In this section we discuss the many features of the New Zealand flora that have been interpreted as anti-browsing adpatations (Table 1).

Toxins and other chemical defences, and low nutrient status

Batcheler (1989) suggested that because a number of low-statuted, fast-growing plants normally found on seral sites have toxins (e.g., Sophora, Coriaria, Brachyglottis repunda, Solanum have alkaloid defences), this is evidence for plant defence against moa browsing. His hypothesis, following from a similar line of reasoning by Greenwood and Atkinson (1977), is that moa browsed preferentially on nutrient-rich sites and on the lower strata of forest, and therefore high levels of defence could be expected in plants that habitually grew there. Low nutrient page 4 quality of some leaves has also been suggested as an anti-moa defence, as for instance the thick, tough, nutrient-poor leaves of Pseudopanax crassifolius (Mitchell 1980).

First of all, there is no reason to assume that toxins, carbon-based chemical defences, and low nutrient and fibrous leaves are specifically anti-moa browsing adaptations; they are effective against a wide range of herbivores, including invertebrates. All the indigenous species of flowering plants in New Zealand toxic to livestock are related to toxic species elsewhere, and no toxin restricted to New Zealand has yet been recognised (Connor 1977). The New Zealand situation appears to be a particular example of a universal trend.

Secondly, toxins may be a preferred means of defence for plants growing on nutrient-rich sites, and the concentration of toxic plants on such sites must be seen in terms of general plant defence strategies. The faster-growing a plant, the less profitable it is for it in terms of absolute growth rate to invest in defence (Coley et al. 1985). Pioneers, for example, accumulate nutrients more quickly than mature forest species, and allocate a great proportion of their resources to growth, but little to defence (Dirzo 1984). As fast-growing plants on nutrient-rich sites have more nitrogen to spare for defence than do slow-growing plants on poor sites, successional plants often possess nitrogen-based toxins such as alkaloids, which have a high turnover within the plant and can be produced in variable quantities according to need (Rhoades 1983). On the other hand, slow-growing plants in low-light, nutrient-poor, or stressed environments rarely have nitrogen-based toxins for defence, but rely on carbon-based chemicals such as polyphenolic compounds and lignins, which are less energy-demanding and use mimimal amounts of scarce nutrients. Nutritive qulity of leaves is usually low in species growing on poor soils, and such leaves tend to be sclerophyllous and are retained longer, possibly to minimise nutrient loss through herbivory, environmental stress, and nutrient leaching (Chabot and Hicks 1982). However, although low nutrient status leads to less browsing through herbivore avoidance, the long life of the leaves means the browsing that does occur has a great impact.

In general, therefore, plants growing in low-nutrient and environmentally stressed sites have well-defended leaves that are not browsed to the same extent as those from plants growing on richer sites. Fast-growing plants invest more in growth than defence, and rely on growing through their vulnerable phases, or deploy facultative defences such as highly effective low-bulk toxins. These defence strategies in the New Zealand flora are not necessarily a consequence of moa-browsing but a generalised adaptation to all herbivory.

Reduced apparency and mimicry as browsing defences

The risk of attack on a given plant by a herbivore is referred to as its “apparency”. Plants can reduce their apparency by modifications that variously make the plant look dead, as if it has already been attacked, like a non-plant object such as a stone, like another distasteful plant, or render it less visible. To a greater or lesser extent all plants employ an apparency strategy, as nearly every change in their external appearance or life history must alter the ease with which they are located and attacked by herbivores.

Mottled leaves. Atkinson and Greenwood (1989) suggested that mottled leaves in seedling plants may have reduced discovery by browsing moa (e.g., Parsonsia capsularis, and Pseudopanax crassifolius juveniles). The irregular blotching breaks up the outline of the leaf, making it difficult to discern. In New Zealand leaf mosttling seems to be a common feature of many shrubs and trees. Mottling is also more common in exposed, sunlit leaves than in shaded leaves (for instance in Pseudowintera colorata). A striking feature of many mottled leaves is their page 5 resemblance to dead or senescing leaves. Stone (1979) suggested that the brown fuyenile leaves of some understorey palms in Malaysia escape browsing because of this similarity.

