* Present address and correspondence: H. van den Berg, FAO ICP, PO Box 22, Peradeniya, Sri Lanka (fax 94 74 476155; e-mail firstname.lastname@example.org).
1.Leptocorisa oratorius, which feeds on the panicle of rice, is a major target for insecticide applications by farmers in Indonesia. The management of this insect has been insufficiently addressed because of unknown damage relationships in modern rice varieties. Hence, a study was conducted to assess the impact of L. oratorius on the yield of variety IR64.
2. Sources of bias typically associated with studies on yield loss (spray effects, caging, plant compensation for damage) were avoided by studying the effect of differences in natural infestation levels of L. oratorius on rice yield at a large number of field sites. Several parameters, including the number of spikelets m−2 prior to panicle emergence (i.e. prior to attack by L. oratorius), were measured at each site to account for local conditions. The effect of these variables on yield was analysed in a statistical model. After removing non-significant variables from the model, the effect unambiguously attributable to L. oratorius density was determined.
3. The median density of L. oratorius in farmers' fields (pooled from panicle emergence until the milky stage of rice) was 0·10 adults hill−1 (i.e. 2·6 adults m−2) in the dry season (75 sites) and 0·14 adults hill−1 (3·5 adults m−2) in the wet season (94 sites).
4. The analysis showed that L. oratorius populations caused no measurable reduction in yield of rice. If a linear relationship in the range of 2–10% yield loss for every added adult per hill was assumed, under the measured field populations of L. oratorius in East Java the median loss would have been in the range of 0·2–1·0% in the dry season of 1997 and 0·2–1·3% in the wet season of 1997/98.
5. Additional field cage experiments were conducted to compare the susceptibility of different stages of rice development to L. oratorius attack and to study yield loss components. In field cages with a bug/panicle ratio substantially higher than observed in the field, L. oratorius reduced yield when feeding at peak flowering of rice, indicating that rice is most susceptible to L. oratorius attack at flowering. The presence of L. oratorius at peak flowering and at the peak milky stage, moreover, caused empty seed. There was, however, no indication that L. oratorius caused partially filled seed. Further study is needed of the possible mechanism by which the plant, under moderate attack from L. oratorius, redirects assimilates from damaged spikelets to those that would have gone unfilled.
6. In conclusion, L. oratorius did not cause an important reduction in rice yield in this study, and we conclude that the application of insecticides targeted at L. oratorius is mostly unwarranted. Farmers need practical training on the biology of L. oratorius and its feeding on rice to appreciate why chemical control is generally not required.
In the Indonesian Integrated Pest Management (IPM) Training Programme, farmers are trained in season-long ‘field schools’ to make crop health management decisions based on their own observations of the crop ecosystem (Gallagher 1992). Through participatory hands-on exercises, farmers learn about the role of natural enemies and plant compensation for damage to leaves or stems. After training, farmers reduce insecticide applications while maintaining or improving yields (FAO 1993).
A major target for insecticide applications both before and after training is the rice bug Leptocorisa oratorius (F.) (Hemiptera: Alydidae) (Pincus 1991; FAO 1993), a common and conspicuous insect that emits a strong odour. The training addresses the management of this insect only to a limited extent. Leptocorisa oratorius, misidentified as L. acuta (Thunb.) in the literature prior to 1965, has been reported from south, east and south-east Asia through to Australia (Ahmad 1965). Although several species of Leptocorisa are found in rice, L. oratorius is by far the most dominant species associated with lowland paddy fields in Indonesia (van der Goot 1949; Siwi & van Doesburg 1984).
Adults and nymphs pierce developing rice spikelets and feed on the ovary [in the case of preflowering or flowering spikelets; hereafter referred to as (pre)flowering spikelets], liquid endosperm (in the case of milky spikelets) or solid endosperm (in the case of dough spikelets). The point of insertion may be visible by white exudate that turns into a brown spot. A number of common fungi, however, also cause seed discolorations (van der Goot 1949; Lee et al. 1986) which are often mistaken for L. oratorius attack by farmers and agriculture extension workers. Upon feeding, the bug secretes a stylet-sheath head on the surface of the spikelet, which can be made visible by staining (Litsinger et al. 1981). Sugimoto & Nugaliyadde (1995), however, found that the stylet-sheath head was often absent from attacked spikelets.
