Recei ed: 16 Ma ch 2023 Re ised: 11 Sep embe 2023 Accep ed: 11 Oc obe 2023
DOI: 10.1111/cobi.14206
CONTRIBUTED PAPERS
A global me a-analysis o he impac s o o es agmen a ion on
bio ic mu ualisms and an agonisms
To ah Siegel1Ainhoa Mag ach2William F. Lau ance3Da id Lu he 1
1Geo ge Mason Uni e si y, Fai ax, Vi ginia, USA
2Basque Cen e o Clima e Change, Leioa, Spain
3Cen e o T opical En i onmen al and
Sus ainabili y Science, James Cook Uni e si y,
Cai ns, Queensland, Aus alia
Co espondence
To ah Siegel, 4400 Uni e si y D i e, Fai ax, VA
22030, USA. Email: o [email p o ec ed]
A icle Impac S a emen : Fo es agmen a ion
dis up s key ecosys em unc ions by nega i ely
a ec ing mu ualisms and s eng hening pa asi ic
in e ac ions.
Abs ac
Fo es agmen a ion is a g a e h ea o biodi e si y. Fo es s a e becoming inc easingly
agmen ed wi h mo e han 70% now <1 km om o es edge. Al hough much is known
abou he e ec s o o es agmen a ion on indi idual species, much less is unde s ood
abou i s e ec s on species in e ac ions (i.e., mu ualisms, an agonisms, e c.). In 2014, a
p e ious me a-analysis assessed he impac s o o es agmen a ion on di e en species
in e ac ions, ac oss 82 s udies. We pooled he p e ious da a wi h da a published in he
las 10 yea s (combined o al 104 s udies and 168 e ec sizes). We compa ed he new
se o publica ions (22 s udies and 32 e ec sizes) wi h he old se o e alua e po en ial
changes in species in e ac ions o e ime gi en he global inc ease in agmen a ion a es.
Mu ualisms we e mo e nega i ely a ec ed by o es agmen a ion han an agonisms (p
<0.0001). Edge e ec s, agmen size, and deg ada ion nega i ely a ec ed mu ualisms,
bu no an agonisms, a di e en inding om he o iginal me a-analysis. Pa asi ic in e ac-
ions inc eased as agmen size dec eased (p<0.0001)—an in iguing esul a a iance
wi h ea lie s udies. New publica ions showed a mo e nega i e mean e ec size o o es
agmen a ion on mu ualisms han old publica ions. Al hough esea ch is s ill limi ed o
some in e ac ions, we iden i ied an impo an scien i ic end: cu en esea ch ends o
ocus on an agonisms. We concluded ha o es agmen a ion dis up s impo an species
in e ac ions and ha his dis up ion has inc eased o e ime.
KEYWORDS
an agonis ic, o es agmen , me a-analy ical, mu ualis ic, species in e ac ions
Me aanálisis Mundial del Impac o de la F agmen ación de Bosques sob e el Mu ualismo y
An agonismo Bió ico
Resumen: La agmen ación del bosque es una amenaza g a e pa a la biodi e sidad. Los
bosques es án más agmen ados, pues más del 70% ienen <1 km a pa i del bo de del
bosque. Aunque hay mucha in o mación del e ec o de la agmen ación sob e las especies,
hay poco conocimien o de sus e ec os sob e las in e acciones en e especies (mu ualismo,
an agonismo, e c.). Un me aanálisis ealizado en 2014 po e aluó en 82 es udios el impac o
de la agmen ación del bosque sob e las di e en es in e acciones en e especies. Jun a-
mos es os da os con da os publicados en los úl imos diez años ( o al combinado de 104
es udios y 168 amaños de e ec o). Compa amos el g upo nue o de publicaciones (22 es u-
dios y 32 amaños de e ec o) con el g upo de pa a e alua los cambios po enciales en
las in e acciones en e especies con el iempo dado el inc emen o mundial de la asa de
agmen ación. El mu ualismo ue más a ec ado nega i amen e po la agmen ación del
bosque que los an agonismos (p<0.0001). El e ec o del bo de, amaño del agmen o y
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion-NonComme cial-NoDe i s License, which pe mi s use and dis ibu ion in any medium, p o ided he
o iginal wo k is p ope ly ci ed, he use is non-comme cial and no modi ica ions o adap a ions a e made.
© 2023 The Au ho s. Conse a ion Biology published by Wiley Pe iodicals LLC on behal o Socie y o Conse a ion Biology.
Conse a ion Biology. 2024;e14206. wileyonlinelib a y.com/jou nal/cobi 1o 10
h ps://doi.o g/10.1111/cobi.14206
2o 10 SIEGEL ET AL.
la deg adación u ie on un e ec o nega i o sob e el mu ualismo, pe o no sob e el an ag-
onismo, un esul ado di e en e al del me aanálisis o iginal. Las in e acciones pa asi a ias
inc emen a on con o me se edujo el amaño del agmen o (p<0.0001)—un esul-
ado in igan e en disc epancia con los p ime os esul ados. Las publicaciones ecien es
mos a on un amaño p omedio de e ec o de la agmen ación del bosque más nega-
i o pa a el mu ualismo que las publicaciones an iguas. Aunque hay poca in es igación
sob e algunas in e acciones, iden i icamos una endencia cien í ica impo an e: la in es i-
gación ac ual iende a en oca se en los an agonismos. Concluimos que la agmen ación del
bosque al e a las in e acciones impo an es en e especies y que es e cambio ha aumen ado
con el iempo.
