Leaf biomechanical properties and the densities of herbivorous insect guilds


†Author to whom correspondence should be addressed. E-mail: paula.peeters@epa.qld.gov.au


  • 1This study investigated relationships between leaf biomechanical properties and the densities of their associated insect herbivores.
  • 2The herbivorous insects associated with 18 co-occurring plant species were sampled over 16 months. Biomechanical properties of new and mature leaves of each plant species were measured using punching, shearing and tearing tests.
  • 3Significant variation in leaf biomechanical properties was found among plant species, and between new and mature leaves.
  • 4Total insect density was significantly negatively correlated with work to tear (r = −0·43, P < 0·05) and work to shear (r = −0·70, P < 0·01).
  • 5Although chewing insect density was significantly correlated with punch strength of mature leaves, sucking insect density was not. While studies of herbivory often include measures of leaf punch strength, this mechanical trait may indicate resistance to chewing insects but not sucking insects.
  • 6We conclude that leaf biomechanical properties are influencing the functional composition of herbivorous insect assemblages in this system.


There is growing evidence that the biomechanical properties of leaves can be a formidable barrier to insect herbivores. Several studies have linked leaf biomechanical properties to levels of insect herbivory (Coley 1983; Lowman & Box 1983; Foggo, Speight & Gregoire 1994), and to insect growth (Feeny 1970; Larsson & Ohmart 1988; Aide & Londono 1989; Stevenson et al. 1993) and survival (Wint 1981; Ohmart, Thomas & Stewart 1987; Larsson & Ohmart 1988; Wheeler, Van & Center 1998).

Leaf biomechanical properties appear to influence the feeding behaviour of chewing insects in a number of ways. Some studies suggest that biomechanical properties can prevent feeding initiation by early larval instars, resulting in starvation (Ohmart et al. 1987; Larsson & Ohmart 1988; Wheeler et al. 1998). However, the smaller chewers can also feed selectively within a leaf, avoiding the toughest tissues (Choong 1996). Nichols-Orians & Schultz (1990) suggested that leaf toughness causes leaf cutter ants to avoid mature leaves in favour of new leaves, even though mature leaves may be more suitable for the growth of their symbiotic fungi. The feeding behaviour of sucking insects may also be influenced by plant biomechanical properties. The penetration of mature leaves by aleyrodid nymphs appears to be inhibited by cuticle thickness (Walker 1985), while increased lignification of stem tissues has been linked to a decline in the feeding ability of cercropid nymphs (Hoffman & McEvoy 1986).

The influence of leaf biomechanical properties on insect distribution in the field is less well known, although Feeny (1970) had evidence that the concentration of winter moth feeding in the spring may relate to the increased leaf toughness of oak leaves later in the year. Correlations between leaf biomechanical properties and insect densities among a range of plant species in the field do not appear to have been investigated, perhaps due to the attention given to the influence of plant chemicals and nutritional quality.

Important insights have been revealed by measures of leaf properties reported in many plant–insect studies, yet some confusion has arisen due to the use of ambiguous terminology for these properties (Vincent 1990; Choong et al. 1992; Wright & Vincent 1996; Sanson et al. 2001). Often the term ‘toughness’ is used to describe the leaf property measured with simple penetrometers. In the terminology of materials engineering, most studies that use simple penetrometers to test leaves are measuring force to fracture or strength but not toughness. Strength is the stress at which a sample fractures, and is calculated as force divided by the area over which the force is applied. By contrast, toughness is the work required to break the sample, and can only be calculated if both the applied force and the displacement of that force are measured. Therefore, toughness cannot be measured by simple penetrometers that do not simultaneously measure the force and displacement.

In the present study, leaf biomechanical properties were examined by shearing, punching and tearing tests using a universal force testing machine, which overcomes limitations associated with simple penetrometers (Sanson et al. 2001). The mechanical tests used were not intended to closely mimic the action of insect mouthparts, as simulation of the wide range of feeding methods employed by different insect herbivores would require many different testing methods. However, the testing methods used here do provide an accurate measure of biomechanical properties for comparisons among plant species and with other studies. Biomechanical properties are divided into ‘structural’ properties that are not normalized to leaf thickness (strength and work) and ‘material’ properties that are normalized to leaf thickness (specific strength and specific work) (Read & Sanson 2003).