Morrill (1997) observed that, when given the choice on plants in the laboratory, adults strongly preferred (pre)flowering spikelets to milky or dough spikelets. Attack of (pre)flowering spikelets causes empty or partially filled seed (van der Goot 1949; Rothschild 1970b; Morrill 1997), or causes dislodgement of spikelets (W.L. Morrill, personal communication). Attack of milky spikelets reduces seed weight (Morrill 1997) and feeding on dough-stage spikelets may cause stained seed. At the site of the present study on Java, however, farmers are concerned about grain yield, seed ripeness and the proportion of empty seed, not about stained seed.
The influence of feeding by L. oratorius on rice yield has been examined in several studies. van der Goot (1949) observed in the field that individual panicles attacked by adults or clusters of nymphs produced normal seed, and suggested that damage occurred only at high densities of adults. In the laboratory, L. oratorius adults were found to probe 1–9 flowering or milky spikelets per day (Rothschild 1970b; Morrill 1997; Litsinger, Gyawali & Wilde 1998). Based on observations of probed spikelets in the field, Rothschild (1970b) estimated that a field population of 25 adults m−2 would cause 25% grain yield loss in a traditional rice variety. The feeding rates recorded by Litsinger, Gyawali & Wilde (1998), however, indicated that damage would be three times lower. Moreover, Morrill (1997) postulated that it is likely that rice plants will compensate for damage to immature spikelets by filling other spikelets, for not all spikelets develop into filled seed. Extrapolating from observations on individual spikelets to the crop level may therefore be invalid.
Other methods used to study the influence of L. oratorius on rice yield are the insecticide check and field cages. Litsinger et al. (1981, 1987) considered the insecticide check unsuitable, because insecticides also affect stemborers and other common plant feeders in the reproductive and ripening stages of rice. Moreover, field cages, used by Morrill (1997) and S. S. Siwi, A. Yasin & D. Sukarna (unpublished data, 1981), possibly influence the feeding behaviour of L. oratorius, or affect plant development.
The aim of this study was to demonstrate the impact of L. oratorius on the yield of a modern rice variety under field conditions. To avoid possible biases due to caging, insecticides or plant compensation, and to ensure relevance of results for the farmer, we studied the effect of differences in natural infestation levels of L. oratorius on rice yield at a large number of field sites. Additional field cage experiments were conducted to determine which stages of rice development were most susceptible to L. oratorius attack and to study yield loss components.
Materials and methods
Field sites were selected in major lowland rice-growing areas of East Java province at locations where season-long farmer field schools were being conducted. Only lowland (median altitude 80 m a.s.l.) rice paddies planted to IR64, the variety grown by 73% of lowland rice farmers in East Java (Anonymous 1996), were selected as study sites.
In the first season, the dry season of May–December 1997, sites (n = 75) were selected from rice-growing areas throughout East Java (Fig. 1a). In the second season, the wet season from December–May 1997/98, sites (n = 94) were selected from an area limited to four bordering districts in East Java (i.e. Nganjuk, Kediri, Tulung Agung and Blitar; Fig. 1b) to reduce environmental variation. Ninety-four per cent of sites were irrigated. Integrated pest management (i.e. management decisions by farmers based on their weekly agroecosystem observations and supplementary experiments; usually with 0 or 1 insecticide application per season) and locally recommended rates of fertilizers (median 132 kg N, 45 kg P, 50 kg K ha−1) were applied in all study plots.
At each site, 10–25 farmers, divided into groups, made weekly visual observations of an average of 45 hills and recorded numbers of rice bug adults and nymphs, as well as other parameters, as part of the season-long training programme. Hills were selected at regular distances along the diagonals of the plots.
By using farmers' observational data, we anticipated that variation due to sampling would be inflated. Yet, the observations by farmers would allow for straightforward extrapolation of the results to aid in the farmer-training programme.
