, Volume 184, Issue 1, pp 127–137 | Cite as

Application of diet theory reveals context-dependent foraging preferences in an herbivorous coral reef fish

Behavioral ecology – original research


Dietary preferences of grazers can drive spatial variability in top-down control of autotroph communities, because diet composition may depend on the relative availability of autotroph species. On Caribbean coral reefs, parrotfish grazing is important in limiting macroalgae, but parrotfish dietary preferences are poorly understood. We applied diet-switching analysis to quantify the foraging preferences of the redband parrotfish (Sparisoma aurofrenatum). At 12 Caribbean reefs, we observed 293 redband parrotfish in 5-min feeding bouts and quantified relative benthic algal cover using quadrats. The primary diet items were macroalgal turfs, Halimeda spp., and foliose macroalgae (primarily Dictyota spp. and Lobophora spp.). When each resource was evaluated independently, there were only weak relationships between resource cover and foraging effort (number of bites taken). Electivity for each resource also showed no pattern, varying from positive (preference for the resource) to negative (avoidance) across sites. However, a diet-switching analysis consisting of pairwise comparisons of relative cover and relative foraging effort revealed clearer patterns: parrotfish (a) preferred Halimeda and macroalgal turfs equally, and those two resources were highly substitutable; (b) preferred Halimeda to foliose macroalgae, but those two resources were complementary; and (c) also preferred turf to foliose macroalgae, and those resources were also complementary. Thus parrotfish grazing rates depend on relative, not absolute, abundance of macroalgal types, due to differences in substitutability among resources. Application of similar analyses may help predict potential changes in foraging effort of benthic grazers over spatial gradients that could inform expectations for reef recovery following the protection of herbivore populations.


Diet choice Diet switching Resource complementarity Foraging theory Sparisoma aurofrenatum 


Grazing herbivores are important to the structure and dynamics of many high-productivity ecosystems (Byrnes et al. 2006; Olff and Ritchie 1998). This is particularly true on Caribbean coral reefs, where there is a trend towards dominance by macroalgae (Burkepile and Hay 2011; Mumby 2009; Norström et al. 2009; Rotjan and Dimond 2010). The decline in corals and rise in macroalgae has been linked to declines in the diversity of reef herbivores, particularly parrotfish (Scaridae) due to overharvesting (Burkepile and Hay 2009; Cheal et al. 2010; Hughes et al. 2003; but see Loh et al. 2015). Parrotfish and other herbivorous reef fishes can limit macroalgal growth and promote reef resilience (Bellwood et al. 2004; Hughes et al. 2007; Burkepile and Hay 2010; Cheal et al. 2010). However, the broad-brush focus on ‘grazers’ and ‘macroalgae’ as monolithic functional groups can obscure subtle but potentially important details of trophic interactions on coral reefs and other high-productivity ecosystems (Burkepile and Hay 2010; Pawlik et al. 2016).

Diet choice by grazers can influence their impact on the ecosystem (Miller et al. 2011; Suding et al. 2004). For example, size- and species-specific foraging preferences of grazing zooplankton can produce orders-of-magnitude variation in primary productivity among freshwater lakes with similar nutrient loading (Carpenter and Kitchell 1984). In both forest and grassland ecosystems, selective grazing by ungulates typically shift plant communities towards dominance by unpalatable species, unless herding or migratory behavior by the ungulates constrains their ability to forage selectively (reviewed by Augustin and McNaughton 1998). On coral reefs, macroalgal selectivity by herbivorous fishes can switch from positive to negative between reefs, and diet composition is not easily predicted by the absolute abundance of a resource (Bruggemann et al. 1994a; Francini-Filho et al. 2010). As a result we cannot reliably predict the expected diet of grazing reef fish on a particular reef, nor how the grazing community will respond to increased macroalgal abundances on degraded reefs (e.g., Burkepile and Hay 2010, 2011).

A key aspect of diet breadth (i.e., how many prey items are included in the diet) and diet choice (i.e., the relative foraging effort for each item) is the nutritional similarity between resources (Oaten and Murdoch 1975; Raubenheimer and Simpson 2003; Simpson and Raubenheimer 2001; van Leeuwen et al. 2013; Visser and Fiksen 2013). One formulation of diet theory, the ‘geometric framework’, predicts that the stoichiometric needs of grazers determine how they select among available resources (Raubenheimer and Simpson 2003). Animals should allocate foraging effort among resources to regulate their relative intake of specific nutritional components such as proteins, carbohydrates, and lipids, and micronutrients such as vitamins and minerals, to achieve a specific nutritional goal (Simpson et al. 2004). However, the predictions of the geometric framework have not been applied to studies of herbivorous coral reef fish diets (Clements et al. 2009).

Coral reefs support a wide diversity of macroalgae species that vary in nutritional value and micronutrient composition (Bruggemann et al. 1994b). When the relative abundance of different macroalgae changes, the nutritional and micronutrient seascape inhabited by grazing fishes also changes. These changes affect what grazing fish choose to eat (Abrams and Matsuda 2003; van Leeuwen et al. 2013), although other factors such as physical and chemical defenses can also influence foraging decisions (Hay et al. 1994; Loh and Pawlik 2014). If grazing reef fish forage according to the geometric framework (a nutritional approach), we can make predictions about the substitutability of different food resources based on the relationship between diet and the abundance of all available food resources. In general, two resources that are nutritionally similar (i.e., substitutable) should be consumed at rates proportional to their relative abundances, whereas two nutritionally dissimilar (i.e., complementary) resources should be consumed at rates based on the current nutritional requirements of the grazer, independent of the relative abundance of the resources (Raubenheimer and Simpson 2003). In other words, as two resources become nutritionally complementary it becomes more necessary for a specified grazer to keep both resources in their diets regardless of resource scarcity.

