arxiv: 2605.06208 · v1 · submitted 2026-05-07 · ⚛️ physics.soc-ph · cond-mat.dis-nn· cond-mat.stat-mech

Recognition: unknown

Two-mode geometry controls multiscale organization in bipartite systems

Giulio Cimini, Ottavia Falconi, Pablo Villegas

Pith reviewed 2026-05-08 04:02 UTC · model grok-4.3

classification ⚛️ physics.soc-ph cond-mat.dis-nncond-mat.stat-mech

keywords bipartite networksmultiscale organizationrenormalizationstructural imbalanceLaplacianrole separationcoarse-grainingcomplex systems

0 comments

The pith

Structural imbalance in bipartite networks reshapes multiscale organization while leaving scaling properties unchanged.

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

The paper introduces a renormalization method that works directly on the two sets of nodes in bipartite networks, avoiding the loss of role separation that occurs in standard projections. Using model networks tuned to critical points, it demonstrates that altering the relative sizes or densities of the two sides changes how structure appears when the network is viewed at coarser scales. At the same time, the scaling exponents that describe overall behavior remain fixed across those changes. When the same procedure is applied to real bipartite data from various domains, it uncovers consistent hierarchical patterns for both node types that do not appear after projection. The contrast shows that the two-mode layout itself acts as a constraint on what multiscale features can be observed.

Core claim

A Laplacian-based renormalization framework is defined that performs coarse-graining while keeping the two node partitions distinct. On critical bipartite ensembles, variation in structural imbalance produces different organization patterns at successive scales, yet the scaling relations stay invariant. In empirical networks the direct method exposes nontrivial multiscale hierarchies for both partitions; performing renormalization after one-mode projection, which restricts diffusion to nearest neighbors, produces qualitatively different structures.

What carries the argument

Laplacian-based renormalization framework that operates directly on the bipartite architecture to enable scale transformations while retaining role differentiation.

If this is right

Scaling properties of bipartite networks remain unchanged under different levels of structural imbalance.
Multiscale hierarchies observed in empirical bipartite data require direct two-mode renormalization to appear correctly.
One-mode projections truncate diffusion paths and therefore produce incorrect multiscale structures.
Role separation is maintained across coarse-grained scales only when the original bipartite geometry is preserved.

Where Pith is reading between the lines

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

The separation of scaling invariance from geometry suggests that bipartite critical behavior may fall into the same universality class irrespective of balance ratios.
Applying the framework to additional real-world bipartite systems could test whether the revealed hierarchies are a general feature of role-separated networks.
Analogous direct renormalization schemes may be required for other systems with strict partition constraints to avoid projection-induced distortions.

Load-bearing premise

The Laplacian-based renormalization correctly preserves the essential multiscale features of bipartite systems without artifacts from the coarse-graining steps.

What would settle it

In controlled critical bipartite ensembles, observing that scaling exponents shift when structural imbalance is varied would disprove the claimed invariance.

Figures

Figures reproduced from arXiv: 2605.06208 by Giulio Cimini, Ottavia Falconi, Pablo Villegas.

**Figure 1.** Figure 1: Two-mode-preserving renormalization in bipartite networks. (a), Diffusion-driven coarse-graining in a bipartite architecture. Node classes are identified by different colors. While standard Laplacian renormalization creates mixed blocks, the bipartite renormalization preserves separation between classes. (b), Illustration of a coarse-graining step of the bipartite renormalization scheme, where the scale is… view at source ↗

**Figure 2.** Figure 2: Critical scaling and geometry in random bipartite networks. (a) Giant-component size P∞ as a function of the average degree ⟨k⟩ (computed on the full bipartite network) for a strongly imbalanced system (α = 0.025) and increasing system size N = NA + NB. (b) Finite-size scaling at criticality for different imbalance values α. The maximum susceptibility scales as χmax(kc, N) ∼ N γ (main panel), while the… view at source ↗

**Figure 3.** Figure 3: Multiscale hierarchy of virus–host interactions. (a) Global dendrogram of viruses obtained from b-LRG, revealing a hierarchical organization across scales. The outer ring indicates viral taxonomic groups. Large-scale branches do not align with taxonomy, indicating that interaction-derived structure captures an organization distinct from phylogenetic similarity. (b) Intermediate-scale zooms of selected bran… view at source ↗

