In a variety of Diptera there are five motomeurones which innervate the wing depressor muscle (DLM) (Ikeda, Koenig & Tsuruhara, 1980; King & Wyman, 1980). During flight these motorneurones fire in a repeating cycle. Each motorneurone fires once per cycle. The firing times of the different motorneurones are separated so that no neurone fires immediately before or after any other (Wyman, 1965, 1969a, b). Wyman and co-workers have proposed that this spacing is due to mutual inhibition between the motorneurones (Wyman, 1969a, b;Harcombe & Wyman, 1977; Tanouye & Wyman, 1981). Koenig & Ikeda (1980) have now challenged this theory. They have presented data on the interspike intervals during flight that they believe do not fit the mutual inhibition hypothesis: they suggest that an ‘excitatory effect’ causes the spacing. The present communication argues that the two sets of data are consistent with each other and with the mutual inhibition theory.

In a variety of Diptera there are five motorneurones which innervate the wing depressor muscle (DLM) (Ikeda, Koenig & Tsuruhara, 1980; King & Wyman, 1980). During flight these motorneurones fire in a repeating cycle. Each motorneurone fires once per cycle. The firing times of the different motorneurones are separated so that no neurone fires immediately before or after any other (Wyman, 1965, 1969a, b). Wyman and co-workers have proposed that this spacing is due to mutual inhibition between the motorneurones (Wyman, 1969a, b;Harcombe & Wyman, 1977; Tanouye & Wyman, 1981). Koenig & Ikeda (1980) have now challenged this theory. They have presented data on the interspike intervals during flight that they believe do not fit the mutual inhibition hypothesis: they suggest that an ‘excitatory effect’ causes the spacing. The present communication argues that the two sets of data are consistent with each other and with the mutual inhibition theory.

The mutual inhibition theory (Harcombe & Wyman, 1977) says that in a normal interspike interval a motoneurone is inhibited once by each of the four other ipsilateral motoneurones. These four inhibitions, plus commonly received excitation, set the interspike interval. It is presumed that an inhibition has about the same duration and effect whenever it occurs in an interspike interval. Although this assumption is not explicitly stated in the 1977 paper, it was made clear in earlier papers when the summation of inhibition bands in multi-unit interactions was discussed (Wyman, 1969b, pp. 310–312; Wyman, 1973, pp. 296–7; Wyman, 1976, pp. 161–163) and a neuromime model incorporating this assumption can generate the proper pattern (Friesen & Wyman, 1980). For all intervals, the interspike interval length is set primarily by the number of inhibitions received, while the timing of the inhibitions is of secondary importance. Koenig and Ikeda argue against a different inhibition theory in which the number of inhibitions is not important, but the timing is. In their theory, inhibition has a significant effect only if it occurs soon before an expected firing of another unit. As Koenig and Ikeda convincingly demonstrate, this new form of the theory does not fit with the data.

Two motorneurones do not always have exactly the same intrinsic frequency, so it is common for one unit to fire slightly earlier (or slightly later) in phase on each successive cycle (‘phase progression’: Wyman, 1969a). In the course of this phase progression, a unit which has been firing just after a slower one, may, on the next cycle, fire synchronously with the slower unit, or even fire earlier than the slower unit.

Almost identical examples and graphic analyses are shown in figs. 8 and 11 of Wyman (1969a) and fig. 2 of Koenig & Ikeda (1980).

The main finding of Koenig and Ikeda is that when a unit progresses to a synchrony with (or an earlier firing than) a slower unit, it has one interspike interval which is shorter than average. At the same time, the slower unit has one longer than average interval. They believe that this finding contradicts the mutual inhibition hypothesis. However, their result is exactly what mutual inhibition predicts.