We can evaluate the relative substitutability vs. complementarity of diet items in a quantitatively rigorous fashion by applying models from prey switching theory (Murdoch 1969; Oaten and Murdoch 1975; Abrams 1990; Abrams and Matsuda 2003; Van Leeuwen et al. 2013). The range of possible foraging patterns on a pair of resources is best illustrated by plotting the log ratio of relative consumption of the two resources vs. the log ratio of their relative abundances (Fig. 1; using the logarithm linearizes the relationship between the two ratios across multiple orders of magnitude). Purely substitutable (i.e., nutritionally equivalent) resources will be consumed in proportion to their relative abundance because there is no advantage to choosing one substitutable resource over another (Raubenheimer and Simpson 2003; van Leeuwen et al. 2013). This will produce a curve with slope = 1 on the consumption–abundance plot (Fig. 1a). Additionally, among two perfectly substitutable resources one of the two may be preferred (and experience proportionately higher grazing) if, for example, it has higher nutritional content per unit mass. This would be reflected by an intercept ≠ 0 on the vertical axis; if the first resource is preferred when the two are at equal abundance, then the intercept would be >0.
Fig. 1

Dietary response to food resources varying in similarity. The vertical axis represents the log ratio of consumption between two food items, R1 and R2. The horizontal axis represents the log ratio of abundance between the two food items. a Two resources with similar nutritional profiles (i.e., substitutable). Each resource is consumed as it is discovered in the environment, resulting in a slope of 1. b Two substitutable resources with similar nutritional profiles, but the forager switches to focus on the more abundant of the two resources, producing a slope >1. At the extremes of the abundance ratio the slope curves back towards a 1:1 relationship due to diminished switching when one resource is extremely rare. Note the vertical axis intercept is >0, indicating a preference for R1 when both resources are in equal abundance. c Two complementary resources that provide unique benefits to the grazer. Here the slope is 0 and intercept is <0, indicating that there is no similarity between R1 and R2 but R2 is preferred when the two are in equal abundance. As either resource becomes rare the grazer must expend energy to locate and consume that rare resource, because it provides necessary nutrients not available via the more abundant resource

In some cases, as one resource becomes more abundant relative to substitutable alternatives it is favorable for the grazer to focus effort on that resource at a greater-than-proportional rate; i.e., switching to the more abundant resource at the expense of less-abundant resources (Murdoch 1969). This leads to a curve with slope >1 on the consumption–abundance plot (Fig. 1b; van Leeuwen et al. 2013). Note that at the extremes of the plot, when one resource is vastly more abundant than the other, it is sometimes not practical for the forager to maintain a correspondingly high preference ratio and the slope of the curve flattens; as a result, this curve sometimes takes on an S-shape.

Finally, non-substitutable complementary resources are not consumed based on their relative abundance because the physiological stoichiometry of the grazer requires a consistent ratio of the two diet items (Abrams 1990; Raubenheimer and Simpson 2003; Behmer and Joern 2008). Therefore, relative foraging effort for one resource should actually increase as it becomes more rare (Fig. 1c; Raubenheimer and Simpson 2003). This would produce a curve with slope <1. The slope would depend on the degree of nutritional difference between the resources; completely complementary resources would have a slope of 0 (Abrams 1993; Rindorf et al. 2006; van Leeuwen et al. 2013). By examining the relative foraging effort on alternative prey items across a wide range of relative abundances, it is possible to deduce the degree of their substitutability (or complementarity), potentially explaining otherwise obscure patterns of preference for individual resources and allowing predictions for grazing patterns on altered landscapes.

We applied diet theory to understand the foraging decisions of a common grazing parrotfish on Caribbean coral reefs. We took an observational approach, recording foraging behavior on multiple reefs that varied widely in relative abundance of potential macroalgal resources. By examining relative consumption across spatial gradients in resources, we were able to discern the relative value of resources to grazers and predict how fluctuations in resource abundance would affect diets, ultimately shaping coral reef community dynamics. Our study represents a case study in the application of dietary switching theory, because we were able to make observations across a wide range of resource abundances.

Materials and methods

Study organism

Parrotfishes (family Scaridae) are generalist grazers that are abundant on coral reefs across the Caribbean. They possess a number of specialized adaptations for grazing on macroalgae that are defended by secondary compounds (e.g., phlorotannins) or inclusion of calcium carbonate, including fused front teeth for scraping, grinding pharyngeal jaws, and a basic gut pH (Crossman et al. 2005; Mumby 2009; Targett and Arnold 1998). These adaptations help parrotfishes access a variety of foods, ranging from relatively protein-rich and undefended macroalgal turfs and associated detrital matter (Bruggemann et al. 1994a; Crossman et al. 2005; Targett and Targett 1990) to more carbohydrate-rich, chemically defended macroalgae including Dictyota spp., Lobophora spp., and Halimeda spp. (Burkepile and Hay 2010; Catano et al. 2015). In general, parrotfishes are more efficient at assimilating proteins and lipids (>90% assimilation efficiency) than carbohydrates (<70% assimilation efficiency; Crossman et al. 2005). As a result, parrotfishes can be expected to focus foraging effort on protein-rich resources, including macroalgal turfs. Parrotfishes will opportunistically consume other resources when they become available, such as palatable, undefended marine sponges (Dunlap and Pawlik 1996); however, such resources are sufficiently rare (Loh and Pawlik 2014) that analyzing foraging preferences for them is impractical.

Redband parrotfish (S. aurofrenatum) are a particularly appropriate organism in which to test diet-switching theory because they have a broad dietary range and are abundant across the Caribbean on reefs that vary widely in macroalgal composition (Loh and Pawlik 2014). Adult redband parrotfish occupy feeding territories approximately 100 m2 in size (Catano et al. 2015). In general, smaller territory sizes are linked to greater resource quality; territories expand as resource quality declines (Mumby and Wabnitz 2002; Catano et al. 2015). These territories generally encompass a large enough patch of reef to minimize any variability in the cover of potential food resources at a given locality (Harris et al. 2015).

Like other parrotfishes, redband parrotfish are protogynous hermaphrodites; they begin their life in the initial phase (most if not all initial-phase fish are female) and older, larger individuals change sex to become terminal phase males. Initial and terminal phases are readily distinguished by coloration.