Variation in the densities of herbivorous insect guilds among co-occurring plant species in Bunyip State Park was documented by Peeters, Read & Sanson (2001). Although Peeters (2002a) suggested that leaf constituents may be influencing this variation, it is unclear whether certain leaf constituents (e.g. fibre) are acting as physiological barriers to insects (e.g. by reducing leaf digestibility) or mechanical barriers (e.g. by making the leaf too strong to chew). Indirect evidence for the importance of leaf mechanical traits was subsequently provided by correlations between insect densities and leaf structural traits (e.g. cuticle thickness and lignified vein area; Peeters 2002b). The present paper seeks to further clarify the influence of leaf biomechanical traits on arboreal insect assemblages in Bunyip State Park by examining correlations between leaf biomechanical properties and herbivorous insect densities.


study site and plant species

The study site is located in Bunyip State Park, Victoria, Australia (145°E, 37°S) in tall open eucalypt forest, at an altitude of 120–220 m. The site experiences a temperate climate (mean winter minimum 2·6 °C, mean summer maximum 24·8 °C) with a median annual rainfall of 1000 mm (see Peeters et al. 2001 for further details).

Eighteen shrub species were selected to represent a variety of leaf types and to maximize contrasts within families and genera (Table 1). All plant species were evergreen, although they varied in the timing and duration of new leaf production (Peeters et al. 2001).

Table 1.  Plant species sampled in this study and their abbreviations (code) used in this paper. Nomenclature follows Walsh & Entwisle (1996, 1999)
MonimiaceaeHedycarya angustifoliaha
MimosaceaeAcacia myrtifoliaam
Acacia genistifoliaag
FabaceaePultenaea muelleripm
Pultenaea weindorferipw
RutaceaeBoronia muelleribom
Correa reflexacr
Phebalium bilobumpb
Zieria arborescensza
RhamnaceaePomaderris asperapa
Spyridium parvifoliumsp
ProteaceaeBanksia marginatabm
Banksia spinulosabs
Grevillea barklyanagb
Hakea sericeahs
Hakea ulicinahu
Lomatia fraserilf
LamiaceaeProstanthera lasianthospl

insect collection

Insects were collected by branch sampling on 12 occasions over 16 months from October 1994 to January 1996. On each occasion, three branches were sampled from three individuals of each plant species, and the presence of new leaves in branch samples was noted. Sampling took place during daylight and in dry weather only. Surface area of branch samples was measured (Peeters 2002a) and insect abundance was converted to insect density per unit surface area of branch. Samples were divided into those containing mature leaves only, and those containing new plus mature leaves. For branch samples, ‘mature’ leaves were defined as fully expanded and hardened, while leaves that were not fully expanded, or fully expanded but not hardened, were considered ‘new’ see Peeters (2002a) for more details.

Insects were placed in herbivore guilds (Table 2) based on insect morphology, feeding method and target tissues as reported by Lewis (1973), Washington & Walker (1990), CSIRO (1991), Sadof & Neal (1993) and Novotny & Wilson (1997). The family Cicadellidae (Hemiptera) has been put into its own guild for two reasons. First, different tribes within the Cicadellidae may target phloem, mesophyll or xylem (Novotny & Wilson 1997). Cicadellids collected in the present study were usually not identified beyond family and thus the plant tissues targeted by individuals are unknown. Furthermore, unlike many Hemiptera, cicadellids probably do not use salivary enzymes for leaf penetration (based on the work of Pollard 1968). Suckers were divided into those that probably use salivary enzymes (insects grouped into meso-, phloem- and xylem feeders), and those that do not (cicadellids) so that the association of these two groups with certain leaf mechanical and structural features could be compared. The term herbivore, as used here, includes insects eating above-ground plant tissues, but excludes insects that specialize on pollen, nectar or fruit. The guilds (sensu Root 1967) formulated specifically for this study represent one example of many categories that could be used to define functional groups of insect herbivores (e.g. Cornell & Kahn 1989; Basset & Burckhardt 1992; Root & Cappuccino 1992). It is acknowledged that the guild system used here is likely to have many flaws, and the assignation of member taxa may be incorrect due to our incomplete knowledge of the biology of arboreal insects associated with Australian plant species. For more details see Peeters et al. (2001).