PALABRAS CLAVE
an agonis a, agmen o de bosque, in e acciones en e especies, me aanálisis, mu ualis a
,
70% 1
,()
2014 ,Mag ach82
10
(104 168 ), (22 32
)Mag ach ,
,,
,(p<0.0001)
,,
(p<0.0001), ,
,
,:
,:
,:;:
:,,,,
INTRODUCTION
Habi a loss and agmen a ion a e c i ical h ea s o Ea h’s
biological di e si y (Haddad e al., 2015; Lau ance e al., 2007,
2011,2018; Newma k e al., 2017) and ecosys em unc ioning
(Haddad e al., 2015). Al hough habi a agmen a ion occu s
in all ecosys ems, o es agmen a ion is o pa icula conce n
because mos o he wo ld’s e es ial species li e in o es s
(FAO & UNEP, 2020; Pillay e al., 2021).
Fo es agmen a ion inc eases he amoun o o es edge and
educes he amoun o in e io o es , whe e mos o es species
eside (Lau ance e al., 2011). Fo example, Taube e al. (2018)
iden i ied o e 130 million o es agmen s ac oss sub opical
and opical egions o he Ame icas, A ica, Asia, and Aus alia.
They ound ha mo e han 70% o he wo ld’s emaining o es s
a e <1 km om he nea es o es edge and ha 20% o emain-
ing o es s a e <100 m om he nea es edge (Haddad e al.,
2015). P oximi y o a o es edge c ea es a mul i ude o nega-
i e ecological e ec s on opical biodi e si y (Be s e al., 2017;
Lau ance e al., 2011, 2018).
Al hough he e ec s o o es agmen a ion on species ich-
ness and communi y composi ion ha e been widely s udied,
less is unde s ood abou he b oade impac s on in e speci ic
in e ac ions (Mag ach e al., 2014). Resea ch has p e iously
assessed species-speci ic in e ac ions in he con ex o ag-
men ed o es s. Some examples include esea ch on seed
dispe sal (e.g., Co dei o & Howe, 2003; C ame e al., 2007),
pollina ion (e.g., Aizen & Feinsinge , 1994; Hadley e al., 2014),
and p eda ion (e.g., Ka ei a, 1987; Huh a e al., 2004). Howe e ,
he e is conside ably less esea ch ha syn hesizes he o e all
impac s o o es agmen a ion on he b oade ca ego ies o
species in e ac ions.
In 2014, Mag ach e al. (2014) syn hesized he da a on
in e speci ic in e ac ions in agmen ed o es s. The au ho s
p oduced a comp ehensi e me a-analysis in which hey assessed
how di e en componen s o o es agmen a ion a ec ed
an agonis ic and mu ualis ic in e ac ions in s udies published
be o e 2012. These componen s included agmen size, edge
e ec s, agmen isola ion, and agmen deg ada ion. F agmen
size is used o examine he o al a ea o a o es agmen
15231739, 0, Downloaded om h ps://conbio.onlinelib a y.wiley.com/doi/10.1111/cobi.14206 by Readcube (Lab i a Inc.), Wiley Online Lib a y on [22/03/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
CONSERVATION BIOLOGY 3o 10
(Lo ejoy, 1980), edge e ec s o assess he impac s o he a i-
icial edge c ea ed du ing agmen a ion (Didham e al., 1998),
agmen isola ion o e alua e he dis ance o a agmen o
o he simila na u al habi a (Lau ance e al., 2011), and ag-
men deg ada ion o examine he quali y o he agmen and
he species in i (Taba elli & Gascon, 2005). Mag ach e al.
(2014) concluded ha o e all mu ualis ic in e ac ions a e mo e
nega i ely a ec ed by o es agmen a ion han an agonis ic
in e ac ions and ha edge e ec s, agmen isola ion, and ag-
men deg ada ion all ha e a nega i e impac on mu ualis ic
in e ac ions. They ound ha agmen size does no app ecia-
bly al e he speci ic ypes o in e ac ions s udied, including seed
dispe sal, pollina ion, myco hizae (a plan oo − ungi in e ac-
ion), p eda ion, pa asi ism, and he bi o y. Howe e , hey ound
ha edge e ec s signi ican ly a ec seed dispe sal. The au ho s
did no e he po en ial limi a ions o hei indings based on he
limi ed numbe o s udies a ailable o each ype o in e ac ion
and he small sample sizes in each s udy.
Mag ach e al. (2014) called o u he esea ch in o he
e ec s o o es agmen a ion on species in e ac ions o be e
unde s and how hese ne wo ks and ecosys ems a e a ec ed.
They no ed ha many ecological in e ac ions had been com-
pa a i ely unde s udied. Fo example, myco hizae had only
3 eco ded in e ac ions, whe eas seed dispe sal and p eda ion
we e much mo e ex ensi ely s udied. They also no ed small
sample sizes we e a p oblem ha could ha e a ec ed he s a is-
ical powe o hei me a-analysis. Now, a ull decade la e , he e
is mo e li e a u e a ailable and mo e esea ch has e ealed new
di ec ions in how species in e ac ions and ela ed ecosys em
unc ions a e a ec ed by o es agmen a ion.
We buil on he esea ch by Mag ach e al. (2014) by summa-
izing and analyzing a decade o new da a collec ed om 2012
o 2022. These new da a should inc ease he s a is ical powe o
he analysis. We asked he ollowing ques ions: Does o es ag-
men a ion a ec mu ualis ic in e ac ions mo e nega i ely han
an agonis ic in e ac ions; do he s udied componen s o o -
es agmen a ion a ec mu ualisms mo e han an agonisms; do
hese componen s di e en ially a ec he 6 ypes o in e spe-
ci ic in e ac ions p e iously de ined by Mag ach e al. (2014);
a e he e changes in he e ec sizes wi hin each g oup be ween
he s udies ound by Mag ach e al. (2014) and he newe pub-
lica ions; and a e he e s ill biases based on he geog aphical
loca ion o esea ch o he ypes o s udied in e ac ions gi en
ha Mag ach e al. (2014) ound mos s udies we e conduc ed
in B azil?
