Introduction

Coral reef fish communities are dominated by small cryptobenthic reef fishes (hereafter cryptobenthics), defined as fishes from families where at least 10% of species do not grow larger than 50 mm (Brandl et al. 2018). These small fishes account for around 50% of the individuals, and 40% of the fish species on coral reefs (Ackerman and Bellwood 2000; Brandl et al. 2018). While cryptobenthics are remarkably abundant and diverse on coral reefs, they also face great risks.

Cryptobenthics experience extreme mortality rates, primarily due to predation. Up to 8% of adult cryptobenthic populations can be consumed each day (Depczynski and Bellwood 2006; Goatley and Bellwood 2016) by almost any fish or invertebrate capable of fitting a cryptobenthic fish in its mouth or mandibles (Goatley et al. 2017). This mortality rate results in cryptobenthics being exceptionally short-lived. The dwarfgoby, Eviota sigillata (Gobiidae), for example, has a maximum lifespan of just 8 weeks, holding the record for the shortest-lived vertebrate (Depczynski and Bellwood 2005). The small body size and limited longevity of cryptobenthics also result in low individual fecundity rates. Quite simply, there is not enough space inside a female cryptobenthic fish to hold many eggs, and not enough time to spawn many clutches before the females are eaten (Depczynski and Bellwood 2006). If cryptobenthics are constantly being eaten, and can only produce a few eggs, how do they maintain such high abundances on coral reefs? Recently, Brandl et al. (2019a, b) suggested that unusual larval dynamics may be the key to the maintenance of cryptobenthic fish populations.

Many large reef fish larvae display highly developed sensory and locomotory abilities (Stobutzki and Bellwood 1997; Leis et al. 2011) and can travel tens to hundreds of kilometres during their pelagic larval phase (Harrison et al. 2012; Williamson et al. 2016). Cryptobenthic larvae, by contrast, appear to have maximised their chance of successful returns by restricting dispersal (Gerlach et al. 2007; Majoris et al. 2019; Rueger et al. 2020), sometimes even completing their larval stage within sheltered island lagoons (Leis et al. 2003). Brandl et al. (2019a) used a meta-analysis to reveal that cryptobenthic larvae dominate near-reef larval assemblages, with cryptobenthics accounting for an average of 66% of reef fish larvae within 10 km of coral reefs. This results in a consistent, localised pool of larvae to replace individuals lost to predation. However, we currently do not know how cryptobenthic larvae remain close to reefs, nor the habitats they occupy near to reefs. With cryptobenthics providing up to 70% of fish tissue consumed on coral reefs (Brandl et al. 2019a, 2019b), it is critical to resolve the dynamics that underpin their larval replenishment.

Comparatively examining the adult and larval stages of cryptobenthics and large reef fishes may provide valuable clues regarding the behaviour, development, and ecology of cryptobenthic larvae and their effects on cryptobenthic populations. As larvae are difficult to observe in the wild, morphological characteristics relating to swimming, feeding, and sensory abilities (fins, mouths, and eyes, respectively) may offer a useful proxy for the ecological strategies and challenges that larvae are subjected to. Our aims, therefore, were to compare the broad morphologies of cryptobenthic and large reef fishes, specifically focussing on differences between larvae. We additionally consider the likely functional and ecological implications of these differences in the context of cryptobenthic larval ecology.

Methods

We compared the morphologies of four groups of fishes: adult cryptobenthics, larval cryptobenthics, adult large reef fishes, and larvae of large reef fishes. Large reef fishes were selected from the 11 families in the ‘consensus list’ of fish families found on coral reefs around the world, presented in Bellwood (1996) and cryptobenthics were selected from the 17 ‘core’ cryptobenthic families from Brandl et al. (2018). To prevent over-representation of speciose families (e.g. Gobiidae and Labridae), we randomly selected up to 16 genera from each family and up to 5 species from each genus as representatives of the families.