Consider a simple model of mutual inhibition (Fig. 1) in which two neurones receive a common input and so depolarize at the same rate (0·2 mV/ms) towards a firing threshold of − 50 mV. When a neurone fires, its membrane potential is repolarized to − 70 mV. The units inhibit each other, but with a slight assymmetry: unit 1 inhibits unit 2 with an i.p.s.p. of 10 mV, while unit 2 inhibits unit 1 with an i.p.s.p. of 8 mV. Most interspike intervals of unit 1 are 140 ms long while most interspike intervals of unit 2 are 150 ms long. Unit 1, being the faster unit, fires slightly earlier in phase on each successive cycle until it fires synchronously with unit 2. Note that in the interval just before the synchrony, unit 1 does not receive an inhibition from unit two, therefore its interval is shorter than usual. In the interval just after the synchrony, unit 2 receives two inhibitions from unit 1 (one from the synchronous spike of unit 1 and the other from the next spike of unit one). This interval of unit 2 is therefore longer than usual. Real examples of this pattern of short interval, synchrony, long interval may be seen in the third cycle in fig. 2 A of Koenig & Ikeda (1980) and in the lower trace of fig. 11 in Wyman (1969a). In the former record, unit 1 is the faster unit and progresses earlier in phase on each cycle (fig. 2C). In the third cycle (fig. 2A) the two units fire synchronously. The second interval of unit 1 does not include a firing of unit 2 ; not receiving that inhibition, the interval is a short one. The third interval of unit 2 is long because unit 1 inhibits unit 2 twice in that interval.

Fig. 1.

(A) Membrane potentials and spike times of two model neurones exchanging constant effect inhibition. (B) Phase of unit i spikes in intervals of unit 2 for the above units. Compare with Koenig & Ikeda (1980, fig. 2C).

Fig. 1.

(A) Membrane potentials and spike times of two model neurones exchanging constant effect inhibition. (B) Phase of unit i spikes in intervals of unit 2 for the above units. Compare with Koenig & Ikeda (1980, fig. 2C).

Exact synchrony is not required for the pattern to develop. Sometimes phase progression causes a unit, which had been firing just after a slower unit, to fire earlier than the slow unit. The slow unit will thus receive two inhibitions in one interval, and the interval will be prolonged. Examples may be seen in the top trace of fig. 11 of Wyman (1969a) and fig. 2(B) of Koenig & Ikeda (1980). In the latter trace the third interval of the faster unit 1 is short because it contains no inhibitions from unit 2. The third interval of the slower unit 2 is long because it contains two inhibitions from unit 1.

In all the above cases, when one unit fires twice in a single interval of another unit, the second firing occurs just before the other unit would have fired. This should be the major reason for the observation that when one unit fires late in another unit’s cycle, the inhibited unit’s next firing is frequently delayed and its interval prolonged (Wyman, 1969a, pp. 303–304; 1976, p. 161 ; Harcombe & Wyman, 1977, p. 1070; Friesen & Wyman, 1980, p. 42). The prolongation is a direct consequence of the constant effect model and does not imply (as Koenig and Ikeda do) that the inhibition is quantitatively different at different times in the cycle.

The phenomenon of shortened intervals of a fast unit when it skips ahead and prolonged intervals of a slow unit when it falls behind explains the remainder of the data collected by Koenig and Ikeda. Their figs. 7, 9 show the same type of phenomenon as fig. 2, discussed above. Figs. 4, 5 show that the ‘interacting intervals’ of the-faster unit 1 are short, while those for the slower unit 2 are long. Fig. 10 shows the same results for units 3 and 4. In each case the number of inhibitions recieved by a unit in an interval predicts whether the interval will be long, short or normal. In fact, all the data of the Koenig and Ikeda paper may be taken as a very nice confirmation that the predictions of the inhibition theory are borne out in detail.

Koenig and Ikeda ‘do not wish to propose a detailed hypothesis’ for how the flight pattern might be generated. However, they suggest that electrical coupling might account for some of their observations. Wyman (1973, p. 303) has argued that such coupling might be the mechanism for the sharing of excitatory input, but that it must be weak. Strong coupling would cause a large number of synchronous spikes ; in fact, only 1-5% of spikes actually occur in synchrony (Harcombe & Wyman, 1978, p. 274; Koenig & Ikeda, 1980, figs. 1, 2C). Koenig and Ikeda suggest that electrotonic transmission of after hyperpolarization could account for the lengthened intervals. It isn’t clear whether the after hyperpolarization following a single spike, after passing weak electrotonic coupling, could cause the 20-30 ms delays seen in these prolonged intervals. They also suggest that synchronies might be due to ephaptic effects on conduction velocity of concurrent spikes in anatomically closely apposed axons. However, the axonal conduction time is only a few tenths of a millisecond (Tayouye & Wyman, 1980, 1981); it is hard to see how changes in these short times can contribute significantly to sychronization. Synchronies are probably more easily explained as occurring when two units recover nearly simultaneously from the inhibition of the previously firing unit. If both units are then near threshold, an e.p.s.p., received by both units from the same source, could push both over threshold simultaneously.

