Pith. sign in

REVIEW 2 major objections 4 minor 22 references

The paper claims that FC2, the fly's goal-holding neurons in the fan-shaped body, normalize an externally set goal rather than selecting among competing goals, and that connectome wiring rules out a within-FC2 winner-take-all.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 13:47 UTC pith:XFRMS7KW

load-bearing objection Solid connectome work with an over-strong title: the structural decomposition is a real contribution, but the normalization-not-selection claim is bounded, not proven. the 2 major comments →

arxiv 2607.18969 v1 pith:XFRMS7KW submitted 2026-07-21 q-bio.NC cs.NE

How the fly holds a single goal: normalization, not selection, in Drosophila FC2

Gioele Nanni , Christopher Lee This is my paper
classification q-bio.NC cs.NE
keywords Drosophilafan-shaped bodyFC2goal representationwinner-take-alldivisive normalizationconnectomeFB5A
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

A walking fly holds a goal direction as a bump of activity across FC2 neurons in the fan-shaped body. Earlier work found that activating one FC2 column suppresses others over distance, which read like a mechanism for choosing a single goal. This paper traces the wiring and argues the opposite: the inhibition is almost entirely a uniform global signal from four FB5A cells, the circuit lacks the local recurrent excitation a winner-take-all needs, and no dynamical model of the circuit latches onto a winner at reference gains. FC2 therefore normalizes and sharpens a goal set elsewhere, with the actual choice happening upstream in a recurrent h-delta network. If right, this moves the search for the fly's decision from FC2 to its inputs, and it turns a proposed silencing experiment into a decisive test.

Core claim

The paper's central claim is that the fly's FC2 goal neurons do not select among competing goals; they normalize and sharpen a goal set elsewhere. Tracing the connectome, the authors decompose FC2's feedback inhibition into a uniform FB5A component, an h-delta component that couples opposite bearings (anti-local), and a negligible direct component; the wiring lacks the local recurrent excitation a winner-take-all needs. No connectome-parameterized model latches at reference gains, across five dynamical families and a spiking network; h-delta selection is bounded, not excluded. FB5A is read as a global normalizer and the upstream h-delta network as the goal's substrate.

What carries the argument

Key machinery: a seeded-basin bistability test run on a decomposition of FC2's feedback inhibition. The inhibition splits into a uniform FB5A term (four cells, flat in bearing) and an anti-local h-delta term (coupling opposite bearings); direct FC2-to-FC2 connections are negligible. The test drives a model with two competing goals, seeds the dynamics left or right, and measures the basin gap between settled states: a large gap indicates history-dependent latching (a winner-take-all), a near-zero gap a single input-determined fixed point. Because the FC2 ring lacks local recurrent excitation, no connectome-parameterized family — subtractive, divisive, spiking, h-delta, or the local route run

Load-bearing premise

The conclusion that FC2 does not select rests on the assumption that the biological gain of h-delta-mediated mutual inhibition is at or near the connectome-scaled reference value; if that gain is several-fold higher, a within-FC2 two-goal selector is not excluded.

What would settle it

Silence or scale h-delta interneurons while presenting two competing goals: if the settled FC2 bump depends on the initial state (history-dependent latching) once h-delta gain rises past a modest threshold, the normalization claim is falsified and a within-FC2 two-goal winner-take-all exists.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • FC2 does not host a within-FC2 ring-attractor winner-take-all: no connectome-parameterized model family latches at reference gains, and the wiring replicates in a second connectome.
  • The reported distance-dependent inhibition decomposes into a uniform FB5A floor plus a modest h-delta anti-local rise, so the single-bump behavior is normalization, not spatial veto.
  • The goal itself is set upstream: the connectome nominates an h-deltaC-led recurrent network, valence-gated through a mushroom-body output pathway, and excludes a proposed persistent-goal attractor that supplies under 0.2 percent of FC2's input.
  • Silencing FB5A during two-cue imaging should preserve the relative activation of competing columns while raising overall activity; a local selector would instead change which column dominates.
  • The h-delta mutual-inhibition route is the bounded open alternative: at anatomically unconstrained high gain it crosses the no-latch bound, so a within-FC2 two-goal selector is not excluded.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If correct, the result separates goal choice from goal maintenance anatomically: a fly with intact upstream h-delta but silenced FB5A should still choose the right goal while showing a broader or noisier goal bump.
  • The bounded h-delta alternative suggests a tunable continuum: the same circuit could normalize at low h-delta gain and select at high gain, so measuring or manipulating that gain would place FC2 on that continuum.
  • The uniform-inhibition motif may generalize: any population that must hold a single sharp bump without choosing a winner — including other fan-shaped-body columns — could use an FB5A-like normalizer, making this a family-wide pattern rather than a mushroom-body special case.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. This paper uses two Drosophila connectomes (FlyWire and hemibrain) to decompose the distance-dependent inhibition among FC2 fan-shaped-body neurons into a uniform FB5A component, an anti-local hDelta component, and negligible direct FC2–FC2 connectivity. It argues that the absence of local recurrent excitation rules out a within-FC2 ring-attractor winner-take-all, and supports this with bistability tests across model families, a spiking LIF network, and feedforward sharpening sweeps. The paper then nominates the hDeltaC-led recurrent network as the upstream goal-holding substrate, excludes the hDeltaK–PFG attractor as FC2's source, and proposes a two-axis FB5A-silencing experiment. The authors explicitly bound rather than exclude an hDelta-mediated two-goal selector and flag the transmitter uncertainty.

