The influence of host and non-host companion plants on the behaviour of pest insects in field crops




Companion plants grown as ‘trap crops’ or ‘intercrops’ can be used to reduce insect infestations in field crops. The ways in which such reductions are achieved are being described currently using either a chemical approach, based on the ‘push-pull strategy’, or a biological approach, based on the ‘appropriate/inappropriate landing theory’. The chemical approach suggests that insect numbers are reduced by chemicals from the intercrop ‘repelling’ insects from the main crop, and by chemicals from the trap-crop ‘attracting’ insects away from the main crop. This approach is based on the assumptions that (1) plants release detectable amounts of volatile chemicals, and (2) insects ‘respond’ while still some distance away from the emitting plant. We discuss whether the above assumptions can be justified using the ‘appropriate/inappropriate landing theory’. Our tenet is that specialist insects respond only to the volatile chemicals released by their host plants and that these are released in such small quantities that, even with a heightened response to such chemicals, specialist insects can only detect them when a few metres from the emitting plant. We can find no robust evidence in the literature that plant chemicals ‘attract’ insects from more than 5 m and believe that ‘trap crops’ function simply as ‘interception barriers’. We can also find no evidence that insects are ‘repelled’ from landing on non-host plants. Instead, we believe that ‘intercrops’ disrupt host-plant finding by providing insects with a choice of host (appropriate) and non-host (inappropriate) plant leaves on which to land, as our research has shown that, for intercropping to be effective, insects must land on the non-host plants. Work is needed to determine whether non-host plants are repellent (chemical approach) or ‘non-stimulating’ (biological approach) to insects.


A common non-insecticidal method for reducing the numbers of pest-insects in field crops is to make the growing system more diverse by introducing ‘companion plants’ (Franck, 1983). Insect numbers are usually lowered by growing a non-host companion plant within the main crop, often described as ‘intercropping’ (Vandermeer, 1989), or by growing an alternative host plant within or around the main crop as a ‘trap-crop’ (Hokkanen, 1991; Shelton & Badenes-Perez, 2006). We have adopted ‘companion planting’ as our general term, as it applies equally well: (1) to systems in which both plants are harvested, true ‘intercropping’ (Vandermeer, 1989), (2) to systems in which one plant is harvested and the other is ‘sacrificial’ (Altieri, 1994), and (3) to ‘trap crops’ (Hokkanen, 1991). It also makes direct comparisons with the early articles easier to follow. We have relied heavily on such articles as some of their untested assumptions (Kennedy, 1977) need further study, particularly those that underpin current research.

According to Aiyer (1949), lower numbers of specialist insects are found on the main crop because the companion plants: (1) make the host plants harder to find (disruptive-crop hypothesis), (2) act as alternative host plants (trap-crop hypothesis), or (3) serve as a repellent to the pest (see Vandermeer, 1989). The reverse way of looking at the ‘disruptive-crop hypothesis’, that host plants are easier to find when grown in pure or dense stands, forms the basis of the ‘resource concentration hypothesis’ (Root, 1973). Vandermeer (1989) suggests that all three mechanisms have been demonstrated in one system or another, but with hindsight we feel this may not be true.

At present, companion plants are being used successfully as ‘trap crops’ and ‘intercrops’ in what is described as a ‘push-pull strategy’ (Cook et al., 2007) for reducing spotted stem-borer (Chilo partellus Swinhoe) damage in maize (Zea mays L.) crops in Kenya (Khan et al., 1997; Khan & Pickett, 2004). The authors suggest that the companion plants used as the intercrop make the maize plants ‘unattractive or unsuitable to the pests (push)’ and that the plants in the trap-crop ‘lure the insects away from the maize crop (pull)’. Hence, instead of having to spray synthetic chemicals (Pyke et al., 1987), the ‘attractant’ effect can be produced with a trap-crop and the ‘repellent’ effect with an intercrop (Khan & Pickett, 2004). Here, we describe how we believe trap crops and intercrops produce their effects, based on information collected during the development of the ‘appropriate/inappropriate landing theory’ (Finch & Collier, 2000) (the biological approach), and contrast this with suggestions from the current push-pull strategy which are based almost exclusively on chemical cues (the chemical approach) (Pickett et al., 1997; Khan & Pickett, 2004).

