Rethinking competence in marine life cycles: ontogenetic changes in the settlement response of sand dollar larvae exposed to turbulence

Complex life cycles have evolved independently numerous times in marine animals as well as in disparate algae. Such life histories typically involve a dispersive immature stage followed by settlement and metamorphosis to an adult stage on the sea floor. One commonality among animals exhibiting transitions of this type is that their larvae pass through a ‘precompetent’ period in which they do not respond to localized settlement cues, before entering a ‘competent’ period, during which cues can induce settlement. Despite the widespread existence of these two phases, relatively little is known about how larvae transition between them. Moreover, recent studies have blurred the distinction between the phases by demonstrating that fluid turbulence can spark precocious activation of competence. Here, we further investigate this phenomenon by exploring how larval interactions with turbulence change across ontogeny, focusing on offspring of the sand dollar Dendraster excentricus(Eschscholtz). Our data indicate that larvae exhibit increased responsiveness to turbulence as they get older. We also demonstrate a likely cost to precocious competence: the resulting juveniles are smaller. Based upon these findings, we outline a new, testable conception of competence that has the potential to reshape our understanding of larval dispersal and connectivity among marine populations.

1. Background

1.1 Settlement, precompetence and competence in marine life cycles

A common pattern among a wide diversity of multicellular marine organisms—including vertebrates, invertebrates and some algae—is a dispersive life-history stage in the plankton followed by settlement to the sea floor [1–3]. In marine invertebrates, these dispersing stages are typically microscopic larvae, which undergo what can be a quite dramatic metamorphosis as they settle and begin the juvenile phase.

One feature that many taxa with this type of complex life history have in common (even across kingdoms) is that the transition to the sea floor at settlement is generally irreversible. In this sense, the ‘decision’ of where and when to settle is a crucial one, and larvae (and corresponding dispersive stages in non-animals) are predicted to use a range of environmental signals to help them accurately discern an appropriate place to settle [4]. These features can be chemical (e.g. compounds released by conspecifics or a favoured food source) or physical (e.g. the sound of waves breaking on the shore, light intensity or substratum roughness). Complicating matters, most dispersive organisms likely employ a combination of such signals [4–9].

A second feature typical of marine larvae is that they tend to develop through a precompetent period, during which they often feed and grow, but are thought to be incapable of settling and completing metamorphosis until they attain competence to settle under appropriate conditions [2,10]. Although the timing of the onset of the precompetent period tends not to be clearly defined, it is generally assumed that the selective advantage of the precompetent period is to ensure that the individual is sufficiently well developed to survive and thrive on the sea floor. For example, sea stars with feeding larvae can develop for a month or longer before attaining competence, at which time their adhesive structures mature, allowing a firm initial attachment at settlement [11]. By contrast (and curiously so), multicellular algal propagules are not known to divide and grow during dispersal, and this constraint may explain their non-existent or short precompetent periods [1].

1.2 Turbulence: a new class of settlement cue

Recently, we discovered that the trajectory of the precompetent period for larvae of the purple sea urchin (Echinodermata: Echinoidea: Strongylocentrotidae) Strongylocentrotus purpuratus (Stimpson) is not ‘pre-programmed’ as has been envisioned [12–15]. Instead, nominally precompetent urchin larvae (defined operationally by the absence of response to a strong settlement cue) will become competent to settle after a mere 3 min exposure to fluid turbulence [16]. Because turbulence levels under breaking waves on rocky shores differ from those in other ocean habitats [17], one implication of this finding is that turbulence might operate as a broad (metre to kilometre) ‘habitat-scale’ cue that gives larvae a general indication that they are in the right neighbourhood for settlement, with settlement itself only coming after larvae subsequently detect more localized cues (i.e. centimetre to millimetre scales) at the sea floor. Specifically, our results with purple urchins raise the possibility that larvae in the field respond to turbulent conditions that are diagnostic of a particular habitat type, such as exposed intertidal shoals versus calmer embayments. A second, potentially even more important implication of turbulence as an inducer of competence is that the precompetent/competent boundary is malleable, in that it can be modified by short-term environmental conditions.

An additional, accompanying issue is that larvae during the precompetent period may be forced to trade off some aspect of juvenile growth and/or survival against the likelihood of encountering suitable juvenile habitat (see [18,19] for a similar argument in competent larvae). This scenario may apply well to echinoderms in particular, as much of juvenile ontogeny is already underway during the precompetent larval period [20,21]. Thus, if an echinoderm larva waits to settle until the nominal competent stage, then it will presumably be able to complete more of juvenile development before settlement, increasing its likelihood of survival on the sea floor. On the other hand, if the larva waits and grows but misses the opportunity to reach good habitat it may forgo the chance to settle at all [22].

