Montgomery et al. (Reference Montgomery, Hamel and Mercier2019) demonstrated through a meta-analysis that the current idea that planktotrophic larvae of marine invertebrates swim faster than their non-feeding counterparts should be challenged. They demonstrated using quantitative analysis that the ciliated lecithotrophic larvae of sessile organisms (especially sponges) swim faster than those of non-sessile animals (polychaetes and echinoderms). The amount of data collected by the authors is impressive and the outcomes of this dataset are also noteworthy. However, the fact that the ciliated larvae of sponges swim faster than the other phyla caught our attention. We have studied larvae of several species of sponges in the Mediterranean Sea and the Western Atlantic Ocean and the results shown in the work are very discordant to our own experiences and to the majority of literature about sponge larval behaviour. We, therefore, decided to write this reply to expand a little the discussion about the speed of the larvae of sponges based on Montgomery et al.'s (Reference Montgomery, Hamel and Mercier2019) findings.

Montgomery et al. (Reference Montgomery, Hamel and Mercier2019) provided a supplementary table with the dataset used for their analysis. In the first step, we checked the original publications indicated by Montgomery et al. (Reference Montgomery, Hamel and Mercier2019) for the phylum Porifera to compile their dataset, because we found the values too high. Our review indicated three inconsistencies. The speed of Tedania ignis presented in their supplementary table (9 mm s⁻¹) was converted incorrectly (the original speed was an average of 0.098 cm s⁻¹ in Maldonado & Young, Reference Maldonado and Young1996). A similar situation was observed for Coscinoderma mattewsi: they report that the speed of the larvae of the species is 45 mm s⁻¹, while the original data presented is 0.45 cm s⁻¹ (Wahab et al., Reference Wahab, de Nys and Whalan2011). Finally, the speed of the larvae of Haliclona tubifera was indicated as 2.8 mm s⁻¹, while the original speed reported by Woollacott (Reference Woollacott1990) is 3.6 mm s⁻¹. We did not check the original data for the other phyla, though, as we focused only in Porifera.

These three different entries in the original dataset reduced the average ± SD of the swimming speed in Porifera from 11.01 ± 14.43 mm s⁻¹ to 6.68 ± 9.21 mm s⁻¹ (Figure 1A). Nonetheless, the result of a one-way ANOVA taking in consideration only the speed of the larvae among the phyla with this corrected dataset still indicated that sponges presented the fastest larvae (P < 0.0001; Figure 1A), although a reduction in the difference among the means of the phyla could be observed (Figure 1C,D). In addition to these mistakes when converting the values of the speed of the sponge larvae, we also added five entries to the original dataset: Haliclona cf. permolis (speed = 0.4 mm s⁻¹; Elliot et al., Reference Elliot, MacDonald and Leys2004); Sigmadocia caerulea (speed = 4 mm s⁻¹; Maldonado et al., Reference Maldonado, George, Young and Vaquerizo1997); Halichondria melanodocia (speed = 2.2 mm s⁻¹; Woollacott, Reference Woollacott1990); Haliclona tubifera (speed = 1.7 mm s⁻¹; Maldonado & Young, Reference Maldonado and Young1996) [Demospongiae] and Sycon coactum (speed = 0.09 mm s⁻¹; Elliot et al., Reference Elliot, MacDonald and Leys2004) [Calcarea]. This complemented dataset decreased even further the average of Porifera larvae speed to 5.11 ± 9.21 mm s⁻¹ (Figure 1B), but it still was significantly different than the other phyla (P = 0.0004). However, with the complemented dataset, Porifera is no longer different than cnidarian and bryozoan larvae (Figure 1E).

Fig. 1. Effects of the changes of the data about the speed of the larvae of Porifera on the dataset of Montgomery et al. (Reference Montgomery, Hamel and Mercier2019). (A) Box-plot showing the variation on the speed of the ciliated larvae of the different phyla. Each open blue dot indicates a single entry (mostly species), the box indicates upper and lower quartiles, horizontal line indicates the median. Outliers are represented as filled blue dots. The red diamond indicates the average in each phylum. Note that the correction of three values in the original dataset reduced the average of the speed of Porifera (PoriferaMOD). (B) Box-plot (as in A) comparing the data related to Porifera from the original, corrected and corrected + 5 new data of larval speed of sponges (see text). (C–E) 95% family-wise confidence level of the Tukey post-hoc test (the difference is significant when the lines do not cross zero): (C) Original dataset; (D) Corrected dataset; (E) Corrected dataset with the additional data (see text).