Suggestions emanating from the push-pull strategy

In the Kenyan system, each 0.1-ha field of maize is grown within a 1-m-wide ‘stockade’ of 2-m-tall Napier (elephant) grass (Pennisetum purpureum Schumacher) and is protected by being ‘hidden’ partially by silverleaf desmodium [Desmodium uncinatum (Jacq.) DC], whose long stems trail over surrounding vegetation (Hacker, 1992). The success of the system depends on the correct choice of ‘repellent’ (intercrop) and ‘attractive’ (trap-crop) plants (Pickett et al., 1997).

Repellent plants

Describing plants as ‘repellent’ has arisen from finding fewer pest-insects on maize plants intercropped with D. uncinatum and then assuming that D. uncinatum must have repellent properties.

It was shown more recently, however, that even highly aromatic companion plants, such as marigolds (Tagetes erecta L.) and mint (Mentha piperita L.), used frequently by organic growers to protect their valued crop, were not ‘repellent’ to either the onion fly [Delia antiqua (Meigen)] or the cabbage root fly [Delia radicum (L.)] (Finch et al., 2003). Instead, they produced their effects by disrupting the normal ‘chain’ of behaviours (Zohren, 1968; Atkins, 1980) of both flies (Finch et al., 2003). The first ‘link’ in this chain occurred when volatile chemicals from the host plant stimulated receptive insects to land (Schoonhoven et al., 2005). The leaves on which the flies landed were then governed by visual stimuli (Prokopy et al., 1983). When flies landed on the leaf of a possible host plant, the flies became highly active, but stayed for only about 1 min on each leaf (Finch et al., 2003). When they landed on leaves of non-host plants, of which 18 different ones were tested (Finch et al., 2003), they became relatively inactive, and stayed for 2–5 min. Hence, chemicals on leaves of non-host plants did not ‘deter’ flies from staying. In addition, if the released chemicals were truly ‘repellent’ in the accepted sense of the word (Dethier et al., 1960), insects would never land on companion (non-host) plants, but many do (Billiald et al., 2005; Couty et al., 2006).

Attractive plants

It is accepted by those researchers who have worked intimately on this subject (Wigglesworth, 1939; Kennedy, 1977; Chapman, 1982) that the distance insects can be ‘attracted’ (Dethier et al., 1960) directly to a source of odour in the field is only a few centimetres. To the majority of researchers, however, the word ‘attractant’ (Fraenkel & Gunn, 1961) conveys distances much greater than ‘a few centimetres’. In essence, the chemicals described as ‘attractants’ are really ‘releasers’ (Kennedy, 1977; Baker, 1986) for other behaviours, such as: (1) turning into the wind (anemotaxis) and then flying upwind (a kinetic action) to the source of the odour (Baker, 1986) or (2) stimulating the insect to land on some nearby visually acceptable object, such as a plant (Chapman et al., 1981; Prokopy et al., 1983; Bernays & Chapman, 1994). It can be difficult to separate the two behaviours when the time between stimulation and landing is only a few seconds or part of a second (Elkington & Cardé, 1984), as insects invariably turn into the wind when landing (Kennedy & Thomas, 1974).

The major problem is in quantifying the distances over which insects are influenced by volatile plant chemicals. The five manuscripts listed in Schoonhoven et al. (2005) suggest distances of ‘attraction’ of between 20 (Evans & Allen-Williams, 1993) and 100 m (Judd & Borden, 1989). However, we can find no robust proof that chemical stimuli are effective over distances of more than 5 m (Finch & Skinner, 1982), as the current estimates have been obtained from insects that normally fly into the wind. Hence, there is no way of separating the part of the upwind flight due to wind alone (optomotor anemotaxis) from the part due to wind plus odour (odour-conditioned optomotor anemotaxis). Our tenet is that it is only when an insect flies within a few metres of a host plant that it perceives the volatile and visual cues that stimulate it to land (Finch & Skinner, 1982). We also believe that the ‘stay or leave’ approach to host-plant finding suggested by Thorsteinson (1960) and supported by Beck (1965) could be closer to the truth than people may be willing to admit.