Based on hese ques ions, we p edic ed ha o es agmen-
a ion would mo e s ongly and nega i ely a ec mu ualisms
because an agonis ic in e ac ions a e hypo hesized o be mo e
adap able o en i onmen al p essu es (B ockhu s & Koskella,
2013) due o a highe connec edness o species in hese ne -
wo ks (Baumga ne , 2020); edge e ec s, agmen isola ion,
and agmen deg ada ion would con inue o a ec mu ualisms
mo e han an agonisms because habi a dis u bance inc eases
he ulne abili y o specialized mu ualisms (Aizen e al., 2012;
Kie s e al., 2010); he 4 componen s o o es agmen a ion
p e iously e alua ed by Mag ach e al. (2014) would inc ease
in a ia ion wi h la ge sample sizes as a esul o he addi ion
o new s udies; and compa ed wi h he publica ions ound by
Ma g ach e al. (2014), newe s udies would show an inc ease
in e ec size gi en inc easing o es agmen a ion wo ldwide
(Fische e al., 2021).
METHODS
Li e a u e sea ch and inclusion c i e ia
We eplica ed mos o he quan i a i e me hods o he li e a u e
e iew as de ailed in Mag ach e al. (2014). A b ie o e iew is
gi en he e o explica e some key di e ences. We sea ched he
ISI Web o Knowledge, Science Di ec , and CAB Abs ac s o
pee - e iewed publica ions ha assessed he impac o o es
agmen a ion on species in e ac ions (sea ch code in Mag ach
e al., 2014 suppo ing in o ma ion). Mag ach e al. (2014) also
sea ched Scopus as a ou h da abase; howe e , ins i u ional
access o Scopus was una ailable o us. We limi ed ou sea ch
pe iod o esea ch published om 2012 o 2022. We sough
o add new da a and eassess Mag ach e al.’s (2014) indings;
hus, we excluded all pape s om hei o iginal imeline (be o e
2012). Because ou sea ch o e lapped pa o 2012, pape s
in hei o iginal s udy ha appea ed du ing ou sea ch we e
excluded.
We excluded pape s o which he o e all ou come o an eco-
logical in e ac ion (e.g., ui se ) was no measu ed, he s udy
a ea was no o es agmen s, and he s udy a ea was an island
o island sys em. Fo unique ci cums ances, we applied he ol-
lowing c i e ia: i 2 di e en species and hei in e ac ions we e
analyzed in a single pape , hen each in e ac ion was measu ed as
a sepa a e eco d; i an in e ac ion was epea edly s udied o e
ime, we used only da a om he mos ecen s udy pe iod; i a
single s udy analyzed 2 di e en ou comes o he same in e ac-
ion, hen only 1 ou come was used o e alua ion; and i a pape
used an expe imen al ea men , we analyzed only he con ol
da a. We excluded li e a u e e iews because mos included da a
al eady e alua ed in Mag ach e al. (2014). Publica ions ha
delinea ed small clumps o ees (<10 indi iduals) as habi a
pa ches we e no conside ed o be wi hin ou de ini ion o o -
es agmen s. Finally, a ew publica ions e alua ed agmen shape
as a componen o o es agmen a ion. We excluded shape as
a componen o o es agmen a ion because i was no pa
o he o iginal Mag ach e al. (2014) pape and ew esea che s
ha e measu ed i .
Da a analyses
Ou ini ial sea ch ecalled 3930 publica ions. A e emo ing
duplica ions, we had 1813 pape s o e iew. The ini ial ounds
o elimina ion based on he a o emen ioned sea ch c i e ia
esul ed in 121 publica ions o ull- ex e iew. A e eading
each publica ion, we u he excluded pape s based on ou c i-
e ia o he a ailabili y o da a needed o ou analyses (Ha ison,
2011). Fo ele an pape s ha did no epo he necessa y da a,
he co esponding au ho was con ac ed o ob ain he in o ma-
15231739, 0, Downloaded om h ps://conbio.onlinelib a y.wiley.com/doi/10.1111/cobi.14206 by Readcube (Lab i a Inc.), Wiley Online Lib a y on [22/03/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
4o 10 SIEGEL ET AL.
ion. I au ho s did no espond o did no wish o sha e hei
da a, he pape was no included. In ou sea ch om 2012 o
2022, we ound 23 s udies in which 36 in e speci ic in e ac ions
in agmen ed o es s we e analyzed. Howe e , 3 in e ac ions
we e emo ed p io o analyses because insec galling and an -
plan mu ualisms we e in es iga ed; hese mu ualisms did no i
in o he 6 ypes o in e ac ions o iginally analyzed in Mag ach
e al. (2014) o sample sizes we e insu icien o analyses (n<2
publica ions).
F om each pape ha i ou c i e ia, we ex ac ed in o ma-
ion on he na u e o he in e ac ion (mu ualism o an agonism),
he ype o in e ac ion, he ype o o es whe e he s udy
ook place, he coun y whe e he esea ch was conduc ed, he
sample size, and ele an s a is ics. Al hough we eco ded all
in e ac ions, we ocused on 6 main ypes: p eda ion, pa asi ism,
he bi o y, seed dispe sal, pollina ion, and myco hizae. We also
eco ded he componen o o es agmen a ion ha he au ho
o au ho s assessed: agmen size, agmen isola ion, o es
deg ada ion, o edge e ec s (Appendix S1).
In acco dance wi h he me hods ou lined in Mag ach e al.