Some insects use colour vision and leaf patterns to locate food plants, although not as commonly as they use smell and taste (Prokopy and Owens 1983). Smith (1986) has suggested that mottling of the leaves of Byttneria aculeata (a subcanopy vine in Panama) mimics effects of leaf miners, and thus discourages ovipositing and leaf damage. He showed that mottling was more common in clearings than in shady sites, possibly because of the trade-off between reduced photosynthesis and decreased insect damage in mottled leaves, and the reverse in unmottled leaves. Givinish (1990) has shown that in northeastern North America mottled leaves are more common in herbs of shaded forest understoreys than in any other growth form, and that they are essentially absent in trees, shrubs, herbs or vines of sunny sites. He suggested that mottling camouflages from vertebrate herbivores the foliage of particularly vulnerable phenological groups (evergreens; spring ephemerals) in light-dappled understoreys. There is therefore a strong possibility that some leaf mottling is directed against herbivores, but investigation of mottled plants in New Zealand would be needed to establish whether this is so here.

Dark coloration. Dark bronze and purple leaves are common in the New Zealand flora, in particular in new foliage and the exposed foliage of high-altitude plants. Blanc (1989: quoted in Givinish 1990) suggested that dark-coloured foliage held close to the ground could be camouflaged against herbivores. In nearly all plants he examined, dark-coloured understorey leaves were within 20 cm of the ground surface. If the foliage is held higher the apparency advantage is largely lost as the leaves are viewed against a contrasting background rather than the dark soil. However, most New Zealand examples of dark-leaved understorey plants (Table 1) I hold their foliage well over 20 cm above the ground, and furthermore, the coloration of the leaves varies seasonally.

It is also possible that herbivory is not involved in the New Zealand situation. Other possible adaptive explanations for dark leaves include modification of the heat balance of the leaf, protection from ultra-violet radiation, response to frost damage, and modification of the photosynthetic capacity of the leaf. Without more investigation we cannot choose between these explanations.

Mimicry. When a plant strongly resembles another object, living or non-living, less palatable than itself, it is often regarded as a mimic. Many cases of mimicry have been alleged but are exceptionally difficult to prove, as assessment of the similarity is usually subjective (see Edmunds 1990 for a lucid discussion). Mimicry by living leaves and stems of unpalatable objects such as stones and dead twigs is essentially different from the better known examples of camouflage mimicry in the insect world. When an insect mimics a leaf, twig or stone, only in a very superficial way does it take on the characteristics of that object. It remains an insect, heterotrophic and mobile, and its energy exchanges and life-style are but marginally affected. However, when a plant mimics an inanimate object, it is changed totally. A leaf or stem that resembles a dead twig, for instance, is very different from a green leafy stem. Its changed shape, ridigity, and pigmentation will substantially alter its heat absorption, gas exchange, light energy capture, construction costs, and competitive relationships. We should therefore be cautious in assuming that a given plant is mimicking an inanimate object, however close the resemblance; its resemblance may be a consequence of environmental factors. Atkinson and Greenwood cited several examples of the various types of mimicry. None of them is convincing.

Some New Zealand plants resemble dead twigs (Table, 1). Muehlenbeckia ephedroides resembles plants from arid regions that have a similar grey pubescence and leafless habit. It is typical of exposed unstable substrates such as river beds page 6 and gravel beaches, which in common with arid regions are largely unvegetated, subject to intense solar radiation, and have a tendency to droughtiness. It could be as cogently argued that Muehlenbeckia's leafless grey stems are an adaptation to the harshness of the local environment and its prostrate habit a response to frequent flooding and abrasion.