Significance. The paper's strengths are notable: the structural decomposition is replicated in an independent connectome (far/near ratios 2.504 vs 2.429; modulation 5% vs 1.2%), the dynamical no-latch result is checked with positive controls on both idealized and real geometry, and the manuscript is unusually transparent about its assumptions and open alternatives, with machine-checked code and a claims ledger. The hDeltaK–PFG exclusion, if completed, would be a useful separation of parallel goal modules. The central conceptual claim—if upheld—reframes FC2 as an APL-like normalizer rather than a ring-attractor selector, which is of interest to the central-complex and navigation communities.

major comments (2)
  1. [Abstract; §4.2; §4.6; §6] The headline 'normalization, not selection' overstates the evidence. In §4.2 the hDelta route family has a basin gap of 24° at the reference gain, but when the anatomically unconstrained scalar gain is swept above ≈6 it crosses the paper's own 30° no-latch bound and peaks at 38.4° (gain 8). Because 30° is a chosen threshold, not a measured one, and because §4.6 shows the fly commits in the antipodal two-goal configuration where hDelta coupling peaks, a within-FC2 hDelta-mediated selector is bounded but not excluded in the regime that matters for the central claim. The authors flag this honestly in §4.2 and §6, but the abstract and title present normalization as the account. Please temper the abstract, title, and conclusion to state that the connectome excludes a ring-attractor WTA and places an upper bound on an hDelta-mediated selector, with normalization as the favored account continge
  2. [§4.4, Fig. 8] The 'decisive exclusion' of the hDeltaK–PFG attractor as FC2's source is based on direct synapse counts: hDeltaK and PFGs supply <0.2% of FC2's input. However, the text reports that hDeltaK projects to FB6A and ExR3 (besides 41% to PFGs). If FB6A or ExR3 provides input to FC2—FB-tangential cells constitute 55% of FC2 input—the attractor could reach FC2 indirectly, and the double-dissociation prediction in §4.4 would not follow. The Methods state that two-hop routes were traced only for MBON→FB5AB→{hDeltaC, FC2}; for hDeltaK/PFG the text says 'in/out of FC2,' which appears to be direct only. Please report whether hDeltaK→FB6A→FC2 or hDeltaK→ExR3→FC2 two-hop paths exist (with synapse counts), or explicitly limit the exclusion to direct input. As written, the 'decisive exclusion' is incomplete.
minor comments (4)
  1. [§3.3, Eq. (1)] Clarify that `pool` is a vector of FC2→FB5A synaptic weights and `w_i` the FB5A→FC2 weight for cell i; the current notation `pool·x` is easy to misread as a scalar pooling operation.
  2. [Figure 4] The LIF network result (basin gap 0°) is discussed in the text but not shown as a bar. Either add it or add a caption note explaining why it is omitted.
  3. [§4.4] When reporting 'hDeltaJ the largest single hDelta input to FC2 (6.1%),' specify the denominator (FC2 total input synapses) to avoid confusion with Table 1.
  4. [§1 vs §4.3] FB5AA/FB5AB are first mentioned in §4.3; define the sibling nomenclature at first use to avoid confusion with FB5A.