1.3 Turbulence-induced settlement and ontogeny

A scenario of selective costs and benefits as outlined above leads us to consider a series of hypotheses concerning how larval behaviour might change during the precompetent period in response to turbulence exposure (figure 1). These hypotheses are based on the conjecture that younger larvae could benefit from further time in the plankton, and thus might be more selective than more advanced larvae in their response to information about habitat quality. The following hypotheses (as well as the accompanying null hypothesis) pertain to how turbulence might induce competence in larvae as a function of their age:

• H0: Larvae at a certain point during the precompetent stage begin to respond to turbulence, but the nature of this response does not change as the larvae continue to develop towards competence. In other words, any increases in settlement through ontogeny could arise simply because a greater fraction of larvae transition from precompetence to competence: there would be no difference between younger and older larvae in their reaction to turbulence (null hypothesis; figure 1a).

• H1: Larvae react more strongly to a given stimulus as they get older. According to this hypothesis, an increasing proportion of precompetent larvae respond to turbulence and become competent to settle, indicating increased responsiveness to turbulence as ontogeny proceeds (figure 1b).

• H2: Larvae respond to a lower level of stimulus as they get older. According to this hypothesis, older precompetent larvae respond to lower levels of turbulence than do younger precompetent larvae, indicating increased sensitivity to turbulence in older larvae, and increased selectivity of younger larvae to turbulence cues (figure 1c).

A corollary to these aforementioned scenarios concerns likely trade-offs associated with settling earlier in ontogeny. Specifically, we predict that precompetent larvae settling after exposure to turbulence do so at some cost relative to larvae that develop to nominal competence before settling.

Here we test each of these hypotheses and predictions in turn, using larvae of the Northeast Pacific sand dollar (Echinoidea: Clypeasteridae) Dendraster excentricus (Eschscholtz). We discuss the implications of our findings for life-history theory and for our understanding of the physiological ‘decision-making’ by larvae as they approach their competent period and subsequent commitment to settlement on the sea floor.

2. Material and methods

2.1 Experiment overview, definitions of terms and study species

In our prior work [16], we examined turbulence-induced precocious competence in advanced precompetent larvae of the purple sea urchin. Here, we extend our taxonomic scope to focus on larvae of the sand dollar D. excentricus. D. excentricus adults live gregariously in intertidal or shallow subtidal beds throughout much of the Northeast Pacific, in both exposed and protected locations [23]. Their planktotrophic larvae have a very wide temperature tolerance range—from at least 8°C through 24°C [24] (J. Hodin 2014, unpublished data)—and when fed ad libitum at approximately 20–22°C, they develop synchronously and reach competence (figure 2) between 12 and 14 days after fertilization. Under these conditions, invagination of the echinus rudiment—the first stage of development of the incipient juvenile (comparable with soft tissue stage i in purple urchins [21])—occurs in 6-arm larvae between days 2 and 3, and the first calcification of juvenile structures within the rudiment occurs in 8-arm larvae on about day 6 (data not shown).

Here we provide definitions of four relevant terms regarding life cycle transitions in the ocean. We define metamorphosis as a more-or-less radical morphological transformation between two multicellular life stages [25]. In sea urchins, the morphological transformation (larva to juvenile) begins several days or even weeks before the habitat shift (planktonic to benthic). Specifically, juvenile structures in sea urchins begin to form over a protracted period within and on the left side of the functioning, feeding larva (e.g. [21]). Therefore, we refer to sea urchin metamorphosis as the entire period from the first formation of juvenile structures (‘rudiment’) in the larva until the juvenile begins to feed on the sea floor. The most dramatic event of metamorphosis is settlement: the (in most cases irreversible) transition that the larva makes from the planktonic to the benthic zone. We define sand dollar settlement in more precise terms for our experiments in §2.6. Competence is generally defined operationally as the stage at which a larva first becomes responsive to localized settlement cues; before this stage, the larva is thus defined as precompetent. We will provisionally adopt these latter two definitions for the purpose of the paper, though our data suggest reconsideration of these concepts; we provide new definitions in the Discussion.

