Volume 21, Issue 6 p. 1137-1145
Free Access

Explaining isotope trophic-step fractionation: why herbivorous fish are different


Corresponding Author


School of Marine Science and Technology, Newcastle University, Newcastle upon Tyne, NE1 4RU, UK,

Present address: FRS Marine Laboratory, 375 Victoria Road, Aberdeen AB11 9DB, UK.

†Author to whom correspondence should be addressed. E-mail: [email protected]Search for more papers by this author


Centre for Environment, Fisheries and Aquaculture Science (CEFAS), Pakefield Road, Lowestoft, NR33 0HT, UK

Search for more papers by this author


School of Marine Science and Technology, Newcastle University, Newcastle upon Tyne, NE1 4RU, UK,

Search for more papers by this author
First published: 21 August 2007
Citations: 96


  • 1

    An assumed constant trophic fractionation of 15N/14N between consumer and diet (usually 3·4 for diet–muscle tissue differences) allows inferences to be made about feeding interactions and trophic level in food web studies. However, considerable variability surrounds this constant, which may conceal subtle differences about the trophodynamics of consumers.

  • 2

    The feeding ecologies of herbivores and carnivores differ in terms of diet quality (in C : N terms) and food processing mechanisms, which may affect fractionation.

  • 3

    We present a new model that explores how consumer feeding rates, excretion rates and diet quality determine the 15N/14N ratios in the consumer's tissues and hence influence the magnitude of trophic fractionation.

  • 4

    Three herbivorous reef fish Acanthurus sohal, Zebrasoma xanthurum and Pomacentrus arabicus were chosen as study organisms. Empirical estimates of diet–tissue stable isotope fractionation were made in the field, and model parameters were derived from feeding observations and literature data.

  • 5

    The trophic fractionation values of A. sohal, Z. xanthurum and P. arabicus were 4·69, 4·47 and 5·25, respectively, by empirical measurement, and 4·41, 4·30 and 5·68, respectively, by model, indicating that herbivores have a higher trophic fractionation than the currently accepted value of 3·4.

  • 6

    The model was most sensitive to the excretion rate, which may differ between herbivores and carnivorous animals. This model is the first to determine stable isotope signatures of a consumer's diet mixture without applying a constant fractionation value.


Stable isotope ratios are increasingly used to explore feeding relationships in aquatic ecosystems, and there have been many attempts to develop isotope ‘mixing models’ to help quantify the potential importance of particular feeding linkages, given the isotope signature of the consumer and potential diets (e.g. Ben-David & Schell 2001; Koch & Phillips 2002; Phillips & Koch 2002). All these models assume that a stepwise enrichment occurs in the heavier isotope (13C or 15N) with each trophic level and that this enrichment is constant (typically 3·4 for 15N and 1 for 13C) irrespective of the animal's biology and feeding behaviour. However, the magnitude of this per trophic-step isotope fractionation (ΔN or ΔC) can be affected by many factors including nutritional stress, diet quality, body size, excretory mechanisms and feeding rate (Hobson & Welch 1995; Ponsard & Averbuch 1999; Overman & Parrish 2001; Pinnegar, Campbell & Polunin 2001; Vanderklift & Ponsard 2003). While most fishes approximate the often-cited 3·4 δ15N trophic-step enrichment (see Vander Zanden & Rasmussen 2001; Sweeting et al. 2007), herbivorous fishes seem to deviate from this pattern, for reasons that remain unclear.

Polunin & Pinnegar (2002) report that the herbivorous parrotfish Sparisoma spp. and the Mediterranean sparid Sarpa salpa exhibit δ15N values much higher than would be expected if they were feeding solely on plant material (based on data from Keegan & DeNiro 1988; Pinnegar & Polunin 2000). Similarly, there is enrichment greater than 3·4 between herbivorous reef fish from French Polynesia (Acanthurus nigrofuscus and Stegastes nigricans) and the turf algae on which they feed (Table 1).