**Figure 4.** Figure 4: One-mode projection distorts multiscale organization. Adjusted mutual information (AMI) between clusters obtained from the empirical bipartite dendrogram constructed by b-LRG and those derived from degree-preserving bipartite null models (edge-swap CM and BiCM, again constructed by b-LRG) or one-mode projections (unweighted and weighted, constructed by standard LRG), as a function of the number of clusters… view at source ↗

**Figure 1.** Figure 1: Finite-size scaling analysis of Random Bipartite Networks with α = 0.25. a) Fraction of nodes in the giant component P∞ (solid lines) and susceptibility χ (dashed lines) as a function of the mean connectivity ⟨k⟩ for different system sizes N (2 10 –2 18). The vertical dotted line marks the analytical percolation threshold ⟨k⟩c. As N increases, the transition sharpens, and the susceptibility peak grows. b) … view at source ↗

**Figure 2.** Figure 2: Finite-size scaling analysis of Random Bipartite Networks with α = 0.11. a) Fraction of nodes in the giant component P∞ (solid lines) and susceptibility χ (dashed lines) as a function of the mean connectivity ⟨k⟩ for different system sizes N (2 10 –2 18). The vertical dotted line marks the analytical percolation threshold ⟨k⟩c. As N increases, the transition sharpens, and the susceptibility peak grows. b) … view at source ↗

**Figure 3.** Figure 3: Multiscale geometry and Bipartite Laplacian Renormalization Group (b-LRG) flow of critical bipartite networks. a) Symmetric bipartite network (α = 1). The inset shows the spectral specific heat C(τ ) for different system sizes, exhibiting a broad plateau consistent with scale-invariant diffusion. The backbone displays a homogeneous, tree-like structure that is preserved under coarse-graining (grey arrow). … view at source ↗

**Figure 4.** Figure 4: Multiscale hierarchical organization of host species derived from bipartite diffusion geometry. The dendrogram is obtained from the b-LRG applied to the host–virus bipartite network. The horizontal axis represents the normalized diffusion distance ∆/∆max. At large scales, the hierarchy separates major host groups into broad clusters. Four major structural blocks emerge at large scales, corresponding to an… view at source ↗

**Figure 5.** Figure 5: Distortion of multiscale hierarchy under one-mode projection. a–c, Viral dendrograms obtained from: (a) bipartite geometry, and from (b) unweighted (c) weighted projections. d–f, Corresponding host dendrograms. Bipartite representations preserve multiscale branching, whereas projections compress the hierarchy into dense structures, reflecting the loss of higher-order interaction pathways view at source ↗

**Figure 6.** Figure 6: Davis Southern Women network. Bipartite network of women and events. Node colors indicate the four-group partition obtained from the diffusion-based hierarchical analysis. In this regime, the diffusion geometry is largely determined by shared-neighbor overlap. Because correlations are primarily encoded in short alternating paths, projection does not substantially alter the effective geometry view at source ↗

**Figure 7.** Figure 7: Information-theoretic comparison for the Davis Southern Women network. Normalized (NMI) and Adjusted Mutual Information (AMI) between the empirical bipartite hierarchy, null models, and one-mode projections. The weighted projection maintains high NMI and AMI across resolutions, indicating that local overlap captures most of the structural organization in this system. 3.2. Projection fidelity across complem… view at source ↗

**Figure 8.** Figure 8: (b,e)), the weighted projection shows high AMI values across a broad range of resolutions. This reflects the strong redundancy of export portfolios, which creates dense overlap patterns and makes short-range correlations dominant. In this regime, projection approximates the bipartite geometry. However, the unweighted projection fails to recover the structure, indicating that edge weights are essential to … view at source ↗

**Figure 9.** Figure 9: Hierarchical organization of the Barrett network. Bipartite diffusion dendrogram for plants and pollinators. This separates structurally specialized plants (e.g., Cypripedium acaule) into peripheral niches distinct from the generalist core. Similarly, diffusion-based hierarchy isolates a central core of highly connected generalist species (e.g., Bombus spp.) from specialized sub-lineages of solitary bees. … view at source ↗