The suggestion that volatile chemicals from trap crops ‘attract’ stem-borer moths from some distance (Khan & Pickett, 2004) can be interpreted differently. Under field conditions, insects appear to proceed simply in a preferred flight direction and are ‘forced’ to land, the ‘stop’ (a kinesis) of Dethier et al. (1960), whenever they pass over vegetation releasing volatile chemicals that characterize their host plants. Insects generally fly within a boundary zone close to the ground and use their eyes to sense the apparent movement of the objects over which they fly (Kennedy, 1939; Wright, 1962; Kennedy & Thomas, 1974; David, 1986). When phytophagous insects encounter higher vegetation, such as the 2-m-tall Napier grass used in the Kenyan system, they turn back or fly over the barrier, as insects rarely attempt to go through (Lewis & Stevenson, 1966; Johnson, 1969). Hence, Napier grass may in part be a physical obstacle (Kareiva, 1983), or ‘interception barrier’, that the moths have to negotiate to reach the maize plants. It is also a true trap-crop, as the spotted stemborer prefers Napier grass to maize (van den Berg et al., 2001; Khan et al., 2006) and so the volatile chemicals from Napier grass will also stimulate the moths to land and lay eggs. The Napier grass obviously ‘stops’ many moths from reaching the maize, but this is not the same as ‘attracting’ moths from a distance.

Any host plant will function as a ‘trap-crop’ provided it is grown around the main crop, as pest insects cannot pass over host plants without being stimulated to land and lay eggs. Hence, ‘interception barriers’ ensure that the numbers of eggs laid on subsequent host plants are reduced. For example, when blocks of preferred host plants were grown within fields of cotton (Gossypium hirsutum L.), the numbers of bollworm [Helicoverpa armigera (Hübner) and Helicoverpa punctigera Wallengren] eggs laid on the plants were never more, and often less, than on the surrounding cotton plants (Jallow & Zalucki, 1996). It was suggested that this greater ‘preference’ for the most abundant host plant, cotton, supports the hypothesis that ‘learning’ may play an important role in host-plant selection in the field (Cunningham et al., 1999). An equally plausible suggestion is that fewer eggs were laid on the ‘preferred’ plants because the surrounding cotton acted as a ‘trap-crop’. The notion that insects ‘search’ specifically for ‘preferred’ (highly stimulating) host plants is not easy to accept, as it implies that the insects have some form of prior ‘knowledge’ that ‘preferred’ plants actually exist.

If highly stimulating host plants are to be used as trap crops, they need to be grown around, rather than within, the valued crop (Khan et al., 2001). The logistics of surrounding large fields of cotton in Australia with a trap-crop would be extremely daunting, as the ‘barrier’ itself has to be wide to produce a measurable effect in large fields (Hokkanen, 1991). Once again the Kenyan situation is unique, as each area cropped is small (0.1 ha) and so ‘trap cropping’ can produce an effect, albeit less pronounced than that from intercropping (Khan et al., 2001).

Unattractive and unsuitable resources

The suggestion that companion plants can make a valued resource ‘unattractive’ or ‘unsuitable’ (Khan & Pickett, 2004), depends upon whether the words are being used as descriptive or technical terms (Painter, 1951; Dethier et al., 1960; Beck, 1965). If the latter, then the companion plants may have made the maize plants visually less ‘apparent’ (Feeny, 1976) but this is not the same as making them chemically ‘unattractive’. Their chemical profile should remain unchanged unless host-plant chemicals are altered in some way by the neighbouring non-host plants. In anthropomorphic terms, this would mean that the only time a maize plant smells like maize is when it is growing in bare soil or amongst other maize plants. Although such a mechanism has been proposed (Theunissen, 1994), we have difficulty in supporting it, as if host plants absorbed chemicals that made them smell and taste like the surrounding natural vegetation, specialist insects would never ‘find’ them, but many do. The second suggestion that maize (host) plants growing in intercrops also become ‘unsuitable’ because they are landed on less frequently (Khan & Pickett, 2004), can be interpreted in other ways. The maize plants may not be ‘unsuitable’, but simply ‘no more suitable’ as landing sites than the leaves of the neighbouring non-host plant (here D. uncinatum) (Finch & Collier, 2000). During the final approach to plants, visual stimuli appear to inhibit orientation to chemical stimuli (Kay, 1976; Kennedy, 1978), and so the insects are simply shared between the two plant types. This raises an even greater conundrum. If volatile chemicals trigger the landing behaviour when the insect is 1–5 m from the plant, is chemotaxis or ‘attraction’ in the strict sense of the word (Dethier et al., 1960), which operates only at distances of <1 m, ever used by flying insects?