Table 2.  Insect guilds
Mobile phloem feederspmHemiptera: Psylloidea, Aphidoidea, Fulgoroidea, Cicadelloidea (excluding Cicadellidae).
Sessile phloem feederspsHemiptera: Coccoidea, Aleyrodoidea.
Mobile mesophyll feedersmmHemiptera: Heteroptera.
Sessile mesophyll feedersmsHemiptera: Diaspididae.
Xylem feedersxHemiptera: Cercopoidea.
CicadellidscicHemiptera: Cicadellidae.
Shallow suckers/chewerssscThysanoptera; Diptera: larvae; grubs of unknown order.
External chewersxcLepidoptera: all herbivorous larvae except miners; Coleoptera: herbivorous adults (excluding Curculionoidea), and all larvae.
Internal chewersicLepidoptera: mining larvae; miners of unknown order.
Rostrum chewersrcColeoptera: Curculionoidea (adults).

leaf collection for biomechanical tests

The term ‘leaf’ is used here to include leaflets (Boronia muelleri and Zieria arborescens), phyllodes (Acacia genistifolia and A. myrtifolia) and simple leaves (all other species). Five replicate new leaves and five replicate mature leaves were collected from replicate plants of each shrub species on the same day in December 1995, and both leaves and plants were selected haphazardly. Leaves were collected in the morning during continuous light rain, were removed with scissors or secateurs and were placed in plastic ziplock bags with moistened paper towelling. Samples were placed in cooled, insulated containers, and were transferred to the laboratory for processing later that day. ‘New’ leaves were fully expanded but not hardened, while ‘mature’ leaves were selected from the previous year's flush. Although we would have preferred to collect and analyse leaf traits every time we collected insects, this was unfortunately not possible because of the time required to process both insect and leaf samples. However, we believe that biomechanical data from newly expanded leaves and mature leaves from last year's flush of a plant species will give a good indication of the range of leaf biomechanical properties that would be encountered by an insect feeding on expanded leaves of that species at other times of the year. Most plant species sampled were shrubs of less than 3 m height with a relatively open and sparse growth habit, growing in relatively sunny positions. This meant that most leaves collected in branch samples, and those collected for testing of biomechanical properties had traits generally associated with ‘sun’ leaves. Exceptions to this were some species that grew in south-facing gullies in more shaded situations (L. fraseri, H. angustifolia, P. lasianthos, Z. arborescens and P. aspera). The branch and leaf samples of these species were more likely to contain leaves with traits associated with ‘shade leaves’. As a result, there was consistency in the type of leaf (sun or shade, or intermediate) collected within a plant species, but there was unavoidable variation in the type of leaf selected across plant species. This variation in leaf traits across species was a necessary and desirable part of the study.

analysis of leaf biomechanical properties

Shearing, punching and tearing tests (discussed in Aranwela, Sanson & Read 1999; Sanson et al. 2001; Read & Sanson 2003) were used to measure seven leaf biomechanical properties: punch strength, specific punch strength, work to punch, specific work to punch, work to shear, work to tear and specific work to tear (calculation and units described in Read & Sanson 2003). All tests were undertaken using a universal force testing machine (Chatillon Universal Tension and Compression Tester, model UTSE-2), which was modified to produce a data output of up to 800 points per second. Biomechanical properties were derived from force-displacement curves using leaf software (M. Logan, Monash University). A blank run was used to record background friction between every 30 test runs, and was subtracted from the force–displacement curves of the test runs.

Shearing tests were performed using a guillotine with silver-steel gauge plates with the cutting edge of the upper blade at 12° to the horizontal, while the lower blade was horizontal, with a relief angle of c. 5°. The widest part of each leaf was placed on the lower blade, which was raised past the upper blade at a displacement speed of 0·5 mm s−1. This created a shearing action and cut the leaf transversely into two parts.

