Journal of Chemical Ecology, Vol. 26, No. 2, 2000
TASTE SENSITIVITY OF INSECT HERBIVORES TO
DETERRENTS IS GREATER IN SPECIALISTS THAN IN
GENERALISTS: A BEHAVIORAL TEST OF THE
HYPOTHESIS WITH TWO CLOSELY RELATED
CATERPILLARS
E. A. BERNAYS,1,* S. OPPENHEIM,2 R. F. CHAPMAN,3
H. KWON,1 and F. GOULD2
1
Entomology Department, University of Arizona
Tucson, Arizona
2
Entomology Department, North Carolina State University
Raleigh, North Carolina
3
Division of Neurobiology, University of Arizona
Tucson, Arizona
(Received June 3, 1999; accepted October 12, 1999)
Abstract Sensitivity of caterpillars of Heliothis virescens, a generalist,
and H. subflexa, a specialist, to eight different plant secondary compounds
was examined behaviorally. The compounds were nicotine hydrogen tartrate,
hordenine, caffeine, sinigrin, linamarin, arbutin, chlorogenic acid, and salicin.
All compounds deterred feeding, at least at the higher concentrations, but
the generalist was less affected than the specialist. Thus the hypothesis that
specialists have greater sensitivity to deterrents than generalists was supported.
In most cases deterrence occurred on first encounter, indicating that the
response was sensory; in some cases short-term postingestive effects also
appeared to play a role. The larger quantities of deterrent-containing food
ingested by H. virescens sometimes resulted in measurable postingestive
effects during the second control test. This did not occur in H. subflexa, which
more commonly rejected the deterrent-containing food on first contact. The
contrast between the species is discussed in relation to tradeoffs involved in
different diet breadths.
Key Words Heliothis virescens, Heliothis subflexa, caterpillar, diet breadth,
deterrent compound, feeding behavior, postingestive toxicity, plant secondary
metabolite.
*To whom correspondence should be addressed.
547
0098-0331
/00/0200-0547$18.00/0 © 2000 Plenum Publishing Corporation
548 BERNAYS ET AL.
INTRODUCTION
Food selection by specialist herbivores may depend on specific chemical sign
stimuli from host plants as well as high levels of sensitivity to secondary metabo-
lites in nonhosts, making them unacceptable (reviewed by Bernays and Chap-
man, 1994; Schoonhoven et al., 1998). The available evidence for comparisons
within taxa suggests that sensitivity to secondary metabolites is greater among
the specialists than the generalists. For example, Jermy (1966) showed that
among five beetles the more extreme specialists were more deterred by nonhosts.
Among caterpillars, Spodoptera exempta, which is restricted to grasses, was
more deterred by certain nonnutrient compounds than the generalist Spodoptera
littoralis (Blaney et al., 1987, 1988), and among grasshoppers, the grass-feed-
ing Locusta migratoria was much more deterred than the generalist Schistocerca
gregaria by 79 of 82 plant secondary metabolites tested (Bernays and Chapman,
1987).
These findings led Bernays and Chapman (1994) to suggest that general
differences in sensitivity to secondary metabolites might be important steps in
changes in diet breadth. The species examined in these studies were not closely
related, however, and the differences observed may be confounded by differ-
ences not directly connected with food selection. In addition, these studies all
involved  end-product assays, in which insects are presented with a food sub-
strate of some kind, with or without the test chemical, and the amounts eaten
after variable periods measured. We believe that more can be revealed about
potential mechanisms by continuous observations over short periods (Bernays
and Weiss, 1996). For example, insects may habituate to the deterrent chemicals
during a longer-term choice test. Usher et al. (1988) found that some compounds
were deterrent to the caterpillar, Pseudaletia unipuncta, for a brief period only.
It is also possible that apparent deterrents are not tasted and are readily ingested
initially, and then rapidly occurring, noxious postingestive effects prevent further
feeding. This has been found with several species of caterpillar (Glendinning and
Slansky, 1995; Glendinning, 1996).