**Figure 10.** Figure 10: Information-theoretic comparison for the Barrett network. Normalized (NMI) and Adjusted Mutual Information (AMI) between the empirical bipartite hierarchy, null models, and one-mode projections. Null models collapse to near-zero AMI, while projections retain a partial and resolution-dependent similarity. Under AMI, degree-preserving null models yield values close to zero across resolutions, indicating tha… view at source ↗

**Figure 11.** Figure 11: Bipartite interaction network of the Bartomeus ecosystem. The topology is dominated by a dense, nested core of highly connected generalist species, resulting in strong overlap redundancy across both node types. The bipartite diffusion geometry reveals a core-dominated structure in both classes. The central module is populated by ubiquitous, highly connected species, including pollinators such as Apis mell… view at source ↗

**Figure 12.** Figure 12: Hierarchical organization of the Bartomeus network. The diffusion-based hierarchy reveals a dense generalist core and peripheral branches associated with more specialized species, reflecting a strongly redundant interaction structure. The information-theoretic comparison (reported in Supplementary view at source ↗

**Figure 13.** Figure 13: Information-theoretic comparison for the Bartomeus plant-pollinator network. NMI and AMI between the bipartite hierarchy, null models, and projections. The weighted projection closely approximates the plant hierarchy due to strong overlap redundancy, while the projection remains incomplete for the network of pollinators. This asymmetry identifies a structural criterion for projection validity: it depends … view at source ↗

**Figure 14.** Figure 14: ). For dispersers, a dense generalist core emerges, comprising highly connected species such as Pycnonotus barbatus, Turdus pelios, and generalist primates (e.g., Papio anubis, Cercopithecus ascanius), while specialized species occupy peripheral branches. A similar large-scale structure is observed for plants, with a dominant core of frequently consumed species. 10 −20 10 −14 10 −8 10 −2 Δ/Δmax Snowy-crow… view at source ↗

**Figure 15.** Figure 15: AMI and NMI of the Schleuning network. Normalized (NMI) and Adjusted Mutual Information (AMI) highlight the severe limitations of one-mode reductions. While the weighted projection captures the redundant structure of the dense disperser core (high AMI), it severely truncates the multiscale geometry of the seeds. Furthermore, the rising AMI of null models (BiCM, Edge-swap CM) at finer scales for the seeds … view at source ↗

read the original abstract

Many complex systems are organized around complementary roles and naturally described as bipartite networks. Unveiling their multiscale structure presents a fundamental challenge because coarse-graining procedures must preserve role separation, whereas standard approaches collapse it via one-mode projections. Here we introduce a Laplacian-based renormalization framework that operates directly on the bipartite architecture, enabling scale transformations while retaining role differentiation. Using controlled bipartite ensembles at criticality, we show that structural imbalance systematically reshapes organization across scales while leaving scaling properties invariant, revealing a separation between universality and geometry. Applying the coarse-graining framework to empirical bipartite networks, we uncover nontrivial multiscale hierarchies for both roles. In contrast, renormalization performed after one-mode projection -- which truncates diffusion paths to nearest neighbors -- yields qualitatively different structures. Our results identify two-mode geometry as a fundamental constraint for revealing multiscale organization in systems with role separation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a direct Laplacian renormalization for bipartite graphs that keeps roles separate and separates geometry-driven reorganization from invariant scaling.

read the letter

The main thing to know is that this work builds a renormalization procedure that stays on the bipartite structure instead of projecting to one mode first. Using the Laplacian, it coarse-grains while preserving the two distinct roles, then shows on critical ensembles that imbalance changes the organization pattern across scales but leaves the scaling exponents untouched. On empirical cases the method finds nontrivial hierarchies on each side that disappear after projection, which truncates the diffusion paths.

Referee Report

0 major / 3 minor

Summary. The paper introduces a Laplacian-based renormalization framework that operates directly on bipartite networks to enable coarse-graining while preserving the separation between the two modes, avoiding artifacts from one-mode projections. Using controlled bipartite ensembles tuned to criticality, it demonstrates that structural imbalance reshapes multiscale organization across scales but leaves scaling properties invariant, thereby separating geometric effects from universal scaling behavior. The framework is then applied to empirical bipartite networks, uncovering nontrivial multiscale hierarchies for both roles, with explicit contrasts showing that post-projection renormalization produces qualitatively different structures.