Study sites

During 2012–2013, we observed redband parrotfish feeding at 13 reef sites spread across the Eastern Yucatan peninsula (7 sites, May 2012: Cancun (Isla Mujeres), Cozumel (Paraiso Bajo), South Cozumel, Akumal, North Banco Chinchorro, Mid Banco Chinchorro, and South Banco Chinchorro), and the Southern Bahamas (6 sites, July 2013: Danger Reef (Exumas), Little Inagua, Great Inagua (Charmicle Bay), Aklins Island, Mayaguana, and Concepcion). The study sites consisted of either spur and groove reefs or patch reefs; map and site details are given in Online Resources 1 and 2.

Field foraging observations

Across all study sites, we followed a total of 293 redband parrotfish for 5-min intervals between 9 A.M. and 4 P.M. using SCUBA at depths ranging from 4.5 to 21 m. The number of fish per site ranged from 15 to 68 individuals depending on fish density and number of dives (Online Resource 2). During each 5-min interval, we recorded the sexual phase, visually estimated the total length of each focal fish to the nearest cm (length ranged from 8 to 25 cm) and recorded the number of bites taken on each type of food resource. Bites taken by redband parrotfish were used as a proxy for resource consumption because a bite represents a unit of foraging effort.

For foraging observations, we used food resource categories similar to those used by Burkepile and Hay (2011), which were in turn based on the functional groupings developed by Steneck and Dethier (1994). The food categories included macroalgal ‘turf’, defined as filamentous or articulated coralline algae <3 cm in length, including any other macroalgae, detrital matter, or crustose coralline algae associated with the turf. Clearly distinguishable bites on crustose coralline algae alone were counted separately. ‘Foliose’ macroalgae was defined as non-filamentous, non-calcareous macroalgae >3 cm in length; in our observations this category almost exclusively consisted of Dictyota spp. and Lobophora spp. The other major macroalgal category was Halimeda spp., which was both common and distinctive enough to be a separate category (hereafter referred to simply as Halimeda). We also enumerated bites on ‘other’ diet items, including bites taken in the water column, sponges, corals, fecal matter, gorgonians and sand, but these represented <2% of the total.

Resource cover was recorded at each study site using a point-intercept method. We used a 1 × 1 m quadrat frame containing an equally spaced 5 × 5 string grid forming 25 individual intersection points; in each quadrat we recorded the identity of organisms underneath each intersection point, using the same categories as in the foraging observations. Each study site was sampled between 7 and 25 times (number of quadrats), depending on the number of dives available at each site.

The majority of all observed bites were on items in the turf, foliose, or Halimeda categories (see “Results”), so our analysis focused exclusively on those three diet categories.

Data analysis

We first examined diet choice using Vanderploeg and Scavia’s Relativized Electivity Index (Vanderploeg and Scavia 1979; Lechowicz 1982). This index is calculated by first finding the selectivity coefficient for diet item i, W i :
$$W_{i} = \frac{{r_{i} /p_{i} }}{{\mathop \sum \nolimits r_{i} /p_{i} }},$$
where r i is the proportion of bites taken in each category i and p i is the proportional cover of each category i. The index W i ranges from 0 (total avoidance) to 1 (total preference). The relativized index is then
$$E_{i} = \frac{{W_{i} - 1/n}}{{W_{i} + 1/n}},$$
where n represents the number of diet categories available (in our case n = 3). The values of E i range from −1 (total avoidance) to 1 (total preference).

Next we tested for a direct effect of resource abundance (proportional cover of each food resource group) on bite rate using a generalized linear mixed model (GLMM; logit link, Poisson error distribution). We used site as a random effect to account for potential variation in foraging effort due to site-specific factors (e.g., depth, swell, light conditions).

We also tested for the effects of relative resource cover on relative consumption, using the diet-switching framework (Fig. 1; van Leeuwen et al. 2013). We took a pairwise approach to this analysis for each of the three major resource types; therefore, the categories for relative resource cover and relative foraging effort were (1) turf/Halimeda, (2) Halimeda/foliose, and (3) turf/foliose. The relationship between relative resource cover and relative foraging effort was estimated using a linear mixed-model (LMM) regression with site as a random effect. The resulting slope and intercept were used to characterize resource similarity (slope) and resource preference (intercept), respectively. A slope ≥1 indicates substitutable resources, with switching occurring if slope >1 (Fig. 1a, b). A slope <1 indicates complementary resources; theoretically a slope approaching 0 would indicate no nutritional substitutability between two resources (Fig. 1c). Because we were primarily interested in deviations from perfect substitutability, our null hypothesis was slope = 1, and we report p values that test that null hypothesis (note the difference from typical linear regressions that test differences from a slope of zero). The intercept of the line reveals resource preference: intercept = 0 indicates neutral preference for the two resources, while intercept >0 indicates preference for the resource in the numerator (and vice versa); therefore, we tested the null hypothesis that the intercept = 0. Data from Cozumel (Paraiso Bajo) were removed from these analyses because the ratio of macroalgal turf cover to Halimeda spp. cover was much greater (295:1) than any other study site, making it a high-leverage and potentially misleading outlier.

Finally, we used one additional analysis to quantify the substitutability between each pair of the three resources. If resources are fully substitutable, then a scatterplot of the pairwise consumption rates should form a triangular distribution, with the outer edge of the triangle representing the maximal consumption rate of both resources (Fig. 2a). The triangular shape represents the substitution of one resource for the other. Alternatively, complementary, non-substitutable resources should produce a rectangular distribution, with consumption rates that do not depend on the other resource (Fig. 2b). We used a quantile regression (Scharf et al. 1998) to quantify the shape (triangle vs. rectangle) of the pairwise consumption relationship between each pair of the three resources. Quantile regression results may be sensitive to which quantile is used, so we performed the regression on the 90th, 75th and 60th quantile of the bite data to capture the range of outcomes. Multiple quantiles were represented in the analysis to illustrate uncertainty in substitutability between each resource combination. A bootstrap resampling procedure (10,000 replications) was used to calculate the standard errors of quantile regression coefficients (Scharf et al. 1998).
Fig. 2