Koenig and Ikeda dismiss the antidromic stimulation experiments of Harcombe & Wyman (1977) on the following basis. They correctly note that antidromic stimuli were not delivered with completely equal efficiency at all times after an orthodromic spike.* They claim that ‘the effective stimuli occurred primarily or totally when no other unit would have been firing.’ However, a glance at fig. 3B of Harcombe & Wyman (1977) shows that there are many effective stimuli at all times after the preceding spike. From Koenig and Ikeda’s fig. 1, it can be seen that other units fire at all phases from 5 % to 95 % of a given unit’s interspike interval. Thus in Harcombe and Wyman’s fig. 3 B, where the total N was 538 effective stimuli, about 480 effective stimuli were delivered when other units could have been firing. The 1977 paper reports many antidromic stimulation experiments, all with comparably large N’s. We can see no reason for dismissing this data.

In summary, mutual inhibition does explain and predict all of the flight data. Electrotonic coupling might also explain some of the flight data, but in the absence of any direct evidence, it can only be concluded that such coupling, if present, must be weak. Koenig and Ikeda do not explain the results of the antidromic stimulation experiments and give no valid reasons for ignoring these experiments. On the basis of the antidromic stimulation data and in view of its consistency with all extant flight data, the mutual inhibition theory remains the best explanation of flight pattern generation in Drosophila.

Friesen
,
W. O.
&
Wyman
,
R. J.
(
1980
).
Analysis of Drosophila motor neuron activity patterns with neural analogs
.
Biol. Cybernetics
38
,
41
50
.
Harcombe
,
E. S.
&
Wyman
,
R. J.
(
1977
).
Output pattern generation by Drosophila flight motoneurons
.
J. Neurophysiol
.
40
,
1066
1077
.
Harcombe
,
E. S.
&
Wyman
,
R. J.
(
1978
).
The cyclically repetitive firing sequence of Drosophila flight motoneurons
.
J. comp. Physiol. A
133
,
271
279
.
Ikeda
,
K.
,
Koenig
,
J. H.
&
Tsuruhara
,
T.
(
1980
).
Organization of identified axons innervating the dorsal longitudinal flight muscle of Drosophila melanogaster
.
J. Neurocytol
.
9
,
799
823
.
King
,
D. G.
&
Wyman
,
R. J.
(
1980
).
Anatomy of the giant fiber pathway in Drosophila I. Three thoracic components of the pathway
.
J. Neurocytol
.
9
,
753
770
.
Koenig
,
J. H.
&
Ikeda
,
K.
(
1980
).
Neural interactions controlling timing of flight muscle activity in Drosophila
.
J. exp. Biol
.
87
,
121
136
.
Tanouye
,
M. A.
&
Wyman
,
R. J.
(
1980
).
Motor outputs of-giant nerve fiber in Drosophila
.
J. Neurophysiol
.
44
,
405
421
.
Tanouye
,
M. A.
&
Wyman
,
R. J.
(
1981
).
Inhibition between flight motoneurons in Drosophila
.
J. comp. Physiol
. (In the Press.)
Wyman
,
R. J.
(
1965
).
Probabilistic characterization of simultaneous nerve impulse sequences controlling dipteran flight
.
Biophys. J
.
5
,
447
471
.
Wyman
,
R. J.
(
1969a
).
Lateral inhibition in a motor output system. I. Reciprocal inhibition in the dipteran flight motor system
.
J. Neurophysiol
.
33
,
297
306
.
Wyman
,
R. J.
(
1969b
).
Lateral inhibition in a motor output system. II. Diverse forms of patterning
.
J. Neurophysiol
.
33
,
307
314
.
Wyman
,
P.. J.
(
1973
).
Neural circuits patterning dipteran flight motoneurone output
.
In Neurobiology of Invertebrates
(ed.
J.
Salanki
), pp.
289
309
.
Budapest
:
Hungarian Academy of Sciences
.
Wyman
,
R. J.
(
1976
).
A simple network for the study of neurogenetics
.
In Simpler Networks and Behavior
(ed.
J. C.
Fentress
), pp.
153
166
.
Sunderland, Mass
:
Sinauer Associates
.
*

It should be noted that while randomly placed stimuli are expected to fall equally at all phases of a cycle, they are not expected to fall equally at all latencies. This is because a short latency stimulus can occur in an interspike interval of any length, but a long latency stimulus can occur only in a long interspike interval. Thus, there should be fewer stimuli at long latencies. See fig. 5 in Wyman (1965) for a complete explanation.