Significance. If the central claims hold, the work provides a valuable tool for analyzing multiscale organization in bipartite systems with role separation, such as ecological or social networks, by isolating the effects of two-mode geometry. Strengths include the use of controlled critical ensembles for testing invariance and direct empirical comparisons against projection-based methods, which supply internal consistency checks. The separation of universality from geometry is a potentially impactful conceptual contribution to network science.

minor comments (3)

The description of the bipartite Laplacian renormalization procedure would benefit from an explicit statement of how the coarse-graining operator is constructed to ensure it does not truncate diffusion paths in a manner analogous to projections; a short derivation or pseudocode would aid reproducibility.
In the empirical section, the specific metrics used to identify 'nontrivial multiscale hierarchies' (e.g., community detection scores, participation coefficients, or renormalization flow diagrams) should be defined with reference to the relevant equations or figures.
The abstract states that scaling properties remain invariant; a brief parenthetical example of the invariant quantity (such as a particular exponent or correlation function) would help readers immediately grasp the distinction from the geometry-driven reorganization.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their supportive summary and recommendation of minor revision. We are pleased that the conceptual separation between two-mode geometry and scaling invariance, along with the direct comparisons to projection-based renormalization, is viewed as a potentially impactful contribution.

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained

full rationale

The paper introduces a Laplacian-based renormalization framework operating directly on bipartite graphs, then applies it to independently generated critical ensembles and empirical networks to demonstrate invariant scaling alongside geometry-driven reorganization. Claims rest on explicit contrasts with one-mode projections and controlled tests rather than any fitted parameter renamed as prediction, self-referential definition, or load-bearing self-citation chain. No equation or step reduces by construction to its own input; the separation of universality from two-mode geometry follows from the framework's application to external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based on abstract only; the framework assumes a suitable Laplacian can be defined on bipartite graphs to support scale transformations while retaining role separation. No free parameters, additional axioms, or invented entities are described.

axioms (1)

domain assumption A Laplacian operator exists that enables renormalization directly on bipartite graphs while preserving bipartiteness and role differentiation.
Invoked by the introduction of the renormalization framework.

pith-pipeline@v0.9.0 · 5449 in / 1165 out tokens · 83984 ms · 2026-05-08T04:02:15.105140+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

56 extracted references

[1]

H. A. Simon, Proc. Am. Philos. Soc.106, 467 (1962)

1962
[2]

Bascompte, P

J. Bascompte, P. Jordano, C. J. Melián, and J. M. Ole- sen, Proc. Natl. Acad. Sci. U.S.A.100, 9383 (2003)

2003
[3]

Latapy, C

M. Latapy, C. Magnien, and N. Del Vecchio, Soc. Netw. 30, 31 (2008)

2008
[4]

Latora, V

V. Latora, V. Nicosia, and G. Russo,Complex Networks: Principles, Methods and Applications(Cambridge Uni- versity Press, New York, NY, USA, 2017)

2017
[5]

S. P. Borgatti and M. G. Everett, Soc. Netw.19, 243 (1997)

1997
[6]

Holme, F

P. Holme, F. Liljeros, C. R. Edling, and B. J. Kim, Phys. Rev. E68, 056107 (2003)

2003
[7]

Guillaume and M

J.-L. Guillaume and M. Latapy, Information Processing Letters90, 215 (2004)

2004
[8]

Bastolla, M

U. Bastolla, M. A. Fortuna, A. Pascual-García, A. Fer- rera, B. Luque, and J. Bascompte, Nature458, 1018 (2009)

2009
[9]

C. A. Hidalgo, B. Klinger, A.-L. Barabási, and R. Haus- mann, Science317, 482 (2007)

2007
[10]

Jeong, B

H. Jeong, B. Tombor, R. Albert, Z. N. Oltvai, and A.-L. Barabási, Nature407, 651 (2000)

2000
[11]

M. E. J. Newman, Phys. Rev. E64, 016131 (2001)

2001
[12]

Bascompte and P

J. Bascompte and P. Jordano, Annu. Rev. Ecol. Evol. Syst.38, 567 (2007)

2007
[13]

P. R. Guimarães, Annu. Rev. Ecol. Evol. Syst.51, 433 (2020)

2020
[14]

