while at the same time dispersal may affect and maintain such phenotypic variability (Stamps, 2007) (feedback loop 2). It will thus be important to not only understand variability in dispersal related traits (see below), but also the processes that maintain this variability (Lowe and McPeek, 2014).

It is increasingly becoming apparent that dispersal polymorphisms are not random across a population, but often associated with particular phenotypes (Cote et al., 2010; Ronce and Clobert, 2012). Dispersal strategies at the individual level may coevolve with variations in morphological, behavioral, and life history traits forming dispersal syndromes (Clobert et al., 2009; Ronce and Clobert, 2012). The accurate assessment, interpretation and extrapolation of these syndromes is hampered by the often strong covariation among different traits and the intra-specific variability thereof depending on internal and external conditions (Ronce and Clobert, 2012). Nevertheless, the identification of syndromes that could serve as dispersal predictors was shown to be a useful proxy in the forecast of dispersal patterns over a range of terrestrial and semi-terrestrial species (Stevens et al., 2013, 2014). Proxies for dispersal would be especially useful in marine systems, where dispersal itself is exceedingly difficult and costly to measure. There is evidence that dispersal distance in various terrestrial organisms may be related to intra-specific morphological predictors, such as wingspan in birds (Dawideit et al., 2009) and butterflies (Sekar, 2012), leg morphology in salamanders (Lowe and McPeek, 2011), or body size in small mammals (O'Riain et al., 1996; Selonen et al., 2012). Up to this point, no specific morphological predictors of dispersal (e.g., larval fin aspect ratios) have been identified and validated in any marine organism [but see Section Costs of dispersal: maternal body size as a potential predictor for natal homing (Beldade et al., 2012)].

Similarly, behavioral traits can have a strong influence on dispersal strategies. Intra-specific variations in personality traits are commonly observed in various taxa and may result in inter-individual differences in the propensity toward dispersal (Stamps, 2007; Baguette et al., 2013). “Boldness” in particular is often shown to relate positively to dispersal tendencies (Cote et al., 2010). In coral reef fishes, larval behavior is believed to be an important factor mediating hydrodynamic advection (Kingsford et al., 2002; Paris et al., 2007). Small differences in vertical migration behavior, for instance, can have strong effects on dispersal distance and direction (Sundelöf and Jonsson, 2012). Intrinsic and context-dependent behavioral trade-offs associated with boldness (e.g., food acquisition vs. predator avoidance) may thus have strong effects on growth, survival, and dispersal trajectories (Fiksen and Jørgensen et al., 2007; Vikebø et al., 2007). At the same time, low levels of boldness (among other traits) were shown to result in increased post-settlement mortality in a damselfish (Pomacentrus wardi) (Fuiman et al., 2010). By influencing dispersal pathways as well as settlement and post-settlement selection, variations in behavioral traits might present a selective mechanism balancing the costs and benefits of dispersal in marine organisms.

Dispersal syndromes affect the propensity as well as the ability of individuals to disperse in a given context (Bowler and Benton, 2011; Baguette et al., 2013). How far does and individual want to go and how far can it go? Individual dispersal propensity and ability thus determine intraspecific polymorphisms in dispersal (Clobert et al., 2009). While this concept stems from the terrestrial literature, there are several examples of larval dispersal polymorphisms in marine systems.

Probably the most striking example of a larval dispersal polymorphism is poecilogony, describing the intra-specific variation in larval developmental mode (pelagic, feeding vs. benthic, non-feeding) that has been reported for a number of invertebrate species (Hoagland and Robertson, 1988; Levin and Bridges, 1995; Chia et al., 1996). While it is not fully resolved whether poecilogony represents a genetic polymorphism (Levin et al., 1991), or a plastic response to environmental factors (Krug, 2007; Krug et al., 2012), it clearly has a large impact on patterns of larval dispersal (Knott and McHugh, 2012). In the few real cases of poecilogony described to date, investigators observed two distinct larval morphs with different dispersal potential: (1) small eggs develop into planktotrophic larvae with an obligatory pelagic larval phase, (2) large eggs develop into lecithotrophic larvae that can metamorphose without feeding in the plankton (Vendetti et al., 2012). Poecilogony is believed to present a bet-hedging strategy by parents, providing offspring with variable phenotypes (i.e., long-distance and short-distance types) to match stochastic and unpredictable environments (Knott and McHugh, 2012).