Table 1. Trophic-level fractionation (ΔN) values of herbivorous fish taken from published studies (overall mean ± standard deviation are also reported)
Species ΔN () Study
Sarpa salpa 4·90 Pinnegar & Polunin (2000)
Sarpa salpa 7·22 Jennings et al. (1997)
Sparisoma spp. 4·10 Keegan & DeNiro (1988)
Stegastes 6·07 N.V.C. Polunin, unpublished data
Stegastes nigricans 4·60 N.V.C. Polunin, unpublished data
Acanthurus lineatus 2·79 N.V.C. Polunin, unpublished data
Plectrogyphidodon lacrymatus 4·30 N.V.C. Polunin, unpublished data
Acanthurus lineatus 3·77 N.V.C. Polunin, unpublished data
Segastes (yellow) 5·25 N.V.C. Polunin, unpublished data
Mean: 4·78 ± 1·3

Several explanations have been proposed for this phenomenon, notably the possibility of ‘isotopic routing’ (see Gannes, O’Brien & Del Rio 1997) whereby the nitrogen consumed by herbivorous fishes comes predominantly from a very minor animal-derived component of the diet, since most marine algae are extremely poor sources of nitrogen and phosphorus (Atkinson & Smith 1983). Many herbivorous fishes are known to actively seek out animal-derived proteinaceous material to supplement their diet (Robertson 1982), but whether or not this would alter the δ15N signature of the animal sufficiently remains a matter of conjecture.

Vander Zanden & Rasmussen (2001) reviewed 35 trophic enrichment (ΔN) estimates to examine whether the mean or variance varied systematically with taxon, habitat or food type. Marked differences in ΔN were noted between carnivores and herbivores. In particular, carnivores, especially those in the wild, exhibited very tight clustering of per-trophic-level fractionation values around 3·4, whereas herbivores exhibited hugely variable per-tropic-level fractionation values ranging from –0·7 in the amphipod Amphithoe valida to +9·2 in the brine shrimp Artemia salina.

Trophic enrichment of δ15N is generally attributed to fractionation during amino acid deamination and transamination (Vander Zanden & Rasmussen 2001), whereby 14N amine groups are preferentially removed to produce isotopically light metabolites, leaving the remaining nitrogen pool enriched in 15N (referred to as ‘metabolic fractionation’; Gannes et al. 1997). Alternatively, fractionation can result from isotopic discrimination during nitrogen assimilation (referred to as ‘assimilative fractionation’). Metabolic fractionation is expected to be the dominant process for carnivores because the animal-derived nitrogen is biochemically more homogeneous and dominated by proteins. For herbivores, both assimilative and metabolic factors are likely to affect fractionation (Vander Zanden et al. 2001).

The food processing mode of herbivores differs from other teleost fish in that gut retention times can be very short (typically 4–5 h, Polunin, Harmelinvivien & Galzin 1995) and gut adaptations to digest structurally complex algal material are present in many species, for example, a very long alimentary tract in most herbivorous species (Elliott & Bellwood 2003), gut fermentation in some kyphosids (Clements & Choat 1997) and very low gut pH in some pomacentrids (Lobel 1981). The absorption efficiency of essential nutrients (e.g. nitrogen, phosphorus and carbon) can be markedly lower in herbivores than in carnivorous species, even within the same family (Polunin et al. 1995). Therefore, in order to meet their bioenergetic needs, herbivorous fishes tend to have higher feeding and excretion rates than carnivores.