We assumed that all rice bugs were L. oratorius, which was confirmed by observing this species' characteristic ventrolateral spots on the abdominal segments (Siwi & van Doesburg 1984) on several occasions, but other species, particularly L. acuta, occasionally might have been present. Leptocorisa oratorius adults are readily observed in the field, but nymphs, especially first and second instars, may be overlooked or miscounted by farmers. We therefore omitted nymphs from the analysis. Nevertheless, mean adult density per site was strongly correlated with the recorded nymphal density per site for data of both seasons combined (r2 = 0·56; n = 169).
In the wet season, programme staff conducted additional observations on soil moisture, plant parameters and crop yield. Soil moisture, recorded weekly from panicle emergence until the milky stage, was ranked as (1) dry, (2) humid (‘can enter the field with shoes on’), (3) wet or muddy (‘need to remove shoes’) and (4) flooded; the average rank (1–4) of three sampling occasions was used as a variable for analysis. Just before panicle emergence, 10 booting tillers (picked arbitrarily from 10 hills selected at 2-m intervals along the diagonal of the field) were clipped at soil level in each site. After drying, the number of spikelets per panicle (which is determined well before panicle emergence; De Datta 1981) was counted in the laboratory. At crop maturity, the number of panicles per hill was counted from 20 hills selected at 2-m intervals along an ‘X’ pattern in the field at each site. The number of hills was counted for three 2·5 × 2·5-m crop-cut samples taken along the diagonal of the plot at crop maturity at each site. Grain yield of samples was measured after 1 day of drying.
To provide a measure of initial conditions of plant growth prior to L. oratorius attack, we estimated the number of spikelets per unit area at booting of rice by multiplying the mean number of spikelets per panicle, panicles per hill, and hills per 6·25 m2. The latter two parameters were recorded at crop maturity, but it was assumed that L. oratorius did not affect the number of panicles or hills.
The mean density of L. oratorius adults on three sampling occasions from panicle emergence until the milky stage was used as the variable ‘L. oratorius density’. The effect of L. oratorius density on crop yield in the wet season was evaluated in analysis of per-site data. We used glim (generalized linear interactive modelling), which fits models to data by using maximum likelihood techniques and provides the deviance as a measure of goodness-of-fit (Aitkin et al. 1989). In the maximal model, the influence of the main effects of the explanatory variables ‘district’, ‘planting date’, ‘soil moisture’, ‘initial plant growth conditions’, ‘altitude’ and L. oratorius density on the response variable ‘yield’ were assessed. Interaction terms were not considered. After removing non-significant terms from the model, the minimal adequate model consisted of district (four levels) and planting date; the influence of L. oratorius density was evaluated by addition to the model. The plot of the fitted values vs. residuals showed no pattern, and the probability plot of ordered residuals vs. expected order statistics was a reasonable straight line, indicating that the model structure and normal error specification were satisfactory (macro mcheck; J.A. Nelder in Crawley 1993).
Seed weight (excluding empty seed) and percentage empty seed were determined at crop maturity. In the dry season, a sample of 10 arbitrarily selected panicles was obtained per site at crop maturity. In the wet season, a sample of harvested seed was obtained. After drying, the number of filled (including partially filled) and empty seed and the weight of filled seed were determined in the laboratory for a subsample of 17 ml of seed per site. For analysis of ‘percentage empty seed’, binomial errors were specified with the individual sample sizes as the binomial denominator. Weighted regression was conducted using sample sizes as weights; a logit-link function ensured linearity. William's procedure (Collett 1991) was used to adjust for over-dispersion and the resulting ‘scaled’ deviances were tested with the F ratio.
Field cage experiments
Cages were used to compare the susceptibility of stages of rice development to L. oratorius attack, using a high bug/panicle ratio to ensure measurable damage. Two series of experiments were conducted at one farmer's field site in Malang district, East Java, during the rainy seasons of 1996/97 and 1997/98. Rice variety IR64 (International Rice Research Institute, Los Baños, Philippines) was planted in hills at 0·2 × 0·2-m spacing. When rice tillers were booting, individual hills, chosen in a regular pattern in the field, were covered with nylon mesh cages (0·3 × 0·3 × 1·2 m, L × W × H). Arthropods were removed from the cages and the lower margins of the nets were sealed in the mud. At peak flowering, the number of panicles was fixed to five per cage; the surplus of booting or heading tillers was clipped at the stem base. At subsequent observations, any newly emerged panicle was clipped.