Another well-documented dispersal polymorphism in marine invertebrates is inter-individual variation in gregariousness (Toonen and Pawlik, 1994, 2001a). Toonen and Pawlik demonstrated in numerous experiments that larvae of the tubeworm Hydroides dianthus differ in their tendency to settle in the presence (aggregators) or absence (founders) of conspecifics (Toonen and Pawlik, 1994, 2001a,b). In addition to specific settlement preferences, aggregators stay competent to settle for longer time periods than do founders. This variability among sibling larvae probably has strong emergent effect on the patterns of dispersal in H. dianthus, resulting in the colonization of nearby, uninhabited habitats by founders and potential long-distance connectivity with already occupied habitats by aggregators (Toonen and Pawlik, 1994, 2001a).

More recently, Mizrahi et al. (2014) discovered a potential dispersal polymorphism in a scleractinian coral, Tubastrea coccinea. Larvae of this species usually settle and metamorphose after 3 days in the plankton. As an alternative strategy, planulae were observed to metamorphose into mobile polyps in the plankton and form clusters of different sizes. Singular and clustered polyps exhibited longer pelagic survival than planulae, potentially presenting an adaptation to increase dispersal capacity (Mizrahi et al., 2014).

Amphidromy describes the dispersal of eggs and larvae through estuary or marine environments before juveniles return to a freshwater stream (McDowall, 2007, 2010). Marine dispersal, however, is often facultative and some individuals may stay in freshwater for their entire life cycle. Using otolith microchemistry, two recent studies independently identified strong dispersal polymorphisms within different species of freshwater gobies (Hogan et al., 2014; Huey et al., 2014). In both studies, more than 60% of the sampled individuals completed their entire life cycle within streams, while the remaining 40% exhibited amphidromous migrations. Hogan et al. (2014) found that in Awaous stamineus, individuals with marine dispersal exhibited clear growth advantages over purely freshwater morphs. These benefits likely persisted into adult life. The authors conclude that different dispersal strategies in A. stamineus are generated by variable cost-benefit ratios of amphidromy vs. retention in the natal stream. While staying within a stream reduces the potential immediate costs of dispersal (see Section Costs of dispersal), marine dispersal confers potential long-term fitness benefits (Hogan et al., 2014).

As Toonen and Pawlik (2001a, p. 2450) state, “all larvae are not created equal.” While true dispersal polymorphisms have been identified only in a handful of species, the marine literature holds numerous examples of inter-individual variability in traits related to dispersal (Raimondi and Keough, 1990). Virtually any study testing discrete settlement preferences, for instance, involves larvae that readily chose the control treatment (Figure 2). Yet, this variability in larval behavior is rarely discussed in the overall conclusions and most authors simply report a species-wide attraction (or repulsion) to specific cues. Rather than representing experimental error or chance, however, individually variable settlement preferences and behaviors may be indicative of important adaptations to stochastic environmental conditions (Toonen and Pawlik, 2001a; Bowler and Benton, 2005; Clobert et al., 2009).

Owing to its potential importance during the dispersal process, swimming abilities of reef fish larvae have received considerable attention over the past two decades (e.g., Stobutzki and Bellwood, 1994, 1997; Fisher et al., 2000; Fisher and Bellwood, 2002). Fisher et al. (2005) found that there is large intra-specific variation (28%) in critical swimming speeds among larvae of 89 species of reef fishes. These differences could not be explained by body size or residual weight, indicating that there is genuine genetic or maternally induced variability in the ability or propensity toward swimming behavior within populations. When further dissecting the data from Fisher and colleagues it becomes clear that there is large inter-individual variability in swimming performance even among clutches and among siblings of a single clutch (Figure 3). The extent of the capacity to actively maneuver during the pelagic phase has obvious implications for patterns of dispersal (Fisher, 2005; Leis, 2007), and any variation in such behavior is likely to have critical effects on metapopulation connectivity. Similarly, variability in sensory capacities (e.g., olfactory, auditory) will play a vital role in the ability to navigate back to suitable habitats after the pelagic phase. To date, despite its likely critical importance for connectivity patterns, little research has directly focused on inter-individual variation in navigational abilities.

The literature holds many more examples of natural intra-specific variability in traits that are potentially important for dispersal, even though this variability is not always explicitly explored in the respective studies. To discuss all of these examples goes beyond the scope of this review (e.g., boldness: Coleman and Wilson, 1998; hormone levels in eggs and size at hatching: McCormick, 1999; growth rates and larval duration: Searcy and Sponaugle, 2000; ability to delay metamorphosis: Marshall and Keough, 2003), yet clearly, inter-individual phenotypic heterogeneity is ubiquitous in marine organisms and potentially has important emergent consequences at the population/species/community level.