Schematic illustrating substitution vs. complementarity. a Consumption (e.g., number of bites taken) of two substitutable resources. As the consumption of resource R2 increases consumption of resource R1 decreases. This results in a triangular distribution of the data and a negative slope at the maximal rate of consumption. b Consumption between two complementary resources. Increased consumption of resource R2 does not affect consumption of resource R1, resulting in a rectangular distribution of the data

Our initial analyses showed that there was no effect of size or sexual phase on foraging preferences, so our reported results include all individuals pooled together. All analyses were performed in R 3.0.1 (R Core Team 2016). Mixed-model analyses (GLMM and LMM) were performed using the lme4 package (Bates et al. 2015). For GLMMs and LMMs we calculated the amount of variance explained by the fixed effects (marginal r 2) following Nakagawa and Schielzeth (2013) using the sem.model.fits function in the piecewiseSEM package (Lefcheck 2015).


Macroalgal turf and foliose macroalgae dominated the benthic cover at all study sites (Online Resource 3). The absolute cover of macroalgal turf ranged from 15.4% (Great Inagua, Carmichael Bay) to 47% (Cozumel). Fleshy brown algae ranged from 3% (Cozumel) to 56% (Akumal); and Halimeda spp. ranged from 0.1% (Cozumel) to 21% (Middle Banco Chinchorro; Online Resource 3).

Redband parrotfish focused most of their foraging effort on turf, Halimeda, and foliose macroalgae; together these categories made up on average >97% of observed bites. However, both absolute and relative bite rates varied considerably among sites (Online Resource 4).

Resource selection, measured by the Vanderploeg and Scavia Relativized Index, showed that redband parrotfish selected against foliose macroalgae at all sites (E i  < 0; Fig. 3). Electivity was neutral to positive (E i  ≥ 0) for macroalgal turf at all study sites except South Banco Chinchorro, where electivity was negative. Selectivity for Halimeda was highly variable, both among and within study sites. At South Cozumel, North Banco Chinchorro, and Mayaguana, electivity for Halimeda ranged from nearly −1 to 1 for fish within the same site. At Cancun there was highly negative electivity for Halimeda, while electivity was neutral or positive at the remaining sites (Fig. 3).
Fig. 3

Boxplots showing redband parrotfish resource electivity (Vanderploeg and Scavia’s relativized index, E i ) by study site for the three main diet items: foliose macroalgae (left bars brown), Halimeda (center bars dark green), and turf macroalgae (right bars light green). Electivity is represented on the vertical axis: values >0 represent active selection disproportionate to abundance, values <0 represent resource avoidance. Box indicates interquartile range; horizontal line indicates median; vertical lines indicate 95% quantile range; points represent observations outside the 95% quantile range. Sample size for each site indicated in the corresponding panel

The effect of resource abundance cover on diet choice varied among diet items, but generally relative benthic cover of a resource was not strongly related to the bite rate on that resource (Fig. 4). The percent cover of turf had a significant positive relationship with bite rate (Poisson GLMM; df = 281, p = 0.014; Fig. 4a, Online Resource 5), but explained relatively little variation in the data (marginal r 2 = 0.25). The percent cover of Halimeda (Poisson GLMM; df = 281, p = 0.067, marginal r 2 = 0.14; Fig. 4b, Online Resource 5) and fleshy algae (Poisson GLMM; df = 281, p = 0.46, marginal r 2 = 0.03; Fig. 4c, Online Resource 5) did not have any relationships with the number of bites on those respective resources.
Fig. 4

Foraging response (number of bites taken in 5 min) of redband parrotfish as a function of resource cover for a macroalgal turf, b Halimeda, and c foliose macroalgae. Resource cover is expressed as a proportion of the total benthic cover. Curves are fits of a Poisson GLMM (n 293); solid lines indicate a relationship with a slope significantly different from zero (p < 0.05), and a dashed curve indicates slopes not significantly different from zero. Marker shape indicates the sexual phase of the parrotfish: initial phase (open circle) or terminal phase (filled diamond)

The relationship between foraging effort and the relative cover of each pair of resources afforded a clearer view of redband parrotfish diet choices than the analyses that focused on each individual resource (Fig. 5). Redband parrotfish did not exhibit a foraging preference for either Halimeda or turf when the two resources had equal cover, as indicated by a regression intercept not statistically different from zero (linear mixed model [LMM]; intercept = 0.53 ± 0.06; p = 0.07). Relative foraging effort on Halimeda increased proportionally when Halimeda cover increased relative to turf cover, and vice versa, as indicated by a regression slope not statistically different from 1 (LMM; slope = 0.91 ± 0.21; r 2 = 0.41; p = 0.34 for null hypothesis of slope = 1; Fig. 5a). The random effect of site explained 20% of the variance in that regression, and the marginal r 2 = 0.30.
Fig. 5

Relative foraging effort of redband parrotfish in response to relative resource cover. The horizontal axis represents the log10 ratio of the proportional benthic cover of the two resources being compared. The vertical axis represents the log10(x + 1) ratio of bites taken by parrotfish on the two resources over a 5-min period. Each pair of resources was compared: a Halimeda and macroalgal turf, b Halimeda and foliose macroalgae, and c turf and foliose macroalgae. Curves are fits of a linear mixed model (n = 293); solid lines indicate a relationship with a slope significantly different from zero (p < 0.05), and a dashed curve indicates slopes not significantly different from zero

Redband parrotfish consumed proportionally more Halimeda than foliose macroalgae when the two resources had equal cover (LMM; intercept = 0.77 ± 0.34; p = 0.01). However, relative foraging effort on Halimeda did not increase as the abundance of Halimeda increased relative to foliose macroalgae (LMM; slope = 0.26 ± 0.21; p = 0.1 for null hypothesis of slope = 0; p = 3×10−4 for null hypothesis of slope = 1; Fig. 5b). The statistically flat slope indicates complementarity between these resources, suggesting each resource provides a different nutritional benefit to redband parrotfish. The random effect of site explained 30% of the variance in that regression, and the marginal r 2 = 0.09.