The dynamic equilibrium model of Olive et al. (2003) indicates that food quality, feeding rate and excretion rate all influence the level of trophic fractionation. This model differs from other published models (Fry & Arnold 1982; Hesslein, Hallard & Ramlal 1993) as the steady-state parameters that determine ΔN can be derived experimentally. The Olive model explains dynamics of per-trophic-level fractionation and can be used to estimate the isotope signature of a consumer at time t following a shift in diet. However, in most stable isotope studies, the signature of the consumer (δa) is known and a model to estimate the isotopic value for the diet, δfood (and hence ΔN), would be more useful. This can be achieved through a simple rearrangement of the Olive model (model parameters defined in Table 2):

Table 2. Definitions of parameters used in the model [based on Olive et al. (2003)]
Model parameter Definition
ΔN () Trophic-step fractionation of nitrogen
δfood () Isotope signature of the food
δa () Isotope signature of the consumer
R d (%) Daily ration as a proportion of body mass consumed per day (BWD)
Ωin Ratio of mass of element ingested to that in the animal as a whole
q (%) Assimilation efficiency
Z () Rate of change by excretion per day
t (subscript) Time after the diet switch
0 (subscript) Time of the diet switch
image( eqn 1)

The majority of fractionation studies have involved controlled feeding in the laboratory (Power, Guiguer & Barton 2003; Vanderklift & Ponsard 2003). In the field, variability in per-trophic-level fractionation can only be studied where the consumer's diet can be well quantified and the isotope signatures of potential food items easily determined. Herbivores that graze on algal turf communities provide an opportunity in this regard, some species maintain a ‘garden’ of algae which they tend, manipulate and vigorously defend (Hata & Kato 2004). The diet material in the foregut of the fish species is easily distinguishable, and this allows feeding preferences to be easily characterized. One damselfish and two surgeonfish were observed in this study. Pomacentrus arabicus, Acanthurus sohal and Zebrasoma xanthurum co-habit an area of high-latitude reef in Oman, and all feed on the same epilithic algal community. The three species differ in size and exhibit different food processing mechanisms (Z. xanthurum is a gut-fermenting species, the other two are not).

The overall aim of this study was to determine why herbivorous fishes exhibit unusual per-trophic-level isotope enrichment patterns and to evaluate whether a dynamic model of the fractionation process, incorporating absorption, feeding rate and excretion, could help understand observations from the field. Specific hypotheses tested include the following: (i) herbivorous fishes exhibit higher ΔN values because they consume more food each day than carnivorous fishes; (ii) herbivorous fishes exhibit higher ΔN values because they are less efficient at absorbing nitrogen from their food; and (iii) herbivorous fishes exhibit higher ΔN values because they excrete more nitrogen per day compared to carnivorous fishes.

Materials and methods

sample collection and processing

Samples were collected during August and November 2004, and February 2005 from Bandar Kayran, Greater Muscat, Oman (23°31′N, 58°43′E). Three specimens of the herbivorous fish species A. sohal, Z. xanthurum, P. arabicus were collected by spearing during each sampling trip, with the exception of P. arabicus where only one specimen was collected in February 2005. Fish were killed and immediately placed on ice until they were processed in the laboratory (maximum 5 h). All fish species were collected after 11:00 h local time to ensure a full digestive tract. Five potential dietary algae species (Hypnea pannosa, Sarconema filiforme, Gracilaria sp., Ulva lactuca and Dictyopteris sp.) were collected between 2 and 7 m depth for isotope analysis. A minimum of three replicate samples were collected for each algae species. In the laboratory, fork length, total length and weight of fishes were recorded. Fresh diet material was collected from the anterior and posterior of the alimentary canal and frozen (–30 ºC) either for isotope analysis or the determination of absorption efficiencies. Anterior material (the first 10% of the alimentary tract) was considered a proxy for diet and posterior material (the last 10% of the alimentary tract) a proxy for faecal material. A further sample was taken from the anterior of the gut for quantification of food items. Fish dorsal white muscle tissue was dissected from each fish frozen for stable isotope analysis. Algal materials were cleaned by hand removing of detritus, epiphytes and sediment from the samples before freezing. All frozen materials were later freeze-dried and homogenized with a mortar and pestle prior to analysis.