Circularity Check

0 steps flagged

No significant circularity: the no-WTA and decomposition results are connectome-derived and self-contained, with the hDelta and transmitter caveats openly bounded rather than hidden.

full rationale

I walked the derivation chain. The central structural claim — no within-FC2 ring-attractor WTA — rests on measured connectome facts (FB5A uniform over bearing, hDelta anti-local, vDelta weak, direct FC2-FC2 negligible) that are replicated in the hemibrain and do not depend on fitted parameters. The bearing axis is explicitly derived from the FC2/PFL fingerprint, not from FB5A or hDelta (Sec. 3.1), so the bearing-based uniformity/anti-locality tests are not circular; the hemibrain provides an independent anatomical index. The bistability/no-latch result is tested across five model families plus a committed spiking LIF run, and the positive controls (local excitation latches at 78–114°) show the test detects a WTA when present. The divisive-normalization equation is an assumed model family, not a derived consequence, and the paper explicitly disclaims the one result that would be self-definitional: Sec. 4.5 states that the single-goal bump shape 'is how the operating point was chosen (Sec. 3.3), so it is not itself a prediction.' The silencing prediction is swept over a 3×3×3 gain grid, and the amplitude rise is labeled definitional while the robust claim is the preserved tuning. The hDelta route is the one gain-contingent exception; rather than hiding it, the paper bounds it (24° at reference gain, crossing the chosen 30° bound above gain≈6, peaking 38.4°) and explicitly names it the primary open alternative and 'the single measurement that would decide' (Sec. 4.2, Sec. 6). The Limitations section similarly flags FB5A's transmitter as a low-confidence prediction and states the hDelta selector is 'not excluded, only bounded.' No load-bearing step reduces to a self-citation; the normalization/APL citations are external and the model form is a declared assumption. Overall, the paper's structural conclusions are self-contained, and the interpretive 'normalization' label is presented with explicit contingency rather than as a forced derivation.

Axiom & Free-Parameter Ledger

7 free parameters · 8 axioms · 0 invented entities

Central claims rest on connectome-derived connectivity (one FlyWire brain, replicated structurally in hemibrain), with synapse counts used as functional weights throughout. The model has several free gains set by a task-agnostic single-bump objective, one unmeasured hDelta gain that can create partial latching at high values, and a transmitter prediction for FB5A that the paper itself rates low-confidence. No new physical entities are introduced; the 'normalizer' role is an interpretation assigned to existing cells.

free parameters (7)
  • divisive sharpening exponent p = p = 2 at reference operating point; swept 0.7–1.4×
    Gain parameter in Eq. (1); chosen from task-agnostic objective 'single clean bump for single committed goal', not from two-cue competition behavior.
  • global inhibition gain g_inh = reference normalized value; swept ×0.5, ×1, ×2
    FB5A feedback gain in divisive model; set with p and σ by the single-bump objective.
  • semi-saturation constant σ = reference normalized value; swept ×0.5, ×1, ×2
    Denominator constant in Eq. (1); chosen with p and g_inh.
  • hDelta scalar gain in bistability sweep = reference = 1 (connectome-scaled); swept up to ≥10
    Free gain multiplying the FC2→hDelta→FC2 route; latching appears above ≈6 and peaks at 38.4°, so the no-WTA conclusion is gain-contingent for this route.
  • basin-gap decision bounds = no-latch ≤30°, WTA ≥60°
    Chosen thresholds, not measured; §4.2 admits the 30° bound is a chosen threshold.
  • von Mises drive concentration κ = not reported
    Width of feedforward goal bump in Eq. (1); affects two-cue coexistence results, value not stated.
  • two-cue competition parameters = 120° separation, 1:0.7 amplitude ratio
    Task parameters for the silencing prediction; not fit to data but set for the simulation.
axioms (8)
  • domain assumption FlyWire single-brain connectome provides accurate reconstructed synaptic edges
    All wiring sourced from one FlyWire brain; hemibrain replication partially mitigates but transmitter and weights rely on this.
  • domain assumption Synapse counts are a valid proxy for functional connection strength
    Used for all weights and 'connectome-scaled reference coupling'; paper itself flags 'synapse count is a structural proxy for functional inhibition' (§4.1).
  • domain assumption The inferred preferred bearing ψ_i from FC2/PFL spectral fingerprint reflects true allocentric bearing
    All bearing-distance claims depend on this; authors note bearing tests on FB5A/hDelta are not circular, and hemibrain uses anatomical column index as independent check.
  • ad hoc to paper Divisive normalization (Eq. 1) is an appropriate model of FC2 dynamics
    The normalizer role is instantiated by choosing this model form and then demonstrated; paper calls divisive form 'a further declared modelling assumption' (§6).
  • ad hoc to paper A single clean bump for one committed goal is the correct task-agnostic objective for setting gains
    Used to pick operating point; Result 1 is therefore not a prediction, by the paper's own admission.
  • domain assumption FB5A (or the uniform-input substrate) is functionally inhibitory
    Needed for the normalizer interpretation; the paper treats this as a low-confidence prediction of the transmitter classifier, with no experimental verification (§6).
  • standard math A ring-attractor WTA requires local recurrent excitation
    Used to argue absence of local excitation rules out WTA; supported by prior ring-attractor theory (Kim et al. 2017) and positive controls.
  • domain assumption The hDelta and vDelta recurrence signs are treated as connectivity geometry, not transmitter
    The anti-local hDelta coupling is modeled as mutual inhibition for the selector test; sign is undetermined, so the hDelta alternative remains open.