M. J. Barber, Phys. Rev. E76, 066102 (2007)

2007
[15]

Guimerà, M

R. Guimerà, M. Sales-Pardo, and L. A. N. Amaral, Phys. Rev. E76, 036102 (2007)

2007
[16]

D. B. Larremore, A. Clauset, and A. Z. Jacobs, Phys. Rev. E90, 012805 (2014)

2014
[17]

Fortunato, Phys

S. Fortunato, Phys. Rep.486, 75 (2010)

2010
[18]

T. P. Peixoto, Phys. Rev. X4, 011047 (2014)

2014
[19]

T. Zhou, J. Ren, M. c. v. Medo, and Y.-C. Zhang, Phys. Rev. E76, 046115 (2007)

2007
[20]

Kitsak, F

M. Kitsak, F. Papadopoulos, and D. Krioukov, Phys. Rev. E95, 032309 (2017)

2017
[21]

Villegas, T

P. Villegas, T. Gili, G. Caldarelli, and A. Gabrielli, Nat. Phys.19, 445 (2023)

2023
[22]

P.Villegas, A.Gabrielli, A.Poggialini, andT.Gili,Phys. Rev. Res.7, 013065 (2025)

2025
[23]

Villegas, A

P. Villegas, A. Gabrielli, F. Santucci, G. Caldarelli, and T. Gili, Phys. Rev. Res.4, 033196 (2022)

2022
[24]

Caldarelli, A

G. Caldarelli, A. Gabrielli, T. Gili, and P. Villegas, J. Stat. Mech.: Theory Exp.2024, 084002 (2024)

2024
[25]

Poggialini, P

A. Poggialini, P. Villegas, M. A. Muñoz, and A. Gabrielli, Phys. Rev. Lett.134, 057401 (2025)

2025
[26]

De Domenico and J

M. De Domenico and J. Biamonte, Phys. Rev. X6, 041062 (2016)

2016
[27]

Davis, B

A. Davis, B. B. Gardner, and M. R. Gardner,Deep South: A Social Anthropological Study of Caste and Class (University of Chicago Press, Chicago, IL, 1941)

1941
[28]

Cui and X

Y. Cui and X. Wang, Phys. A: Stat. Mech. Appl.407, 7 (2014)

2014
[29]

M. E. J. Newman, S. H. Strogatz, and D. J. Watts, Phys. Rev. E64, 026118 (2001)

2001
[30]

Jankowski, R

R. Jankowski, R. Aliakbarisani, M. Á. Serrano, and M. Boguñá, Comm. Phys.9, 35 (2026)

2026
[31]

J. Alt, R. Ducatez, and A. Knowles, Commun. Math. Phys.405, 9 (2024)

2024
[32]

K. J. Olival, P. R. Hosseini, C. Zambrana-Torrelio, N. Ross, T. L. Bogich, and P. Daszak, Nature546, 646 (2017)

2017
[33]

Saracco, R

F. Saracco, R. Di Clemente, A. Gabrielli, and T. Squar- tini, Sci. Rep.5, 10595 (2015)

2015
[34]

Strona, D

G. Strona, D. Nappo, F. Boccacci, S. Fattorini, and J. San-Miguel-Ayanz, Nat. Commun.5, 4114 (2014)

2014
[35]

Tumminello, S

M. Tumminello, S. Miccichè, F. Lillo, J. Piilo, and R. N. Mantegna, PLoS ONE6, e17994 (2011)

2011
[36]

Gualdi, G

S. Gualdi, G. Cimini, K. Primicerio, R. Di Clemente, and D. Challet, Sci. Rep.6, 39467 (2016)

2016
[37]

Saracco, M

F. Saracco, M. J. Straka, R. Di Clemente, A. Gabrielli, G. Caldarelli, and T. Squartini, New J. Phys.19, 053022 (2017)

2017
[38]

Tamarit, M

I. Tamarit, M. Pereda, and J. A. Cuesta, Phys. Rev. E 102, 042304 (2020)

2020
[39]

Cimini, A

G. Cimini, A. Carra, L. Didomenicantonio, F. Saracco, and T. Squartini, Commun. Phys.5, 76 (2022)

2022
[40]

Neal, Social Networks39, 84 (2014)

Z. Neal, Social Networks39, 84 (2014)

2014
[41]