In this study we employed two very closely related species of the noc-
tuid genus Heliothis the extreme generalist, H. virescens (Graham and Robert-
son, 1970; Neunzig, 1969; Stadelbacher, 1981), and the specialist H. subflexa,
which is restricted to Physalis spp. in nature (Mitchell and Heath, 1987). This
species pair provides a model system that is under study with respect to the evo-
lution of diet breadth (Sheck and Gould, 1996) and with respect to the possible
chemosensory mechanisms underlying the different feeding habits (Bernays and
Chapman unpublished data). We employed continuous observations to test for
subtle differences that might be important in understanding possible physiologi-
cal and ecological processes. The specific questions asked in this work are: (1)
Is H. subflexa more readily deterred by a variety of secondary metabolites than
TASTE SENSITIVITY OF INSECT HERBIVORES 549
H. virescens? (2) Is there behavioral evidence of immediate sensory detection
of the compounds in either species? (3) Are deterrent responses mediated after
ingestion, providing evidence of postingestive or experiential effects in either
species?
METHODS AND MATERIALS
Insects. Heliothis virescens (F.) larvae were obtained from the YDK strain,
which is described in Gould et al. (1995). The H. subflexa (Gn.) larvae were
taken from a colony that was originally collected near Orangeburg, South Car-
olina, in 1997. All larvae of both species were maintained on the same artificial
diet (Sheck and Gould, 1996). Only by using the artificial diet could experience
of insects be controlled, since the plant host of H. subflexa is unsatisfactory for
growth of H. virescens.
Diets. For the deterrence tests, chemicals were incorporated into the standard
artificial diet at a series of concentrations expected to cover the dynamic range of
responses by the insects as well as the naturally occurring concentrations. The use
of artificial diets controlled for all food variables other than the test chemical. The
diet (86% water) was assumed to have the density of water. Thus, for example,
the molecular weight in milligrams was added to 1 g wet weight of diet to make
a concentration called 1 M. The aliquots of diet were warmed to make them liq-
uid and the chemicals thoroughly mixed, before they were placed in a refrigerator
and allowed to reset. The diet when used was semisolid but not gelled and had a
somewhat granular texture. The control diet was similarly heated and reset.
Chemicals. The selection of secondary metabolites for testing was gov-
erned by three factors. First, they were chosen to cover a range of chemical
classes alkaloids (three different subclasses), phenolic glycosides, glucosino-
lates, cyanogenic glycosides, and phenylpropanoids. Second, we wanted to be
sure that they were typical of nonhosts, at least of the specialist H. subflexa.
While difficult to ensure, a wide investigation of the literature indicated that this
was so, although one, chlorogenic acid, is known to occur in some host plants
of the generalist, H. virescens. Third, we were limited by the need to test com-
pounds that were polar, in order to be able to study their electrophysiological
effects on the taste sensilla, where the technique requires dissolving the com-
pounds in water. Finally we were constrained by what is commercially available.
The chemicals, their class, and source are listed in Table 1.
Testing Regime. All insects used in experiments were last-instars in the
middle of the active feeding period. Each was confined in a Petri dish contain-
ing wet filter paper and deprived of food for 1 4 hr in a lighted room with a
controlled temperature of approximately 27 C. High relative humidity proved
to be important in preventing hyperactivity. Each individual was subjected to a
three-part trial to examine its initial readiness to feed on plain diet, the amount
550 BERNAYS ET AL.
TABLE 1. CHEMICALS USED IN EXPERIMENTS
Chemical Chemical class Source
Nicotine hydrogen alkaloid (pyridine) Aldrich
tartrate (NHT)
Hordenine hemisulfate alkaloid (amine) Aldrich
Caffeine alkaloid (methylxanthine) Sigma
Sinigrin glucosinolate Sigma
Linamarin cyanogenic glycoside Sigma
Arbutin phenolic glycoside Aldrich
Chlorogenic acid phenylpropanoid Sigma
Salicin phenolic glycoside Sigma
of feeding on a test diet, and, finally, its readiness to once again feed on the
plain diet. The testing protocol was as follows: (1) Control 1 presentation of a
small lump of plain diet on a piece of aluminum foil approximately 1 cm2. The
first part of the test began when the caterpillar walked into the diet and made
mouthpart contact with it. The caterpillar was allowed 3 min from this contact,
after which the foil with the remaining diet was carefully removed. (2) Test a
similar piece of the test diet containing a particular concentration of one of the
deterrents was in like manner presented to the caterpillar. Three minutes were
again allowed from the time of first contact with the food, before the test diet
was removed. (3) Control 2 plain diet was again presented with 3 min allowed
from the time of contact. Pilot experiments showed that insects deprived of food
for 1 4 hr fed continuously on artificial diet for at least 10 min, so we could
assume that the caterpillars were not satiated even if they fed for the full 9 min.