The confusion lies with the term ‘attractant’, which Miller et al. (2009) suggest can remain unchanged even though we are far from convinced that the interpretations of Baker (1986) and Kennedy (1986) are how ‘attractants’ were perceived by Dethier et al. (1960). When ‘attractant’ was first coined, most authors believed that the wind-borne odour plume from a distant odour source was literally a scent trail along which a flying insect was guided chemotactically in the manner of a walking insect on a terrestrial trail (Shorey, 1973, 1977). At present, ‘attraction’ includes both direct actions of the chemicals, such as chemotaxis, which by definition (Dethier et al., 1960) can only occur over short distances, and indirect actions in which the chemicals simply condition or ‘trigger’ the flying insects to respond by turning (a taxis) into the wind (anemotaxis) or becoming more responsive to local visual stimuli. Apart from the actual turns (taxes), the upwind flights are governed by a combination of optomotor responses (Kennedy, 1977) and various forms of kinetic actions or ‘kineses’ (Fraenkel & Gunn, 1961; Kennedy, 1977). Hence, we are not saying that chemicals are not involved, but simply that the chemicals do not provide the directional information and so are not ‘attractants’ in the designated (Dethier et al., 1960) sense of the word. As Kennedy (1978) cautioned, if the term ‘attractant’ covers both the direct and indirect actions of the chemical, then ‘attractant’ becomes no more than a blanket teleological term signifying an end result. In an attempt to produce an acceptable compromise to this impasse, it was suggested (Roelofs & Cardé, 1977; Kennedy, 1978) that any orientation (attraction) that could not be accounted for by the direct effect of the chemical, that is distances over 1 m, should be described as ‘long-distant’. However, this creates new problems as, by analogy, visual stimuli would then be classed as ‘long-distant’ which most people would find hard to accept. Perhaps it is now too late to use ‘attraction’ as anything other than a ‘blanket term’, despite Kennedy (1986) considering this a retrogressive step.

It is important to remember that landing by chance at regular intervals, or landing more frequently when stimulated by volatile chemicals (Douwes, 1968) are both suitable ways of ‘finding’ a host plant. Finding a host plant is an end-point for only a few insects (Wellington, 1977), as most only stop briefly, to feed or to lay eggs, and then move on to ‘find’ other host plants. Pieris rapae (L.), for example, lands on about 500 plants per hour (Root & Kareiva, 1984; Finch & Collier, 2007), many of which are possible host plants, indicating that even ‘specialist’ insects do not have difficulty in ‘finding’ host plants (Finch & Collier, 2007).

Behaviour of insects in monocultures and polycultures

The ‘resource concentration hypothesis’ (Root, 1973) indicates that phytophagous insects are more likely to find and remain on host plants growing in dense or nearly pure stands (monocultures). The relative ‘trapping effect’ depends upon the size and purity of the plant stand (Pimentel, 1961). When plant stands were made impure by the addition, not substitution, of non-host plants (polyculture) to ensure that the concentration of the resource stayed the same, although similar numbers of insects entered both stands (Root & Kareiva, 1984), more insects left the mixed (polycultures) than the pure (monoculture) stands (Bach, 1980; Root & Kareiva, 1984).