Punching tests utilized a flat-ended, sharp-edged cylindrical steel punch with a 1·2 mm diameter (1·13 mm2 punch area), and a die mounted onto the moving head of the test machine. The die was moved upwards at a constant displacement speed of 0·5 mm s−1, causing the stationary punch to pass through a hole with a clearance of 0·068 mm. The punch was applied halfway along the longitudinal axis of the leaf and halfway between the midrib and the leaf margin. As primary and secondary veins were avoided where possible, the punch test measured the properties of the intercostal lamina and tertiary veins for most leaves of this study. However, secondary veins were included in the punch test for Hakea ulicina and Pultenaea muelleri, due to the close spacing of veins in these species. Punch data could not be collected for A. genistifolia, H. sericea and P. weindorferi due to the narrowness of their leaves.

Tearing tests used a longitudinal strip of lamina cut from the middle of the left-hand side of the leaf (if the leaf lamina was sufficiently big) such that the length was more than eight times the width to counter the effects of necking (Vincent 1990). For smaller- or narrow-leaved species the whole leaf had to be used, rather than a strip of lamina, to achieve the recommended aspect ratio. Test strips were clamped in the force-tester using pneumatic clamps. The strips were notched on the left-hand side to direct the position of fracture to prevent the test strip from breaking at the secured ends. Any effect of notch sensitivity (Vincent 1990; Lucas et al. 1991) was minimized by standardizing the notch length relative to the test strip. Replicate tear data could not be collected for A. genistifolia (new), H. ulicina and P. muelleri (mature) as most test leaves of these species slipped out of the pneumatic clamps without tearing during the tests, or fractured at the clamps rather than the notch.

Leaf width was measured with calipers just prior to mechanical testing for each leaf tested. Average lamina thickness was measured from resin-embedded leaf sections (n = 3; Peeters 2002b) as caliper measurements of fresh leaves were found to overestimate lamina thickness.


Biomechanical variables were compared among species and between new and mature leaves using anova. Principal components analysis (PCA) was performed on biomechanical parameters of new and mature leaves and on mature leaves only. Pearson correlations were used to assess relationships between leaf biomechanical traits, mean guild densities and the first principal component of leaf biomechanical trait PCAs. Mean guild densities were calculated as for Peeters (2002a). Correlations compared (1) new and mature leaf traits with guild densities calculated from samples containing new leaves plus mature leaves, and (2) mature leaf traits with guild densities calculated from samples containing mature leaves only. All analyses were performed using systat v.7 for Windows (SPSS Inc.). Variables were transformed where appropriate to comply with assumptions and the alpha level for all tests was P < 0·05. P-values were not adjusted because we consider that the adjustment of P-values requires a subjective decision (i.e. adjust across the correlations of one paper, across several papers, or across the entire study?) and does more to obscure the results than clarify them (Stewart-Oaten 1995). By presenting unadjusted P-values the reader has the opportunity to make whatever adjustments they deem to be appropriate.


leaf biomechanical properties

Leaf biomechanical properties varied significantly among plant species and between new and mature leaves, although there were also significant interactions between species and age for tearing variables (Figs 1–3, Table 3). All variables were generally higher for mature leaves than new leaves, although the magnitude of the difference varied among plant species (Figs 1–3). Plant species have been grouped in Figs 1–3 to enable comparison between leaf types (kite, box and spine) that are based on the arrangement of their lignified tissues (Grubb 1986; Peeters 2002b). Kite leaves have unfortified tissue supported by a lignified framework of veins, while box leaves have a lignified hypodermis covering the upper surface of the leaf and extending down the revolute leaf margins, and spine-like leaves are in the shape of flattened or cylindrical spines (Grubb 1986; Peeters 2002b). Spine-like leaves tended to have values of structural and material properties that were generally greater than those of other leaf types (Figs 1–3). The box leaves of Banksia marginata also tended to have high values of structural and material properties (Figs 1–3), and higher values than the box leaves of B. spinulosa, except for tearing properties where values for both Banksia species were similar (Figs 2c and 3b). Structural values for most kite leaves were generally low compared with B. marginata and spine-like leaves, with the exception of the work to shear values of Grevillea barklyana and Lomatia fraseri (Figs 1 and 2). Material values for most kite leaves were also generally low compared with B. marginata and spine-like leaves, with the exception of Pultenaea muelleri (Figs 1b and 3).