Once the feeding trial was completed, the caterpillar was discarded.
Each caterpillar was observed continuously until the three-part trial was
finished. The activities of the caterpillars were recorded by a program made
with The Observer software (Noldus, 1991) loaded on laptop computers. The
behaviors recorded were mouthpart contact, feeding, resting, and walking. From
these data we calculated time spent feeding during the 180 sec allowed for each
diet piece, the occurrence of rejection events, the numbers of feeding bouts, and
the proportion of time spent walking vs. resting. No caterpillar was used more
than once. Several caterpillars were observed at the same time by each observer.
Between presentation of the diet pieces, caterpillars had variable periods of inac-
tivity, presumably due, at least in part, to different levels of disturbance. On some
occasions, caterpillars remained motionless for periods of longer than 30 min and
were removed from the experiment. The average time between the end of one
part and the beginning of the next was 11 min.
Observations involving any one chemical occurred in a series of experi-
ments carried out over at least three days. On any one day replicates of every
TASTE SENSITIVITY OF INSECT HERBIVORES 551
concentration were included. In some cases both species could be tested on the
same day, but mostly this was not feasible as their development times differed
considerably. At each concentration of each chemical for each species, the num-
ber of replicates generally varied from 6 to 12 with an average of 8. However,
if several low concentrations proved to have no effect on the amount eaten, the
lowest ones were not always included in the second and later days of the exper-
iment, so that some of the lowest of the ineffective concentrations have as few
as three replicates.
Analyses. For each concentration of each chemical, we used the amount of
time spent feeding on the first piece of plain control diet as a measure against
which the time feeding on the two successive tests was measured. Insects that
did not feed on the first control for at least 120 sec of the total possible 180
sec were not included in the analysis, on the assumption that they were, for
some reason, not fully ready to feed. Only 2% fell into this category. The times
spent feeding on the test chemical diet and on the second control diets were
then calculated as a percentage of the time spent feeding on the initial control
diet. For example, if the insect fed for 160 sec on the first control and 80 sec
on the test diet, its value for the test diet was 50%. These data are presented
graphically. The significance of trends was measured by analysis of covariance
with concentration as the covariate in the statistical program JMP (SAS). Species
differences may be seen directly or as concentration × species interactions. In
one case, there were no effects either way, but at the highest concentration the
two species appeared markedly different. In this case, a t test showed differences
at that concentration.
From the concentration response data, the approximate concentration that
would reduce feeding by 50% was estimated graphically based on an approxi-
mated linear dose response for each insect. There were insufficient data to use
probit analysis. However, the estimated ED50 values (concentrations reducing
feeding by 50%) give an additional way to compare the species. We also estab-
lished the lowest concentration that was effective in reducing feeding by using
paired t tests in which the total time spent feeding by each insect on control 1
and the test diet was compared.
In attempting to further elucidate mechanisms involved in deterrence, we
examined the percentage of individuals rejecting test diet on first encounter at
each concentration and average numbers of separate bouts taken on each of the
test diets.
When not feeding on the test diets, caterpillars were resting or walking.
Experience of some chemicals might elicit relatively more locomotion, which
would translate into movement for longer distances away from such a substrate
in nature, and a species difference would be of interest ecologically. Some differ-
ences between chemicals appeared to be consistent, but because no differences
were observed between the insect species, the data are not presented here.
552 BERNAYS ET AL.
The second control test following the deterrent test was initially included to
ensure that reduced feeding on the test diet did not occur simply because the cater-
pillar was replete. While reduced feeding was not seen with low concentrations
of test compounds, there was evidence of persistent effects of the secondary com-
pounds at higher concentrations in some cases. We therefore analyzed the amounts
eaten in the second control trial and tested for differences between the two species.