(2014), we calcula ed he e ec size o each in e ac ion wi h he
da a a ailable in each publica ion o calcula e Hedges’ d, an unbi-
ased measu emen o he s anda dized mean di e ence be ween
2 ea men s (Hedges, 1983, 1985). Fo hese calcula ions, we
p io i ized he use o aw da a o e o he s a is ical ou pu s,
when a ailable. I a s udy epo ed esul s ela ed o a compo-
nen o agmen a ion as a con inuous a iable, we analyzed he
eplica es o he lowes and highes alues. Fo a icles ha did
no epo ans o mable s a is ics bu p o ided ele an da a
in g aphs o igu es, we used Da aThie III (Tumme s, 2006)o
Think-Cell (Wya & Schödl, 2020) o ex ac he equi ed in o -
ma ion. Mag ach e al. (2014) used Me awin 2.0 o hei da a
ans o ma ions, whe eas we used Geo ge Mason Uni e si y’s
P ac ical Me a-Analysis E ec Size Calcula o (Wilson, 2017a)
o ans o m associa ed numbe s and s a is ics in o Cohen’s d.
F om he e, we used Hedges’ gco ec ion ac o (Appendix S3)
o ans o m Cohen’s d o Hedges’ g(Wilson, 2017b) and hen
ans o med Hedges’ g o Hedges’ dusing Me awin 3.0 (Rosen-
be g, 2022). Consis en wi h Mag ach e al. (2014), a posi i e
e ec size indica ed ha inc easing agmen a ion escala es he
s eng h o an ecological in e ac ion.
A e de e mining Hedges’ d o each in e ac ion, we calcu-
la ed he ela ed a iance (Appendix S4) wi h in o ma ion on
he s udy’s sample size. All he e ec sizes and ela ed a iances
calcula ed om ou s udy we e combined wi h he 136 e ec
sizes o iginally epo ed in Mag ach e al. (2014). The combined
da a we e used in he inal analyses o assess he impac s o o -
es agmen a ion on all published li e a u e a ailable om 1997
o 2022. Hedges’ dand he ela ed a iance we e used o gen-
e a e bias-co ec ed 95% con idence in e als wi h he unc ion
esampling me hod wi h 1000 i e a ions in Me awin 3.0 (Rosen-
be g, 2022). We also calcula ed he mean e ec size o each
ca ego y o in e ac ion. E ec s we e conside ed signi ican i
he p alue was <0.05. I he con idence in e als o e lapped
ze o, we s ill conside ed ha he di e ence be ween he 2 means
was s a is ically signi ican .
As in Mag ach e al. (2014), we used andom e ec s ca -
ego ical me a-analy ic models wi h e ec sizes weigh ed by
hei a iances (Ve e e al., 2013) and boo s apped esampling
p ocedu es wi h 1000 i e a ions o assess he o e all e ec s
o o es agmen a ion on mu ualisms and an agonisms, o
e alua e he e ec o each componen o agmen a ion on
mu ualisms and an agonisms, o de e mine he impac o each
componen o agmen a ion on he 6 di e en ypes o in e -
ac ions, and o assess any di e ences be ween old and new
publica ions o hese g oups. These models we e used o de e -
mine he e ogenei y in e ec sizes based on Q s a is ics, which
we e es ed agains a chi-squa e dis ibu ion wi h n– 1 d , all in
Me awin 3.0. To assess geog aphic biases, we added he numbe
o s udies pe coun y and de e mined whe e he majo i y o he
esea ch was conduc ed.
Finally, we checked o publica ion bias, which o en occu s
in me a-analyses, because esea ch wi h s a is ically signi ican
esul s is mo e likely o be published han esea ch wi hou
s a is ically signi ican esul s (Eas e b ook e al., 1991). Publi-
ca ion bias is a p oblem in me a-analyses because i can lead
o alse posi i e esul s (Van Ae e al., 2019). To check
o publica ion bias, we c ea ed unnel plo s o ou da a and
isually inspec ed hem o asymme y. Funnel plo s a e sca -
e plo s o he ea men e ec s ha ha e been es ima ed
om each publica ion plo ed agains each e ec size, in his
case, Hedges’ d(S e ne e al., 2005). Asymme y in a unnel
plo is an indica o o publica ion bias, and in he absence
o bias, a unnel plo will appea symme ical (S e ne & Ha -
bo d, 2004). We checked ou isual obse a ions o publica ion
bias wi h he R package me a o 3.8-1 (Viech baue , 2010) o
e alua e bias wi h Rosen hal’s ail-sa e numbe , which gene -
a es he numbe o missing s udies a e aging null esul s ha
would need o be added o make he combined e ec sizes
s a is ically insigni ican (O win, 1983; Rosen hal, 1979), and
Egge ’s eg ession es , which s a is ically measu es unnel plo
asymme y (Egge e al., 1997). I bo h s a is ical es s we e sig-
ni ican , hen we assumed no publica ion bias and us ed ha
we we e no missing da a ha would lead us o alse posi i e
esul s.
RESULTS
New publica ions and inal combined da a se
Mag ach e al. (2014) ound 82 s udies published om 1997 o
2012 and an a e age o 5.5 s udies pe yea , whe eas we ound 22
s udies wi h an a e age o 2.3 s udies pe yea o 2012−2022.
O he s udies we ound, 22 assessed 1 o mo e o he 6 ca e-
go ies o species in e ac ions (32 unique in e ac ions). In hese
22 s udies, 7 mu ualis ic in e ac ions and 25 an agonis ic in e ac-
ions we e examined. When we combined he s udies we ound
wi h hose Mag ach e al. (2014) assessed, we analyzed 104 s ud-
ies wi h 168 e ec sizes. The epo ed numbe s include he o al,
combined s udies included in he analyses (bo h hose included
in Mag ach e al.2014, he ea e old publica ions, and pape s we
15231739, 0, Downloaded om h ps://conbio.onlinelib a y.wiley.com/doi/10.1111/cobi.14206 by Readcube (Lab i a Inc.), Wiley Online Lib a y on [22/03/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
CONSERVATION BIOLOGY 5o 10
TABLE 1 Numbe o e ec sizes o di e en ypes o species in e ac ions a e combining hose p e iously epo ed by Mag ach e al. (2014) wi h new e ec
sizes in publica ions om 2012 o 2022.