The reduced twig-like leaves of the juveniles of forest-dwelling Pittosporum obcordatum, are assumed to increase its resemblance to a dead or partially browsed plant (Atkinson and Greenwood 1989:83). However, there are plausible reasons why a juvenile plant may restrict its photosynthetic surfaces, the most likely of which is the necessary diversion of scarce resources to roots (Wright 1992). The prominence of the whitish mark along the midrib is a consequence of the reduction of the reaf blade; it may or may not be significant in apparency.

When one plant is allegedly mimicking another plant, the problem is more complex. It first has to be established that the model is both less palatable to herbivores and more widespread and abundant than the mimic plant, and that the two taxa frequently occur together, Without these conditions, selective pressure would not be sufficient to result in mimicry. A second complexity is that in adopting the form, structure, and colour of the unpalatable model, the leaves of the mimic also begin to resemble the model in essential life functions such as light interception, heat balance, and water relations. It then becomes debatable whether it is environmental or habitat convergence, or true mimicry. Atkinson and Greenwood (1989) mentioned a number of possible examples of plant-plant mimicry. For none of them do they establish a convincing case for mimicry, although it remains a possible interpretation.

Alseuosmia pusilla is a small, usually unbranched shrub, the palatable leaves of which are strikingly similar to the red-mottled, distasteful leaves of Pseudowintera colorata. However, although A. pusilla fulfils the requirement of a mimic, in that it is less common than its model but grows with it, this is not sufficient to establish mimicry. Aside from the red mottling, which is only well developed in sun-exposed leaves, there is nothing distinctive about P. colorata leaves. Red, brown, or yellow mottling of leaves is common in New Zealand, and seems to be a response to leaf damage, especially in the presence of bright light. Other species of Alseuosmia, which have dissimilar leaves to Pseudowintera colorata, also have red and brown mottles. Moa, as they possessed an excellent sense of smell, are unlikely to have been confused by the visual similarity between the two species. Therefore, the most likely explanation of the visual similarity between the two species is chance.

The moa's excellent sense of smell makes the proposed mimicry between spiny-leaved Aciphylla species and some similar appearing, but non-spiny tussock-forming plants improbable. Atkinson and Greenwood (1989) link the spiny-leaved Aciphylla subflabellata and unpalatable Festuca novae-zelandiae as a Mullerian mimicry complex. Likewise, they claimed that the juvenile or young foliage of Pittosporum pimeleoides and Podocarpus acutifolius mimics the unpalatable Cyathodes juniperina. Testing how well these meet the requirement for mimicry would be difficult, and environmental explanations for convergence are probably more convincing.

Tough fibrous stems and leaves

Although numerous plants in the New Zealand flora have tough fibrous leaves, Atkinson and Greenwood (1989) singled out Pseudopanax crassifolius, P.ferox, P. lineare, Cordyline australis, and Phormium tenax, because this feature is both exceptionally well-developed and best expressed when these plants are of a height that could have permitted moa-browsing. The Pseudopanax crassifolius group-the lancewoods - have a long (up to 1 m) fibrous linear serrated leaf, which is page 7 replaced by a shorter, less sclerophyllous, ovate leaf in the adult (Gould 1993). Cordyline and Phormium leaves do not change markedly from juvenile to adult, but they are formidably tough at any stage.

Tough, fibrous stems and leaves do deter herbivores (Choong et al. 1992). However, it is likely that in many plants the anti-herbivory effects are a secondary consequence of their structural function. In the suggested New Zealand examples, the long leaves are held at angles ranging from horizontal to near vertical. The single-stemmed juveniles of Cordyline and the lancewoods gain fully adult leaves only when branching begins. Their leaves alone form the juvenile canopy and must provide the necessary extension; hence they have the strength and toughness characteristic of branches. Phormium has the form of a giant tussock, with individual leaves up to 3 m in length, thick at the base and often quite stiffly erect except for the top few centimetres, and therefore clearly acting as stems. These species have developed these leaves primarily in relation to their particular requirements for canopy formation.