Figure 1.

Strength measurements (mean + SE) for new (open bars) and mature leaves (shaded bars) of 15 plant species measured by punching. Plant species have been grouped by leaf type (kite, box or spine-like). See Table 1 for an explanation of plant species abbreviations. Leaves of B. marginata (new) were not tested.

Figure 2.

Work measurements (mean + SE) for new (open bars) and mature leaves (shaded bars) of 18 plant species. Plant species have been grouped by leaf type (kite, box or spine-like). See Table 1 for an explanation of plant species abbreviations. Leaves of B. marginata (new) were not tested, leaves of P. weindorferi, A. genistifolia and H. sericea could not be tested by punching, and leaves of H. ulicina (mature) could not be tested by tearing.

Figure 3.

Specific work measurements (mean + SE) for new (open bars) and mature leaves (shaded bars) of 18 plant species. Plant species have been grouped by leaf type (kite, box or spine-like). See Table 1 for an explanation of plant species abbreviations. Leaves of B. marginata (new) were not tested. Leaves of H. ulicina (mature) could not be tested by tearing.

Table 3. anova results comparing mechanical properties between new (n) and mature (m) leaves, and among plant species (sp.). All data were loge-transformed
Punch strength
n & msp13 6·4472·72< 0·001
age1 4·0045·18< 0·001
sp × age13 0·10 1·080·38
error111 0·09  
Newsp13 2·8822·83< 0·001
error56 0·13  
Maturesp14 4·1786·07< 0·001
error59 0·05  
Specific punch strength
n & msp13 4·3852·86< 0·001
age1 1·1313·67< 0·001
sp × age13 0·08 0·990·467
error111 0·08  
Newsp13 2·0116·56< 0·001
error56 0·12  
Maturesp14 2·8667·41< 0·001
error59 0·04  
Work to punch
n & msp13 9·7571·69< 0·001
age1 4·3031·57< 0·001
sp × age13 0·10 0·750·709
error111 0·14  
Newsp13 4·7426·26< 0·001
error56 0·18  
Maturesp14 6·1272·0< 0·001
error59 0·09  
Specific work to punch
n & msp13 6·7852·71< 0·001
age1 1·2910·040·002
sp × age13 0·05 0·380·975
error111 0·13  
Newsp13 3·4119·44< 0·001
error56 0·18  
Maturesp14 4·3056·05< 0·001
error59 0·08  
Work to tear
n & msp1311·2727·71< 0·001
age111·3227·83< 0·001
sp × age13 2·27 5·585< 0·001
error102 0·407  
Newsp14 3·18 6·81< 0·001
error52 0·47  
Maturesp1512·7038·80< 0·001
error58 0·33  
Specific work to tear
n & msp13 7·5017·44< 0·001
age1 6·5215·19< 0·001
sp × age13 2·43 5·67< 0·001
error102 0·43  
Newsp14 3·39 7·15< 0·001
error52 0·47  
Maturesp15 8·5423·50< 0·001
error58 0·36  
Work to shear
n & msp1619·3560·81< 0·001
age1 8·6727·26< 0·001
sp × age16 0·42 1·330·187
error128 0·32  
Newsp16 9·0123·88< 0·001
error63 0·38  
Maturesp1711·1844·94< 0·001
error69 0·25  

The first three axes of the principal components analyses of leaf biomechanical traits explained 97% of the variation for new and mature leaves, and 99% of the variation for mature leaves. The first component explained 82% of the variation for new and mature leaves, and 84% of the variation for mature leaves (Table 4). All variables recorded component loadings greater than 0·80 on component 1 (Table 4) and tear variables contributed most to component 2 for both PCAs.