Our overall approach was to rear larvae through their feeding larval stage using standard methods [24], and then, in the days preceding the competent stage, subject them to a brief turbulence exposure of a specified intensity in a Taylor–Couette cell (described further below). Immediately after exposure, we transferred the larvae into excess potassium chloride (KCl) in seawater for 1 h, which induces competent D. excentricus larvae to settle [26] (and can, based on our prior work, induce settlement in turbulence-exposed, precompetent purple sea urchin larvae [16]), and we then quantified settlement as it relates to turbulence intensity. In a subset of trials, we also used an extract of adult sand dollar bed sand as a natural inducer, to ensure that the results we detected were not limited to induction with KCl. Lastly, we examined one potential consequence of precocious settlement by comparing sizes of juveniles that resulted either from settlement of competent larvae or from precompetent larvae that settled early following exposure to turbulence.

2.2 Source populations and collection

We collected adult D. excentricus adults by snorkelling on 27 May 2014 to a subtidal population (approx. 100 m offshore and 1.5 m below the surface at mean lower low water) approximately 30 m east of Municipal Pier 2 in downtown Monterey, CA, USA. We transported the sand dollars to Hopkins Marine Station (‘HMS’; Pacific Grove, CA, USA) and maintained them in sand in running seawater aquaria until spawning.

2.3 Larval culture

We obtained gametes by intracoelomic injection of (depending on the size of the adult) between 0.2 and 0.5 ml 0.55 M KCl in distilled water, and fertilized by standard methods [24]. For each larval culture, we combined approximately equal numbers of embryos from three crosses between one male and each of three females, with each combination resulting in greater than 90% fertilization success. We set up larval cultures in this way on three consecutive days (with different sets of adults each day) on 27–29 May 2014, and then again (with different sets of adults) on 22 August 2014.

In the latter spawning, we transferred a large portion of the unfertilized eggs from each female into individual 250 ml culture flasks containing 0.1 g l⁻¹ampicillin in 0.45 μm Millipore-filtered natural seawater (MFSW) with UV treatment. Storing eggs in this manner at ambient temperature (here 14°C) renders them fertilizable for a week or longer [27,28]. We stored sperm dry at 4°C. Then, on 24 and 25 August, we set up fertilizations as above from this same batch of sperm and stored eggs, following several washes in MFSW to remove any residual antibiotic. Thus, for the September 2014 experiments, we raised three batches of genetically comparable larvae of different ages in parallel.

We began feeding 4-arm plutei 1 day after fertilization. Initial larval densities were 0.5–1 larva ml⁻¹, which we lowered to 0.2–0.3 larva ml⁻¹ on day 5 or 6 (8-arm larvae; early rudiment stages). We fed larvae 5000 cells ml⁻¹ Rhodomonasspp. For the June 2014 experiments, we transported embryos and larvae on 29 May from HMS to Bodega Marine Laboratory (‘BML’; Bodega Bay, CA, USA), where we maintained them as above until experimental trials. For the September 2014 experiments, we similarly transported larvae on 27 August from HMS to BML.

At both HMS and BML, we cultured embryos and larvae in 0.45 μm MFSW with UV treatment, in gallon jars with mechanical stirring [24] at room temperature, which varied from 20 to 22°C.

2.4 Generating turbulence in the laboratory

To generate turbulence intensities (quantified in terms of the energy dissipation rate, in units of W kg⁻¹) ranging from those found in open ocean waters to those arising on wave-battered coasts, we employed a Taylor–Couette cell [29], an apparatus composed of two vertically oriented, coaxial cylinders separated by a 3.5 mm gap that contains seawater (described in greater detail in [16]). We held the stationary inner cylinder, and thus the water in the gap, at 19–21°C by means of a circulating water stream from a temperature-controlled water bath passing through the cylinder's interior. During operation, the outer cylinder rotated at a prescribed speed causing relative motion between the cylinders and thereby shearing the seawater between them. At rotation speeds employed for testing sand dollar larvae, the sheared flow was turbulent [16].

2.5 Larval exposure to turbulent shear

In early June (batches A and B) and early September 2014 (batch C) at BML, we exposed sand dollar larvae to prescribed levels of turbulence, using larvae that were between 7 and 14 days old. We first concentrated larvae by reverse filtration and selected 25–35 larvae into individual 125 ml glass beakers at a density of one larva per 3–4 ml MFSW. Larvae in these experiments were developing synchronously; we only eliminated larvae from consideration if they were either obviously delayed in development (less than 5% of the larvae; data not shown), or in the case of late-precompetent larvae, if they adhered to glassware at the time of selection or during transfer into exposure treatments. Otherwise, we chose larvae haphazardly. Then, we randomly assigned beakers to treatment and replicates, and began the exposures.