Four treatments were established, using a completely randomized design: (1) inclusion of L. oratorius at peak flowering stage of rice; (2) inclusion at peak milky stage; (3) inclusion at peak soft dough stage; and (4) the control without L. oratorius. For inclusion of L. oratorius, one field-collected adult female was put into each cage (i.e. one per five panicles) for a period of 7 days during the appropriate development stage of rice. Eggs and nymphs were removed weekly from the cages. At crop maturity, panicles from each cage were harvested and sun-dried for 3 days. The number of filled and empty seed, the weight of filled seed, and the average seed weight (excluding empty seed) were determined for each cage. If after 7 days a bug was found dead or missing, the cage was omitted from analysis. Valid data were obtained from 20 cages in series 1 (five cages per treatment), and from 30 cages in series 2 (nine, seven, five and nine cages for treatments 1, 2, 3 and 4, respectively).
The response variables ‘yield’ (i.e. weight of filled seed, in g per cage), ‘percentage empty seed’, ‘total number of seed’ and ‘seed weight’ were analysed in a model with the factors treatment (four levels) and series (two levels). Deviance attributable to each factor was determined by deletion from the model with the main effect of each factor. The non-significant interaction term had been removed from the model. The difference between any pair of treatment levels was assessed through estimation by the model; t-tests of differences between parameter estimates (and SE) were conducted with the appropriate degrees of freedom for each combination. Percentage empty seed was analysed as described above.
On average, plant density was 25·9 hills m−2 and hills had 15 panicles in the wet season 1997/98. The number of spikelets per panicle and the number of panicles per hill varied as much within sites as between sites, but the number of hills and grain yield per area unit varied less within sites than between sites (Table 1).
Table 1. Summary of plant observations on rice variety IR64 at 94 sites during the wet season 1997/98; East Java
Standard deviation of samples within sites, averaged over all sites.
Number of samples within sites.
Spikelets per panicle
Panicles per hill
Hills per 6·25 m2
Yield (103 g) per 6·25 m2
Spikelets within a panicle mature at different times (Sikder & Das Gupta 1976). Figure 2 shows that the flowering spikelets were present in the crop from peak panicle emergence until peak soft dough stage.
The incidence of L. oratorius was distributed fairly uniformly over East Java (Fig. 1a). The districts Kediri and Blitar had more sites with above-average densities than the districts Nganjuk and Tulung Agung (Fig. 1b).
Leptocorisa oratorius adult densities generally increased during panicle emergence of rice, and densities were highest during the flowering and milky stages (Fig. 3). In the wet season of 1997/98, densities declined as rice matured, presumably because adults moved to fields at earlier growth stages. Nymphal densities (not presented for reasons discussed earlier) were highest at the milky and dough stages of rice development.
The median density of L. oratorius adults (pooled over the three sampling occasions from panicle emergence until milky stage) was 0·10 (mean 0·14) adults hill−1[i.e. 2·6 (mean 3·6) adults m−2] in the dry season and 0·14 (mean 0·23) adults hill−1[i.e. 3·5 (mean 6·0) adults m−2] in the wet season. The majority of sites (92% and 81% during the dry season and wet season, respectively) had less than 0·3 adults hill−1 (Fig. 4). Densities above 1 adult hill−1 were rare; the highest recorded density was 1·34 adult hill−1 (or 34·7 m−2).
Figure 5 shows the relationship between the incidence of L. oratorius and yield at individual sites. In single regression, the factor district explained 35% of the total deviance in yield (F3,90 = 15·1; P < 0·01), indicating that crop yield varied considerably between districts. Planting date explained 16% of total deviance in single regression (F1,92 = 14·6; P < 0·01), whereas in covariance analysis planting date explained 13% of total deviance; the remaining 3% was contained within district. The latter model predicted a 0·3% yield reduction (−0·013 divided by 4·66) for every added day after 3 December 1997 (Table 2). Crops were planted slightly later in Blitar than in other districts in the wet season.