stable isotope determination

For each fish captured, c. 1 mg of homogenized muscle tissue was accurately weighed into tin capsules. Algal samples were weighed into two aliquots of c. 1 and c. 2 mg to determine δ13C and δ15N separately. Samples were analysed using one of two machines, Automated Nitrogen Carbon Analysis (ANCA) 20-20 isotope ratio mass spectrometer (Scottish Crops Research Institute (SCRI), Dundee, Scotland) or a Thermo-Finnegan mass spectrometer (Scottish Universities Environmental Reactor Centre (SUERC), East Kilbride, Scotland). Internal standards and ecological samples analysed on both machines revealed slight discrepancies between the two machines but allowed data to be aligned to one machine to ensure all results were comparable. Experimental precision based on the standard deviation of the internal standards was 0·2 for both δ15N and δ13C (SCRI), and 0·3 for δ15N and 0·2 for δ13C (SUERC).

stomach contents analysis

Two methods were used to quantify the diets of the fish depending upon the size of the fragments consumed. The stomach contents of A. sohal and Z. xanthurum were described using a line-transect method to describe the relative abundance of algal genera in the stomach. Stomach contents were laid in a transparent tray and viewed under a dissecting microscope, the extent of the transect line covered by each food category being recorded (Choat & Clements 1992). The line transect method was repeated for five transects for each specimen. A total of seven stomachs were analysed for A. sohal and five stomachs of Z. xanthurum. The totals for each transect were expressed as percentages to remove the effect of varying gut volume, then pooled to obtain mean percentages and variances for each fish; a pooled mean for each species was also obtained. Species-specific identification was not possible so algae were pooled by genera.

The stomachs of P. arabicus contained smaller algal fragments than those of the other two species, so a point intercept method was used to quantify diet. The stomach contents were laid out in a tray and algae, detritus, sediment and blue-green algae were recorded where they occurred directly under predetermined points on an intercept line. One-hundred points were recorded and converted into a percentage to remove the effect of varying gut volume. Fragments that could be identified to genera were noted as being present in the diet. This method was repeated for the stomachs of all seven P. arabicus specimens sampled.

q and nitrogen absorption efficiency (NAE)

In the appendix to Olive et al. (2003), the authors demonstrated how q might be derived for a fish, given the information about nitrogen absorption efficiency (NAE). NAE of A. sohal, Z. xanthurum and P. arabicus were determined by ash determination method (Montgomery & Gerking 1980) as follows. About 100 mg of dried homogenized material were weighed into a crucible and combusted in a muffle furnace at 600 °C for 24 h. Samples were then reweighed and the remaining ash expressed as a percentage of the original mass. Bulk nitrogen content of the food, fish and faeces was obtained as the area of N peak from mass spectrometry calibrated vs. tryptophan (r2 = 0·999).

image( eqn 2)
image( eqn 3)

Calculated NAE estimates were compared with 41 other NAE estimates of herbivorous fish (appendix 1).

Ωin and daily ration (rd)

Olive et al. (2003) described Ωin as the ‘dimensionless ratio of the mass of an element (e.g. nitrogen) in the ingested food in relation to the mass of the element in the animal as a whole’. It is closely related to the daily ration (Rd) of the animal, expressed as the weight of food ingested per day as a proportion of the animal's body mass. It is possible to determine the Rd of herbivorous fishes by determining the mean bite size and the number of bites taken in a day. It was not possible to do this for the species in this study; however, bite size and counts have been made for similar grazing species elsewhere (Table 3). Rd values were selected from the literature, where possible from fishes in the same genus (i.e. Acanthurus, Zebrasoma and Pomacentrus). In order to calculate Ωin from Rd, it was necessary to know something about the nitrogen content of the consumer and its food, these data were obtained by mass spectrometry. The wet weight Rd percentage was converted to dry weight and hence the amount of nitrogen consumed per day, Ωin, calculated using eqn 4

