for analyzing lobster agonistic encounters (®ghts). Ani-
mals that had been visually and physically isolated from
other lobsters since the 4th stage (when they begin their
benthic existence), were used in an attempt to eliminate
the in¯uence of experience on ®ghting behavior. Our
analyses demonstrated that lobster ®ghts include three
highly stereotypical components (described above, in the
Introduction): (1) display, during which animals stand
tall and prominently show their claws, which are their
major weapons; (2) limited aggression, in which the claws
are used to grasp and attempt to overturn an opponent;
and (3) high level aggression, in which animals grab
whatever they can of an opponent, and with short up-
ward tail ¯ips attempt to tear o the appendage. At
present we analyze ®ghts by making a videotape of the
entire ®ght (usually 30 min), and then scoring the times
animals approach and are within one body length of
each other (called an encounter). Usually we have ani-
mals ®ght three times: once to establish a hierarchy; a
second time 1 h later to con®rm that a hierarchy is es-
tablished; and a third time either after a variable time
period (to test the ``memory'' of the initial result), or
after some kind of pharmacological manipulation has
been performed. For each encounter we score: who ini-
tiates, who retreats, the duration, and the maximum
intensity on a 0±3 scale (0 one animal continually re-
treating, usually late in a ®ght; 1 display; 2 limited
aggression; 3 high level aggression). A statistical
analysis then identi®es the components changed by
repeated ®ghts, or by pharmacological intervention.
Using this method we demonstrated that the most
signi®cant variable changed during ®ghts in crustaceans
was the duration of individual encounters (Huber et al.
1997a). After a hierarchy was established the average
durations of encounters in lobsters were reduced from
about 30 s in ®rst ®ghts to about 5±10 s after a hier-
archy was established. Since our studies showed a close
linkage between duration and maximum intensity
(Huber and Kravitz 1995; Huber et al. 1997a), we
usually found a statistically signi®cant decrease in in-
tensity as well. We next asked what the consequences
Fig. 4 Schematic of the known pathways involved in activation and
inhibition of A1-5HT neurons. The right side of the ®gure shows
linkages between the A1 cell and the tonic (postural) muscle system. In
addition to the information already shown in Fig. 3, extensor
commands inhibit the ®ring of A1 cells (Ma et al. 1992) and a major
source of inhibitory input to these cells comes from an unidenti®ed
spontaneously active putative GABAergic neuron in the A3 ganglion
(Weiger and Ma 1993). On the left side of the ®gure, known
interactions with the phasic muscle system involved in escape and
®ghting behavior are shown. The LG and MG axons excite A1-5HT
neurons, while pre®ring the A1 cells reduces the magnitude of the
excitatory input from these sources (HoÈrner et al. 1997). The phasic
(fast) muscles activated by the lateral giant (LG) and medial giant
(MG) axons also show enhanced contractility when treated with 5HT.
The synaptic contacts between sensory receptors in the telson (tail) of
the cray®sh and the LG neuron are modulated by 5HT in opposite
directions in dominant and subordinate animals (Yeh et al. 1997). See
text for further details. This ®gure is slightly modi®ed from Fig. 10 of
HoÈrner et al. (1997)
229
would be of acute injections of 5HT on ®ghting be-
havior. Studies were carried out using both cray®sh and
lobster pairs, subordinate cray®sh receiving continuous
infusions of test substances, subordinate lobsters being
removed from the tank and injected with the test sub-
stances. The results with both lobsters and cray®sh
were qualitatively the same: 5HT infusion into subor-
dinate animals, after a variable but lengthy (ca. 45 min)
delay, increased the duration and the maximum inten-
sity reached during subsequent encounters. Subordinate
animals, who hardly ever initiate encounters, could be
seen to advance on the former dominants. The eect
was transient, and in all but a very few cases, was
completely reversible, with the former dominant rees-
tablishing its dominant position once more. Thus, 5HT
injection appeared to increase the willingness of ani-
mals that had just lost ®ghts to engage in combat
again.
In asking why it took so long for the eect to appear,
we considered two options: (1) 5HT, via surface recep-
tors, activated slowly-acting second messenger pathways
that altered the willingness to ®ght; and (2) 5HT was
taken back into serotonergic neurons (Livingstone et al.