While it is not clear to what extent marine larvae can make choices about their dispersal trajectories, i.e., exhibit dispersal propensities, there are some indications that adaptive dispersal polymorphisms are widespread in larvae of marine organisms. The following line of arguments is based on empirical estimates of larval dispersal in populations of coral reef fishes derived from genetic parentage analysis. (1) Within-patch variability I: Two different species of coral reef fishes (A. percula and Chaetodon vagabundus) at Kimbe Island in Papua New Guinea produced individuals returning to the natal reef as well as larvae dispersing to distant patches (in excess of 30 km) within the same cohort (Planes et al., 2009; Berumen et al., 2012). While it remains uncertain whether this pattern is a result of mere physical transport or active behavior, the studies do show that there may be within-patch variability in dispersal strategies in a single generation. (2) Within-patch variability II: Around Kimbe Island, certain anemones were shown to regularly produce larvae that return to the island to settle, while other anemones never do. This pattern was consistent over four sampling periods (2005, 2007, 2009, 2011) (Saenz-Agudelo et al., unpublished data). Local adaptation to small-scale environmental heterogeneity may produce within-patch variability in maternal effects on larval phenotypes (see Section Parental effects) with far-reaching consequences for patterns of dispersal. (3) Among-patch variability: Different locations were shown to exhibit specific and persistent patterns of local recruitment between two species of anemonefish with similar life history characteristics (A. percula and A. bicinctus). While Kimbe Island seems to generate consistently high levels of self-recruitment across different generations (Berumen et al., 2012), a coral reef in the Red Sea was shown to exhibit very limited local recruitment over two consecutive years (Nanninga et al., 2014). (4) Intergenerational variability: Individual offspring of the anemonefish A. polymnus from the same parent pair were shown to disperse in one generation and return to the natal reef in the next (Saenz-Agudelo et al., 2014). While the mechanisms behind these patterns also remain unclear, it may be an indication of the spatial and temporal variation in condition-dependent dispersal strategies. While these arguments are largely speculative, the existence of different intra-specific dispersal strategies or context-dependent dispersal propensities may have important consequences for the way we view, study and model metapopulation connectivity in marine species. Clearly, adaptive dispersal polymorphisms—if real—are not limited to philopatry vs. long distance dispersal, but present a continuum of dispersal distance strategies determined to large extents also by the ambient oceanographic regime and habitat availability. An understanding of the causes and effects of different dispersal strategies might shed new light on the functioning of marine population dynamics.

The parental and geographic origin of an individual may have long-lasting effects on dispersal behaviors and abilities, settlement habitat preferences, and post-settlement performance. Spatial and temporal heterogeneity in environmental factors among and within habitat patches may thus create important variability in dispersal strategies in marine metapopulations. The influence of the natal environment may be direct (e.g., via imprinting) or indirect (e.g., via maternal effects).

The largest conceptual discrepancy between terrestrial and marine dispersal theory is probably anchored in the process of departure. As discussed before, departure in a terrestrial context is often treated as a conscious decision by a mobile individual that exhibits active movement among habitats across a matrix of unsuitable habitat. Condition- and context dependent costs-benefit ratios of dispersal can thus cause intra-specific variability in the propensity toward dispersal (Bowler and Benton, 2005; Clobert et al., 2009). This dispersal propensity—in concert with dispersal ability—determines which individuals disperse, how far they move, and which return to or stay near the natal patch (Baguette et al., 2013). In marine systems, dispersal into a pelagic habitat is often obligatory and neutral forces are thought to largely determine dispersal patterns. It hence remains unclear whether dispersal propensities exist in marine organisms. It was shown in several terrestrial species that the tendency of individuals to disperse may be transferred across generations via maternal effects (Tschirren et al., 2007 [birds]), or epigenetic inheritance (Ahlroth et al., 2010 [arthropods]). This ability to plastically adapt the dispersal strategy of the next generation, depending on current environmental conditions and dispersal histories experienced by parents, may be particularly useful in many marine species where offspring are quickly dispersed by currents and “dispersal decisions” must be made early in ontogeny. The described biophysical differences between terrestrial and marine dispersal may also result in a major discrepancy in the concept of dispersal costs.