In e ac ion ype Old e ec sizes New e ec sizes To al e ec sizes Inc ease (%)*
Mu ualism 58 7 65 12
An agonism 78 25 103 32
P eda ion 44 13 57 30
Pa asi ism 16 3 19 19
He bi o y 19 9 28 47
Seed dispe sal 31 3 34 3
Pollina ion 23 2 25 9
Myco hizae 3 2 5 67
*Change in he numbe o s udies and e ec sizes om he alues o iginally epo ed by Mag ach e al. (2014).
TABLE 2 To al numbe o publica ions and e ec sizes epo ed o each componen o o es agmen a ion a e combining hose p e ious epo ed by
Mag ach e al. (2014) wi h da a om he mos ecen publica ions om 2012 o 2022.
Componen o o es agmen a ion To al publica ions Inc ease (%)*To al e ec sizes Inc ease (%)*
F agmen size 69 21 99 15
Edge e ec s 38 52 50 39
F agmen isola ion 4 33 7 40
F agmen deg ada ion 11 22 12 20
*Change in he numbe o s udies and e ec sizes om he alues o iginally epo ed by Mag ach e al. (2014).
ound ha we e published om 2012 o 2022, he ea e new
publica ions).
O he o al s udies in ou combined da a se , 47 analyzed
mu ualisms and 69 analyzed an agonisms (n=116 because
some s udies epo bo h in e ac ion ypes). No new s udies
we e ound ha analyzed commensalisms (in e ac ion be ween
mu ualisms and an agonisms in which one species bene i s and
he o he species is no a ec ed), so he ca ego y was no
included.
F agmen size was he mos well-s udied componen o o -
es agmen a ion: 69 s udies and 99 in e ac ions we e obse ed.
Edge e ec was second: 38 s udies and 50 in e ac ions we e
obse ed. Fou s udies e alua ed he e ec s o isola ion asso-
cia ed wi h 7 in e ac ions. Ele en s udies add essed o es
deg ada ion associa ed wi h 12 in e ac ions.
The amoun o a ailable esea ch inc eased mo e o an ag-
onisms han mu ualisms. We analyzed 103 e ec sizes o
an agonisms and only 65 o mu ualisms. F om he numbe o
e ec sizes o iginally epo ed in Mag ach e al. (2014), he o al
inc eased by 32% o an agonisms and 12% o mu ualisms.
Fo he 6 ypes o in e ac ions, he inc ease in sample size om
he addi ion o new publica ions a ied d as ically om 3% o
s udies on seed dispe sal o 67% o s udies on myco hizae.
Fo componen s o o es agmen a ion, he numbe o in e ac-
ions epo ed in he li e a u e inc eased he mos o agmen
isola ion (40% mo e e ec sizes) and he leas o agmen size
(15% mo e e ec sizes) (Table 1,2).
Finally, sample sizes appea ed o be inc easing o s udies
assessing he impac o o es agmen a ion on species in e -
ac ions. Sample sizes epo ed by Mag ach e al. (2014) ange
om 4 o 1200, whe eas sample sizes om he new publica ions
anged om 10 o 12,604 (Appendix S5).
E ec s o o es agmen a ion on mu ualisms
and an agonisms
Suppo ing he o iginal indings om Mag ach e al. (2014),
he e ec s o o es agmen a ion emained consis en ly mo e
nega i e on mu ualisms han an agonisms (bias-co ec ed boo -
s apped CIs: mu ualisms, −1.38 o −0.35; an agonisms −0.33
o 0.60; Qbe ween =15.62, d =1, p=0.0001; an agonisms
Qwi hin =232.34, d =102, p<0.0001; mu ualisms Qwi hin =
116.36, d =64, p=0.0001) (Figu e 1).
Consis en wi h Mag ach e al. (2014), no componen o o -
es agmen a ion had a signi ican e ec on he s eng h o
an agonisms ( Qbe ween=0.73, d =3, p=0.87). In addi ion,
3 o he 4 componen s o o es agmen a ion had signi ican
nega i e e ec s on he s eng h o mu ualisms ( Qbe ween=8.93,
d =3, p=0.03). Two o he 3 signi ican componen s o o -
es agmen a ion emained consis en be ween Mag ach e al.
(2014) and he new combined publica ions: agmen deg ada-
ion and edge e ec s. Howe e , Mag ach e al. (2014) p e iously
ound ha agmen isola ion had a signi ican e ec , whe eas
we ound ha agmen size was a hi d signi ican compo-
nen o o es agmen a ion nega i ely al e ing he s eng h o
mu ualisms (Figu e 2).
The key di e ence in he new indings compa ed wi h he
p e ious indings was ha in ou esul s agmen size signi i-
can ly a ec ed mu ualisms (Qwi hin =60.31, d =40, p=0.02)
15231739, 0, Downloaded om h ps://conbio.onlinelib a y.wiley.com/doi/10.1111/cobi.14206 by Readcube (Lab i a Inc.), Wiley Online Lib a y on [22/03/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
6o 10 SIEGEL ET AL.
FIGURE 1 Impac s o o es agmen a ion on mu ualis ic and an agonis ic species in e ac ions o all publica ions (lines, bias-co ec ed boo s apped 95%
con idence in e als; do s, mean e ec size o each ype o in e ac ion).
FIGURE 2 Impac s o he 4 componen s o o es agmen a ion (edge e ec s, size, isola ion, and deg ada ion) on mu ualis ic in e ac ions in Mag ach e al.
(2014) (old publica ions) and combined new and old publica ions (lines, bias-co ec ed boo s apped 95% con idence in e als; do s, mean e ec size o each ype
o in e ac ion).
15231739, 0, Downloaded om h ps://conbio.onlinelib a y.wiley.com/doi/10.1111/cobi.14206 by Readcube (Lab i a Inc.), Wiley Online Lib a y on [22/03/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
CONSERVATION BIOLOGY 7o 10
(Appendix S2), whe eas agmen isola ion did no ha e a sig-
ni ican e ec on mu ualisms (Qwi hin =5.30, d =3, p=0.15).