pith-pipeline@v1.3.0-alltime-deepseek · 14449 in / 16356 out tokens · 156699 ms · 2026-08-01T13:47:49.166049+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of How the fly holds a single goal: normalization, not selection, in Drosophila FC2." pith.science (2026). https://pith.science/paper/XFRMS7KW

@misc{pith2026260718969,
  author       = {Pith},
  title        = {Pith review of: How the fly holds a single goal: normalization, not selection, in Drosophila FC2},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XFRMS7KW}},
  note         = {Machine review of arXiv:2607.18969}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

A walking fly steers toward a goal direction, held as a bump of activity across the FC2 neurons of the fan-shaped body. These neurons also inhibit one another over distance, more strongly the farther apart they are, a feedback proposed to keep the fly on a single goal. We asked, from the connectome, what circuit produces this inhibition, and whether it lets FC2 actively choose one goal among competitors (a winner-take-all) or simply keeps a goal set elsewhere as one clean bump. Tracing the wiring in a single FlyWire brain, we find the inhibition is almost entirely global: four FB5A cells inhibit every FC2 neuron roughly equally, with a smaller, distance-dependent contribution from hDelta interneurons and a negligible direct component. A ring-attractor winner-take-all (the kind the compass uses) requires local recurrent excitation that the FC2 wiring lacks, so this geometry cannot build one; and across a range of dynamical models, including a spiking network, no version of the circuit locks onto a winner at the connectome-scaled reference coupling. FC2 therefore normalizes an externally set goal rather than selecting it, with FB5A likely acting as the global normalizer, much as the APL neuron does in the mushroom body. We are explicit about two open points: a different mechanism, mutual inhibition between two competing goals (which hDelta supplies), could in principle select at very strong coupling, and we bound rather than exclude it; and FB5A's inhibitory identity is a low-confidence prediction of the connectome's transmitter classifier, not yet measured, and likely not GABAergic. We then ask where the goal is actually set: the connectome nominates an upstream hDelta network and rules out the leading proposed alternative, whose neurons supply under 0.2% of FC2's input. Finally, we propose a direct experiment, silencing FB5A while imaging FC2, that would test the account.

Figures

Figures reproduced from arXiv: 2607.18969 by Christopher Lee, Gioele Nanni.