1981; Huber et al. 1997b), thereby increasing the pool of
amine available for release, and the subsequent release
of ``extra'' 5HT contributed to the eect. We believe that
the latter is an important part of the explanation, be-
cause Prozac (¯uoxetine), as in vertebrates, blocks the
uptake of 5HT in lobsters (Huber et al. 1997b), and
when co-injected with 5HT, prevents the behavioral re-
versal. Injections of Prozac alone into subordinate ani-
mals have no eect on ®ghting behavior. It is interesting
that acute treatment with Prozac in patients also is not
eective in the treatment of depression (for review see
Stokes 1993). It takes several weeks for a fully thera-
peutic eect to be seen, suggesting that the ability of
Prozac to block 5HT uptake is only part of the expla-
nation for its eectiveness. We carried out chronic
treatments of cray®sh and lobsters with Prozac using
osmotic mini-pumps glued to the backs of animals that
continuously infused the drug into the cardiac sinus.
Prozac was injected over a 2-week period in this way and
then animals were paired against larger and smaller
opponents to search for drug eects. The eects seen
were not dramatic, but animals that had received
chronic Prozac treatment fought for longer periods of
time than saline-infused control groups (A. Delago,
unpublished observations).
We then lowered 5HT levels in animals by injections
of the neurotoxin 5,7-DHT. In general, this treatment
does not destroy serotonergic neurons in invertebrates.
Instead it seems mostly to deplete neurons of 5HT by
blocking uptake and synthesis and by enhancing release,
while allowing the uptake of 5,7 DHT into the neurons
(reviewed in Cook and Orchard 1993). Injections were
given over a period of several weeks (either six or eight
injections), and animals were then paired against larger
and smaller opponents. Once again three ®ghts were
carried out: a ®rst to establish a hierarchy, a second 1 h
later, and a third 1 day later. The goal was to test not
only whether 5HT-depleted animals would ®ght, but
also how eectively they would ®ght, and whether they
``remembered'' the outcome of an initial encounter.
After the ®nal ®ght, animals were sacri®ced to measure
their amine contents by high performance liquid chro-
matography or for immunocytochemistry to examine
the arbors of processes of the known 5HT neurons in
lobsters. In all cases the 5,7-DHT treatments signi®-
cantly lowered the amine content in animals. Animals
with reduced levels of 5HT still would ®ght, and still
could win ®ghts. In fact, the main change we observed in
preliminary results, was mainly in the duration of
encounters, which were signi®cantly increased.
While many more experiments of these types are re-
quired, the main conclusion that we can oer is that
5HT is not important in determining whether animals
will ®ght, or even if they will win ®ghts, but only for how
long they are willing to ®ght. Moreover, as in the de-
velopmental studies, where elevated or lowered levels of
5HT caused developmental abnormalities, too much or
too little 5HT both seem to cause an increased willing-
ness of animals to ®ght. It remains to be established
whether this translates as 5HT serving mainly a ``moti-
vational'' role in ®ghting behavior in lobsters.
A speculative synthesis: amine neurons
and ®ghting behavior
As in the past, when invertebrate models provided es-
sential information to explain fundamental properties
of neurons like how action potentials are generated (for
review see Hodgkin 1964), how synaptic processes like
inhibition work (Fatt and Katz 1953; Boistel and Fatt
1958; Kuer and Edwards 1958; Otsuka et al. 1966),
and how modulators function (cf. Adams and Levitan
1982; Glusman and Kravitz 1982; Dixon and Atwood
1989; Harris-Warrick and Marder 1991), many of these
same models now provide key insights into how com-
plex sensory systems function (for reviews see Hilde-
brand 1996; Ache et al. 1998; Passaglia et al. 1998) and
to how behaviors are constructed and organized by
nervous systems (for reviews see Altman and Kien
1987; Kravitz 1988; Bicker and Menzel 1989; Harris-
Warrick and Marder 1991; Morton and Chiel 1994;
Katz 1995). Invariably it is the ability to repeatedly ®nd
and identify single large neurons that is an important
part of why these systems have proven so valuable. At
the behavioral level, elegant studies are leading to im-
portant advances in our understanding of behavior.
The systems used for study range from relatively simple
ones, in which the potential exists for all the neurons
involved to be identi®ed and de®ned, like food pro-
cessing mediated by the stomatogastric systems of
decapod crustaceans (for review see Harris Warrick and
Marder 1991), through more complex models, in which
relatively smaller numbers of the neurons involved have
been identi®ed, like swimming (cf. Kristan and Weeks
230