Edge e ec s emained a nega i e e ec on an agonisms and
mu ualisms ( Qbe ween=5.49, d =1, p=0.02; an agonisms
Qwi hin =97.44, d =36, p<0.0001; mu ualisms Qwi hin =49.33,
d =12, p<0.0001; bias-co ec ed CIs: an agonisms −0.67 o
0.70, mu ualisms −2.63 o 0.13).
When compa ing changes in he mean e ec size be ween he
old and new publica ions, we ound a shi in he mean e ec
size o bo h an agonisms and mu ualisms (Appendix S6). Fo
an agonisms, he mean e ec size o he old publica ions was
0.23 (bias-co ec ed boo s apped CIs: −3.00 o 0.74), whe eas
he mean e ec size o he new publica ions was −0.01 (bias-
co ec ed boo s apped CIs −0.81 o 0.75). Fo mu ualisms, he
mean e ec size inc eased be ween old and new publica ions.
Old publica ions assessing mu ualisms had a mean e ec size
o −0.74 (bias-co ec ed boo s apped CIs −1.29 o −0.22),
whe eas new publica ions had a mean e ec size o −1.21 (bias-
co ec ed boo s apped CIs −2.19 o −0.10). Al hough hese
esul s showed an inc ease in he mean e ec size o mu u-
alisms and a dec ease o an agonisms, he p alues o hese
shi s be ween old and new publica ions we e no signi ican (p
>0.05).
Fo he di e en componen s o o es agmen a ion, he
e ec o agmen size on mu ualisms was signi ican ly di e -
en be ween old and new publica ions ( Qbe ween =17.5,d =
1,p<0.001).Old publica ions had a mean e ec size o 0.08
(bias-co ec ed boo s apped CIs −0.44 o 0.65), whe eas new
publica ions had a mean e ec size o −2.36 (bias-co ec ed
boo s apped CIs −3.01 o −1.83) (Appendix S7). Fo he
e ec o deg ada ion on an agonisms, edge e ec s on mu u-
alisms and an agonisms, and agmen size on an agonism, we
ound no signi ican di e ence be ween old and new publica-
ions (p>0.05). Finally, isola ion was incompa able because
he e we e no new publica ions o mu ualisms and he sam-
ple size o old publica ions o an agonisms was also oo
small o analyses (n<2). The e we e also no new pub-
lica ions on he impac o edge e ec s and deg ada ion on
mu ualisms.
E ec s o o es agmen a ion on he 6 ypes o
in e ac ions
Con a y o Mag ach e al. (2014), who ound ha agmen size
did no ha e a signi ican impac on he 6 ca ego ies o in e -
speci ic in e ac ions, he addi ion o new publica ions caused
agmen size o a y signi ican ly be ween he in e ac ions (
Qbe ween=11.15, d =5, p=0.048) and was no ably signi ican
o pa asi ism (Qwi hin =51.76, d =15, p<0.0001) (Figu e 3).
Wi h a mean posi i e e ec size o 0.5, pa asi ism inc eased as
o es agmen a ion inc eased, which indica ed highe amoun s
o pa asi ism in smalle o es agmen s. Al hough Mag ach
e al. (2014) ound ha edge e ec s a ied signi ican ly o he
di e en in e ac ions, ou esul s indica ed ha edge e ec s did
no a y signi ican ly ac oss he 6 ca ego ies o in e ac ions, o
which only 5 we e a ailable o measu emen ( Qbe ween=6.99,
d =4, p=0.14). We we e only able o analyze 19 in e ac ions
o pa asi ism, and his small sample size was likely he cause o
con idence in e als o e lapping ze o.
When we compa ed he old publica ions wi h he new, we
ound a signi ican di e ence be ween he 2 g oups o myc-
o hizae ( Qbe ween =9.68,d =1,p=0.002).The mean
e ec size inc eased om −0.80 o old publica ions (bias-
co ec ed boo s apped CIs −1.43 o 0.17) o −2.80 o new
publica ions (bias-co ec ed boo s apped CIs −2.97 o −3.38).
Fo he o he 5 ypes o in e ac ions, we ound no signi ican
di e ences be ween he new and old publica ions (Figu e 4).
Geog aphic bias
Fo es agmen a ion esea ch emained geog aphically biased.
Mag ach e al. (2014) o iginally ound ha 14 o hei s udies
we e conduc ed in B azil, he highes numbe o all coun ies
included. O he new s udies, mo e han one-qua e we e con-
duc ed in B azil, demons a ing a con inued bias o his a ea.
Howe e , he esea ch om A gen ina doubled in size (inc ease
om 4 o 8 s udies).
Publica ion bias
We ound no e idence o publica ion bias in ou da a se . The e
did no appea o be any unnel plo asymme y, con i med
by Egge ’s es o unnel plo asymme y (Egge ’s es p=
0.76). Rosen hal’s ail-sa e Ncalcula ion also con i med a lack
o publica ion bias (p<0.0001, ail-sa e N=9833 publica ions).
DISCUSSION
We ound new e idence ha mu ualis ic in e ac ions a e mo e
nega i ely a ec ed by o es agmen a ion han an agonis ic
in e ac ions. Mos o ou indings we e consis en wi h Mag ach
e al. (2014), con i ming ha global o es agmen a ion has
con as ing e ec s on di e en ypes o species in e ac ions.