Figure 1
Figure 1. Figure 1: C2, FB5A inhibition is uniform across bearing. The disynaptic FC2 → FB5A → FC2 loop weight is flat as a function of the bearing distance between FC2 cells (modulation 5%): FB5A is a global scaler, not a spatial selector. Measured from the FlyWire connectome (analyze fc2 scoping.py). 4 Results 4.1 The feedback inhibition decomposes into a uniform, an anti-local, and a negligible term The reported distant in… view at source ↗
Figure 2
Figure 2. Figure 2: C3, no local recurrent excitation. Direct FC2→FC2 is negligible; the FC2→h∆→ FC2 disynaptic weight peaks at antipodal bearings (anti-local, near 37.8 → far 94.7, far/near 2.5, dipping mid-range; a vector-summation motif, not neighbour coupling); v∆ is local but weak. The local positive feedback a WTA needs is absent. What the three routes sum to. On the common disynaptic scale, the total FC2→FC2 in￾hibitor… view at source ↗
Figure 3
Figure 3. Figure 3: Two brains agree. The FB5A-uniformity (left) and the h∆ anti-local recurrence (right) reproduce in both the FlyWire and hemibrain connectomes (analyze fc2 scoping hemibrain.py). The dotted line on the right marks the anti-local threshold (far/near ratio 1.5). The hemibrain’s 105◦ bearing bin is empty (partial-volume dataset), hence the gap in that curve. bistability in any connectome-parameterized family (… view at source ↗
Figure 4
Figure 4. Figure 4: C4/C5, no bistability, hence no WTA. Seeded-basin test: the pure global-inhibition families (subtractive, divisive, divisive+noise), the one local route v∆ run as local excitation (2◦ ), and a global-only control ring have |basin gap| ≈ 0 ◦ (one input-determined fixed point). The h∆ anti-local mutual-inhibition family sits at 24◦ at the reference gain, below the 30◦ no-latch bound but the one gain-continge… view at source ↗
Figure 5
Figure 5. Figure 5: No winner-take-all via the feedforward sharpening term. Left: with two equal competing goals, the loser bump keeps ∼half the mass at the operating point (p=2) and collapses only supra-physiologically (p ≳ 12); that collapse is init-independent (a fixed bias, not a bistable choice). Right: the sharpening picks the stronger goal at p0 but at high p the fixed anatomi￾cal bias wins instead (stronger input drop… view at source ↗
Figure 6
Figure 6. Figure 6: FC2’s inputs: composition and per-cell geometry. Left: input composition (FB￾tangential dominant; MBON ≈0 direct). Right: per-cell targeting, PFN cells are narrow (few FC2 targets), FB5A/FB-tangential cells are broad (bearing-uniform), the geometry a normalizer would use (trace fc2 inputs.py). The direct PFN→FC2 edge is minor; the canonical PFN route into FC2 is via h∆ [PITH_FULL_IMAGE:figures/full_fig_p0… view at source ↗
Figure 7
Figure 7. Figure 7: Who supplies FC2’s uniform input. Left: FC2’s uniform-input synapse mass by predicted transmitter, ∼51% inhibitory-capable (GABA + glutamate/GluClα), ∼49% choliner￾gic/other (excitatory). Right: the top uniform inhibitory-capable inputs by synapse count, FB5A (blue) is the largest and the only GABA-predicted one, atop a broad glutamatergic FB-tangential family (green) (analyze fc2 uniform inhibitors.py). r… view at source ↗
Figure 8
Figure 8. Figure 8: Where FC2’s goal is set. Left: FC2’s input is dominated by a broad, non￾directional FB-tangential gate (∼55%); the directional carriers are the recurrent h∆ network and the narrow PFN stream. Right: the Lanz h∆K–PFG persistent-goal attractor is excluded from FC2 (both <0.2%), while learned valence enters via MBON→FB5AB→ {h∆C, FC2} (analyze fc2 selector.py). 4.5 In-silico demonstration Result 1, the single-… view at source ↗
Figure 9
Figure 9. Figure 9: Which FB5A-silencing effect is robust. Left: across a 3 × 3 × 3 gain sweep, the shape/“mush” effect holds in only 30% of combinations (fragile), while the disinhibition (amplitude-rise) effect holds in 100%. Right: the direction of the amplitude effect is always the same (activity rises); its magnitude is a model artifact. The paper’s prediction is the robust one (analyze fc2 prediction.py). • Two goal bum… view at source ↗
Figure 10
Figure 10. Figure 10: The two-cue test distinguishes a uniform FB5A from a local selector. Left: under FB5A silencing, the competing-column ratio is preserved for every uniform functional form (divisive/subtractive/excitatory, | log-dev|<0.01) but distorted by a local-excitation selector control (≈21). Right: the ratio survives inhibition non-uniformity up to α≈0.3; the measured FB5A non￾uniformity (0.04) sits well inside the … view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

22 extracted references · 15 canonical work pages

  1. [4]