Z. P. Neal, A. Cadieux, D. Garlaschelli, N. J. Gotelli, F. Saracco, T. Squartini, S. T. Shutters, W. Ulrich, G. Wang, and G. Strona, PLOS Complex Syst.1, e0000010 (2024)

2024
[43]

Bartomeus, M

I. Bartomeus, M. Vila, and L. Santamaria, Oecologia 155, 761 (2008)

2008
[44]

Schleuning, N

M. Schleuning, N. Blüthgen, M. Flörchinger, J. Braun, H. M. Schaefer, and K. Böhning-Gaese, Ecology92, 26 (2011). 9

2011
[45]

Tacchella, M

A. Tacchella, M. Cristelli, G. Caldarelli, A. Gabrielli, and L. Pietronero, Sci. Rep.2, 723 (2012)

2012
[46]

Caccioli, M

F. Caccioli, M. Shrestha, C. Moore, and J. D. Farmer, J. Bank. Financ.46, 233 (2014)

2014
[47]

Kunegis, inProceedings of the 22nd International Con- ference on World Wide Web Companion, WWW ’13 Companion (Association for Computing Machinery, New York, NY, USA, 2013) pp

J. Kunegis, inProceedings of the 22nd International Con- ference on World Wide Web Companion, WWW ’13 Companion (Association for Computing Machinery, New York, NY, USA, 2013) pp. 1343–1350

2013
[48]

A. A. Bogdanov,Essays in Tektology: The General Sci- ence of Organization(IntersystemsPublications,Seaside, CA, 1996) originally published in Russian (1912–1917). Two-mode geometry controls multiscale organization in bipartite systems Supplementary Materials Ottavia Falconi,1, 2 Giulio Cimini,3, 2 and Pablo Villegas2, 4,∗ 1Physics Department University of...

1996
[49]

Random bipartite network generation 2 1.2

Percolation transition and finite-size scaling in imbalanced bipartite networks 2 1.1. Random bipartite network generation 2 1.2. Analytical percolation threshold via generating functions 2 1.3. Finite-size scaling and universality 3 1.4. b-LRG flow and multiscale geometry 4
[50]

Host multiscale hierarchy and taxonomic correlation 6
[51]

Davis Southern Women: a low-complexity regime 8 3.2

Topological limits and structural dependencies of one-mode projections 8 3.1. Davis Southern Women: a low-complexity regime 8 3.2. Projection fidelity across complementary partitions 9 3.3. Ecological interaction networks 11 3.3.1. Barrett plant–pollinator network: sparse, hierarchy-dominated regime 11 3.3.2. Bartomeus plant–pollinator network: high-overl...
[52]

PERCOLATION TRANSITION AND FINITE-SIZE SCALING IN IMBALANCED BIPARTITE NETWORKS 1.1. Random bipartite network generation To probe how partition asymmetry affects the internal geometry of the critical backbone, we study random bipartite networks composed of two disjoint sets of nodes,AandB, with sizesN A andN B. The total number of nodes is N=N A +NB, and ...
[53]

Here, we extend the analysis to the host partition within the same interaction space

HOST MULTISCALE HIERARCHY AND TAXONOMIC CORRELATION In the main text, we showed that bipartite diffusion geometry reveals a multiscale hierarchy for viral species that is distorted under one-mode projection. Here, we extend the analysis to the host partition within the same interaction space. Unlike projected representations, which compress the branching ...
[54]

TOPOLOGICAL LIMITS AND STRUCTURAL DEPENDENCIES OF ONE-MODE PROJECTIONS To investigate the conditions under which one-mode projections represent a well-behaved approximation, we extend the NMI and AMI analysis to a broader set of empirical bipartite networks. As discussed in the main text, projection acts as a low-order truncation of the diffusion geometry...
[55]

S. C. H. Barrett and K. Helenurm, Can. J. Bot.65, 2036 (1987)

2036
[56]

Bartomeus, M

I. Bartomeus, M. Vila, and L. Santamaria, Oecologia155, 761 (2008)

2008
[57]

Schleuning, N

M. Schleuning, N. Blüthgen, M. Flörchinger, J. Braun, H. M. Schaefer, and K. Böhning-Gaese, Ecology92, 26 (2011)

2011