For each treatment run, we gently poured the entire contents of a 125 ml beaker into a finger bowl, and used a glass Pasteur pipette to introduce all of the larvae into 150 ml of MFSW within the Taylor–Couette cell. We then subjected the entire water volume within the apparatus to a specified level of turbulent shear for 3 min. Immediately following each treatment run, we gently poured the larvae within the Taylor–Couette cell into a 1 l glass beaker already containing approximately 100 ml of MSFW (to minimize additional stimulus to larvae during the pour), and we rinsed the Taylor–Couette cell twice with 19–21°C MFSW to capture any remaining larvae. We generally recovered greater than 90% of the larvae, all of which we used in ensuing settlement assays. We then rinsed the Taylor–Couette cell thoroughly with distilled water to ensure that no living larvae were transferred to subsequent trials, and we initiated the next trial.

In concert with the treatment exposures, we exposed randomly selected batches of larvae to one of two control conditions. For larvae assigned to ‘unmanipulated controls’, we poured the contents of the 125 ml beakers into finger bowls, as described above, and then immediately subjected them to settlement assays. For larvae assigned to ‘handling controls’ (0 W kg⁻¹ (= 0 r.p.m.) treatments), we treated them the same as larvae assigned to the various turbulence exposure treatments, except we did not activate the Taylor–Couette cell during the 3 min that larvae were within it, thereby controlling for manipulations associated with transfer of larvae into and out of the apparatus.

2.6 Quantification of settlement

Following turbulence treatments, we assessed settlement by exposing all larvae from a run to 8 ml of 40 mM excess KCl in MFSW for 1 h at 19–21°C in a single well of a non-tissue culture-treated 6-well plate, followed by recovery in 8 ml MFSW, as in previously employed protocols [16,26,30]. We scored a larva as settled after this 1 h exposure if skin had begun to withdraw from the tips of the larval skeletal rods, and we verified continued withdrawal of skin over the next several hours.

To confirm our scoring of settlement in the KCl exposures, and to ensure that the settled larvae would complete metamorphosis normally, we re-examined all settled larvae following an additional 12–24 h of recovery in MFSW at 19–21°C. This evaluation verified that greater than 95% of all larvae had been scored correctly the previous day, with adoption by 12–24 h of the definitive juvenile morphology, with emergent and active tube feet and spines. We detected no post-settlement mortality. Larvae from all treatments that we had scored as not settled also recovered normally, were swimming within 1 h of recovery, and remained swimming after 24 h with no apparent ill effects of the treatment.

2.7 Natural inducer experiments

In one set of trials examining whether turbulence effects were associated exclusively with KCl induction or would manifest more broadly (e.g. with a natural inducer), we used an extract of surface sand that we collected in August 2014 from a high intertidal sand dollar bed in Alki Beach, Seattle, WA, USA (approx. 1.5 vertical feet above the mean lower low water line), and which we then froze at −20°C in 50 ml aliquots the same day. Our unpublished observations suggest that sand collected from intertidal sand dollar beds of D. excentricus (as found in WA) is a more potent settlement inducer than sand from subtidal populations (as found in CA), regardless of the geographical source of sand dollar larvae tested.

To prepare the sand cue, we thawed 25 ml of the Alki sand, shook it vigorously with 25 ml of MFSW for 30 s, allowed the large sand particulates to settle out for 10 s, and then decanted off the opaque supernatant into a new tube. We then diluted this supernatant 1 : 1 with MFSW, resulting in what we termed a 100% sand extract. We used this extract for settlement tests at various dilutions in MFSW within a few hours of thawing. We conducted sand extract exposure trials in 6-well plates in which the remaining sediment suspended in the sand extract had been allowed to settle out for at least 30 min, rendering the water above reasonably transparent.

For our natural inducer trials, we transferred 25 larvae into the Taylor–Couette cell, which we activated at approximately 6 W kg⁻¹ (500 r.p.m.) for 3 min. Upon recovery of larvae from the Taylor–Couette cell, we immediately exposed them for 60 min to 8 ml of a 50% sand extract in a single well of a 6-well plate, which we then diluted to 35% strength (by the addition of 3.5 ml MFSW) for eight additional hours to allow for the completion of settlement in partially settled larvae, while limiting de novo settlement during this protracted exposure period. After the 9 h total exposure, we scored a larva as ‘settled’ only if its skin was fully withdrawn from its larval arms with juvenile rudiment everted. We compared three replicates of these 6 W kg⁻¹-exposed larvae to three replicates of unmanipulated control larvae, treated in parallel as above.