Table 4.  Component 1 loadings of the PCAs of (a) new and mature leaf mechanical traits and (b) mature leaf mechanical traits
Punch strength0·950·97
Specific punch strength0·930·95
Work to punch0·940·97
Specific work to punch0·970·97
Work to tear0·850·83
Specific work to tear0·800·82
Work to shear0·900·90
Percent of total variance explained82·1684·40

correlations between leaf biomechanical properties and insect density

Analyses of individual biomechanical traits of new and mature leaves revealed significant negative correlations with the densities of five herbivore guilds, total suckers, total chewers and all herbivores (Table 5a). Densities of total chewers were correlated with all biomechanical traits and PC1 of the new and mature leaf PCA (Table 5a). Rostrum chewers were correlated with all biomechanical traits except specific punch strength (Table 5a). All herbivores and external chewers were correlated with work to tear and work to shear (Table 5a). Densities of total suckers were correlated with tearing and shearing variables (Table 5a). Cicadellids were correlated with work to punch and work to shear, while phloem feeders were correlated with work to shear (Table 5a). Densities of mobile phloem feeders were also correlated with PC1 of the new and mature leaf PCA (Table 5a).

Table 5.  Significant results (r-values) of Pearson correlations comparing mean insect densities to mean leaf mechanical traits and PC1 of the principal components analysis of (a) new and mature leaves [n = 29 (punch and tear), 35 (shear)] and (b) mature leaves [n = 15 (punch), 16 (tear), 18 (shear)]. Xylem feeders and internal chewers were not analysed separately due to non-normal or sparse data
 Punch strength§Specific punch strength§Work to punch§Specific work to punch§Work to tear§Specific work to tear§Work to shear§PC1
  • Significance is indicated by asterisks: ***P < 0·001, **P < 0·01, *P < 0·05. ‘–’ = not significant. See Table 2 for insect guild abbreviations. †Transformed using loge(x + 1).

  • Transformed using square root (loge(x + 1)).

  • §

    Transformed using loge(x).

  • Transformed using loge(x + 2).

(a) New and mature leaves
Total suckers−0·50**−0·36*−0·61***
cic & ssc
Total chewers−0·54**−0·52**−0·56**−0·56**−0·49**−0·37*−0·61***−0·42*
All herbivores−0·43*−0·57***
(b) Mature leaves
Total suckers−0·53*−0·64**
cic & ssc
Total chewers−0·65**−0·66**−0·65**−0·68**−0·53*−0·50*−0·64**
All herbivores−0·55*−0·70**

Analyses of individual biomechanical traits of mature leaves revealed significant negative correlations with three herbivore guilds, total suckers, total chewers and all herbivores (Table 5b). Densities of total chewers were correlated with all biomechanical traits (Table 5b). External chewers were correlated with tearing and shearing variables, and rostrum chewers were correlated with work to shear (Table 5b). All herbivores and total suckers were correlated with work to tear and work to shear, and sessile phloem feeders were also correlated with work to shear (Table 5b).


This study reveals significant correlations between leaf biomechanical properties and herbivorous insect densities. These findings are consistent with reports linking low levels of insect herbivory with high levels of leaf biomechanical properties (Coley 1983; Lowman & Box 1983) and also support the premise that some structural traits that appear to influence insect densities probably achieve this by being mechanical barriers to insects (Peeters 2002b). However, the high level of intercorrelation among leaf nutrient, structural and biomechanical traits (Peeters 2001) may mean that correlations between insect densities and biomechanical traits are caused by leaf features other than, or in combination with, biomechanical traits. Therefore any discussion of the results presented in this paper should also consider the correlations between insect guild densities and other leaf traits reported for the same plant species (Table 6).

Table 6.  Summary results of Pearson correlations between insect guild densities and (a) leaf constituents (Peeters 2002a) and (b) leaf structural traits (Peeters 2002b). Leaf age preference of guilds (Peeters et al. 2001) is also indicated
  1. + = significant positive correlation, – = significant negative correlation and shading = no significant correlation (P < 0·05). See Table 2 for insect guild abbreviations.

Leaf age preferenceNewNewNewMatureNew
Leaf constituents
 N   ++
Leaf structure
 PC1 structure (new and mature leaves) +++ 
 Specific leaf weight   
 Cuticle thickness  
 Lignified vein area   
 Distance between lignified tissues+    
 Leaf surface area+    

The negative correlations between chewer density and punch strength (Table 5) suggest that some chewers are simply not strong enough to penetrate the leaf. If this is the case, one would expect smaller insects to be excluded from eating leaves that are very strong, assuming that insect size is an indication of bite strength. This may result in the exclusion of species of small insects and early instars from strong leaves.