    Lina Enell, Yasutaka Hamasaka, Agata Kolodziejczyk, and Dick R

    doi: 10.1016/j.cell.2024.03.016. Lina Enell, Yasutaka Hamasaka, Agata Kolodziejczyk, and Dick R. N¨ assel. GABA signaling com- ponents in Drosophila: Immunocytochemical localization of GABA-B receptors in relation to the GABA-A receptor subunit RDL and a vesicular GABA transporter.Journal of Comparative Neurology, 505(1):18–31,

    work page doi:10.1016/j.cell.2024.03.016 2024

  2. [11]

    doi: 10.1126/science.aal4835. Aaron J. Lanz, Nicholas D. Kathman, Emily Hao, Bard Ermentrout, and Katherine I. Nagel. Disinhibition of a recurrent attractor gates a persistent goal signal for navigation.bioRxiv,

    doi:10.1126/science.aal4835

  3. [12]

    Andrew C

    doi: 10.1101/2025.10.07.681003. Andrew C. Lin, Alexei M. Bygrave, Alix de Calignon, Tzumin Lee, and Gero Miesenb¨ ock. Sparse, decorrelated odor coding in the mushroom body enhances learned odor discrimination.Nature Neuroscience, 17(4):559–568,

    work page doi:10.1101/2025.10.07.681003 2025

  4. [16]

    Andrew M

    doi: 10.1038/s41467-022-32247-7. Andrew M. M. Matheson, Aaron J. Lanz, Ashley M. Medina, Al M. Licata, Timothy A. Currier, Mubarak H. Syed, and Katherine I. Nagel. Addendum: A neural circuit for wind-guided olfactory navigation.Nature Communications, 15:1903,

    work page doi:10.1038/s41467-022-32247-7 1903

  5. [17]

    Peter Mussells Pires, Lingwei Zhang, Victoria Parache, L

    doi: 10.1038/s41467-024-46225-8. Peter Mussells Pires, Lingwei Zhang, Victoria Parache, L. F. Abbott, and Gaby Maimon. Convert- ing an allocentric goal into an egocentric steering signal.Nature, 626(8000):808–818,

    work page doi:10.1038/s41467-024-46225-8

  6. [18]

    doi: 10.1038/s41586-023-07006-3. Louis K. Scheffer, C. Shan Xu, Michal Januszewski, Zhiyuan Lu, Shin-ya Takemura, Kenneth J. Hayworth, Gary B. Huang, Kazunori Shinomiya, Jeremy Maitin-Shepard, Stuart Berg, et al. A connectome and analysis of the adult Drosophila central brain.eLife, 9:e57443,

    work page doi:10.1038/s41586-023-07006-3

  7. [20]

    Johannes D

    doi: 10.1038/s41586-024-07686-5. Johannes D. Seelig and Vivek Jayaraman. Neural dynamics for landmark orientation and angular path integration.Nature, 521(7551):186–191,

    doi:10.1038/s41586-024-07686-5

  8. [22]

    doi: 10.1038/s41586-024-07039-2. 18

    work page doi:10.1038/s41586-024-07039-2

  9. [1992]

    doi: 10.1017/S0952523800009640. Brad K. Hulse, Hannah Haberkern, Romain Franconville, Daniel B. Turner-Evans, Shin-ya Take- mura, Tanya Wolff, Marcella Noorman, Marisa Dreher, Chuntao Dan, Ruchi Parekh, Ann M. Hermundstad, Gerald M. Rubin, and Vivek Jayaraman. A connectome of the Drosophila cen- tral complex reveals network motifs suitable for flexible na...

    doi:10.1017/s0952523800009640

  10. [2007]

    Ysabel Milton Giraldo, Katherine J

    doi: 10.1002/cne.21472. Ysabel Milton Giraldo, Katherine J. Leitch, Ivo G. Ros, Timothy L. Warren, Peter T. Weir, and Michael H. Dickinson. Sun navigation requires compass neurons in Drosophila.Current Biology, 28(17):2845–2852.e4,

    work page doi:10.1002/cne.21472

  11. [2012]

    Sanjoy Dasgupta, Charles F

    doi: 10.1038/nrn3136. Sanjoy Dasgupta, Charles F. Stevens, and Saket Navlakha. A neural algorithm for a fundamental computing problem.Science, 358(6364):793–796,

    doi:10.1038/nrn3136

  12. [2013]