Table 3. Feeding rates Rd (%body weight per day) of tropical-reef-associated herbivorous fish species from literature
Species Location R d Study
Stegastes fuscus Brazil 13·03 Ferriera et al. (1998)
Sparisoma atomarium Brazil 24·94 Ferriera et al. (1998)
Acanthurus bahianus Brazil 23·26 Ferriera et al. (1998)
Stegastes nigricans French Polynesia 30·3 Polunin et al. (1995)
Ctenochaetus striatus French Polynesia 31·1 Polunin et al. (1995)
Acanthurus nigrofuscus French Polynesia 24·2 Polunin et al. (1995)
Scarus sordidus French Polynesia 36·0 Polunin et al. (1995)
Zebrasoma scopas French Polynesia 31·8 Polunin et al. (1995)
Stegastes apicalis (Summer) Great Barrier Reef 20·5 Klumpp & Polunin (1989)
Stegastes apicalis (Winter) Great Barrier Reef 11·1 Klumpp & Polunin (1989)
Atrosalarius sp. Great Barrier Reef 8·5 Klumpp & Polunin (1989)
Mean 20·8
image(eqn 4)

Where Nfood is the percentage nitrogen content of the food and Na is the nitrogen content of the consumer.

z, isotope discrimination associated with excretion

Z is the instantaneous rate of change in the isotope signature of an animal (in δ units per day) due to excretion. This term is very difficult to measure experimentally, although Olive et al. (2003) show how Z might be estimated as the instantaneous rate of 15N enrichment at the onset of starvation. Following the onset of starvation (when Ωin is zero), animals become progressively enriched in 15N with time, due to continued export of depleted (high in 14N) excretory products.

It was not possible to conduct starvation experiments for every fish in this study, and thus an estimate for Z was determined using data from existing laboratory feeding experiments (e.g. Harvey et al. 2002). Given a known isotope signature of the consumer at time ta), together with known values for Ωin and q, it is possible to re-arrange eqn 1 to solve for Z:

image( eqn 5)

For each data point, an estimate of Z may be derived, with the arithmetic mean providing an overall estimate, given all the available data. This procedure was attempted for nine sets of existing data (Table 4). There were no suitable diet switch experiments to derive data for herbivorous fish so the mean value of these studies, 0·14 (Table 4), was used to express Z in the model for all three fish species.

Table 4. Calculation of Z using data from diet switch experiments on fish in the laboratory (parameters are defined in Table 2)
Species Diet Mean weight (g) R d Ωin q δfood δa Z Study
Rhinogobius sp. Aquatic insects 0·41 0·072 0·03 0·9481 4·8 13·8 0·475 Maruyama et al. (2001)
Salvelinus namaycush Pellet 125 0·02 0·013 0·9251 9·51 10·23 0·079 Harvey et al. (2002)
Coregonus nasus Pellet 26 0·012 0·041 0·948 9·7 11·6 0·053 Hesslein et al. (1993)
Dicentrarchus labrax Sandeels 40 0·04 0·028 0·9251 12·16 17·83 0·131 Sweeting (2004)
Dicentrarchus labrax Dab 47 0·04 0·046 0·9251 13·87 17·48 0·1447 Sweeting (2004)
Dicentrarchus labrax Sandeels 33·1 0·032 0·026 0·9251 13·16 16·41 0·14 Barnes (2006)
Fundulus heteroclitus Tuna 1–2 0·04 0·026 0·9251 15·6 7·0 0·047 Logan et al. (2006)
Fundulus heteroclitus Tuna 1–2 0·03 0·019 0·9251 15·6 7·0 0·083 Logan et al. (2006)
Chromis chromis Plankton 10·25 0·046 0·023 0·9251 3·365 6·657 0·133 Pinnegar (2000)
Mean 0·14


stable isotope analysis

There was no significant difference in stable isotope signature of any fish species across seasons (GLM: A. Sohal: F2,7 = 2·17, P = 0·209; Z. xanthurum: F2,7 = 2·74, P = 1·43; P. arabicus: F2,6 = 1·43, P = 0·340). Therefore, fish from all seasons were pooled at species level to obtain mean δ13C and δ15N values (Fig. 1). The algae showed greater variability in δ13C than the three fish species. An empirical estimate of ΔN (Table 5) was calculated by subtracting the mean isotopic signature of the algae (Table 6) from the mean isotopic signature of each fish species.