Despite recording numerical differences of this type, the major question that needed to be answered was ‘how do arthropods respond to polycultures compared with monocultures (Andow, 1991)?’ We believe we produced a realistic answer to this question in the ‘appropriate/inappropriate landing theory’ (Finch & Collier, 2000). Earlier opinions on whether this question could be answered were divided. Most authors believed that nature was inherently idiosyncratic (Andow, 1991) and so adopted an ‘empirical’ approach. Although others (Andow, 1991; Altieri, 1994) did use a ‘theoretical approach’, it proved extremely difficult to find anything that was common, amongst a prodigious amount of disparate information, much of which was anecdotal. Hence, it is hardly surprising general principles were not uncovered.

By using a standard system in which Brassica host plants were surrounded by either clover (polycrop situation) or bare soil (monocrop situation) (Kostal & Finch, 1994), we found that host-plant finding by eight pest-insect species, from four insect orders, was affected adversely by the presence of clover (Finch & Kienegger, 1997). Initially, we were disappointed, as we had hoped to write in length about how just one companion planting system could affect different insect orders in different ways. Such a finding, however, would have devalued companion planting, if all companion planting did was to reduce the numbers of one pest-insect species while increasing the numbers of others. The latter is still a widely held belief and has probably prevented many researchers from introducing companion planting into their research programmes. We believe that companion planting is capable of reducing infestations of all pest-insects associated with cultivated crops, provided the plants are sufficiently close for their foliage to intermingle (O’Donnell & Coaker, 1975). The effectiveness of companion plants is not in question. What is in question is whether they produce their effects by being chemically ‘repellent’ (Cook et al., 2007) or chemically ‘neutral’ (Finch et al., 2003).

Opposition to the ‘appropriate/inappropriate landing theory’

The ‘appropriate/inappropriate landing theory’ has been refuted recently by Couty et al. (2006) who wrote ‘With the exception of a test design in which lettuce plants (Lactuca sativa L.) and Chinese cabbage plants (Brassica chinensis L.) were alternated (chequer-board pattern) and one in which lettuce were presented in front of Chinese cabbage plants, females of the diamondback moth, Plutella xylostella L. were able to land “preferentially” on their host plants’. ‘Thus, these results do not support Finch & Collier’s (2000) theory of host selection by crop pests’.

However, the results from the chequer-board pattern show that the female moths landed just as frequently on lettuce (non-host) as on Chinese cabbage (host), indicating that the volatile chemicals did not provide sufficiently directional information to ensure the insects landed ‘preferentially’ on host plants. In addition, cabbage root flies land just as frequently on companion plants made from green card as on living plants (Kostal & Finch, 1994). Hence, once an insect is stimulated to land, the volatile chemicals, central to the chemical approach (Bruce et al., 2005), are unable to override the effect of visual stimuli (Kay, 1976; Kennedy, 1978), even when the ‘companion plant’ is not competing chemically with the host plant.

There was also a strong positional effect in the data of Couty et al. (2006), as they stressed that ‘P. xylostella were more likely to land on the first row of plants (the nearest)’, another aspect central to the ‘appropriate/inappropriate landing theory’. Put simply, if you surround a host plant with non-host plants, as in natural vegetation, the tendency for the insects to land on the surrounding plants, once stimulated to land by volatile chemicals, means that fewer land on the actual host plant. The non-host plants therefore act as what are known as ‘guard’ plants in field experiments and it is customary not to collect data from this row of plants. In the chequer-board pattern (Couty et al., 2006), the numbers of landings on the lettuce:Chinese cabbage plants when the data from the first row were, and were not, included were 37:53 and 22:17, respectively. Rather than contradicting the ‘appropriate/inappropriate landing theory’, the data appear to support it.

The final suggestion that the two sets of data differ because P. xylostella is crepuscular whereas ‘the theory of Finch & Collier (2000) is based on day-flying insects’ (Couty et al., 2006) is hard to accept, as P. xylostella, although the least responsive (Finch & Collier, 2000), was one of the eight insect species used to develop the theory. We are also not sure why crepuscular activity should make a difference, as P. xylostella is highly adapted to the wide range of light intensities that occur during its periods of activity (Hocking, 1953; Chapman, 1982).