Table 2. Analysis of deviance table for yield (103 6·25 m−2) of rice variety IR64 per site (n = 94) in East Java; wet season 1997/98. P indicates the probability associated with the effect of a variable(s). NS =P > 0·05
Model with all insignificant terms removed; deviance attributable to a variable was assessed by deletion of the variable from the model.
Leptocorisa oratorius density was added to the minimal adequate model.
Soil moisture showed a positive relationship with yield in single regression (F1,92 = 4·7; P < 0·05). However, after including district in factorial analysis, soil moisture was not significant (F1,89 = 0·3; P > 0·05). This indicates that although soil moisture was probably important to crop yield it was associated with district (and to a lesser extent with planting date). For example, in Blitar both soil moisture and yield were lower than the average.
The effect of L. oratorius density on crop yield was assessed in several models. Single regression predicted a 12% yield loss at 1 adult hill−1, but the regression was only marginally significant (Table 2). Treating L. oratorius density as a factor (with three categories, as indicated in Fig. 5) slightly improved the description of data; L. oratorius density was non-significant (Table 2).
The minimal adequate model with the main effects of district and planting date explained 54% of total deviance in yield per site (Table 2); non-significant variables had been removed. Initial plant growth conditions provided no significant improvement of the resolution of analysis (F1,88 = 1·3; P > 0·05). Perhaps a high natural variation in percentage empty seed between sites or a sampling error due to selection of panicles in the field (which was conducted by a different person at every site) obscured a relationship between initial plant condition and yield.
Adding L. oratorius density to the minimal adequate model did not improve the fit (F1,88 = 0·0; P > 0·05), indicating that the deviance unambiguously attributable to L. oratorius density was non-significant. Therefore, there was no evidence that L. oratorius influenced the yield of rice. However, L. oratorius density showed a marginal positive correlation with planting date (r2 = 0·04; n = 94; P = 0·054), which may have caused a slight underestimation of its influence on yield.
Regarding L. oratorius densities at separate plant stages, no significant effects on yield were found (L. oratorius densities during panicle emergence, flowering stage and milky stage corresponded with F1, 88 ratios of 2·7, 0·0 and 1·9, respectively).
The seed weight (excluding empty seed) was fairly uniform between sites (dry season, mean 25·7 mg, SD 1·6 mg; wet season, mean 24·8 mg, SD 1·9 mg). Analysis of covariance showed that seed weight differed between seasons (F1,158 = 8·8; P < 0·01), but was not influenced by L. oratorius adult density (F1,158 = 0·0; P > 0·05).
The percentage empty seed varied greatly between sites (from 3% to 33% in the dry season, and from an unknown figure to 41% in the wet season). Unfortunately, empty seed had been removed from an unknown number of samples of the wet season and, thus, percentage empty seed could only be analysed for the dry season. None of the variables district, date, or L. oratorius density caused a significant reduction in deviance in percentage empty seed (P > 0·05). Likewise, single regression of L. oratorius density failed to show a relationship with percentage empty seed (F1,63 = 1·3; P > 0·05).
Field cage experiments
At a high bug/panicle ratio in field cages, L. oratorius reduced yield more when feeding during the flowering stage than during the dough stage (Fig. 6a). Factorial analysis showed that yield (weight of filled seed per cage) was significantly affected by treatment (F3,45 = 4·0; P < 0·001). The treatment × series interaction was not significant (F3,42 = 0·9; P > 0·05), indicating that treatment had a similar effect in both series.
Similar analysis with percentage empty seed as the response variable showed a significant effect of treatment (F3,45 = 7·4; P < 0·01). Percentage empty seed in the flowering-stage treatment was higher than that in the dough-stage treatment or in the control, and percentage empty seed in the milky-stage treatment was higher than that in the control (Fig. 6b). This indicates that the presence of L. oratorius during flowering and, to a lesser extent, during the milky stage caused empty seed.