2.8 Juvenile size

In early September at BML, we gathered all juveniles that had settled in 40 mM excess KCl following 3 min, Taylor–Couette cell exposure trials at approximately 2 W kg⁻¹ (290 r.p.m.) on day 11 (24 August fertilization) and photographed them after 24 h at 20× with a Leica (Solms, Germany) MC170 HD camera mounted on a Leica M125 dissecting microscope, using Leica Application Suite software (v. 4.5). We did the same for larvae that had settled following 1 h of 40 mM excess KCl exposure on day 14 (i.e. nominally competent larvae that had not been spun in the Taylor–Couette cell; 22 August fertilization). We measured the test diameters in two orthogonal dimensions of these juveniles from the photos using ImageJ (v. 1.49D). We calculated the test area for each juvenile, approximating each test as an ellipse in cross section. We captured additional images of larvae (figure 2) using the same set-up mounted on a Leica DM1000 compound microscope.

2.9 Statistical analyses

For the ontogeny comparisons from days 7 to 12 at 6 W kg⁻¹ (500 r.p.m.) versus unmanipulated controls, we used an ANCOVA, with developmental day as a covariate, and a Tukey's HSD test for post hoc comparisons. We employed a single batch of larvae (batch A) for this experiment.

Using the dose–response data on our three larval batches (‘A’, ‘B’ and ‘C’), we evaluated the hypotheses outlined in figure 1 as follows.

To quantify potential shifts in larval ‘responsiveness’ (H1, figure 1b), we determined the maximum (max.) and minimum (min.) proportion of larvae that settled in response to exposure to KCl following any turbulence treatment (including handling controls: 0 W kg⁻¹, 0 r.p.m.) within each replicate dose response from days 9–11 (N=15). We considered this minimum proportion as the background settlement rate (i.e. the proportion of larvae that were competent) within that set. To calculate the maximum proportion of larvae that became competent and settled in response to turbulence, we performed the following calculation for each replicate dose response:

The resulting value represents the fraction of formally precompetent larvae (i.e. the quantity (1 − min.)) that became competent to settle as a consequence of turbulence exposure. Then, to assess statistical confidence in the relationship between maximum proportion of larvae that became competent to settle in response to turbulence and developmental day (9–11), we used a mixed linear model with batch (A–C) as a random effect. In the model, we allowed batch effects to vary both the intercept and slope of the regression, and used an unpaired t-test (with Satterthwaite approximations to degrees of freedom) to assess whether developmental day is a significant predictor of the regression.

To quantify potential shifts in ‘sensitivity’ (H2, figure 1c), we used a generalized mixed linear model to compare the larval response to turbulence on days 9 and 11. We began with a basic logistic model including only turbulence intensity and developmental day (Akaike information criterion (AIC) score: 542.3), and then subsequently added the following additional variables in turn to arrive at the best-supported model based on AIC score [31]: random effects (batch and runs nested within batch; AIC score: 481.8), and a quadratic term for turbulence intensity (to approximate the apparent drop off at high turbulence intensities; AIC score: 465.7). The model failed to improve further when we added either an additional variable for the interaction between turbulence intensity and age (AIC score: 467.1), or for an interaction between the quadratic term and age (AIC score: 466.7). As expected, the model also failed to improve with inclusion of both of these interaction terms (AIC score: 468.3).

In the best-supported model, the quadratic variable was also highly significant (p<0.001). Therefore, for further analysis, we chose this latter model, which included the following variables: turbulence intensity, developmental day, the quadratic term for turbulence intensity and random effects.

Using this model, we generated best curve fits, producing mean response curves on days 9 and 11 with ±95% confidence intervals (CIs). To determine if these inflection point estimates on days 9 and 11 were statistically different, we ran a non-parametric bootstrap algorithm (with random effects as above) to generate 10 000 inflection point estimates. We then calculated the 95% confidence range for these inflection point estimates on days 9 and 11 using the boot.ci function in R. This same function allowed us to identify the maximum confidence range (X) of these inflection point estimates on days 9 versus 11 that showed no overlap. We report the p value for these inflection point estimates as 1−X.