Several studies have suggested that early instar larvae of some chewers are unable to feed on leaves because of their biomechanical properties (Ohmart et al. 1987; Larsson & Ohmart 1988; Wheeler et al. 1998). Pappers et al. (2001) found that beetles originating from plant species with strong leaves had larger eggs, larger head capsules as first instar larvae, larger adult body size, and adults also had disproportionately bigger mandibles than those of the same species originating from plant species with leaves that were less strong. In addition, Braby (1994) reported that caterpillars that hatched from heavier eggs tended to have larger head capsules, and also had a significantly better chance of surviving on strong leaves than smaller caterpillars that hatched from lighter eggs. This suggests that the superior strength of larger caterpillars enabled them to eat the stronger leaves. The advantage of heavier egg size was diminished when the caterpillars were reared on softer leaves (Braby 1994). Therefore, it is likely that strength limitations may result in strong leaves being avoided by chewing insects, leading to lower densities of chewers. However, accurate measures of the bite strength of herbivorous chewing insects are needed to determine whether these insects are constrained by leaf strength.

Even if chewers are strong enough to penetrate strong leaves, they may not be able to expend the energy needed to process leaves that are very tough. Toughness, in the materials engineering sense (rather than the general sense used in many ecological papers), is a measure of the resistance of a material to crack propagation. Hence, it is a measure of the work that must be done to fracture a leaf, and the work needed to bite and remove leaf tissue from some leaves may be beyond the energy budgets of some chewing insects. Roces & Lighton (1995) found that extremely high levels of metabolic activity were associated with the leaf-cutting activity of leaf-cutter ants. Choong (1996) found that veins were the toughest structure within Castanopsis fissa leaves, and that these were avoided by feeding caterpillars. In the present study, significant negative correlations between chewer densities and both work and specific work (Table 5), suggest that a lack of expendable energy may be limiting the densities of chewing insects on tough leaves. Furthermore, some studies have found that chewing insects capable of feeding on strong leaves none the less suffer from lower larval and pupal weights (Feeny 1970), perhaps due to the energy spent in food acquisition. In addition, chewing insects that process strong leaves may also experience increased mandible wear (Raupp 1985) and increased risk of predation and parasitism (Mueller & Dearing 1994).

Other leaf traits may also be contributing to the correlations between leaf biomechanical properties and chewer densities. As high leaf strength is often associated with low leaf nitrogen levels (Feeny 1970; Coley 1983; Read et al. 2000; Read & Sanson 2003), the energy expenditure of insects chewing strong leaves may be even less sustainable due to the poor food quality of these leaves. In addition, other leaf traits not measured in this study may also be contributing to the patterns observed, including secondary chemicals that inhibit digestion (Swain 1979).

External chewers showed a significant preference for new leaves and appear to be responding to the structural and mechanical attributes of leaves, rather than nutritional aspects alone (Tables 5 and 6). Choong (1996) also observed that caterpillars preferred new leaves despite the higher levels of total soluble phenolics they contained, and suggested that the higher toughness of mature leaves contributed to the feeding patterns observed.

Rostrum chewers also appear to be responding more to leaf structural and mechanical traits than leaf nutrient content (Tables 5 and 6). However, the negative correlations with cellulose, lignin and fibre (Table 6) may indicate that rostrum chewers require leaf tissue of higher quality, and they have less capacity to cope with fibre as a diluent, compared with external chewers. As rostrum chewers are adult insects, their gut size is restricted by a hard exoskeleton, the presence of reproductive organs, and the need for flight. Because of this, rostrum chewers are likely to need small quantities of high-quality food. Other adult Coleoptera may also have this constraint and are a small component of the external chewer guild. However, the larvae that make up the largest component of the external chewer guild have a softer exoskeleton, can expand their gut, and have no requirement for flight or reproduction. These insects can probably obtain the nutrients they need by ingesting large quantities of low-nutrient food.