We then conducted a search for ‘Randall style’ images of each species from online databases and publications (primarily Robertson and Allen 2015; FishwisePro 2017; Froese and Pauly 2019). For adult specimens, a single, representative image or scientific illustration was selected for each species where they met the following criteria: 1) the image was a lateral view of a fish on a flat surface (angled views and misshapen specimens were excluded); 2) the specimen was in good condition with fins intact; 3) a scale was included with the image, whether as a measuring device in the frame or an associated measurement of the specimen; 4) the specimen was an adult (at least 50% maximum reported length), lacking juvenile or extreme secondary sexual characteristics. Images or scientific illustrations of larvae were selected following criteria 1–3, as above, but to ensure that all specimens were directly comparable, we only selected late-stage (transitional) pelagic larvae. Fewer images of fish larvae were available, so over-representation was not an issue. Overall, we compiled images of 372 species of cryptobenthic adults, 57 species of cryptobenthic larvae, 282 species of large reef fish adults, and 35 species of large reef fish larvae (ESM 1).

Morphological analyses closely followed those used in Bellwood et al. (2014). From each image, we measured the standard length and seven functionally relevant traits associated with body shape and the sizes of fins, eyes, and mouths (Bellwood et al. 2014; Fig. 1). To correct for differences in size between fishes, all trait measurements were converted to proportions of standard length. Then, allometric relationships were identified by plotting the proportional values against standard length. Any trait with a significant linear regression (p ≤ 0.05) was considered to display allometry. Residual values were calculated for allometric traits, and these values were used in all subsequent analyses.

Fig. 1
figure 1

Morphological measurements collected from each specimen were taken along or between pairs of reference lines plotted on the image. Distances between lines A, B represented the snout length (along the anteroposterior axis); B, C eye diameter (along the anteroposterior axis); D, E the length of the caudal fin (along the anteroposterior axis); F, G the maximum body depth; H the lower jaw length (anterior tip of dentary to articulation point), and I the longest dorsal fin ray. All measurements were standardised against A–D the standard length

Full size image

The data were visualised using a metric multidimensional scaling ordination (mMDS) based on a Euclidean distance matrix of normalised data. This analysis was selected as it accurately preserves multidimensional distances in the ordinations, allowing for further assessment of the morphospaces (Clarke et al. 2014). The relative importance of morphological traits in shaping the morphospaces was calculated using Pearson correlations and plotted as vectors alongside the ordination. Morphological differences among the four categories of fishes (adult vs. larval cryptobenthics, adult vs. larval large reef fishes) were assessed using a permutational multivariate analysis of variance (PERMANOVA).

To test for morphological differences among the four categories of fishes in more detail, we compared the dispersion of taxa within the morphospaces and the area of morphospace occupied. To assess the dispersion of taxa through the morphospace, we applied a permutational analysis of multivariate dispersion (PERMDISP), in which average distance from the centroid was used as the measure of disparity. Finally, we used a random permutation testing approach to compare relative morphospace areas in the mMDS, irrespective of sample sizes. Specifically, we computed the convex hull area delineated by a random subset of 10 specimens in each category and repeated this calculation over 999 permutations for each category of fishes (following Villéger et al. 2011; Bellwood et al. 2014; Brandl and Bellwood 2014). We then compared the morphospace areas occupied by each category against the morphospace occupied by a random subset of ten points drawn from all four categories. To determine whether morphospace occupation of the four categories deviated significantly from random draws, p values were computed by calculating the fraction of random permutations greater than or equal to the empirical values.

Results

The mMDS (Fig. 2) and PERMANOVA display clear differences among all categories (adult cryptobenthics, larval cryptobenthics, adult large reef fishes, and larvae of large reef fishes; PERMANOVA, Pseudo-F3,757 = 40.46, p(perm) ≤ 0.0001, all pairwise comparisons p(perm) ≤ 0.0001). The PERMDISP results (F3,757 = 23.79, p(perm) ≤ 0.0001) indicate that care should be taken interpreting the PERMANOVA (homogenous dispersion of data is an assumption of PERMANOVA); however, the pairwise tests show that adult and larval large reef fishes, and adult cryptobenthic reef fishes have a similar disparity, supporting the significant difference between these groups (Fig. 3a).