Comparing ‘biological’ and ‘chemical’ approaches

Biological approach (e.g., ‘appropriate/inappropriate landing theory’)  Our tenet concerning the ‘biological approach’ is that insects encounter stimulating chemicals on and around host plants, but not on and around non-host plants (Finch & Collier, 2000; Morley et al., 2005). Hence, the biological approach involves ‘stimulating’ plants (host plants) and the opposite which are ‘non-stimulating’ plants (non-host plants), or plants that are effectively ‘chemically neutral’. Obviously, non-host plants release small amounts of their own characteristic chemicals on and around their leaf surfaces (Schraudolf & Bergman, 1965) but, as specialist insects are not ‘sensitive’ (adapted) to such chemicals, they go undetected. In the biological approach, the diversity of chemicals released by natural vegetation does not create problems, as the only secondary plant compounds the insect needs to ‘respond’ to are those that characterize its host plants. Hence, only one specific receptor, such as the glucosinolate receptor identified in larvae (Schoonhoven, 1967) and adults (Ma & Schoonhoven, 1973) of Pieris brassicae (L.), and in adults of D. radicum (Städler, 1978), needs to have evolved.

Despite only small amounts of volatile chemicals being released from host plants during normal growth (Schraudolf & Bergman, 1965; Cole & Finch, 1979), specialist insects are now able to detect them. The amounts of chemical released appear sufficient only to stimulate flying insects to land, but not to stay and lay eggs. The latter is achieved mainly by stimulation from non-volatile chemicals (Schoonhoven et al., 2005). Hence, insects that cannot bite the leaf to assess its chemical content have developed elaborate behaviours, such as ‘drumming’ (Ilse, 1956) or probing rapidly with the highly sensitive labellum (Wilczek, 1967; Dethier, 1976), to obtain sufficient chemicals to induce them to stay (Städler & Buser, 1984; Havukkala & Virtanen, 1985).

In contrast, when specialist insects land on non-host plants, they are not stimulated to bite, drum, or probe and so effectively never encounter the chemicals that characterize such plants (Finch & Collier, 2007). Hence, it does not matter how repellent or toxic the chemicals are that ‘lurk’ beneath the surfaces of non-host plants, specialist insects will never encounter them. Consequently, a wide range of non-host plants can be used to disrupt host-plant finding by phytophagous insects (Finch et al., 2003), as the effectiveness of such plants depends on their physical (size, shape, leafiness, etc.) rather than their chemical attributes. The data in the Kenyan push-pull project appear to support this finding, as cowpea [Vigna sinensis (L.) Walp.] and Molasses grass (Melinis minutiflora P. Beauv.) also reduced pest insect numbers (Omolo et al., 1993; Kfir et al., 2002; Khan & Pickett, 2004) despite their chemistry differing markedly from that of D. uncinatum.

Chemical approach (e.g., ‘push-pull strategy’)  For the chemical approach to operate, insects have to be able to ‘recognize’ a vast number of chemicals from a wide and diverse range of non-host plants. The conundrums are highlighted by Visser, namely that ‘It is hardly feasible that specific receptors would have evolved in sufficiently large numbers to cope with the extreme chemical diversity found in non-host plant species (Visser, 1983) and that, to date, receptors for repellents have not been found on the antennae of insects (Visser, 1986)’. Despite such reservations, many authors (e.g., Bruce et al., 2005; Couty et al., 2006) still favour the chemical approach although for it to be effective, all plants would have to be constantly releasing chemicals onto their leaf surfaces in amounts sufficient to influence insect behaviour directly. Host plants would release chemicals that ‘attract/arrest’ their specialist insects, whereas non-host plants would release chemicals that ‘repel/deter’ such insects (Finch & Collier, 2000). However, plants that are hosts to some insects are non-hosts to others, and vice versa. So, for the chemical approach to be effective every phytophagous insect would have to be capable of ‘recognizing’ the plant chemicals found on the surfaces of every plant. Is this possible? There is no evidence that large amounts of any chemical are released onto the surfaces of undamaged plants (Juniper & Southwood, 1986), and it invariably takes several days to collect sufficient volatile chemicals for biological testing. For example, Blight (1990) trapped the chemicals released by plants over a 6-day period, involving 17 000 l of air, to get sufficient chemical to test for ‘repellent’ properties. In contrast, an insect approaching a plant must obtain its chemical stimulation during a time span of several seconds to perhaps a fraction of a second (Elkington & Cardé, 1984) from a volume of air that may not exceed 1 l. Unfortunately, no-one tests chemicals at these natural rates because the insects invariably do not respond, but is this ‘no-response’ actually the meaningful result?