Similarly, redband parrotfish consumed proportionally more macroalgal turf than foliose macroalgae at equal cover (LMM; intercept = 1.33 ± 0.29; p = 2 × 10−6), but increased their relative foraging on turf at a less-than-proportional rate when the abundance of turf relative to foliose macroalgae increased (LMM; slope = 0.57 ± 0.37; p = 0.06 for null hypothesis of slope = 0; p = 0.13 for null hypothesis of slope = 1; Fig. 5c). Note that the slope was not statistically different from either 0 or 1 at the 0.05 level, but the evidence points to a value that is closer to 1 than 0, suggesting some degree of substitutability between these two resources. The random effect of site explained 48% of the variance in that regression, and the marginal r 2 = 0.11.

Analysis of pairwise consumption rates revealed similar patterns in the degree to which the three resources were substitutable. Foraging effort between turf and Halimeda displayed a statistically significant inverse relationship (Fig. 6a; Online Resource 6). The negative slopes indicate a triangular distribution in feeding patterns, which suggests that redband parrotfish substituted between the two resources. The degree of substitutability between Halimeda and foliose macroalgae was quite opposite. The 90th quantile regression was positive, indicating that as parrotfish increased foraging effort on Halimeda they also increased foraging on foliose macroalgae. The lower quantiles had slopes not statistically different from zero, indicating a rectangular distribution of foraging effort between the two resources. Together these patterns suggest that the two resources are complementary (Fig. 6b; Online Resource 6). For the relationship between turf and foliose algae, the 90th and 60th quantile regressions had slopes not statistically different from zero, and the 75th quantile regression was significantly negative but very shallow. These patterns are indicative of a rectangular, complementary relationship with a weak indication of some degree of substitutability (Fig. 6c, Online Resource 6).
Fig. 6

Quantile regression to evaluate substitutability of resources. Horizontal and vertical axes indicate number of redband parrotfish bites per 5 min. Quantile regression reveals the shape of the outer edge of the distribution, indicating either a triangle (negative slope; substitutable resource) or rectangle (flat or positive slope; complimentary resources). We regressed the 60th (lower curve), 75th (middle curve) and 90th (upper curve) quantiles of each dataset in order to illustrate the degree of consistency in the shape of the distribution, and thus our confidence in the underlying substitutability patterns. Solid lines indicate a relationship with a slope significantly different from zero (p < 0.05), and a dashed curve indicates slopes not significantly different from zero (n = 293). a Macroalgal turf and Halimeda; b Halimeda and foliose macroalgae; c foliose macroalgae and macroalgal turf


Our results demonstrate that the dietary preferences of a reef herbivore, the redband parrotfish, depend on the relative cover of all available food resources. Foraging preferences of redband parrotfish did not correlate with the cover of each individual algal resource. Moreover, feeding electivity varied widely across study sites, with parrotfish exhibiting both positive and negative electivity for the same resource. Clearer patterns of resource use emerged only when we examined parrotfish foraging in the context of the relative abundances of all major diet items, as predicted by the theoretical framework developed by van Leeuwen et al. (2013).

By comparing relative foraging rates as a function of relative resource cover, we detected clear patterns of overall resource preference that were consistent across space: redband parrotfish preferred Halimeda spp. and macroalgal turfs equally, and preferred both of those diet items over foliose macroalgae. Additionally, by considering pairwise relative consumption, we deduced that turf macroalgae and Halimeda were substitutable resources, and redband parrotfish foraged on those two resource in proportion to their relative abundance at a given site. Conversely, foraging on Halimeda and foliose macroalgae exhibited a pattern typical of complementary resources: although redband parrotfish preferred Halimeda, they always consumed a consistent ratio of the two resources regardless of their relative abundance. The pattern of foraging on turf and foliose macroalgae was similar: a preference for turf, but foraging at a consistent ratio regardless of relative abundance. There was weak evidence for some degree of substitutability between turf and foliose macroalgae (e.g., the slope of the consumption–abundance plot was nearly different from zero, with p = 0.06; Fig. 5a), but in general the foraging patterns on those two resources were not statistically distinguishable from complementarity.

Predictions of the geometric framework

There is substantial evidence that animals will regulate their intake of different resources to achieve the desired balance of nutrition or nutrients (Raubenheimer and Simpson 2003, Simpson et al. 2004). This suggests that differences in the nutritional makeup of turf, foliose macroalgae, and Halimeda are an important factor determining grazer preferences for them (Abrams and Matsuda 2003; Raubenheimer and Simpson 2003). Prior work suggests that parrotfishes prefer protein-rich resources (Targett and Targett 1990; Crossman et al. 2005; Francini-Filho et al. 2010), so it is reasonable to presume that protein content is an important aspect of the resource preferences we observed. In general, macroalgal turfs have higher protein content than foliose macroalgae (Lourenço et al. 2002; McDermid and Stuercke 2003), which is consistent with the grazing preference observed in our study. We are not aware of any studies characterizing protein content of Halimeda relative to macroalgal turfs or foliose macroalgae, but our observational results predict it should be more similar to the former than the latter. It is clear, however, that protein content does not fully explain parrotfish foraging preferences; the complementarity between the two preferred resources and foliose macroalgae implies that the latter provides some additional important nutritional component. Additional research on the nutritional composition of algal resources would be required to determine the observed complementarity. One additional factor affecting resource preference could be that Halimeda spp. are calcareous, and foraging on them may provide redband parrotfish with abrasive substances that assist in the function of their pharyngeal mill. However, we are not aware of a specific experimental test of that hypothesis.