Fig. 2
figure 2

Morphospace obtained from the metric multidimensional scaling (mMDS) ordination of cryptobenthic reef fish (blue) and large reef fish (grey) families based on seven morphometric measurements (Fig. 1). A single ordination was conducted including all categories of fishes; adult and larval morphospaces have been separated for ease of visualisation. a The morphospaces of adult fishes as convex hulls. b Vectors (calculated using Pearson’s correlations) reveal how individual traits drive the morphospace occupation. Asterisks show allometric traits, which were corrected using residuals. c The same mMDS with larval fishes as convex hulls. Cryptobenthic larvae display lower morphological diversity than other groups

Full size image
Fig. 3
figure 3

a Violin plots of deviations from the centroids of each category of fishes, calculated using the PERMDISP. Solid horizontal lines represent the medians, while dashed lines represent the 2.5 and 97.5 percentiles for each category (i.e. 95% of data fall within this band). ** indicates that cryptobenthic larvae formed a separate grouping from all other categories in the pairwise tests (p < 0.0001 for all comparisons). b Sample-size independent measures of relative morphospace occupation (i.e. estimated sizes of convex hulls). For each category of fishes, ten points were drawn from the 2D mMDS ordination, and the area delineated by these points was calculated. This process was repeated over 999 permutations for each category. Data from each category (darker, right) were compared to similar, random draws across all categories (i.e. the total data set; pale, left). **indicates where the morphospace occupied by one of the categories differs from that of the random draws (p = 0.004)

Full size image

This result indicates that there is comparable variability in body shapes among adult large reef fishes and cryptobenthic fishes. On average, cryptobenthics are somewhat shallower-bodied than large reef fishes; dimension 1 (Fig. 2a,b) is characterised by a transition from deep-bodied, long-finned forms (such as the batfish Platax pinnatus), to elongate fishes with relatively small, or absent fins (e.g. pipefishes [Syngnathidae]). However, adult cryptobenthic fishes include species with similar extreme forms to those in larger reef fishes. Species with highly elongated dorsal fins or filaments occupy the upper-left portions of morphospace (e.g. Callionymus draconis and Paraplesiops meleagris).

The PERMDISP also highlights limited dispersion in cryptobenthic larvae (Fig. 3a). The median deviation from the centroid for cryptobenthic larvae (0.95) was just 50% that of any other group (1.9 for large larvae), and they differed significantly from all other categories of fishes (Fig. 3a). This pattern is mirrored by the morphospace areas in the mMDS (Fig. 3b). The relative area of morphospace occupied by cryptobenthic larvae is just 20% of the area of the next largest group; adult cryptobenthics (medians 0.016 and 0.082, respectively). Cryptobenthic larvae were also the only category to differ significantly from area occupations obtained from random draws across all categories (p = 0.004; Fig. 3b).

Overall, cryptobenthic larvae are relatively uniform, situated close to the origin of the ordination. This region of the morphospace is characterised by fishes that lack ornamentation with elongate forms and relatively small fins. This area is entirely encompassed within the adult cryptobenthic morphospace, meaning that larval cryptobenthic fishes have few novel characteristics in their general body shape. They are merely ‘simple’ elongate morphotypes, a body form that is retained in some adult cryptobenthic fishes.

Discussion

Our findings indicate that adult cryptobenthics, and both adult and larval large reef fishes exhibit comparable variability in body shape around distinct ‘average’ morphologies (i.e. the extent of variation around centroids is comparable, but centroid position differs). Cryptobenthic larvae, however, display highly constrained morphologies. They occupy a smaller area of morphospace and are more densely distributed in this morphospace than the other categories of fishes in this study.