Secondly and maybe more important, is related to the swimming behaviour of the sponge larvae. It is clear in the work how the authors obtained the speed of the larvae used in their meta-analysis (‘All speeds used in the analysis were absolute speeds and were not standardized to body size. Wherever possible, horizontal swimming speeds (mm s⁻¹) were used, but these data were not always reported in the literature. When multiple swimming directions were reported, the fastest reported speed was used (e.g. among upward and downward swimming speeds)’). However, it is well known that sponge larvae do not swim constantly and the reported speeds are likely bursts of the larvae that cease soon after it starts. For instance, Uriz et al. (Reference Uriz, Turon and Mariani2008) reported that ‘[larvae of Scopalina lophyropoda] swam erratically and alternated very short periods of relatively fast (up to 15 mm s⁻¹) directional swimming with other periods of slow (1–3 mm s⁻¹) erratic swimming’ and Wahab et al. (Reference Wahab, de Nys and Whalan2011) observed that the highest speed of the larvae of C. mattewsi (4.5 mm s⁻¹) was observed only after larval release, subsequently slowing down. The most comprehensive investigation of sponge larvae behaviour, comprising more than 20 taxa, did not measure quantitatively the speed of these propagules. Nonetheless, almost all larvae in that study were considered to swim weakly, even under light stimuli (Mariani et al., Reference Mariani, Uriz and Turon2005). Montgomery et al.'s (Reference Montgomery, Hamel and Mercier2019) data suggest that sponges have the fastest ciliated propagules among the investigated group, and such strong swimming behaviour should have not passed unattended by Mariani et al. (Reference Mariani, Uriz and Turon2005). In addition, it is important to note that most swimming speeds for sponge larvae were recorded under laboratory conditions, while in the wild, larvae when seen are reported to act merely as passive particles drifting with the current until they reach proximity to their settlement surfaces and start actively swimming (Maldonado, Reference Maldonado2006).

We understand the challenges that come with analysing such a diverse literature and some generalizations have to be made to obtain the big picture, compromising the idiosyncrasies of each system. However, we think that a discussion about the behaviour of these larvae is still necessary, mostly because of the implications this could have for larval dispersal modelling in the future. As mentioned before, most of the sponge larvae swim intermittently, and the majority of their movements are rotating around their own axis and describing small circles clock or counter-clockwise (Maldonado, Reference Maldonado2006). A constant and strong swimming would be highly energetically costly, which is not in accordance with the energetic reservoir of the sponge larvae. Due to their reduced size, the sponge larva depends entirely on its ciliary movements to swim and at the same time when the cilia stop, the larva is stuck in the water (Maldonado, Reference Maldonado2006). Although Montgomery et al. (Reference Montgomery, Hamel and Mercier2019) state that the shape of the larvae of Porifera is mostly ‘spheroid’ (generating less ‘drag’) and that sponge larvae lack mineralized parts, the truth is that the majority of sponge larvae are bullet-shaped and, especially in Demospongiae, many have spicules made of silica. Therefore, the main characteristics used by Montgomery et al. (Reference Montgomery, Hamel and Mercier2019) to explain this unexpected result do not seem to be actual and widespread characteristics of Porifera larvae.

Finally, Montgomery et al. (Reference Montgomery, Hamel and Mercier2019) state ‘Clearly there is something unique about the swimming mechanics of sponge larvae that deserves further attention’. Although we agree with the need for further investigations on the swimming mechanics of sponge larvae, we believe that the discussion presented in their work is misleading, as the large differences pointed out by them are likely artefacts of the data obtained for the study. The number of entries for the phylum Porifera is limited (only 11 entries) if compared to the other phyla (except Bryozoa) and it might be that the results obtained are not the real average speed of sponges (as showed above).

As models to predict the dispersal of marine larvae usually consider the propagule speed, hydrographic conditions and planktonic duration, it is necessary to have in mind the different types of behaviour (Mariani et al., Reference Mariani, Uriz and Turon2005). The absence of such a concern in their study compromises the results and the discussion of Montgomery et al. (Reference Montgomery, Hamel and Mercier2019) and we fear that their conclusions might be propagated without proper discussion. The objective of meta-analysis studies is always to provoke discussions and to point further research topics aiming to refine the knowledge in a given field. The work of Montgomery et al. (Reference Montgomery, Hamel and Mercier2019) indicates that more data about sponge larval speed are necessary and, in addition, new ways to ‘normalize’ the speed of larvae in the investigated phyla against their behaviour will be important to make this comparison more sound.