RESULTS
All eight chemicals deterred H. subflexa more than H. virescens at one or
more concentrations (Figure 1). Analyses of covariance show that for seven of
the eight sets of trials there was either a direct species effect or an interac-
tion between concentration and species (Table 2). In the case of the exception,
arbutin, the levels of variation were relatively high and the dose response curves
less regular. However, even here, the highest concentration, when tested sepa-
rately, was significantly more deterrent to H. subflexa than H. virescens (Stu-
dent s t test, P < 0.03). In every case, the lowest concentration at which a dif-
ference from the control diet was observed was lower in H. subflexa than in H.
virescens (asterisk in Figure 1), often by an order of magnitude. H. subflexa was
significantly deterred even by the lowest (0.0001 M) concentration of nicotine
hydrogen tartrate (NHT), caffeine, and sinigrin.
The doses that caused 50% reduction in feeding (ED50) range over four
orders of magnitude, with caffeine and sinigrin being the most deterrent to both
species (Table 3). The approximate ED50 for H. subflexa was lower than for H.
virescens for all of the eight chemicals. In the case of linamarin, the difference
was greater by more than an order of magnitude, but it was less for the other
seven chemicals.
The concentration response curves shown in Figure 1 (and the associated
ANCOVA results) might suggest that the two insect species differ qualitatively
in response to the chemicals. However, the concentrations tested do not always
cover the dynamic range of the response, and, in the case of H. virescens, two
chemicals, linamarin and salicin, had relatively little effect at the concentrations
tested.
FIG. 1. (Opposite) Effects of different concentrations of eight plant secondary metabolites
on feeding by H. virescens and H. subflexa. Time feeding in the three minute test with
diet plus chemical is shown as a percent of the time feeding on the prior test with plain
diet (see text for further details). The vertical bars represent standard errors. The asterisk
indicates the lowest concentration at which each species showed a significant difference
from the control, using t tests at P < 0.05.
TASTE SENSITIVITY OF INSECT HERBIVORES 553
554 BERNAYS ET AL.
TABLE 2. ANCOVA RESULTS RELATING TO EFFECTS OF DIFFERENT CONCENTRATIONS OF
CHEMICALS ON TWO SPECIES OF CATERPILLARSa
df F P
NHT
Species 15.0926 0.0256
Concentration 173.1037 < 0.0001
S × C 10.4435 0.5066
Hordenine
Species 12.7759 0.0992
Concentration 163.6933 < 0.0001
S × C 113.4948 0.0004
Caffeine
Species 17.6466 0.0077
Concentration 19.5542 0.0031
S × C 10.5998 0.4420
Sinigrin
Species 19.0280 0.0034
Concentration 115.8268 < 0.0001
S × C 10.0207 0.8860
Linamarin
Species 132.9033 < 0.0001
Concentration 117.9068 < 0.0001
S × C 13.9247 0.004
Arbutin
Species 10.8670 0.3546
Concentration 118.4699 < 0.0001
S × C 12.2557 0.1371
Chlorogenic acid
Species 124.1418 0.003
Concentration 188.8303 < 0.001
S × C 15.9041 0.0167
Salicin
Species 15.7395 0.0184
Concentration 141.8952 < 0.0001
S × C 120.3922 < 0.0001
a
Probability values in bold indicate where species differences occur.
When caterpillars encountered the control diet for the first time in the tests,
rejection without ingestion was rare (less than 1% of all trials). In most cases, the
caterpillar initiated feeding and then fed continuously for the 3 min allowed. In
some cases, feeding was interrupted by gaps of varying length (15% of all trials).
With diets containing test chemicals, rejection with no feeding often occurred
at the higher concentrations. The caterpillar in such cases would either remain
stationary, occasionally moving the head from side to side, or it would walk away
from the diet. In either case, further mouthpart contact with the food sometimes
TASTE SENSITIVITY OF INSECT HERBIVORES 555
TABLE 3. ESTIMATED CONCENTRATION (M) OF CHEMICAL THAT REDUCED FEEDING
DURATION TO HALF THAT ON CONTROL DIET (ED50)
Chemical H. virescens H. subflexa
NHT 0.01 0.008
Hordenine 0.40.08
Caffeine 0.001 0.0003
Sinigrin 0.002 0.0003
Linamarin > 0.10.009
Arbutin 0.08 0.02
Chlorogenic acid 0.10.02
Salicin > 0.1 0.05
occurred, resulting in either an additional rejection, or a short bout of ingestion.
In some cases, usually at the lower concentrations of chemicals, there was no
initial rejection, but one or more bouts of ingestion of varying length. Contact
followed by rejection also often occurred once or more, before a later period of
ingestion.