Reduced leaf area

The claim that plants reduce the size of their leaves or the total area per plant, or dispense with them either entirely or seasonally primarily because of herbivory (Atkinson and Greenwood 1989, Batcheler 1989) has little merit.

Batcheler (1989) suggested that deciduousness in the New Zealand flora is related to herbivory. However, deciduousness in New Zealand species is linked to cool winters. For example: Fuchsia excorticata and Aristotelia serrata, both mainly deciduous in the South Island, are evergreen in the warmer parts of the North Island (Dawson 1988). This pattern is unlikely to result from greater moa-browsing pressure in colder areas.

The New Zealand brooms (Carmichaelia, Chordospartium, Corallospartium) follow a typical broom pattern of reducing or dispensing with leaves in favour of photosynthetic stems, an adaptation therefore unlikely to be specifically directed at moa. It is, moreover, an adaptation that probably developed to reduce water-loss rather than herbivory. However, as with many such adaptations, it undoubtedly also reduces the palatability of the plant to a range of herbivores.

Reduction in total leaf area, or loss of leaves for a period, we believe therefore is best explained by the plant being unable to afford the cost of leaves if their income in form of photosynthate is low versus outgoings such as respiration, increased root area, or stem mass. Reduction in leaf area because of an elevated risk of herbivory seems inexplicable, given that there are other ways to minimise herbivory that do not involve sacrificing photosynthetic potential.

Spines

Spines are usually a defence against animals. Spininess is in particular a characteristic of arid-land plants, and documented evidence of avoidance or lower usage of such plants by vertebrate herbivores suggests that it evolved as a defence against them (see discussion in Janzen 1986:616–620). A strong point in Greenwood and Atkinson's (1977) hypothesis is that the New Zealand flora has few spiny plants, and they made a convincing case for this being because birds have horny beaks and therefore are undeterred by spines. However, they made an exception for Aciphylla, regarding their stiff sharp pointed leaves as effective anti-browsing devices, presumably because the spines threatened the eyes of browsing birds.

Strong stiff leaves with thick cuticles will tend to be spine-like. Hence, adaptation to drying windy climates by a tussock species could bring with it spininess as a mere byproduct. As previously pointed out by McGlone and Webb (1981),

1. Nutritional test

Moa should have been attracted to plants of high nutrient status. Therefore, if a plant with a putative anti-browse feature is also nutritionally attractive to birds, it will have passed this test.

Lee and Johnson (1984) measured the concentration of several nutrients in leaves of six divaricate and four non-divaricate Coprosma species. Nitrogen, phosphorus, calcium, and sodium were all present in higher concentration in divaricating species, and only potassium was more abundant in non-divaricating species. A canonical discriminant analysis of the data showed that the leaves fell into two groups, largely on the basis of nitrogen, phosphorus and sodium, in which small-leaved divaricating species were separated from large and small-leaved non-divaricating species.

However, if we take the two most important nutrients, nitrogen and phosphorus, and plot the data (Fig 1), a less clear-cut picture emerges. The two most weakly divaricating species of the group (C. rotundifolia and C. rubra) are separated from the rest by their high levels of nitrogen and phosphorus. The other group of divaricate and non-divaricate species have substantially lower nitrogen and phosphorus levels. The least nutritious leaves among the divaricating plants, belong to the most tightly divaricating species (C. propinqua and C. crassifolia). It would be hard to argue, despite their lower nutrient levels, that any of these Coprosma spp. are unattractive to vertebrates. Coprosma lucida and C. grandifolia, the two largest and thickest leaved species, are so palatable to present day browsers that they are completely eaten out by deer in some areas (Allen et al. 1984, Wilson 1987).