Regardless of whether an individual displays philopatry or dispersal, most marine larval stages have an obligate pelagic period (Leis, 2006). As discussed, during this period in open water, even slight variations in larval condition, behavior, and life history traits may strongly influence dispersal trajectories (Fiksen and Jørgensen, 2011). Reciprocally, environmental conditions encountered during the pelagic phase may have profound effects on larval condition, development, and settlement success. Such matrix effects (sensu Shima and Swearer, 2009) include food availability (e.g., Meekan et al., 2003; Phillips, 2004; Donelson et al., 2009), variations in temperature (e.g., Meekan et al., 2003; Sponaugle et al., 2006; Grorud-Colvert and Sponaugle, 2011; Rankin and Sponaugle, 2011), solar radiation, wave height, wind regime, and rainfall (Bergenius et al., 2005; Macpherson and Raventos, 2005). Recent research demonstrated that the pelagic dispersal matrix has a significant influence on patterns of dispersal and settlement success (Shima et al., 2010). Using information stored in the otoliths of the common triplefin (Forsterygion lapillum), Shima and Swearer (2009, 2010) were able to link higher recruit quality (a composite measure of growth-related traits) to retention in local waters compared to long-distance dispersal. While their results were probably a result of strong heterogeneity in the dispersal matrix (nutrient rich near-shore vs. nutrient poor off-shore habitat), the studies strikingly establish the potential importance of different dispersal pathways for recruit condition. More recently, Shulzitski (2012) showed that residency in meso-scale eddies during the pelagic stage enhanced larval growth in four out of five reef fish species examined. Larval trajectories and local recruitment might hence not only be influenced by the interplay of hydrodynamic forcing with larval behavior (Leis, 2002), but also by individual dispersal histories and experiences in the pelagic environment.

Whether larval quality is a function of the natal environment (parental effects), dispersal history (matrix effects), or a combination thereof will have vast implications for metapopulation connectivity and our ability to accurately predict it (Shima and Swearer, 2009). Yet, few studies have so far addressed this issue and the ones that did yielded somewhat conflicting results (parental effects: Donelson et al., 2009; pelagic conditions: Shima and Swearer, 2009). Contrasting outcomes are to be expected, however, where inter- and intra-specific phenotypic variability is high and methodologies differ. Monro et al. (2010), for instance, demonstrated that selection for offspring size in bryozoans varies greatly between laboratory and field conditions and even among cohorts in the latter. Care should hence be taken when interpreting observed patterns under controlled laboratory conditions, as they may not reflect real field responses. So far the relative importance of parental effects over matrix effects on recruit quality remains largely unknown.

In marine organisms with complex life cycles, the dispersal phase has long been known to represent a population bottleneck involving exceedingly high rates of mortality during the larval stage due to predation, starvation, or advection to unsuitable habitats (Houde, 1989; Leggett and Deblois, 1994). At the same time, even small fluctuations in larval mortality can potentially cause large variations in recruit cohorts (Caley et al., 1996; Houde, 1997). Any inter-individual variability in settlement preferences (Figure 2), sensory or locomotory abilities (Figure 3), or physiological condition (e.g., Burgess and Marshall, 2011) among late stage larvae may thus have critical effects on connectivity, because low quality individuals are less likely to successfully settle. Marshall and Keough (2003) demonstrated that the ability to delay metamorphosis varied significantly among individuals with different growth patterns in three species of marine invertebrates. While this ability to extend the larval duration increases an individual's capacity to search for suitable habitat, it may also come at a direct deferred cost after settlement (Marshall et al., 2003). Similarly, inter-individual size variation in larvae of the sea-urchin Heliocidaris erythrogramma were recently fond to confer different dispersal potential (Marshall and Steinberg, 2014). Smaller larvae settled earlier and more frequently in less preferred habitat than larger conspecifics, indicating that maternal effects on offspring size may have an important influence on settlement across life history stages. Finding, reaching, and colonizing suitable habitat after dispersal is a critical process in the life cycle, and the phenotype and condition of settlers in combination with the settlement environment may have long lasting effects on individual fitness.

The process of metamorphosis and recruitment into the juvenile and adult population represents another critical bottleneck in marine populations, regularly involving the loss of more than half of the cohort immediately after settlement (Caley, 1998; Doherty et al., 2004; Almany and Webster, 2006). Post-settlement selection is thus a key process in recruitment regulation and may have a greater influence on population growth and dynamics than mere patch connectivity (Figueira, 2009; Carson et al., 2011; López-Duarte et al., 2012). Predictions of population dynamics based solely on colonizer numbers may hence not effectively reflect realized connectivity, especially if incoming recruits die non-randomly shortly after settlement (Pineda et al., 2010; Burgess and Marshall, 2011). In other words, selective post-settlement processes may decouple patterns of larval dispersal from true reproductive connectivity, the fraction of recruits that survives to reproduce (Pineda et al., 2007; Marshall and Morgan, 2011; Burgess et al., 2012). Early life-history mortality is often selective with respect to specific traits (Perez and Munch, 2010) and “high quality” individuals should theoretically contribute proportionately more to the population than those in poor condition (Suthers, 1998). The phenotype and physiological condition of an individual in turn may be the result of various processes during different life history stages in the natal and the dispersal environment interacting with the organism's genotype (Stamps, 2007; Figure 4). These processes may carry over across the life cycle (Burton and Metcalfe, 2014) to affect dispersal strategies in the next generation.

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.