Rejection at first encounter, unequivocally indicating a sensory response,
was observed with all eight chemicals (Figure 2). There were clear differences
between the two species for hordenine, caffeine, linamarin, arbutin, chlorogenic
acid, and salicin, and in no case was H. virescens more sensitive than H. subflexa
at first encounter. However, for any one chemical, data are generally insufficient
to ascribe statistical value to the differences. There were big differences among
chemicals. With caffeine and sinigrin, most individuals of both H. subflexa and
H. virescens rejected the test diets at first encounter across most concentrations
except the very lowest, where feeding was not significantly different from the
first control (in the case of H. virescens). On the other hand, with hordenine,
and especially NHT, there was little rejection behavior except at the highest
concentrations for a given level of deterrence than with other chemicals only
half of all individuals rejected the diet containing the highest concentration at
first encounter, although final feeding times were very short. Across all trials,
chemicals, and doses, there was a strong correlation between mean depression
in feeding time and percentage of individuals that showed rejection responses
on first encounter (H. virescens, N 374, r 0.75, P < 0.01; H. subflexa, N
325, r 0.81, P < 0.01).
When ingestion was initiated, short bouts of ingestion occurred at the higher
concentrations of chemicals. At all concentrations that reduced feeding, there
were individuals that took some additional bouts after the first one. At extremely
deterrent concentrations, one short bout was most common, and at concentrations
that deterred feeding very little, one long bout was most common. With concen-
trations having intermediate effects, there tended to be relatively more bouts,
556 BERNAYS ET AL.
TASTE SENSITIVITY OF INSECT HERBIVORES 557
especially in H. subflexa. Thus, in the range of 30 70% reduction in feeding
(collapsed across chemicals), H. subflexa averaged 2.4 bouts, while H. virescens
averaged 1.5 bouts (Mann-Whitney U test, P < 0.05). In other words, the reduced
period of feeding in H. subflexa resulted from more short bouts while in H.
virescens it resulted from fewer longer bouts. As with rejection at first encounter,
caffeine and sinigrin were unusual, and in both species the average number of
bouts across all concentrations was three.
When offered the control diet following experience on the test diet, a major-
ity of insects fed for over 75% of the time spent feeding on the first control (Fig-
ure 3). However, some chemicals affected the amount of feeding in the second
control (Figure 3, Table 4). There were significant negative correlations between
the estimated amount of chemical ingested by an individual caterpillar during the
deterrent test and the length of time feeding during the second control with NHT,
caffeine, linamarin, chlorogenic acid, or salicin for H. virescens but not for H.
subflexa (Table 5). However, feeding by H. subflexa was depressed in the second
control after feeding (or exposure to) NHT and chlorogenic acid (Figure 3). In
both these cases, a similar suppression of feeding occurred in caterpillars that
had experienced very different test concentrations of the chemicals and eaten
very different amounts of them.
DISCUSSION
In this study, the specialist noctuid, H. subflexa, was found to be more
deterred than the generalist, H. virescens, by the eight plant secondary metabo-
lites tested. This is consistent with the model of Bernays and Chapman (1994),
which suggests that differences in taste sensitivity to deterrent compounds could
account for the difference in host range. Phylogenetic studies indicate that the
common ancestor of the two species examined here was a generalist (Mitter et
al., 1993; Fang et al., 1997), so that an important step towards restriction of diet
breath could have been increased sensitivity to secondary metabolites in plants.
Simple genetic differences in the sensitivity of the deterrent receptor cells of
Bombyx mori in relation to diet breadth (Asaoka, 1994) imply that the effect
could be peripheral, although preliminary work so far with the Heliothis species
suggests that the differences are not peripheral and must therefore be in the cen-
tral nervous system (Bernays and Chapman, unpublished).
This study involved the use of artificial diets rather than host plant parts
FIG. 2. (Opposite) Percentage of individual insects tested that rejected the deterrent-
containing diets on first encounter at different concentrations.