We argue that the high nutritional status of small thin leaves is probably related to their high ratio of photosynthetic tissue to supporting tissue, and possibly to higher photosynthetic activity than in the larger leaves. Therefore, acquisition of a small thin leaf or of a large, thick leaf could alter the nutritional balance regardless of any other factor. Highly active leaves may be more attractive, and therefore browsed more often, but this is off-set by higher growth-rates and turn-over, as discussed above.

As they often occur as successional plants on nutrient-rich sites, but often grow slowly, and allegedly devote a high proportion of their resources to defence, divaricating plants must be seen as highly anomalous. If browsing was so intense on nutrient-rich sites in New Zealand, it would seem that high levels of defensive chemicals, in particular toxins, would be a more optimal solution, as it would demand minimal change in the structure of the plants, be more flexible, and permit faster growth.

Is this test a good one? What if nearly all divaricating plants had leaves of exceptionally low nutrient content? Would this lead to the abandonment of the browsing hypothesis? We think not. For instance, Atkinson and Greenwood (1989:88) include among adaptations to moa browsing “…linear and fibrous juvenile leaves of low nutrient value of Pseudopanax crassifolius.” If, as in the thin-leaved divaricating shrubs, the leaves of P. crassifolius juveniles had high nutrient status, this could be easily explained as the reason for their tough, fibrous nature. Low nutrient status, however, here is seen as simply another defensive strategy. A test used in this way cannot be useful.

2. Life-cycle test

The anti-browsing adaptation should be best developed at that part of the plants life-cycle when it is most exposed to moa browsing. The anti-browsing feature should not be retained when the plant is out of reach of moa.

This test is based on the assumption that terrestrial vertebrate browsing necessitates a greater defence investment in foliage borne low on the plant. This is a reasonable assumption with some plants and in some habitats. For instance, in Europe, holly (Ilex aquifolium) has more abundant spines on leaves close to the ground and leaves on heavily browsed shoots have longer spines (Peterken and Lloyd 1967), and in Africa thorny plants are intrinsically thornier below 3–5 m from the ground (Milewski et al. 1991) and browsed shoots tend to become more densely spiny. However, in some situations juvenile spininess is reduced and lost in the adult form without the presence of large vertebrate herbivores. For example, in the Hawaiian Islands certain Cyanea species have juveniles that are densely armed with stout, sharp prickles, while the adults are smoother, and less prickly (Lammers 1990). The prickles are regarded as a defence against phytophagous land snails and insects. Presumably, either attacks by animals are more intense close to the ground or the effect of the browsing is more severe on the juveniles. Regardless, the greater defence of the juveniles is not in any way related to the size per se of the browsing animal. At a more general level, Lowman (1985) found that leaves closer to the ground in tropical Australian forests were grazed more heavily by insects. It would seem reasonable to suppose that similar sorts of mollusc and insect attack occurred in New Zealand, and that for whatever reason, it was more intense closer to the ground. There was a large array of flighted and climbing browsing birds and insects in pre-historic New Zealand. The prevalence or better development of a putative browsing defence when the plant is small, which is then lost or reduced at greater heights, cannot be automatically attributed to the browsing height range of moa.

Even if it is accepted that moa browsing was a decisive factor in the evolution of plants, it is by no means clear that many allegedly browse-resistant plants pass the ‘life-cycle test’. Although most divaricating plants are of low stature, nearly 20% regularly attain heights of more than 4 m, and some can grow to 12 m, thus beyond the reach of even the tallest moa. Unless the divaricating plant form is a juvenile phase only, the degree of alleged defence does not change, no matter how tall the plant. For instance, in tall divaricating plants, such as Coprosma crassifolia. Melicope simplex and Myrsine divaricata, the uppermost foliage may be 5 more metres high, and invariably more strongly divaricate than at ground level. If divarication involves a cost to the plant, we could expect it to be reduced in intensity as the risk of attack lessened.