In addition to nutritional content of macroalgae, chemical or physical defenses play an important role in affecting dietary behavior. Both foliose macroalgae and Halimeda spp. are known to employ chemical defenses (Hay and Fenical 1987; Hay et al. 1994; Paul and van Alstyne 1988; Targett and Arnold 1998). However, in the context of our observations of this particular parrotfish species, it appears the effects of macroalgal defenses were minimal, or were not integral to the foraging responses of the fish. According to Targett and Arnold (1998), phlorotannins from brown algae (such as Lobophora spp. in our foliose category) would not be an effective deterrent for parrotfishes due to the basic pH of their guts. Under basic conditions, phlorotannins do not bond with the free amino groups of proteins, allowing protein assimilation by parrotfishes. Fishes with similar gut types have also been observed to consume phlorotannin-rich brown algae (Targett and Arnold 1998). Further, chemically defended Halimeda spp. were substitutable with undefended macroalgal turf, suggesting that the chemical defenses employed by Halimeda spp. (primarily the diterpenoids halimedatrial and halimedatetraacetate; Hay and Fenical 1987) were not a strong deterrent (Paul and van Alstyne 1988). Hay et al. (1994) also observed that calcium carbonate, a potential physical and chemical defense found in Halimeda spp., did not deter herbivory by parrotfishes.

Implications for coral reef ecology

Our observations of redband parrotfish diet preferences generally agree with previous studies of parrotfish species that reported foraging primarily on macroalgal turf and to some extent Halimeda spp. (Targett and Targett 1990; Bruggemann et al. 1994a; Francini-Filho et al. 2010). Redband parrotfish have also been observed to consume foliose macroalgae (including Dictyota and Lobophora) inside experimental enclosures in the Florida Keys, similar to our observations, though—similar to our results—it appears that category is not a preferred diet item (Burkepile and Hay 2008).

Across our study sites, redband parrotfish took far fewer bites from foliose macroalgae relative to its abundance when compared with other resources. Foliose algae such as Dictyota spp. and Lobophora spp. are among the most abundant macroalgae on Caribbean coral reefs and negatively affect the growth rates and fecundity of existing hard corals (Foster et al. 2008). It is widely accepted that reducing the cover of foliose macroalgae such as these is important for the resilience and recovery of Caribbean coral reefs, and grazing fishes are often prescribed as a solution for algal overgrowth (e.g., Bellwood et al. 2004, Hughes et al. 2007). However, caging experiments conducted by Burkepile and Hay (2010) showed that redband parrotfish were unable to prevent those foliose macroalgae from growing within their enclosures. Our data support the idea that redband parrotfish prefer other resources, such as macroalgal turfs, and should not be expected to significantly reduce the standing crop of foliose macroalgae, although other parrotfish species may prefer this particular resource. Nonetheless, redband parrotfish will continue to feed on foliose macroalgae as a complementary resource, despite the preference for turfs and Halimeda, and would be an important component of the overall grazing pressure on foliose macroalgae in a full community (as implied by the results of Burkepile and Hay 2010). This reinforces the lesson that the diet preferences of the individual species making up the grazing fish community must be taken into account in order to predict their combined influence on the benthic community (Adam et al. 2015). Our results support the findings of Burkepile and Hay (2010, 2011) and the recommendations of Adam et al. (2015) that a diverse assemblage of reef herbivores is essential to maintaining foraging pressure on the complete suite of reef macroalgae.


Raubenheimer et al. (2009) indicated that a priority goal of nutritional ecology is the application of organism-based models, such as the geometric framework, to specific studies in community ecology. We have taken a first step in that direction, using a quantitative accounting of relative resource cover to help explain the foraging patterns of a generalist grazer across a heterogeneous seascape. Our study provides testable predictions about resource preferences and nutritional relationships, as well as the expected grazing behavior of fish confronted with a mixture of algal resources. Viewing food resources as nutritional vectors can inform how an individual grazer will respond to changes in food resource abundance over spatial gradients. This approach could be applied to generalist grazers in any ecosystem, but in the context of coral reef ecology, the predictions that arise from the foregoing analysis could be used in resource and conservation management.



We thank W. Freshwater for advice on algal biology and helpful comments on the manuscript, and T.-L. Loh, S. McMurray, L. Deignan, I. Conti-Jerpe, M. Marty, A. Dingeldein, and C. Marino for assistance in the field. We also acknowledge the staff at UNCW and NURC for logistical support. Constructive suggestions from two anonymous reviewers improved the manuscript.

Author contribution statement

JH and JRP conceived and designed the study. JH collected the data. JH and JWW analyzed the data. JH, JWW, and JRP wrote the manuscript.

Compliance with ethical standards


This study was funded by the National Science Foundation (OCE-0550468, 1029515).

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical approval

All applicable institutional and national guidelines for the care and use of animals were followed.

Supplementary material

442_2017_3855_MOESM1_ESM.docx (390 kb)
Supplementary material 1 (DOCX 389 kb)
442_2017_3855_MOESM2_ESM.docx (95 kb)
Supplementary material 2 (DOCX 94 kb)
442_2017_3855_MOESM3_ESM.docx (293 kb)
Supplementary material 3 (DOCX 293 kb)
442_2017_3855_MOESM4_ESM.docx (269 kb)
Supplementary material 4 (DOCX 269 kb)
442_2017_3855_MOESM5_ESM.docx (82 kb)
Supplementary material 5 (DOCX 81 kb)
442_2017_3855_MOESM6_ESM.docx (69 kb)
Supplementary material 6 (DOCX 68 kb)