The total number of seed per hill was not influenced by treatment in factorial analysis (F3,45 = 0·5; P > 0·05), suggesting that L. oratorius did not cause important seed dislodgement at a high bug/panicle ratio. Moreover, seed weight was not affected by treatment (F3,45 = 0·4; P > 0·05) and, therefore, there was no evidence that L. oratorius caused spikelets to be partially filled.
Next, we established whether the weight of small yield samples per cage (5–20 g) would be measurable at the farmer level, for the programme's training purposes. In the cage studies of 1996/97, yield measurements using a kitchen scale and a jeweller's scale were compared with those using an electrobalance. The kitchen scale, which is available in every village, was inaccurate in comparison with the electrobalance (r2 = 0·74; n = 20), whereas the jeweller's scale was sufficiently accurate (r2 = 0·999; n = 20) but is generally unavailable to farmers.
As an alternative to weight measurements, we examined whether yield could be estimated adequately by counts of empty or filled seed. The number of empty seed showed a poor negative correlation with yield (r2 = 0·17; n = 20), but the number of filled seed was strongly correlated with yield (r2 = 0·91; n = 20). Hence, the most suitable method for farmers to determine yield per cage is to count the number of filled seed per cage. Based on these results, a modified cage exercise for farmer training was developed that avoided the use of replication by counting the number of spikelets per cage at booting of rice and the number of filled spikelets per cage at ripening.
The field observations showed that L. oratorius populations caused no measurable reduction in yield of a modern rice variety with its high number of spikelets per unit area. If we assume a linear relationship in the range of 2–10% yield loss for every added adult per hill, under the measured field populations of L. oratorius in East Java the median loss would have been in the range of 0·2–1·0% (mean 0·2–1·4%) in the dry season 1997 and 0·2–1·3% (mean 0·5–2·3%) in the wet season 1997/98. The presence of a convex relationship due to plant compensation at moderate densities of L. oratorius would imply a lower loss than these estimates. Hence, we conclude that L. oratorius feeding does not cause an important reduction in rice yields in East Java. Chemical applications targeted at natural L. oratorius populations are mostly unwarranted, particularly when the environmental and human health costs are considered. Our results further suggest that the economic threshold level of 2–4 adults m−2 (Dyck et al. 1981), which has been widely recommended in Asia, is inappropriate for modern rice varieties in East Java.
In contrast to the field observational data, the field cage experiments demonstrated a clear yield loss following attack during the flowering of rice. However, the L. oratorius density was substantially higher (equivalent to 3 hill−1) than observed at any field site (the highest field density at flowering observed at any site was 1·8 adults hill−1). Moreover, only females were used, which feed 30–100% more than males (Rothschild 1970b; Morrill 1997), whereas the sex ratio is close to 1 (Rothschild 1970a). Although caged adults may spend more time on the muslin wall than on the panicles (Rothschild 1970b), this does not necessarily imply that they feed more under free conditions. In fact, free adults are highly mobile (Rothschild 1970a; Sands 1977; K.L. Heong, personal communication) and may visit non-rice habitats for feeding. Moreover, caging provides shade comparable to overcast weather, which has been reported to extend the feeding period of adults during the daytime (Uichanco 1921).
The percentage empty seed of IR64 at maturity was highly variable and depended largely on factors other than L. oratorius. The rice bug caused empty seed, presumably by attacking (pre)flowering spikelets that were subsequently aborted, but the effect was measurable only at the high bug/panicle ratio in the field cage experiments. Possibly, at moderate densities of L. oratorius, plants redirect assimilates from attacked spikelets to those that would have gone unfilled, as has been suggested by Morrill (1997), but this requires further study.
The seed weight (excluding empty seed) was unaffected by L. oratorius, both in the field observations and in the cage experiments. Rothschild (1970b) reported that weight loss of attacked spikelets was 40–60%, but this pooled average presumably included spikelets that remained empty after attack. Morrill (1997), who distinguished between different development stages of spikelets, found that mean weight loss after attack was larger for (pre)flowering spikelets than for dough spikelets; milky spikelets had an intermediate weight loss. However, attack of milky and dough spikelets might be relatively rare events in the field because (pre)flowering spikelets are strongly preferred over milky or dough spikelets (Morrill 1997) and are present in the crop over an extended period (Fig. 2). We conclude that, at the field level, L. oratorius caused spikelets to remain empty, but there was no evidence that it caused spikelets to be partially filled, probably due to the bug's preference for young spikelets.