For the natural inducer data, we used unpaired t-tests to test for differences between turbulence-exposed larvae (6 W kg⁻¹ (500 r.p.m.)) and unmanipulated controls. We did the same for the corresponding KCl induction data.

For the juvenile size data, we compared the mean test areas of day 12 juveniles (with turbulence exposure at 2 W kg⁻¹ (290 r.p.m.)) with the day 15 juveniles (without exposure). Because we did not replicate the turbulence exposures or post-settlement recovery conditions, we tested for bias in the data by performing 10 000 comparisons between sets of 10 randomly chosen juveniles from each treatment, all 10 000 of which had p<0.005 (not shown).

For all data, we assessed normality using q–q plots. We performed all statistical analyses using R (v. 3.1.0).

3. Results

3.1 Ontogenetic scope of the response to turbulence

Sand dollar larvae at days 7–12 after fertilization settled approximately 2 days earlier and at higher rates overall if we first exposed them to intense turbulence in a Taylor–Couette cell (figure 3; ANCOVA, F_1,10=31.25, p<0.001). As in purple sea urchins [16], turbulence exposure does not itself induce settlement in sand dollars (data not shown). Instead, larvae exposed to turbulence must be presented with a localized cue (e.g. excess potassium chloride in seawater or a natural inducer) before they settle and transform to the juvenile stage.

The data in figure 3 thus demonstrate an increase in turbulence-induced settlement as ontogeny proceeds, but this outcome could emerge in several ways (figure 1); our analysis of the dose response data, as explained above and detailed below, allows us to distinguish among these different possibilities.

3.2 Changes in turbulence-induced settlement through the precompetent period

We reared several independent batches of larvae until they were between 7 and 11 days old, and then exposed them to turbulence intensities spanning conditions characteristic of quiescent waters, to calmer offshore flows, to violently breaking waves (i.e. turbulent energy dissipation rates from 0 to in excess of 10 W kg⁻¹ (see [16,17])). Owing to substantial variation among batches in how larvae responded to turbulence across ontogeny (figure 4), we undertook a series of analyses to test H1 and H2 (figure 1) using our data from the three batches considered together, as we detail in the following sections.

3.2.1 Shifts in larval responsiveness to turbulence (H1)

When we used these data to test H1, our results suggest that larvae indeed exhibited increased ‘responsiveness’ to turbulence exposure as ontogeny proceeds (figure 1b). In particular, we see an increase as larvae get older in the peak proportion of nominally precompetent larvae that nevertheless settled across all turbulence intensities when subsequently given a settlement inducer (figure 5). In our analysis of these data, age of larvae was a significant predictor of the regression (t-test; t_7.9=3.251; p<0.02), thus supporting H1.

3.2.2 Shifts in larval sensitivity to turbulence (H2)

Our next analysis tested whether larvae became more sensitive to a given level of turbulence as ontogeny proceeds (H2; see figure 1c), by examining the position of inflection points. For this analysis, we fit the combined settlement data for days 9 and 11 (i.e. figure 4a–c) to a logistic curve using a mixed linear model (figure 6a). The model with the most support was one with a quadratic term included to account for the apparent decline in settlement seen at higher turbulence intensities in the majority of runs shown in figure 4 (see Material and methods for details on and AIC scores of the various model combinations). In the logistic-quadratic regression, the slope (±s.e.m.) was 0.43 (±0.06; p<0.001) log odds of settling per unit increase of W kg⁻¹, the quadratic term was −0.02 (±0.00; p<0.001) log odds of settling per squared unit increase of W kg⁻¹, and the intercepts (±s.e.m.) were as follows: day 9, −3.35 (±0.30; p<0.001); day 11, −1.61 (±0.37; p<0.001) log odds of settling.

Using this best-supported model, we generated best-fit curves ±95% CIs for the data from days 9 and 11, and from a non-parametric bootstrap of the data, calculated and compared the inflection points (figure 6). The 95% CIs on the inflection point for the response on day 11 (mean: 2.2 W kg⁻¹; 95% confidence range: 0.7–3.1 W kg⁻¹) showed slight overlap with the range on day 9 (mean: 4.5 W kg⁻¹; 95% confidence range: 2.7–7.1 W kg⁻¹), as seen graphically in figure 6b. These data indicate a trend (p=0.085) in which nominally precompetent larvae become more sensitive to turbulence as ontogeny proceeds, thus lending borderline support to H2.