558 BERNAYS ET AL.
TASTE SENSITIVITY OF INSECT HERBIVORES 559
TABLE 4. ANCOVA RESULTS RELATING TO EFFECTS OF DIFFERENT CONCENTRATIONS
OF CHEMICALS ON SUBSEQUENT FEEDING ON CONTROL DIET FOR TWO SPECIES OF
CATERPILLARSa
df F P
NHT
Species 10.0362 0.8494
Concentration 15.7980 0.0174
S × C 14.1288 0.0442
Hordenine
Species 10.1937 0.6611
Concentration 11.0694 0.3042
S × C 10.5280 0.4696
Caffeine
Species 10.1303 0.7195
Concentration 12.8262 0.0984
S × C 14.0111 0.0501
Sinigrin
Species 10.3259 0.5694
Concentration 10.0592 0.8083
S × C 10.0443 0.8337
Linamarin
Species 10.0001 > 0.9908
Concentration 112.5905 < 0.001
S × C 19.5071 0.0030
Arbutin
Species 10.3579 0.5520
Concentration 10.2353 0.6295
S × C 10.0609 0.8060
Chlorogenic acid
Species 15.8096 0.0177
Concentration 14.7875 0.0310
S × C 15.9378 0.0166
Salicin
Species 10.5547 0.4581
Concentration 12.7442 0.1007
S × C 16.0607 0.0155
a
Probability values less than 0.05 are in bold type.
FIG. 3. (Opposite) Effects of diets with added chemicals on subsequent feeding on control
diet. Time feeding in the 3-min second control is shown as a percent of the time feeding
on the initial test (first control) with plain diet. The vertical bars represent standard errors.
Concentrations are those experienced by the larvae on the test diets immediately prior to
feeding on the control diet.
560 BERNAYS ET AL.
TABLE 5. CORRELATION BETWEEN ESTIMATED AMOUNTS OF CHEMICAL INGESTED AND
TIME FEEDING IN CONTROL 2a
Correlation coefficient
Chemical H. virescens H. subflexa
NHT -0.552 0.175
Hordenine 0.156 0.191
Caffeine -0.449 0.01
Sinigrin 0.011 0.102
Linamarin -0.619 -0.08
Arbutin 0.056 0.133
Chlorogenic acid -0.352 -0.075
Salicin -0.524 -0.077
a
Bold indicates significant correlation.
in order to control for the many possible confounding variables. It is possible that
sensitivity levels to deterrents were, thus, less overall, as has been found for other
caterpillar species, especially specialists (e.g., Huang and Renwick, 1995). We
believe that if this were the case in the present study with species of Heliothis,
then rearing on plants would be expected to heighten the levels of deterrence in
H. subflexa more than in H. virescens, thus increasing the differences found.
Reduced feeding on deterrent diets was a consequence either of rejection
without any ingestion or of rejection following one or more shortened periods
of feeding. In either case, the insect might repeatedly encounter and reject its
food within the 3-min period.
We take rejection without ingestion to indicate that the deterrent compound
was detected by chemoreceptors on the mouthparts or antennae. Rejection of
sinigrin- and caffeine-containing diets nearly always occurred when an insect
(or either species) first encountered the food. On diets containing the other com-
pounds, only a proportion of the insects rejected on first encounter, with the num-
ber increasing with the concentration of deterrent. Overall, H. subflexa was more
likely to reject at first encounter than was H. virescens, with differences being
most marked for linamarin and salicin. NHT elicited relatively fewer rejection
responses even at concentrations that greatly reduce feeding. Thus, even where
NHT reduced feeding time to 10% of the control, half the individuals began
eating it immediately on first contact, suggesting that its sensory effects were
minimal.
Rejection following a period of feeding might result from the accumu-
lation of sensory information, since deterrent receptors sometimes adapt rela-
tively slowly (Schoonhoven et al., 1998). There could also be postingestive feed-
backs, which can occur very quickly. In Manduca sexta, for example, caterpillars
feed readily on food substrates containing NHT, and feeding stops as a result
TASTE SENSITIVITY OF INSECT HERBIVORES 561
of rapidly occurring postingestive toxicity (Glendinning and Slansky, 1994;
Glendinning, 1996). Such feedbacks are strongly suggested in those instances
where feeding on the second control is reduced relative to the first control (see
below).