  1. Abrams PA (1990) Mixed responses to resource densities and their implications for character displacement. Evol Ecol 4:93–102CrossRefGoogle Scholar
  2. Abrams PA (1993) Why predation should not be proportional to predator density. Ecology 74:726–733CrossRefGoogle Scholar
  3. Abrams PA, Matsuda H (2003) Population dynamical consequences of reduced predator switching at low total prey densities. Popul Ecol 45:175–185CrossRefGoogle Scholar
  4. Adam TC, Burkepile DE, Ruttenberg BI, Paddack MJ (2015) Herbivory and the resilience of Caribbean coral reefs: knowledge gaps and implications for management. Mar Ecol Prog Ser 520:1–20CrossRefGoogle Scholar
  5. Augustin DJ, McNaughton SJ (1998) Ungulate effects on the species composition of plan communities: herbivore selectivity and plant tolerance. J Wildlife Manag 62:1165–1183CrossRefGoogle Scholar
  6. Bates D, Mächler M, Bolker B, Walker S (2015) Fitting linear mixed-effects models using lme4. J Stat Soft 67:1–48CrossRefGoogle Scholar
  7. Behmer ST, Joern A (2008) Coexisting generalist herbivores occupy unique nutritional feeding niches. Proc Natl Acad Sci USA 105:1977–1982CrossRefPubMedPubMedCentralGoogle Scholar
  8. Bellwood DR, Hughes TP, Folke C, Nyström M (2004) Confronting the coral reef crisis. Nature 249:827–833CrossRefGoogle Scholar
  9. Bruggemann JH, van Oppen MJH, Breeman AM (1994a) Foraging by the stoplight parrotfish Sparisoma viride. I. Food selection in different, socially-determined habitats. Mar Ecol Prog Ser 106:41–55CrossRefGoogle Scholar
  10. Bruggemann JH, Begeman J, Bosma EM, Verburg P, Breeman AM (1994b) Foraging by the stoplight parrotfish Sparisoma viride. II. Intake and assimilation of food, protein and energy. Mar Ecol Prog Ser 106:57–71CrossRefGoogle Scholar
  11. Burkepile DE, Hay ME (2008) Herbivore species richness and feeding complementarity affect community structure and function on a coral reef. Proc Natl Acad Sci USA 105:16201–16206CrossRefPubMedPubMedCentralGoogle Scholar
  12. Burkepile DE, Hay ME (2009) Nutrient versus herbivore control of macroalgal community development and coral growth on a Caribbean reef. Mar Ecol Prog Ser 389:71–84CrossRefGoogle Scholar
  13. Burkepile DE, Hay ME (2010) Impact of herbivore identity on algal succession and coral growth on a Caribbean reef. PLoS One 5:1–9CrossRefGoogle Scholar
  14. Burkepile DE, Hay ME (2011) Feeding complementarity versus redundancy among herbivorous fishes on a Caribbean reef. Coral Reefs 30:351–362CrossRefGoogle Scholar
  15. Byrnes J, Stachowicz JJ, Hultgren KM, Hughes AR, Olyarnik SV, Thornber CS (2006) Predator diversity strengthens trophic cascades in kelp forests by modifying herbivore behaviour. Ecol Lett 9:61–71PubMedGoogle Scholar
  16. Carpenter SR, Kitchell JF (1984) Plankton community structure and limnetic primary production. Am Nat 124:159–172CrossRefGoogle Scholar
  17. Catano LB, Gunn BK, Kelley MC, Burkepile DE (2015) Predation risk, resource quality, and reef structural complexity shape territoriality in a coral reef herbivore. PLoS ONE 10:e0118764CrossRefPubMedPubMedCentralGoogle Scholar
  18. Cheal AJ, MacNeil MA, Cripps E, Emslie MJ, Jonker M, Schaffelke B, Sweatman H (2010) Coral–macroalgal phase shifts or reef resilience: links with diversity and functional roles of herbivorous fishes on the Great Barrier Reef. Coral Reefs 29:1005–1015CrossRefGoogle Scholar
  19. Clements KD, Raubenheimer D, Choat JH (2009) Nutritional ecology of marine herbivorous fishes: ten years on. Funct Ecol 23:79–92CrossRefGoogle Scholar
  20. Crossman DJ, Choat JH, Clements KD (2005) Nutritional ecology of nominally herbivorous fishes on coral reefs. Mar Ecol Prog Ser 296:129–142CrossRefGoogle Scholar
  21. Dunlap M, Pawlik JR (1996) Video-monitored predation by Caribbean reef fishes on an array of mangrove and reef sponges. Mar Biol 126:117–123CrossRefGoogle Scholar
  22. Foster NL, Box SJ, Mumby PJ (2008) Competitive effects of macroalgae on the fecundity of the reef-building coral Montastraea annularis. Mar Ecol Prog Ser 367:143–152CrossRefGoogle Scholar
  23. Francini-Filho RB, Ferreira CM, Oliveira E, Coni C, de Moura RL, Kaufman L (2010) Foraging activity of roving herbivorous reef fish (Acanthuridae and Scaridae) in eastern Brazil: influence of resource availability and interference competition. J Mar Biol Ass UK 90:481–492CrossRefGoogle Scholar
  24. Harris JL, Lewis LS, Smith JE (2015) Quantifying scales of spatial variability in algal turf assemblages on coral reefs. Mar Ecol Prog Ser 532:41–57CrossRefGoogle Scholar
  25. Hay ME, Fenical W (1987) Marine plant-herbivore interactions: the ecology of chemical defense. Ann Rev Ecol Syst 19:111–145CrossRefGoogle Scholar
  26. Hay ME, Kappel QE, Fenical W (1994) Synergisms in plant defenses against herbivores: interactions of chemistry, calcification, and plant quality. Ecology 75:1714–1726CrossRefGoogle Scholar
  27. Hughes TP, Baird AH, Bellwood DR, Card M, Connolly SR, Folke C, Grosberg R, Hoegh-Guldberg O, Jackson JBC, Kleypas J, Lough JM, Marshall P, Nyström M, Palumbi SR, Pandolfi JM, Rosen B, Roughgarden J (2003) Climate change, human impacts, and the resilience of coral reefs. Science 301:929–933CrossRefPubMedGoogle Scholar
  28. Hughes TP, Bellwood DR, Folke CS, McCook LJ, Pandolfi JM (2007) No-take areas, herbivory and coral reef resilience. Trends Ecol Evol 22:1–3CrossRefPubMedGoogle Scholar
  29. Lechowicz MJ (1982) The sampling characteristics of electivity indices. Oecologia 52:22–30CrossRefPubMedGoogle Scholar