The degree of divarication often alters markedly with the aspect of the plant. Those parts of the plant exposed to the sun or the prevailing wind are usually strongly divaricate, and leafy shoots are often protected by one or more layers of near leafless branches. Sheltered, or shaded parts are mostly less strongly divaricate, are usually leafy to the exterior, and the leaves are larger. If a divaricating plant is growing completely in shade, leaf size and leafiness approximate more closely to those of the interior of an open-grown plant, and it is rare for divarication to be as well-developed as in the open. Many (for example, Coprosma rhamnoides) develop spreading, sparsely leaved plagiotropic arrays of thin branches at intervals up the stem.

In windshorn plants degree of divarication is primarily related to exposure not to ease of access by herbivores. For example, in exposed plants of Hoheria angustifolia the small-leaved divaricating juvenile foliage will extend further up the windward side of the plant, with adult foliage at the same level on the relatively protected lee side. In extremely windy sites, adult foliage often does not extend above the level of juvenile foliage, but the stem with adult leaves grows parallel to the ground on the lee side (McGlone, pers. obs.). Similar cases can be seen when divaricating plants grow in the shelter of rocks. Why divarication should be so intense a few centimetres above the shelter of a rock, and relaxed in the lee of the page 12 rock, is not easily explained by reference to browsing.

Most divaricating plants therefore fail the life cycle test, in that they are heavily defended when apparently there is no need, such in the uppermost layers out of the reach of moa, and weakly defended, if at all, underneath a canopy well within the reach of any moa. In most open-grown plants, the lower parts of the plant and the side away from environmental exposure are inexplicably left weakly guarded.

The height at which the tough-leaved single-stemmed juvenile of Pseudopanax crassifolius begins to change into the less coriaceous and branched adult tree certainly is at about the level where moa browsing would have been greatly reduced or absent. However, measurement in a range of habitats on the height to the beginning of canopy branching (which is a conservative indication of the height at which the tough juvenile foliage is replaced by adult leaves) show a variable range of heights for changeover (Table 2). Changeover occurs at a height of 1–2 m lower in the open when the competing vegetation is low-growing than under tall dark forest canopies. Under very dark, high canopies juveniles can reach as high as 7 m, with only a small cluster of leaves at the top of a stem often less than 5 cm in diameter. Juveniles in the open tend to have abundant leaves down the length of relatively thick stems. If moa-browsing is the main reason for the juvenile form, why should it be retained at heights well above the risk of attack? We suggest that the single-stemmed juvenile is an adaptation designed to carry the stem apex to a height where branching will spread the canopy above competing foliage. There fore, the lower the light level, the longer the juvenile form is retained.

3. Site distribution test

Plants with anti-browsing features should reach their greatest abundance on fertile sites where moa are certain to have browsed most intensively, and be unimportant on sites inaccessible to moa.

Divaricating plants are more abundant on landforms characterised by soils of moderate to high fertility. They are often abundant on alluvial lowland soils, but are not so common on soils derived from mafic rocks (Lee 1992), or on acid or leached soils. However, it is possible to exaggerate the relationship. For instance, on Stewart Island, the variety and abundance of divaricating plants are greatest on alluvial flats, and shrubs such as Pseudopanax anomalus and trees with divaricating juveniles such as Prumnopity taxifolia and Plagianthus regius are virtually confined to these sites (Wilson 1987). However, despite the acid soils of low fertility that characterise the island, some divaricating shrubs are both widespread and common (e.g., Myrsine divaricata, Coprosma propinqua, Coprosma rhamnoides, and Carpodetus serratus). In the eastern and central North Island regions apparently fertile alluvial flats, containing the greatest number of divaricating species known in New Zealand, are subject to severe frosts and/or substantial seasonal water table fluctuations, conditions which most broadleaved species cannot tolerate. The relationship between divarication and soil fertility is therefore complex and of little assistance in testing the moa-browsing hypothesis.