On deterrent foods that were eaten to some extent, individual caterpillars
often fed briefly, then rested or walked away, returned, and had another feeding
bout. In H. subflexa, peak numbers of bouts tended to occur at intermediate levels
of feeding reduction, but in H. virescens, there was no clear pattern except that
when the feeding occurred for more than half of the test time, there was most
commonly only one bout. In both species the compounds that elicited the initial
rejection responses most were those in which we noted repeated bouts of feeding
most (Spearman s rank correlation, 0.9 and 0.85, P < 0.01), suggesting that the
two are related.
Insects on the second control were expected to feed just as well as they
did on the first control because the total possible period of feeding allowed was
not sufficient to cause satiety. In the case of H. subflexa, this was invariably
so. However, H. virescens ate less of the second control after ingesting five of
the eight compounds during the test. Its reduced behavioral sensitivity, allowing
ingestion of greater quantities of potential toxins, may therefore involve a cost.
However, we cannot evaluate whether the effects were due to postingestive tox-
icity, leading to a general disinclination to feed, or whether there may have been
a learned aversion to the diet after a noxious experience. Such learning has been
demonstrated in other insect herbivores (Bernays, 1995).
At first sight it may appear that the specialist is better able to avoid poten-
tially noxious plant secondary metabolites than the generalist, since it appears
that some ingested chemicals may have deleterious effects on the generalist.
Thus the specialist, with its greater sensitivity, is protected from deleterious
postingestive effects of plant secondary metabolites. However, the generalist,
having ingested the larger quantities of toxin, might be induced to produce high
levels of specific detoxification enzymes, enabling it to feed with impunity on
the same material at a later time. Such induction of enzymes can occur within
hours (Lindroth, 1991; Glendinning and Slansky, 1995; Snyder and Glendinning,
1996). Thus the reduced sensitivity has a twofold impact allowing feeding to
occur in the first instance, and then, when postingestive feedbacks allow only
limited intake, induction of detoxifying enzymes can increase the future toler-
ance to the toxin, such that feeding can then occur for longer.
We presume that there is likely to be a trade-off in these two strategies.
Deterrence is greater for the specialist and protects caterpillars from a multitude
of potentially poisonous plant chemicals, as well as contributing to the limited
host range. Even a short-lived toxicity could be risky, and this is likely to be
avoided by the sensitivity exhibited by H. subflexa. By contrast, the general-
ist can utilize a wide host range, enabling it to have an extensive geographic
562 BERNAYS ET AL.
range, multiple generations on different hosts, and at least some hosts when pre-
ferred plants are absent. Presumably the breadth of opportunity offsets the risks
of short-term deleterious effects of feeding on various secondary metabolites
before the induction of appropriate detoxification pathways.
Acknowledgments We thank Neil Vickers for providing some of the H. virescens and we
thank John Glendinning and Michael S. Singer for critically reading a draft of this manuscript.
REFERENCES
ASAOKA, K. 1994. Different spectrum in responses of deterrent receptor cells in Sawa-J, a strain of
the silkworm, Bombyx mori, with abnormal feeding habit. Zool. Sci. II (suppl): 102.
BERNAYS, E. A. 1995. Effects of experience in host-plant selection, pp. 47 64, in R. Cardé and W.
Bell (eds.). Chemical Ecology, Chapman and Hall, New York.
BERNAYS, E. A., and CHAPMAN, R. F. 1987. The evolution of deterrent responses in plant-feeding
insects, pp. 150 174, in R. F. Chapman, E. A. Bernays, and J. G. Stoffolano (eds.). Perspectives
in Chemoreception and Behavior. Springer-Verlag, New York.
BERNAYS, E. A., and CHAPMAN, R. F. 1994. Host Plant Selection by Phytophagous Insects. Chapman
and Hall, New York.
BERNAYS, E. A., and WEISS, E. A. 1996. Induced food preferences in caterpillars: The need to
identify mechanisms. Entomol. Exp. Appl. 78:1 8.
BLANEY, W. M., SIMMONDS, M. S. J., LEY, S. V., and KATZ, R. B. 1987. An electrophysiological
and behavioural study of insect antifeedant properties of natural and synthetic drimane-related
compounds. Physiol. Entomol. 12:281 291.
BLANEY, W. M., SIMMONDS, M. S. J., LEY, S. V., and JONES, P. S. 1988. Insect antifeedants: A
behavioural and electrophysiological investigation of natural and synthetically derived clero-
dane diterpenoids. Entomol. Exp. Appl. 46:267 274.