The preference of cicadellids for new leaves appears to be more closely related to leaf biomechanical and structural properties than to leaf constituents (Tables 5 and 6). The negative correlation recorded between cicadellids and work may reflect the feeding methods of these insects, as cicadellids and shallow suckers/chewers use physical force to penetrate the plant, and probably do not utilize salivary enzymes (Pollard 1968; Lewis 1973). Thus, leaves measured as tough may deter these insects more than most other sucker guilds. However, there are no correlations to indicate that this is the case for shallow suckers/chewers.

Although sessile phloem feeders appear to be responding to leaf structural (Table 6) and mechanical traits (Table 5), this should be questioned given their biology. Sessile phloem feeders use salivary enzymes for leaf penetration and their stylet diameters are extremely small (approximately 3 µm for some aleyrodids and coccids compared with 9–10 µm for some cicadellids; Pollard 1968; Miles 1972). Therefore, the combination of narrow stylets and enzymatic digestion is likely to enable the penetration of strong and tough leaves. Unlike mobile suckers, sessile suckers tend to use one stylet track for long periods, hence the energetic cost of penetrating the leaf would be repaid many times over by the food reward gained. As sessile phloem feeders target phloem tissue and feed on phloem sap, the density and distribution of these insects may be strongly influenced by phloem sap quality and plant photosynthetic rate. This is supported by the correlation between sessile phloem feeders and leaf mass : area (= specific leaf weight, Table 6), and correlations linking leaf mass : area ratios and photosynthetic rate (Reich et al. 1991; Reich, Walters & Ellsworth 1992; Poorter & Evans 1998; Wright, Reich & Westoby 2001). In addition, the positive correlation between leaf mass : area and work to shear (Read & Sanson 2003), may indicate that work to shear is negatively correlated with photosynthetic rate per leaf mass, and this may underlie the negative correlation between sessile phloem feeders and work to shear.

Mobile phloem feeders were correlated with leaf constituents (Table 6) and work to shear (Table 5), but were not correlated with any leaf structural trait (Table 6). Mobile phloem feeders also use salivary enzymes for leaf penetration and have very narrow stylets. Therefore, as was suggested for sessile phloem feeders, the correlation between mobile phloem feeders and work to shear is more likely to result from a response to plant photosynthetic rate than a direct response to leaf mechanical properties. These observations indicate that leaf structural and mechanical properties, at the resolution measured in this study, do not have a strong influence on mobile phloem feeders.

Densities of all insect herbivores were negatively correlated with work to tear and work to shear, but were not correlated with punch strength (Table 5). This potentially indicates that punch strength, which is often used as a measure of ‘leaf toughness’ in studies of herbivory, provides an inferior estimate of the potential densities of insect herbivores on a plant compared with work to tear and work to shear. However, this seems unlikely as many leaf biomechanical properties are strongly correlated (Read & Sanson 2003), and this result may actually be because some of the toughest leaves were omitted from the punching test. None the less, it is worth noting that while chewing insect density was significantly correlated with mature leaf punch strength, sucking insect density was not (Table 5). Therefore leaf punch strength may indicate resistance to chewing insects but not sucking insects.

Although the importance of leaf biomechanical traits to insect herbivores cannot be conclusively determined from this study, these findings suggest that leaf biomechanical properties may have a large influence on the within- and between-species distribution of some groups of herbivorous insects in the field. However, it is unlikely that leaf biomechanical properties have evolved in response to herbivore pressure alone, but rather have arisen from complex factors in addition to herbivores, including plant physiology and other agents of physical damage.


Thanks to F. Clissold for assisting with leaf collection and mechanical testing; to A. Yen, P. McQuillan, M. Malipatel, P. Gullan, B. Thomas and M. Fletcher for insect identification; to F. Clissold, R. Carpenter and T. Peeters for assistance with insect collection; to G. Quinn and R. McNally for statistical advice; to M. Logan for writing the software, and to P. Domelow for help in the workshop. The Department of Natural Resources and Environment (Victoria) granted permission to work in Bunyip State Park. The support of an Australian Postgraduate Award is gratefully acknowledged by P. Peeters.