High infestation levels of L. oratorius have occasionally been associated with situations of large losses or crop failure (Srivastava & Saxena 1967; Rothschild 1970a), but causal evidence has been scant. van der Goot (1949) reported that high adult densities occur either when rice is planted synchronously near extensive uncultivated habitats with wild grasses that support L. oratorius populations, or when small patches of late-planted rice attract adults from surrounding earlier-planted rice fields. Indeed, L. oratorius generally appears to be more abundant in rice in sparsely cultivated areas than, for instance, on the intensively cultivated island of Java. Already in the 1930s, van der Goot (1949) speculated that the uncultivated area on Java was too small to sustain L. oratorius populations able to cause high infestation levels over large areas of synchronous-planted rice.
On rice, L. oratorius was found to suffer 93% mortality from eggs to the fifth nymphal stage in Sarawak, Malaysia; most mortality took place during the egg and first nymphal stages and was attributed to egg parasitism and predation (Rothschild 1970a, 1970c). Others have reported lower levels of egg parasitism (Corbett 1930; van der Goot 1949; Sands 1977; Morrill & Almazon 1990). Rothschild (1970a) further detected high adult mortality in the field. In East Java, we occasionally observed epizootics of Beauveria bassiana (Bals.) Vuill. when the densities of L. oratorius were high and the weather conditions humid. Hence, natural enemies may play an important role in suppressing L. oratorius, and their role should be considered in the management of the rice bug. The role of predation, particularly on young stages, however, requires further field assessment.
Nymphs were not considered in our analysis, but densities of adults and nymphs per site were highly correlated. Moreover, late-instar nymphs generally developed only after the most susceptible stage of rice had passed and, thus, their influence on crop yield was probably limited, but this requires further study. Nymphs may build up earlier if early flowering graminaceous weeds are present in the field (Morrill, Pen-Elec & Almazon 1990; Sugimoto & Nugaliyadde 1996). Nymphs can continue to increase throughout the ripening stage of rice, and nymphs developed from eggs laid from the milky stage onwards starve after harvest of rice (Rothschild 1970a). Consequently, the role of predation and parasitism would be most crucial during panicle emergence and flowering, when young bug stages could reach the adult stage before the time of harvest.
Typically, conspicuous insects or those causing obvious feeding marks are readily treated with chemicals by farmers, although more often than not the perceived loss is higher than the actual loss. This appears also to be the case for L. oratorius in East Java. Therefore, training is required to improve farmers' perceptions of L. oratorius; being conspicuous makes the insect a suitable object for training.
Because experiential learning by farmers has been shown to have a positive impact on farmers' practices (FAO 1993), simplistic messages on how to control L. oratorius (e.g. to spray at 0·5 or 1 adult hill−1) are best avoided. Instead, exercises have been developed to help farmers better understand the biology of L. oratorius and its feeding on rice. These exercises include observations on adult feeding behaviour in the field, the development period of immature stages, egg parasitism and predation, and modified cage studies to assess damage. The training was aimed at helping farmers understand why insecticide applications are generally not required or desirable against natural populations of L. oratorius, as demonstrated by our data.
We are grateful to the Programme Field Leaders Abdul Chanan, Fathoni, Mastur, G. Prawito, G. Rochmawan, Santoso, Sarjuli, Sugeng, A. Sukani, Sumarwan, B.H. Susetyo, Suyadi, Suyoto and R. Yekti for co-ordinating field activities, and to the numerous programme field trainers and IPM farmers for their participation in the field observations. We also thank P.M. Santoso and S. Hartono for their assistance in the laboratory, and W. Setiwan, Sugiat, Khojin and Kasiyono for their assistance in the field cage experiments. We thank K.L. Heong and W.L. Morrill for valuable discussions, and W.L. Morrill, P.A.C. Ooi, B.M. Shepard, J.K. Waage and an anonymous referee for their comments on the manuscript.
Received 17 December 1998; revision received 15 March 2000