The variability in body shapes of large reef fish adults and their larvae is well documented (Webb 1984; Goatley and Bellwood 2009; Llopiz and Cowen 2009; Katz and Hale 2016; Larouche et al. 2020) and reflected in the broad variety of functions that these fishes perform (Goatley et al. 2016; Bellwood et al. 2017, 2019; Harborne et al. 2017). The diversity of forms in adult cryptobenthic fishes is, however, a little unexpected. Miniaturisation is often associated with reduction and simplification of forms, and homoplasious (convergent) adaptations to address the shared challenges of small body sizes (Hanken and Wake 1993; Britz and Conway 2009). Furthermore, major groups of cryptobenthics fishes are frequently considered to display limited morphological diversity (e.g. Eviota gobies; Thacker 2011; Tornabene et al. 2013a, b). Our findings highlight that the diversity of cryptobenthic fishes is reflected not only in their taxonomy (Greenfield 2017; Brandl et al. 2018) but also in their functional morphologies (Kotrschal 1989; Herler 2007). Future ecological studies of cryptobenthic fishes may, therefore, benefit from using a morphological approach to determine their trophic position and functional role on coral reefs (e.g. Herler 2007; Bellwood et al. 2014, 2015).

The most striking finding of this study, however, is the morphological constraint displayed by larval cryptobenthics. Cryptobenthic larvae are defined by relatively uniform, simple morphologies, with elongate bodies and relatively small fins. This result is particularly surprising when the differences in reproductive modes between large reef fishes and cryptobenthic fishes are considered. Fishes which guard benthic eggs (the primary mode of reproduction in cryptobenthics; Brandl et al. 2018) usually have larger eggs, more developed larvae, and shorter planktonic larval durations than fishes which release their eggs directly into the water column (i.e. most large reef fishes; Flegler-Balon 1989; Sponaugle et al. 2002; Riginos et al. 2014). Tropical cryptobenthic fishes appear to break this ‘rule’. The size of their eggs and larvae, and their planktonic larval durations are all comparable to large reef fishes (Leis and Rennis 1984; Stobutzki and Bellwood 1997; Privitera 2001; Depczynski and Bellwood 2006; Victor et al. 2010). Indeed, the late-stage cryptobenthic larvae measured herein are superficially similar in form to early post-flexion larvae of most other reef fishes (cf. Leis and Rennis 1984).

While many large fish species have relatively elongate larvae, which undergo ontogenetic shifts to deeper-bodied adults (Katz and Hale 2016), the morphological constraints displayed by cryptobenthic larvae are exceptional and may be the result of ecological or evolutionary pressures. The question is, do these morphological constraints offer any insights into how cryptobenthic fishes maintain their numerical dominance of both reef fish communities (Ackerman and Bellwood 2000; Brandl et al. 2018) and near-reef ichthyoplankton assemblages (Brandl et al. 2019a)?

Functionally, the traits displayed by cryptobenthic larvae are an indicator of limited swimming abilities. The relatively small fin area of cryptobenthics is unlikely to be of benefit in acceleration, and the elongate bodies are likely to result in poor manoeuvrability (Webb 1984; Larouche et al. 2020). This interpretation corresponds well with the existing studies of cryptobenthic larvae, which have shown poor swimming abilities in gobies (Gobiidae Majoris et al. 2019), cardinalfishes (Apogonidae; although the genus Gymnapogon have strong-swimming larvae; Stobutzki and Bellwood 1997; Fisher and Hogan 2007; Leis et al. 2015), clingfishes (Gobiesocidae; Faria and Gonçalves 2010), some benthic blennies (Blenniidae; although larvae of the semi-pelagic nemophine blennies are relatively strong swimmers; Leis et al. 2007; Liu et al. 2018) and other ‘borderline’ cryptobenthic taxa (cf. Brandl et al. 2018), including the frogfishes (Antennariidae; Fisher and Hogan 2007).