Greenwood and Atkinson (1977), Batcheler (1989) and Atkinson and Greenwood (1989) pointed out that few if any epiphytes are divaricating, despite the their dry branch and rock environment. They saw this as contradicting claims that the divaricating habit is related to drought avoidance (see McGlone and Webb 1981), and also as demonstrating that the divaricating syndrome was rare when the risk of moa browsing was low. Some divaricating plants have developed an ability to grow in dry areas, and the small expendable leaves, near leafless exterior, and self-shading habit of many divaricating plants could help explain this. Keen (1970) concluded on the basis of experimental studies that the small-leaved divaricating page 13 plant species she examined were not more resistant to absolute drought than their large-leaved relatives but were better adapted to survive and grow in conditions of physiological drought, such as those induced by drying winds. The epiphytic environment is so demanding and specialised that few vascular plants of any sort are found as epiphytes, and in New Zealand they are most abundant when rainfall is consistent and high. Whether the absence of epiphytic divaricating plants is significant or not is therefore difficult to determine. Dawson (1988) listed only seven shrubs and small trees regularly found on tall trees. All but one have thick fleshy leaves, and the tree species have roots that eventually reach the ground. Therefore, in keeping with this highly specialised habitat, nearly all have well-developed water storage capabilities or obtain water from the ground. There is a parallel among the herbaceous monocotyledons. It is not grasses adapted to dry, open sites that have colonised tree branches, but Astelia solandri and Collospermum, which form deep soil ‘nests’ or store water in tanks. The five species of epiphytic orchids are specialists, with fleshy spreading rhizomes. A number of orchids adapted to dry droughty sites do not occur on trees.

Greenwood and Atkinson (1977) claimed that divaricating plants are rare on steep cliffs, where moa would have found it impossible to browse. Like tree brunches, cliffs present a relatively specialised environment. Many cliffs are unstable, and seem to have a large opportunistic rather than cliff-specific element in their vegetative cover. Furthermore, cliffs vary greatly in all sorts of characteristics, including accessibility. Sheltered moist cool and relatively stable cliff sites are not favourable to divaricating plants, but sunny dry windy and exposed cliffs on Banks Peninsula have an abundance of divaricating plants (Hugh Wilson, pers comm).

We have discussed above why toxic foliage in seral vegetation is not necessarily a defence against moa browsing, but a more general adaptation to herbivore attacks on palatable, fast-growing plants. The argument could be reversed. Fast-growing but browse-sensitive plants characteristic of fertile soils, for example, Aristotelia serrata and Fuchsia excorticata, could be equally well taken as evidence that moa did not have a substantial effect on the vegetation on such soils.

4. Geographic distribution test

Areas known not to have had moa should have reduced numbers of species and smaller populations of plants with strongly developed anti-herbivore defences, than in areas that had moa.

We are certain that moa did not occur on most offshore islands more than a few kilometres from the mainland, although Stewart Island had moa (Anderson 1989). This test therefore contrasts islands with mainland sites.

Greenwood and Atkinson (1977) showed that the percentage of divaricating plants (in relation to the woody flora) on moa-free islands was lower than on comparable areas of coastal mainland (4.7% versus 10%). In general, the coastal mainland sites had more woody species (average 47) than the island sites (32). However, the Chatham Islands have a percentage of divaricating plants (8%) not much lower than the New Zealand flora as a whole, and higher than two of the mainland coastal areas chosen as comparisons. The low numbers of species involved and differences in climatic conditions between island and mainland sites (which are only rough proxies for islands) make this comparison of doubtful significance.

Of more significance are species or closely related species that occur on islands and the mainland, but differ in degree of divarication. Greenwood and Atkinson (1977) noted that a non-divaricating relative of Myrsine divaricata and a non-divaricating form of Coprosma propinqua occur on the Poor Knights Islands and page 14 the Chatham Islands, respectively. As well, Plagianthus ragius and Sophora microphylla have non-divaricating juveniles on the Chatham Islands, (Greenwood 1992). However, when examined in detail, these cases seem less convincing.