Once the insect lands, the amount of plant surface it is in contact with is also small. Hence, to accumulate sufficient stimulation to stay, some insects become extremely active on the leaf surface. Why then should insects invest the same amount of time and energy to ‘detect’ that leaves of non-host plants are unsuitable? The evidence suggests that they do not, as some insects become relatively inactive when they land on leaves of non-host plants (Morley et al., 2005). Others, for example P. rapae (Finch & Collier, 2007), fly away almost immediately, but this does not mean that have been ‘repelled’ or ‘deterred’, but simply that they have not been stimulated to stay.

Unfortunately, when a chemical extracted from a plant ‘repels’ insects under laboratory conditions, there has been a tendency to assume that the intact plant must also ‘repel’ the same insects under field conditions. However, the chemicals found in plant extracts are released in the field only when a plant is damaged. For most of the time the precursors of these chemicals are maintained within the cells and organelles of the plant and so cannot influence insect behaviour, much the same way as synthetic chemicals cannot influence insect behaviour until someone opens the ‘bottle’. In the plant situation, it is often a case of opening two ‘bottles’ and mixing the contents, precursor and enzyme, so that the resulting chemicals can cauterize the wound and provide a major line of chemical defence against bacteria, fungi, insects, and mammals (Feeny, 1977; Städler & Buser, 1984). Consequently, no matter how potent chemicals extracted from certain plants might appear in laboratory studies, such chemicals are not released naturally onto the surfaces of intact plants (Städler, 1986) in amounts sufficient to influence the behaviour of non-specialist insects. This appears to be supported by the work of Khan et al. (2000), who showed that the chemicals present in the companion plant D. uncinatum increased parasitoid ‘foraging’ in laboratory tests and yet they detected ‘no increase in parasitism in their intercropped plots’. They concluded that other chemicals, such as α-cedrene (Turlings et al., 1990), could be interfering with the increased parasitoid foraging effect; but a lack of sufficient amounts of the stimulating chemicals on intact plant surfaces seems an equally plausible explanation. As stated earlier (Finch et al., 2003), we believe that surfaces of non-host plants are not ‘hostile’ to insects, but chemically neutral and that this enables predatory and parasitoid insects to move freely over most leaf surfaces. By the same token, ‘specialist insects’ are adapted only to the small amounts of chemicals released by their host plants, but not by other plants, which could explain why chemicals from non-host plants have to be applied in large amounts to affect the behaviour of non-adapted insects. In simple terms, the chemical approach is a ‘black’ (repellent/deterrent) and ‘white’ (attractant/arrestant) system, and the biological approach a ‘grey’ (non-stimulating) and ‘white’ system.


The push-pull strategy has stimulated us to ask whether ‘repellent’ companion plants would really help in crop protection. The answer depends upon the distances over which ‘repellent’ chemicals influence the behaviour of insects. The consensus appears to be that they are effective over only a few centimetres (Chapman, 1982; Schoonhoven et al., 2005). Hence, as crop plants are usually spaced 50–100 cm apart between the rows, then for ‘repellent’ plants to be effective they would have to be grown, as they are at present, in the inter-row spaces. The system would then be that the insect would be stimulated to land by the volatile chemicals from the host plants (valued crop) and then, instead of landing in similar numbers on the two plant types, would be ‘repelled’ from landing on the companion plants. Presumably, this would deflect the insects onto the valued crop and so nullify the crop-protection benefit of companion planting.