Details are in the caption following the image

δ13C and δ15N isotope plot of herbivores and algae. Pomacentrus arabicus (triangle), Acanthurus sohal (circle) and Zebrasoma xanthurum (square). Algae genera are in squares. A1, Dictyopteris; A2, Gracilaria; A3, Hypnea; A4, Sarconema; A5, Ulva; A6, Pterocladia; A7, Turbinaria.

Table 5. Mean fork length (FL), δ15N (± 1 SD) and ΔN (calculated by subtracting the mean δ15N of algae from the δ15N of the animal) of three herbivorous fish
Species Mean FL (mm) n δ15N ΔN
Acanthurus sohal 296 9 13·44 ± 0·65 4·92
Zebrasoma xanthurum 205 9 13·04 ± 0·60 4·52
Pomacentrus arabicus 117 7 14·19 ± 0·90 5·67
Table 6. Mean δ15N and %N (± 1 SD) values for seven algae genera
Code Genus n δ15N %N
A1 Dictyopteris 3 8·54 ± 0·12 4·28 ± 0·11
A2 Gracilaria 4 9·00 ± 0·22 1·95 ± 0·33
A3 Hypnea 6 8·37 ± 0·26 4·34 ± 0·11
A4 Sarconema 5 8·38 ± 0·82 2·86 ± 0·52
A5 Ulva 3 8·59 ± 0·38 3·65 ± 0·32
A6 Pterocladia 3 9·64 ± 0·09 3·24 ± 0·45
A7 Turbinaria 5 7·08 ± 0·21 1·43 ± 0·19
Mean: 8·51 ± 0·30 Mean: 3·11 ± 0·29

stomach contents analysis

The fish consumed over 30 different species of algae, the dominant dietary algae genera differing among fish species: Hypnea spp. and Phylophora spp. were most frequent in A. sohal, while Feldmannia spp. and Pterocladia spp. were dominant in Z. xanthurum (see Table S1 in Supplementary material). Pomacentrus arabicus had a diet that was 35% detritus and 60% algae.

absorption efficiencies and q

Total and elemental absorption efficiencies for the three fish species compare well with values taken from the literature (Table 7 and Table S2 in Supplementary material). Zebrasoma xanthurum and A. sohal had measured NAE values of 67·07% and 72·28%, respectively, and these were used in the model. There was insufficient sample size to conduct ash analysis for P. arabicus, thus the mean literature NAE of 67·63% was used for this species (Table 7 and Table S2 in Supplementary material).

Table 7. Input parameters and results of fractionation model for three herbivorous species and one planktivore
Species Weight (g) R d δ15N q Ωin Z δ food ΔN
Acanthurus sohal 500·00 20 13·09 0·7228 0·021 0·141 9·03 4·41
Zebrasoma xanthurum 250·00 21 13·16 0·6707 0·018 0·141 8·74 4·30
Pomacentrus arabicus 60·00 17 14·62 0·6763 0·017 0·141 8·73 5·68
Chromis chromis * 20·26 4·68 6·90 0·93 0·031 0·110 3·64 3·26

model outputs

The model predicted δfood to be 9·03 for A. sohal, 8·74 for Z. xanthurum and 8·73 for P. arabicus (Table 7), giving ΔN values of 4·41%, 4·30% and 5·68%, respectively. These are much closer to the empirically derived ΔN estimates than to the usually accepted value of 3·4.

model sensitivity

In order to determine how feeding rate, absorption efficiency and excretion rate influence model estimation of δfood, sensitivity analysis was conducted whereby values of Ωin, q and Z were varied systematically and the impact on δfood noted (Fig. 2). To compare how this may differ between a herbivorous fish and a carnivorous fish, this analysis was repeated for Chromis chromis, a small planktivorous fish from the Mediterranean. All parameters were taken from Pinnegar (2000) shown here in Table 7.