    Cheng Lyu, L

    doi: 10.1073/pnas.1220560110. Cheng Lyu, L. F. Abbott, and Gaby Maimon. Building an allocentric travelling-direction signal via vector computation.Nature, 601(7891):92–97,

    work page doi:10.1073/pnas.1220560110

  13. [2014]

    doi: 10.1038/nn.3660. Wendy W. Liu and Rachel I. Wilson. Glutamate is an inhibitory neurotransmitter in the Drosophila olfactory system.Proceedings of the National Academy of Sciences, 110(25):10294–10299,

    work page doi:10.1038/nn.3660

  14. [2015]

    Franziska Toepfer, Reinhard Wolf, and Martin Heisenberg

    doi: 10.1038/nature14446. Franziska Toepfer, Reinhard Wolf, and Martin Heisenberg. Multi-stability with ambiguous visual stimuli in Drosophila orientation behavior.PLOS Biology, 16(2):e2003113,

    work page doi:10.1038/nature14446

  15. [2017]

    Sven Dorkenwald, Arie Matsliah, Amy R

    doi: 10.1126/science.aam9868. Sven Dorkenwald, Arie Matsliah, Amy R. Sterling, Philipp Schlegel, Szi-chieh Yu, Claire E. McKel- lar, et al. Neuronal wiring diagram of an adult brain.Nature, 634(8032):124–138,

    work page doi:10.1126/science.aam9868

  16. [2018]

    Jonathan Green, Vikram Vijayan, Peter Mussells Pires, Atsuko Adachi, and Gaby Maimon

    doi: 10.1016/j.cub.2018.07.002. Jonathan Green, Vikram Vijayan, Peter Mussells Pires, Atsuko Adachi, and Gaby Maimon. A neural heading estimate is compared with an internal goal to guide oriented navigation.Nature Neuroscience, 22(9):1460–1468,

    work page doi:10.1016/j.cub.2018.07.002 2018

  17. [2019]

    doi: 10.1038/s41593-019-0444-x. David J. Heeger. Normalization of cell responses in cat striate cortex.Visual Neuroscience, 9(2): 181–197,

    work page doi:10.1038/s41593-019-0444-x

  18. [2020]

    Philipp Schlegel, Yijie Yin, Alexander S

    doi: 10.7554/eLife.57443. Philipp Schlegel, Yijie Yin, Alexander S. Bates, Sven Dorkenwald, Katharina Eichler, Paul Brooks, et al. Whole-brain annotation and multi-connectome cell typing of Drosophila.Nature, 634 (8032):139–152,

    doi:10.7554/elife.57443

  19. [2021]

    Joseph D

    doi: 10.7554/eLife.66039. Joseph D. Jones, Brandon L. Holder, Andrew C. Montgomery, Chloe V. McAdams, Emily He, Anna E. Burns, Kiran R. Eiken, Alex Vogt, Adriana I. Velarde, Alexandra J. Elder, Jennifer A. McEllin, and Stephane Dissel. The dorsal fan-shaped body is a neurochemically heterogeneous sleep-regulating center in Drosophila.PLOS Biology, 23(3):e3003014,

    work page doi:10.7554/elife.66039

  20. [2022]

    Andrew M

    doi: 10.1038/s41586-021-04067-0. Andrew M. M. Matheson, Aaron J. Lanz, Ashley M. Medina, Al M. Licata, Timothy A. Currier, Mubarak H. Syed, and Katherine I. Nagel. A neural circuit for wind-guided olfactory navigation. Nature Communications, 13:4613,

    work page doi:10.1038/s41586-021-04067-0

  21. [2024]

    Nils Eckstein, Alexander S

    doi: 10.1038/s41586-024-07558-y. Nils Eckstein, Alexander S. Bates, Andrew Champion, Michelle Du, Yijie Yin, Philipp Schlegel, Alicia Kun-Yang Lu, Thomson Rymer, Samantha Finley-May, Tyler Paterson, Ruchi Parekh, Sven Dorkenwald, Arie Matsliah, Szi-chieh Yu, Claire McKellar, Amy Sterling, Katharina Eichler, Marta Costa, Sebastian Seung, Mala Murthy, Volke...

    doi:10.1038/s41586-024-07558-y

  22. [2025]

    pbio.3003014

    doi: 10.1371/journal. pbio.3003014. 17 Sung Soo Kim, Herv´ e Rouault, Shaul Druckmann, and Vivek Jayaraman. Ring attractor dynamics in the Drosophila central brain.Science, 356(6340):849–853,

    doi:10.1371/journal