Sophora microphylla also has non-divaricating juvenile populations on the mainland, and the divaricating form is best expressed only in the far south and south-east of the South Island (Godley 1979) and in the east of the North Island. Myrsine divaricata is abundant on the subantarctic Auckland Island, which Greenwood and Atkinson (1977) attribute to the relatively recent colonization of the subantarctics from the mainland at the end of the Last Glaciation. Even allowing that this is so, it still does not alter the fact that one of the most consistently and tightly divaricated species, and one highly resistant to browsing by ungulates at present, has successfully invaded browse-free islands and remained unaltered through at least several thousand years. We, on the other hand, would stress the windy cool climates of the subantarctic islands in contrast to the mild, humid climates of the Poor Knights Islands. Corokia cotoneaster is strongly divaricating with small exterior leaves in inland districts of the North and South Island; under the milder conditions of coastal Taranaki, a form of this species is far less divaricating and has larger leaves borne on the exposed tips of the branches. A further counter example is Phormium, which is widespread on many islands, including Norfolk Island. If the tough, and not easily browsed Phormium persists in the absence of any browsing pressure at all, it is a strong argument for its leaf morphology not being primarily a browsing defence. Island plants in general tend to have larger leaves than typical mainland forms, and Greenwood (1992) pointed out that a range of woody species on the Chatham Islands have larger leaves than mainland forms, especially as juveniles. The comparative rarity of strongly divaricating plants on islands could therefore simply be related to greater success of broadleaved plants in the milder, less variable climates of offshore islands (Dawson 1988).

In general, we are dubious the geographic distribution test is satisfactory. Islands differ so much from mainland situations the only convincing test would be similar islands with and without moa. What evidence there is appears to support the view that it is not prehistoric moa presence or absence but current climatic conditions which control the numbers and abundance of divaricating plants on islands. We would suggest that the other putative anti-moa adaptations would likewise show no consistent geographic relationship to the past distribution of moa.

5. The ‘absence of herbivore’ test

Where vertebrate terrestrial herbivores are or were absent, heavily defended plants should be at a competitive disadvantage in relation to less well defended plants.

This is not a test proposed by Atkinson and Greenwood (1989) but one that should be considered. A major weakness of the moa-browsing hypothesis is the way that divaricating plants and other plants with supposed anti-browsing defences remained abundant in the absence of substantial browse pressure and in the presence of fast-growing competitors. A fair assumption is that anti-browsing strategies come with some cost to the plant. Nevertheless, divaricating plants, and other plants with supposed anti-browsing features, can be successful in the absence of any browsing pressure, and in the presence of ‘normal’ large-leaved, fast-growing competitors. With the extenction of the moa some 400 years ago, and major depletion starting about 600 years ago (Anderson 1989), non-defended plants have had between 300 and 500 years without significant browsing to overwhelm their defended competitors. That this did not happen is clear from the abundance of divaricating plants, spiny plants, plants with tough leaves, and dark page 15 or mottled coloration at the time of European settlement.

A possible counter to this argument is that although defended against browsing, these plants are better adapted to their characteristic environments than other species, and therefore will not be outcompeted, even in the absence of browsing. If this is granted, it leaves the problem of what it is about plants with these defence strategies that makes them so suitable for these environments.

The argument can be reversed: if browsing pressure was strong enough to induce in the vegetation such a wide range of anti-browsing strategies, it follows that before the extinction of the moa, many palatable and weakly defended plant species should have been restricted by browsing. This contention is not supported by the fossil evidence. Pollen of highly palatable species of Griselinia, Pseudopanax, Aristotelia, and Fuchsia are not often abundant in the fossil record because of poorly-distributed grains, but as far as can be told, they neither increase or decrease with the elimination of the moa. At some pre-human sites they are at times abundant (McGlone, Neall and Clarkson 1988), and this suggests that moa-browsing pressure was certainly not sufficient to restrict them.