Another difficulty is not how to get more repellent chemical into a particular plant, (Khan et al., 2000) but how to get more out. Only chemicals that are present on the surfaces of plant leaves in high amounts stand any chance of disrupting pest-insect behaviour, as specialist insects do not respond to small amounts of non-host plant chemicals. As far as we know, plants do not have a mechanism for secreting large amounts of chemicals, other than waxes and sugars (e.g., extra-floral nectaries), onto their leaf surfaces (Juniper & Southwood, 1986). There might be a good reason for this, as the breakdown products of the secondary plant compounds are often phytotoxic even to the producing plant (Feeny, 1977; Taiz & Zeiger, 2010).

The suggestion that the underlying mechanism in intercropping is based on deterrent/repellent chemicals is not easy to accept, as the disruptive effect against specialist insects can be produced without chemicals on the plant-surface (‘neutral’ surfaces) by using artificial companion plants (Kostal & Finch, 1994). We have argued earlier that the surfaces of non-host plants are chemically ‘neutral’ to specialist insects as practically any non-host plant will produce the disruptive effect (Finch et al., 2003). Hence, surrounding a host plant with just one species of companion plant (Finch & Kienegger, 1997) is as effective as allowing crops to become surrounded by a wide range of ‘weed’ species (Smith, 1976). The overall disruptive effect does not appear to be produced by plant chemicals, but simply by providing the insects with alternative green surfaces on which to land (Finch & Collier, 2000).

We believe like Fabre (1923) and Kennedy (1986) that it is extremely important to observe what the insects actually do. For example, once P. rapae lands on a companion plant it flies off immediately (Jones, 1977; Root & Kareiva, 1984), whereas D. radicum spends considerable time before moving on (Morley et al., 2005). Volatile host-plant chemicals stimulate the geometrid moth Cidaria albulata L. to turn more frequently and land on every plant instead of every third plant (Douwes, 1968). This ensures that once in a patch of host plants the moth ‘finds’ quite a few of them. This was described as ‘preferentially’ landing on host plants (Kennedy, 1977), but when you look at the actual figures (Douwes, 1968; Schoonhoven et al., 2005; Finch & Collier, 2007), once within the plant stand the moth in fact landed on 15 host plants and eight non-host plants, values that represented the ratio of host:non-host plants in the particular plant stand. Therefore, in its natural habitat the moth was unable to separate host plants from non-host plants by either visual or chemical stimuli and so landed on both. The observations of Vandermeer (1989) on the tobacco hornworm, Manduca quinquemaculata (Haworth), are even more intriguing, as once this moth landed on a plant it flew off the plant and hovered 10–20 cm above the plant surface. It repeated this landing and hovering behaviour 5–10 times on both host and non-host plants before moving off. The only difference was that when it landed on host plants (tomato, Solanum lycopersicum L.) it laid eggs, whereas when it landed on companion plants (bean, Vicia faba L.) it did not. The intriguing factor about the behaviour of both the tobacco hornworm moth (Vandermeer, 1989) and the cabbage root fly (Morley et al., 2005) is that although both spent considerable time on the surface of the companion plant, presumably in a cloud of host-plant odour, there were no lateral flights onto the host plants even though such plants were often only a few centimetres away and hence near enough for true chemotaxis (Fraenkel & Gunn, 1961) to operate.

In their 2005 book, Schoonhoven et al. state that ‘In many specialized herbivores no evident orientation can be demonstrated when the insects are at some distance from their host plant and it appears that to find a suitable plant they must literally bump into it’. The more we have looked at the literature, the more we are convinced that volatile plant chemicals only stimulate receptive flying insects to land and have little if any effect once the insects have landed. Hence, the major difference between specialist and generalist insects is that characteristic volatile chemicals ensure that specialist insects land on fewer ‘inappropriate’ plants (Finch & Collier, 2000) and so ‘bump into suitable (‘appropriate’) plants’ (Schoonhoven et al., 2005) more often.

It is obvious from the few examples cited that observing in detail what the insects actually do (Fabre, 1923) could be extremely rewarding. Supporting the sentiments of Wellington (1977), we believe that it really is time to ‘put the insect back into insect ecology’ particularly in studies that relate to host-plant finding.


We dedicate this manuscript to our mentor the late John Kennedy who taught both of us much more than we ever realized.