As soon as they leave the reef, planktonic larvae have to contend with tidal and oceanic currents. If larvae drifted as passive particles, in 21 days (a relatively conservative planktonic larval duration; Depczynski and Bellwood 2006; Victor et al. 2010) they could disperse 245 km at 13.5 cm s−1 (an estimate of average flow rates near Lizard Island on the Great Barrier Reef; Frith et al. 1986), or 756 km at 1.5 km h−1 (42 cm s−1; an estimate of average flow rates throughout the Caribbean; Taylor and Hellberg 2003). While it may initially seem that larvae would need strong-swimming abilities to hold their position against these currents (Gerlach et al. 2007; Almany et al. 2017), multiple lines of evidence suggest that this is not the case in cryptobenthic larvae, which can maintain localised populations despite limited swimming abilities (Gerlach et al. 2007; Farnsworth et al. 2010; D’Aloia et al. 2015; Majoris et al. 2019).

Larvae with poor swimming abilities, including the cryptobenthics, likely use behavioural adaptations to remain close to reefs. One likely behaviour is flow-refuging (avoiding areas with rapid water movement), whether by remaining close to the benthos, where the friction of water passing over a solid surface reduces flow (Carpenter and Williams 1993; Fisher and Bellwood 2003) or by using down-current eddies created by objects from the size of coral heads, up to islands (Burgess et al. 2007). In tropical systems, fish larvae are known to use large down-current eddies to maintain position near islands or reefs (Leis 1986; Burgess et al. 2007). However, beyond identifying that some goby and apogonid larvae may be more common in deeper waters (e.g. Kingsford 2001; Fisher and Bellwood 2002), no studies have assessed whether larvae use fine-scale flow-refuging (e.g. staying near the benthos or benthic structures) to remain near coral reefs.

In temperate and subtropical systems around the world, cryptobenthic larvae in the Blenniidae, Bythitidae, Chaenopsidae, Gobiesocidae, Gobiidae, Labrisomidae, and Tripterygiidae are regularly found to remain close to the seabed in rocky reef habitats (e.g. Brogan 1994; Tilney et al. 1996; Beldade et al. 2006; Borges et al. 2007), and even using microtopographic features (e.g. rocks) to hold position and shelter from currents (Breitburg et al. 1995). Tropical confamilials of these temperate cryptobenthic larvae may also display this behaviour to maintain their position near their natal reefs. However, to confirm whether cryptobenthic larvae use this fine-scale benthic flow-refuging, we still require information regarding the distribution of cryptobenthic larvae near coral reefs.

If flow-refuging is the key to the retention of cryptobenthic larvae near reefs (Brandl et al. 2019a), they must have an ability to detect areas with reduced flow. We suspect that cryptobenthics may use their well-developed chemo- and mechanoreceptors—present even in early-stage larvae (Hu et al. 2019; Majoris et al. 2020)—to sense, and hold position in, low-flow environments. Beyond the sensory abilities to detect such areas, flow-refuging requires few morphological adaptations. This makes specialised structures obsolete, reducing energetically and developmentally costly pressures to extend the morphology of cryptobenthic larvae beyond basic ‘ancestral’ forms. It is not possible to establish if the larvae are morphologically constrained or if simple cryptobenthic larvae merely reflect a lack of pressure to diversify, but the functional consequences are the same: cryptobenthic fishes must compensate for poorly developed bodies and concomitant limited swimming abilities.

From an ecological perspective, the findings of our study further highlight the clear differences between cryptobenthic and large reef fishes. While the definition of a cryptobenthic reef fish presented in Brandl et al. (2018) is based on an arbitrary adult size threshold (i.e. 10% of species within a family not exceeding 50 mm in length), we see clear differences between the two groups (large reef fishes and cryptobenthics) in terms of their larval dynamics (Brandl et al. 2019a) and, from this study, their larval morphologies.

In the face of high predation rates on coral reefs, cryptobenthics rely heavily on the survival of their larvae. Indeed, larval phases represent up to 42% of the maximum total lifespans of some cryptobenthic species (Depczynski and Bellwood 2006). Survival in vast numbers (Brandl et al. 2019a, b) and settlement throughout the year (Lefèvre et al. 2016) reveal the exceptional ability of cryptobenthics to survive in the larval phase. While the morphological diversity seen in large reef fishes and adult cryptobenthics is exceptional, the simple, highly uniform, bodies of cryptobenthic larvae belie their unseen yet equally exceptional abilities, to survive and remain close to reefs.