Howe e , hese e ec s ended o be mo e nega i e o mu u-
alis ic in e ac ions, which could exace ba e ecosys em decay
a e agmen isola ion. A key di e ence in ou combined
esul s compa ed wi h Mag ach e al.’s (2014), which is mos
likely caused by a la ge sample size, is ha agmen size neg-
a i ely a ec ed mu ualis ic in e ac ions bu no an agonis ic
in e ac ions, such ha smalle agmen s had a disp opo ion-
a ely nega i e e ec on mu ualis ic in e ac ions compa ed wi h
la ge agmen s o wi h an agonis ic in e ac ions. This nega i e
impac o o es agmen a ion on mu ualisms was held when
we compa ed old and new publica ions because he mean e ec
size was la ge and mo e nega i e o new publica ions. We saw
his wi h he shi o myco hizal in e ac ions. When assessing
he o e all impac s o o es agmen a ion, he mean e ec size
in ensi ied om −0.80 o he old publica ions o −2.80 o he
new publica ions. The sample size o his speci ic in e ac ion
was small, bu he e we e s ill 2 key ends in he da a: an inc ease
15231739, 0, Downloaded om h ps://conbio.onlinelib a y.wiley.com/doi/10.1111/cobi.14206 by Readcube (Lab i a Inc.), Wiley Online Lib a y on [22/03/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
8o 10 SIEGEL ET AL.
FIGURE 3 Impac o agmen size on he 6 di e en ypes o species in e ac ions (lines, no mal 95% con idence in e als; a ows, bias-co ec ed
boo s apped 95% con idence in e als; do s, mean e ec sizes o each ype o in e ac ion).
FIGURE 4 Impac s o o es agmen a ion on he 6 ypes o species in e ac ions g ouped by Mag ach e al. (2014) included publica ions (old) and new
publica ions (lines, bias-co ec ed boo s apped 95% con idence in e als; do s, mean e ec size o each ype o in e ac ion).
in he nega i e e ec s o o es agmen a ion on mu ualisms
and a s eng hening o ce ain ypes o an agonisms.
Mu ualis ic in e ac ions a e hypo hesized o be less adap -
able o en i onmen al p essu es han an agonis ic in e ac ions
(B ockhu s & Koskella, 2013). This is due o a highe con-
nec edness o species in an agonis ic ne wo ks, which has
been shown ia modeling o dec ease he e ec s o dis u -
bance and help s abilize ecological communi ies (Baumga ne ,
2020). Con e sely, mu ualisms a e hough o be mo e highly
coe ol ed and in e dependen compa ed wi h an agonisms
(Kawaki a e al., 2010). Habi a dis u bance can inc ease he ul-
ne abili y o he mo e specialized mu ualisms (Aizen e al., 2012;
Kie s e al., 2010) because hei coe ol ed coun e pa s migh
no pe sis in newly dis u bed ecosys ems (De ic o e al., 2008).
In he con ex o agmen size, la ge o es agmen s can
sus ain mo e biodi e si y han small agmen s (Chase e al.,
2020; Lau ance e al., 2002, 2011; Phillips e al., 2018) and con-
ain mo e habi a o bu e agains abio ic s esso s, such as
edge e ec s (Gascon e al., 2000). La ge amoun s o biodi e si y
in la ge o es agmen s inc ease he p obabili y ha mul iple
15231739, 0, Downloaded om h ps://conbio.onlinelib a y.wiley.com/doi/10.1111/cobi.14206 by Readcube (Lab i a Inc.), Wiley Online Lib a y on [22/03/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
CONSERVATION BIOLOGY 9o 10
mu ualis ic pa ne s a e p esen and ha hei popula ions a e
s able. In deg aded habi a (e.g., small o es agmen s), he
ecosys em decays mo e quickly (Chase e al., 2020), and species
a e mo e likely o be isola ed o ex i pa ed, which esul s in
mo e biodi e si y loss and, subsequen ly, he loss o he special-
ized pa ne s needed o sus ain mu ualis ic in e ac ions (Gibson
e al., 2013; Halley e al., 2016). Mo e biodi e si y in a la ge ag-
men , o he in ac o es , also ensu es ha one pa ne does no
become oo concen a ed in an a ea, esul ing in a shi owa d
an agonis ic beha io s ha ha m he mu ualism. Thus, ela i ely
la ge agmen s compa ed wi h small agmen s a e mo e likely
o main ain mu ualis ic in e ac ions.
When compa ing he e ec s o agmen size in he old e -
sus new publica ions, we ound ha new publica ions signaled
a g ea e nega i e e ec o agmen size on mu ualis ic in e -
ac ions. Old publica ions had a mean e ec size ha was mo e
han 3 imes smalle han he new publica ions. I is well es ab-
lished ha o es s a e becoming inc easingly agmen ed, wi h
mo e han 70% o emaining o es s now wi hin 1 km o o es
edge (Haddad e al., 2015). Addi ionally, species loss in o es
agmen s can ake decades o mani es (Tilman e al., 1994;
Gibson e al., 2013; Wea n e al., 2012). This lag has he po en ial
o signi ican ly delay he ul ima e impac s o o es dis u bance
on biodi e si y. As o es agmen a ion con inues o in ensi y
(Fishche e al., 2021), ou esul s help demons a e ha mu u-
alisms a e one o he mo e sensi i e g oups acing he nega i e
impac s o habi a dis u bance.
When compa ing all publica ions, agmen size a ec ed he
s eng h o pa asi ic in e ac ions, such ha pa asi ism inc eased
in s eng h in small o es agmen s ela i e o la ge agmen s.
Habi a agmen a ion al e s he landscape o c ea e imp o ed
condi ions o some hos −pa asi e in e ac ions (F oeschke
e al., 2013; Gillespie & Chapman, 2008). Fo example, o es
agmen a ion is expec ed o inc ease he occu ence o nes
and b ood pa asi ism because gene alis nes p eda o s a e be -
e able o en e he o es h ough he adjoining de o es ed
land (Lloyd e al., 2005). Highe pa asi ism a es in small o -
es agmen s compa ed wi h la ge o es agmen s ha e also
been a ibu ed o a la ge a ie y o hos s and ood esou ces
a ailable in mo e he e ogenous, agmen ed landscapes (Mon-
many & Aide, 2009). Finally, small agmen s may also indica e
a eas o highe human ac i i y, which inc eases pa asi ism a es
in wild species (Hussain e al., 2013; Mbo a & McPeek, 2009).