  30. Lefcheck JS (2015) piecewiseSEM: Piecewise structural equation modeling in R for ecology, evolution, and systematics. Methods Ecol Evol 7:573–579CrossRefGoogle Scholar
  31. Loh T-L, Pawlik JR (2014) Chemical defenses and resource trade-offs structure sponge communities on Caribbean coral reefs. Proc Natl Acad Sci USA 111:4151–4156CrossRefPubMedPubMedCentralGoogle Scholar
  32. Loh T-L, McMurray SE, Henkel TP, Vicente J, Pawlik JR (2015) Indirect effects of overfishing on Caribbean reefs: sponges overgrow reef-building corals. PeerJ 3:e901CrossRefPubMedPubMedCentralGoogle Scholar
  33. Lourenço SO, Barbarino E, De-Paula JC, Pereira LOS, Lanfer Marquez UM (2002) Amino acid composition, protein content and calculation of nitrogen-to-protein conversion factors for 19 tropical seaweeds. Phycol Res 50:233–241CrossRefGoogle Scholar
  34. McDermid KJ, Stuercke B (2003) Nutritional composition of edible Hawaiian seaweeds. J Appl Phycol 15:513–524CrossRefGoogle Scholar
  35. Miller AD, Roxburgh SH, Shea K (2011) How frequency and intensity shape diversity-disturbance relationship. Proc Natl Acad Sci USA 108:5643–5648CrossRefPubMedPubMedCentralGoogle Scholar
  36. Mumby PJ (2009) Herbivory versus corallivory: are parrotfish good or bad for Caribbean coral reefs? Coral Reefs 28:683–690CrossRefGoogle Scholar
  37. Mumby PJ, Wabnitz C (2002) Spatial patterns of aggression, territory size, and harem size in five sympatric Caribbean parrotfish species. Environ Biol Fish 63:265–279CrossRefGoogle Scholar
  38. Murdoch WW (1969) Switching in general predators: experiments on predator specificity and stability of prey populations. Ecol Monogr 39:335–354CrossRefGoogle Scholar
  39. Nakagawa S, Schielzeth H (2013) A general and simple method for obtaining R 2 from generalized linear mixed-effects models. Methods Ecol Evol 4:133–142CrossRefGoogle Scholar
  40. Norström AV, Nyström M, Lokrantz J, Folke C (2009) Alternative states on coral reefs: beyond coral–macroalgal phase shifts. Mar Ecol Prog Ser 376:295–306CrossRefGoogle Scholar
  41. Oaten A, Murdoch WW (1975) Switching functional response and stability in predator-prey systems. Am Nat 109:299–318CrossRefGoogle Scholar
  42. Olff H, Ritchie ME (1998) Effects of herbivores on grassland plant diversity. Trends Ecol Evol 13:261–265CrossRefPubMedGoogle Scholar
  43. Paul VJ, van Alstyne KL (1988) Chemical defense and chemical variation in some tropical Pacific species of Halimeda. Coral Reefs 6:263–269CrossRefGoogle Scholar
  44. Pawlik JR, Burkepile DE, Vega Thurber R (2016) A vicious circle? Altered carbon and nutrient cycling may explain the low resilience of Caribbean coral reefs. Bioscience 66:470–476CrossRefGoogle Scholar
  45. R Core Team (2016) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org
  46. Raubenheimer D, Simpson SJ (2003) Nutrient balancing in grasshoppers: behavioural and physiological correlates of dietary breadth. J Exp Biol 206:1669–1681CrossRefPubMedGoogle Scholar
  47. Raubenheimer D, Simpson SJ, Mayntz D (2009) Nutrition, ecology and nutritional ecology: toward an integrated framework. Funct Ecol 23:4–16CrossRefGoogle Scholar
  48. Rindorf A, Gislason H, Lewy P (2006) Prey switching of cod and whiting in the north sea. Mar Ecol Prog Ser 325:243–253CrossRefGoogle Scholar
  49. Rotjan RD, Dimond JL (2010) Discriminating causes from consequences of persistent parrotfish corallivory. J Exp Mar Biol Ecol 390:188–195CrossRefGoogle Scholar
  50. Scharf FS, Juanes F, Sutherland M (1998) Inferring ecological relationships from the edges of scatter diagrams: comparison of regression techniques. Ecology 79:448–460CrossRefGoogle Scholar
  51. Simpson SJ, Raubenheimer D (2001) A framework for the study of macronutrient intake in fish. Aquac Res 32:421–432CrossRefGoogle Scholar
  52. Simpson SJ, Sibly RM, Lee KP, Behmer ST, Raubenheimer D (2004) Optimal foraging when regulating intake of multiple nutrients. Anim Behav 68:1299–1311CrossRefGoogle Scholar
  53. Steneck RS, Dethier MN (1994) A functional group approach to the structure of algal-dominated communities. Oikos 69:476–498CrossRefGoogle Scholar
  54. Suding KN, Gross KL, Houseman GR (2004) Alternative states and positive feedbacks in restoration ecology. Trends Ecol Evol 19:46–53CrossRefPubMedGoogle Scholar
  55. Targett NM, Arnold TM (1998) Predicting the effects of brown algal phlorotannins on marine herbivores in tropical and temperate oceans. J Phycol 34:195–205CrossRefGoogle Scholar
  56. Targett TE, Targett NM (1990) Energetics of food selection by the herbivorous parrotfish Sparisoma radians: roles of assimilation efficiency, gut evacuation rate, and algal secondary metabolites. Mar Ecol Prog Ser 66:13–21CrossRefGoogle Scholar
  57. van Leeuwen E, Brännström Å, Jansen VAA, Dieckmann U, Rossberg AG (2013) A generalized functional response for predators that switch between multiple prey species. J Theor Biol 328:89–98CrossRefPubMedGoogle Scholar
  58. Vanderploeg HA, Scavia D (1979) Calculation and use of selectivity coefficients of feeding: zooplankton grazing. Ecol Model 7:135–149CrossRefGoogle Scholar
  59. Visser A, Fiksen Ø (2013) Optimal foraging in marine ecosystem models: selectivity, profitability and switching. Mar Ecol Prog Ser 473:91–101CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2017

Authors and Affiliations

  1. 1.Department of Biology and Marine Biology, Center for Marine ScienceUniversity of North Carolina WilmingtonWilmingtonUSA

Personalised recommendations