Details are in the caption following the image

Sensitivity of δfood to varying model parameters. (a) q (absorption efficiency), (b) Rd (% body weight ingested per day) and (c) Z (enrichment due to excretion) for Acanthurus sohal (filled squares) and Chromis chromis (open squares), default parameters as in Table 7. δa for A. sohal is shown in each plot as a solid black line, and the arrows indicate the value of the parameter used in the model.

absorption efficiency

There was relatively little change in δfood when q was 0·5–1·0 (50%–100% efficiency; Fig. 2). At lower values of q, there was a dramatic decrease in δfood, with values becoming negative for values of q less than 0·35 for the herbivore and less than 0·47 for the carnivorous fish.

daily ration

Small changes in low levels of daily ration had the greatest impact on δfood. When the consumption rate was greater than 30% BWD, A. sohal showed little change in δfood; in comparison, C. chromis showed little change once BWD was > 7% (Fig. 2). When Rd was < 30% BWD, the model predicted very different values for δfood, with only a little increase in Rd for A. sohal, similar effects but to a lesser degree were seen for C. chromis at BWD values < 7% (Fig. 2).

excretion rate

There appeared to be a linear change in δfood with Z for both species (Fig. 2). However, the rate at which Z influenced δfood was reduced in the carnivore, between the values 0·1 and 0·2, Zδfood changed by c. 4 for C. chromis and by c. 8 for A. sohal.


In their natural settings, herbivorous fish trophic-step 15N/14N fractionation was significantly higher than 3·4. This contrasts with the studies that show herbivore trophic fractionation to be lower than the commonly cited ΔN of 3·4 (Owens 1987; McCutchan et al. 2003; Vanderklift et al. 2003). Previous studies have attributed high ΔN exhibited by animals consuming low-quality diets (high C : N ratio) to unknown animal material in the diet (assuming the diet contained more protein than was observed Pinnegar & Polunin 2001) or assumed that the animal was undergoing nutritional stress (Adams & Sterner 2000). The wild animals in this study were evidently in good condition, and visual inspection revealed no additional protein among materials in the diet other than from the algae. However, by using a dynamic model, incorporating absorption, feeding rate and excretion, we repeatedly predicted higher ΔN values for herbivorous fish. Previous fractionation models have used bioenergetics to determine δfood with ΔN based on literature values (Harvey et al. 2002). This is the first model to our knowledge to calculate both δfood and ΔN.

Herbivorous fish consume around 20% of their body weight per day compared to only 3%–4% for carnivorous fish (Horn 1989). In this model, food consumption rate was incorporated into the term Ωin, along with N content of the diet and the absorption efficiency, q. The increased Rd in herbivores would contribute to a greater ΔN if the %N of the diet and q were not much smaller than those of carnivores. However, these factors are related such that to meet the bioenergetic needs of an animal feeding on a low-N food, it is necessary to have a high Rd (Fris & Horn 1993; Choat, Clements & Robbins 2002; Choat, Robbins & Clements 2004). Conversely, an organism with an N-rich diet will not feed as much; Ωin of herbivorous and carnivorous species may thus be broadly similar, corresponding with the dietary requirements of herbivores and carnivores being significantly different (Pandian & Marian 1985). The Rd values used in the model were taken, where possible, from species of the same genus with similar feeding habits because a slight change in Rd can result in a significantly different prediction for δfood (Fig. 2). Rd may also be affected by a change in temperature as fish can alter their metabolic rates to suit their environment (Klumpp & McKinnon 1989), which may lead to seasonal variations in the observed ΔN. NAE calculations for A. sohal and Z. xanthurum were similar to those of other herbivorous fish from the literature. The ash marker method (Montgomery & Gerking 1980) used to determine NAE may, however, lead to some inaccuracies as it is based on the assumption that ash is indigestible and that all organic matter is absorbed by the fish. Herbivorous species with gizzard-like stomachs (e.g. Z. xanthurum) have previously been found to have negative assimilation efficiencies for some macronutrients, thought to be a result of high levels of inorganic materials retained in their guts (Crossman, Choat & Clements 2005). The gut material of Z. xanthurum used for ash analysis in this study was taken from the immediate anterior and posterior of the intestine to minimise excess inorganics not present in the diet material. Sediment or inorganic matter was present in the diet; however, the mean NAE obtained from ash analysis was within the published range. NAE has been positively correlated with the N content of food (Pandian & Marian 1985); hence, in the model, if N content of the food were to decrease, a decrease in NAE would be expected. Body weight, food ration and temperature significantly influence absorption (Pandian et al. 1985), yet, NAE may vary with the size of the fish (Lassuy 1984); this was not the case in this study. When q in the model was below the value of 0·5 (50% efficiency), the predicted value for δfood would decrease significantly. The q parameter was found to influence the difference between δa and δfood in a way opposite to Olive et al. (2003) whereby when q < 1, the isotopic ratio for the animal would be depleted relative to the isotopic ratio of the food (Olive et al. 2003). Fish in this study had a q value of > 0·5.