Ou upda ed me a-analysis highligh s he impo ance o con-
inuously pooling and assessing da a on c i ically impo an
ecological ques ions. We ound a clea shi be ween old and
new publica ions ha showed ha inc easing global o es
agmen a ion may be al e ing key species in e ac ions. New
esea ch mus con inue o expand on indings om p e ious
analyses o e ine p io conclusions and p o ide new insigh s
(e.g., Zome e al., 2014). Fu u e in es iga ions migh conside a
b oade di e si y o in e ac ion ypes o cap u e he nuances o
changes o mu ualisms and an agonisms. Addi ional da a om
inaccessible s udies may also help bols e hese indings. Ou
esul s ea i m ha al hough o es agmen a ion has a y-
ing e ec s on di e en ypes o ecological in e ac ions; o e all,
hose impac s a e mo e nega i e o mu ualisms han an ago-
nisms. In addi ion, some an agonisms appea o be inc easing
in s eng h. The combined educ ion in mu ualis ic in e ac ions
and inc ease in an agonis ic in e ac ions may accele a e he a e
o ecosys em decay in agmen ed o es s o e ime. Such esul s
show he impo ance o eassessing ecological ques ions p e i-
ously answe ed wi h mo e limi ed da a because he impac s o
human dis u bance can shi wi h ime.
ACKNOWLEDGMENTS
This esea ch is in memo y o T. Lo ejoy o his men o ship and
encou agemen o pu sue a be e unde s anding o agmen ed
ecosys ems.
ORCID
To ah Siegel h ps://o cid.o g/0000-0003-2589-2618
Da id Lu he h ps://o cid.o g/0000-0003-3331-6186
REFERENCES
Aizen, M. A., & Feinsinge , P. (1994). Fo es agmen a ion, pollina ion, and
plan ep oduc ion in a Chaco d y o es , A gen ina. Ecology,75, 330–
351.
Aizen, M. A., Saba ino, M., & Tylianakis, J. M. (2012). Specializa ion and a i y
p edic non andom loss o in e ac ions om mu ualis ne wo ks. Science,335,
1486–1489.
Baumga ne , M. T. (2020). Connec ance and nes edness as s abilizing ac o s
in esponse o pulse dis u bances in adap i e an agonis ic ne wo ks. Jou nal
o Theo e ical Biology,486, 110073.
Be s, M., Wol , C., Ripple, W., Phalan, B., Mille s, K., Dua e, A., Bu cha , S.,
& Le i, T. (2017). Global o es loss disp opo iona ely e odes biodi e si y
in in ac landscapes. Na u e,547, 441–444.
B ockhu s , M., & Koskella, B. (2013). Expe imen al coe olu ion o species
in e ac ions. T ends in Ecology & E olu ion,28, 367–375.
Chase, J. M., Blowes, S. A., Knigh , T. M., Ge s ney, K., & May, F. (2020).
Ecosys em decay exace ba es biodi e si y loss wi h habi a loss. Na u e,584,
238–243.
Co dei o, N. J., & Howe, H. F. (2003). Fo es agmen a ion se e s mu ual-
ism be ween seed dispe se s and an endemic A ican ee. P oceedings o he
Na ional Academy o Sciences,100, 14052–14056.
C ame , J. M., Mesqui a, R. C. G., Ben os, T. V., Mose , B., & Williamson,
G. B. (2007). Fo es agmen a ion educes seed dispe sal o Duckeodend on
ces oides, a Cen al Amazon endemic. Bio opica,137, 415–423.
De ic o , V., Jullia d, R., & Jigue , F. (2008). Dis ibu ion o specialis and gene -
alis species along spa ial g adien s o habi a dis u bance and agmen a ion.
Oikos,117, 507–514.
Didham, R. K., Hammond, P. M., Law on, J. H., Eggle on, P., & S o k, N. E.
(1998). Bee le species esponses o opical o es agmen a ion. Ecological
Monog aphs,68, 295–323.
Eas e b ook, P. J., Gopalan, R., Be lin, J. A., & Ma hews, D. R. (1991).
Publica ion bias in clinical esea ch. Lance ,337, 867–872.
Egge , M. S., Smi h, G. D., Schneide , M., & Minde , C. (1997). Bias in me a-
analysis de ec ed by a simple, g aphical es . B i ish Medical Jou nal,315, 629–
634.
FAO & UNEP. (2020). The s a e o he wo ld’s o es s 2020. Fo es s,
biodi e si y and people. Rome, I aly. 188 Pages.
Fische , R., Taube , F., Mulle , M. S., G oen eld, J., Lehman, S., Wiegan, T., &
Hu h, A. (2021). Accele a ed o es agmen a ion leads o c i ical inc ease in
opical o es edge. Science Ad ances,7, eabg7012.
F oeschke, G., an de Mesch , L., McGeoch, M., & Ma hee, S. (2013).
Li e his o y s a egy in luences pa asi e esponses o habi a agmen a ion.
In e na ional Jou nal o Pa asi ology,43, 1109–1118.
Gascon, C., Williamson, B., & Da Fonseca, G. (2000). Receding o es edges and
anishing ese es. Science,288, 1356.
Gibson, L., Lynam, A. J., B adshaw, C. J. A., He, F., Bick o d, D. P., Wood u ,
D. S., Bum ungs i, S., & Lau ance, W. F. (2013). Nea -comple e ex inc ion o
15231739, 0, Downloaded om h ps://conbio.onlinelib a y.wiley.com/doi/10.1111/cobi.14206 by Readcube (Lab i a Inc.), Wiley Online Lib a y on [22/03/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
