1998). In Helisoma neurons, 5HT inhibits growth-cone
elongation in a speci®c subpopulation of cells in vitro,
and in¯uences the growth and morphological appear-
ance of these same neurons in vivo (Haydon et al. 1987;
Goldberg and Kater 1989). These and related ®ndings
have led several investigators to propose that 5HT serves
an organizational role in the nervous system (see Gold-
berg and Kater 1989; Lauder 1990), but precisely what
that role is has not been spelled out. The suggestion
oered here is that the role may be similar to that pro-
posed for steroids by Phoenix et al. (1959). Perhaps the
early arrival of amine neurons in relatively undieren-
tiated areas of the brain helps to carve out later amine
responsive territories. In fact, possibly the organizational
role of steroids only represents a speci®c case of more
general developmental roles served by modulators in
de®ning the areas of the brain they later will in¯uence in
behaviorally important ways.
The activational role of amines
This role, by contrast, deals with how amine neurons
function within already formed responsive territories.
The sequence of highly stereotypical motor acts that
make up ®ghting behavior have been described above.
How this context-dependent sequence is assembled in
the nervous system, and how to determine the parts of
the nervous system that are concerned with the behavior,
are issues of paramount importance. Many models have
been put forward to attempt to explain how complex
behaviors are assembled in 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 the models deal
with how sensory cues assemble the appropriate neuro-
nal circuits to respond to the cues, and with the roles
served by hormonal substances in consolidating those
choices. One particularly elegant illustration of the
complexity of the issues involved comes from studies
from the Cohen laboratory on voluntary and evoked
contractions of the gill mantle in the sea slug, Aplysia
(Wu et al. 1994). These authors used optical recording
methods in the abdominal ganglion to monitor the
population of neurons active during re¯ex withdrawal of
the gill, during respiratory pumping and during small
spontaneous gill contractions. They observed a very
great overlap of the populations of neurons involved in
each of these types of contractions. Even more sur-
prising, they found that over 20% of 900 neurons in the
ganglion were involved in this ``simple'' re¯ex.
The authors discuss two alternative models for how
the ``behavior'' is organized: one is a dedicated circuit
model in which sensory elements activate small sets of
neurons each subserving a fragment of motor behavior;
the other is a distributed organization in which sensory
elements arrange patterns of behavior de novo from
interneuronal pools of cells, depending on the context,
to perform various motor acts. As is usually the case
when two extremes are oered, probably both models
are correct in how behaviors are assembled in nervous
systems.
In ®ghting behavior, the survival of the animal is at
stake. Much of the nervous system probably is involved
in this important and essential behavior, including the
regions controlling the changing motor patterns we ob-
serve, but also including regions concerned with the
regulation of cardiac function (heart rate changes during
®ghts, HernaÂndez-FalcoÂn and Kravitz 1999), respira-
tion, excretion (see above, Breithaupt et al. 1999), and
possibly other physiological processes as well. In un-
published preliminary observations we have found that
losing animals not only refuse to engage other animals in
®ghts, but they also show dramatic reductions in their
feeding behavior. The challenge is to sort out the cir-
cuitries involved, and to ®gure out how and where
amines and amine neurons ®t into these circuits. The
suggestions are that amines (and almost surely other
modulatory substances) serve important roles: in assem-
bling the arrays of neurons that pattern the behavior; in
the smooth transitions between one behavioral pattern and
the next (e.g., limited aggression to high level aggres-
sion); and in the temporal domain concerned with how
long animals are willing to perform the sequences that
make up the behavior.
These suggestions are not original and derive heavily
from explorations on the roles served by amines and
other modulatory substances on the motor output of
systems of neurons like the stomatogastric ganglion of
crustaceans (for review see Harris-Warrick and Marder
1991). This approximately 30 neuron network governs
the processing of food on its way through the stomach,
and much detailed information is available on how it
works. Many modulators in¯uence the activity of the
two circuits that govern the behavior (Christie et al.
1997), but some of the best studied are the amines 5HT,
OCT, and dopamine (Flamm and Harris-Warrick
1986a, b). Each amine causes a unique change in the way
that a circuit functions. The changes are amine speci®c
and concentration dependent, i.e., the same amine can
cause dierent eects depending on its concentration
(Flamm and Harris-Warrick 1986a). Almost all neurons
in the circuit respond to all three amines, but through an
elegant cell by cell analysis, it was demonstrated that
each amine has selective actions on the conductance
properties of the cell under study (Flamm and Harris-
Warrick 1986b). The cellular studies allowed a simple
model to be made that explained the circuit eects of
applied amines. In our understanding of the circuitry
dealing with aggression, we have not yet done any cell-
by-cell analysis of the actions of 5HT on its target
neurons (although much work has been done on amine
actions on neuromuscular junctions ± Glusman and
Kravitz 1982; Dixon and Atwood 1989; Goy and Kra-
vitz 1989). We believe, however, that the same general
principals seen in the neurohormonal modulation of
behavioral output in the stomatogastric system, also will
apply in the more complex behavior as well. Thus,
232