Stomach contents analysis showed Feldmannia, Phylophora and Enteromorpha to be among the most dominant algal genera, but these were not analysed for δ15N and N% content due to cost constraints. Similarly, it was not possible to obtain δ15N value for the detritus fraction of the P. arabicus diet. The omission of these dietary components may have led to errors in empirical estimations of ΔN. However, macroalgae δ15N varied little across genera, and since their nitrogen source was the same, it is unlikely the seven genera used would have significantly biased the mean value. The accuracy of these estimates could be improved by further analysis of all genera and weighting the contribution of each by their relative importance in the diet.

If accurate values are to be applied to trophic fractionation, controlling processes must be well understood. Our sensitivity analysis has highlighted that the value of Z is an important factor in determining the level trophic fractionation; however, Z has not yet been measured directly. Z has been estimated from diet switch experiments where the study animals were not in equilibrium with their diet. All the diet switch species were carnivorous; hence, the mean Z value used in the model would be appropriate for carnivores. Whether herbivores would be more accurately described with a significantly different Z value remains to be tested. There may be differences in Z between herbivores and carnivores as much more ingested N appears to be released as dissolved waste in carnivores than herbivores (Polunin & Koike 1987), and Z is therefore likely to be higher in herbivores. High nitrogen-use efficiency, whereby only a small portion of the ingested N is excreted, is an adaptation in herbivores to deal with low N intake and has been suggested to contribute to low ΔN values (Vanderklift et al. 2003). This may be the case for aphids and certain detritivores (Vanderklift et al. 2003) but does not seem to be the case for herbivorous fish as they are known to exhibit high fractionation values (Table 1). Differential nitrogen excretion has previously been suggested as a factor contributing to variance in ΔN (Minagawa & Wada 1984; Ponsard & Averbuch 1999; Vanderklift et al. 2003), but so far, researchers have failed to reach a consensus view. Excretion rate measurements in a range of species of differing trophic groups would potentially further our understanding of how these processes affect fractionation.

A single mean ΔN may seem useful and convenient in application to food web studies, especially to determine trophic level. However, by applying one value to determine δfood simply reflects the consumer signatures offset by 3·4. This approach may lead to the misinterpretation of the relative importance of potential food sources of a consumer. This study has emphasized the importance of determining fractionation values for consumers on a case-by-case basis. As the Olive model takes into account nutritional functionality, it has the potential to be used for a range of consumers in a food web, giving species-specific fractionation values. The model output – the value for δfood– could be used within isotope ‘mixing models’ (e.g. Phillips 2001) to determine the different proportions that contribute to the diet mixture (Koch & Phillips 2002; Lubetkin & Simenstad 2004).


We thank J. McIlwain who provided essential assistance and support in the field. C. Scrimgeour and R. McGill are also thanked for stable isotope analysis, and B. P. Jupp for algal identification. Staff at Sultan Qaboos University and the Ministry of Agriculture and Fisheries in Oman were instrumental in facilitating field work. This study was funded by a NERC studentship A/2003/11885 with a CASE collaboration with CEFAS.