FANG, Q. Q., CHO, S., REGIER, J. C., MITTER, C., MATTHEWS, M., POOLE, R. W., FRIEDLANDER,
T. P. and ZHAO, S. W. 1997. A new nuclear gene for insect phylogenetics: Dopa decarboxy-
lase is informative of relationships within Heliothinae (Lepidoptera: Noctuidae). Syst. Biol.
46:269 283.
GLENDINNING, J. I. 1996. Is chemosensory input essential for the rapid rejection of toxic foods? J.
Exp. Biol. 199:1523 1534.
GLENDINNING, J. I., and SLANSKY, F. 1994. Interactions of allelochemicals with dietary constituents:
effects on deterrency. Physiol. Entomol. 19:173 186.
GLENDINNING, J. I., and SLANSKY, F. 1995. Consumption of a toxic food by caterpillars increases
with dietary exposure: Support for a role of induced detoxification enzymes. J. Comp. Physiol.
A 176:337 345.
GOULD, F., ANDERSON, A., REYNOLDS, A., BUMGARDNER, L., and MOAR, W. 1995. Selection and
genetic analysis of a Heliothis virescens (Lepidoptera: Noctuidae) strain with high levels of
resistance to Bacillus thuringiensis toxins. J. Econ. Entomol. 88:1545 1559.
GRAHAM, H. M., and ROBERTSON, O. T. 1970. Host plants of Heliothis virescens and H. zea
(Lepidoptera: Noctuidae) in the lower Rio Grande valley, Texas. Ann. Entomol. Soc. Am.
3:1261 1265.
HUANG, X. P., and RENWICK, J. A. A. 1995. Chemical and experiential basis for rejection of Tropae-
olum majus by Pieris rapae larvae. J. Chem. Ecol. 21:1601 1618.
JERMY, T. 1966. Feeding inhibitors and food preference in chewing phytophagous insects. Entomol.
Exp. Appl. 9:1 12.
LINDROTH, R. L. 1991. Differential toxicity of plant allelochemicals to insects: roles of enzymic
TASTE SENSITIVITY OF INSECT HERBIVORES 563
detoxication systems, pp. 1 34, in E. A. Bernays (ed.). Insect Plant Interactions, Vol. III. CRC
Press, Boca Raton, Florida.
MITCHELL, E. R., and HEATH, R. R. 1987. Heliothis subflexa (Gn.). (Lepidoptera: Noctuidae):
Demonstration of oviposition stimulant from groundcherry using novel bioassay. J. Chem. Ecol.
13:1849 1858.
MITTER, C., POOLE, P. W., and MATTHEWS, M. 1993. Biosystematics of the Heliothinae (Lepidoptera:
Noctuidae). Annu. Rev. Entomol. 38:207 225.
NEUNZIG, H. H. 1969. The biology of the tobacco budworm and the corn earworm in North Carolina.
Tech. Bull No. 196. North Carolina Agriculture Experiment Station, Raleigh.
NOLDUS, L. P. J. J. 1991. The observer: a software system for collection and analysis of observational
data. Behav. Res. Methods Instrum. Comput. 23:415 429.
SCHOONHOVEN, L. M., JERMY, T., and VAN LOON, J. J. A. 1998. Insect Plant Biology. Chapman and
Hall, London.
SHECK, A., and GOULD, F. 1996. The genetic basis of differences in growth and behavior of special-
ist and generalist herbivore species: Selection on hybrids of Heliothis virescens and Heliothis
subflexa. Evolution 50:831 841.
SNYDER, M. J., and GLENDINNING, J. I. 1996. Causal connection between detoxification enzyme
activity and consumption of a toxic plant compound. J. Comp. Physiol. A 179:255 261.
SOKAL, R. R., and ROHLF, F. J. 1995. Biometry, 3rd ed. W. H. Freeman, New York.
STADELBACHER, E. A. 1981. Role of early-season wild and naturalized host plants in the buildup
of the F1 generation of Heliothis zea and H. virescens in the delta of the Mississippi. Environ.
Entomol. 10:766 770.
USHER, B. F., BERNAYS, E. A., and BARBEHENN, R. V. 1988. Antifeedant tests with larvae of Pseu-
daletia unipuncta: Variability of behavioral responses. Entomol. Exp. Appl. 48:203 212.