Access to this full-text is provided by Wiley.
Content available from Ecology and Evolution
This content is subject to copyright. Terms and conditions apply.
1 of 12
Ecology and Evolution, 2024; 14:e70391
https://doi.org/10.1002/ece3.70391
Ecology and Evolution
RESEARCH ARTICLE OPEN ACCESS
Developmental Social Experience Changes Behavior in a
Threatening Environment in Corydoras Catfish
MunirSiddiqui1 | AustinChiang1 | EthanLac1 | JesseKern1 | GeraldWilkinson1 | ArneJungwirth2 |
JamesAllen3 | RivaJ.Riley1
1Department of Biology, University of Maryland, College Park, Maryland, USA | 2Konrad Lorenz Institute of Ethology, University of Veterinary Medicine
Vienna, Wien, Austria | 3Social and P ublic Health Sciences Unit, School of Health and Wellbeing, University of Glasgow, Glasgow, UK
Correspondence: Riva J. Riley (riva.riley@gmail.com)
Received: 3 May 2024 | Revised: 3 September 2024 | Accepted: 17 September 202 4
Funding: This work was supported by the President's Postdoctoral Fellowship and the AGEP Promise Academy fellowship.
Keywords: behavioral ecology| behavioral evolution| Corydoras| developmental social experience| social behavior
ABSTRACT
Coordinated responses to threats are important for predator evasion in many species. This study examines the effect of de-
velopmental social experience on antipredator behavior and group cohesion in a highly gregarious catfish that communicates
via tactile interaction, Corydoras aeneus. We reared fish either in a mixed- age group of age- matched peers and adult C. aeneu s
(mixed- age condition, or MAC), or with age- matched peers only (same- age condition, or SAC). A startle test was conducted
with small groups of subadults from either social rearing condition. Prior to any startle events, SAC subadults had increased
tactile communication compared to MAC subadults, but SAC individuals were overall less active. SAC fish exhibited a stronger
antipredator response to startles, and were more likely to freeze or take refuge in cover in response to a startle than MAC fish.
MAC fish tended to respond to startle events by maintaining or decreasing their cohesion, whereas SAC fish tended to main-
tain or increase their cohesion. These behavioral differences are attributed to MAC fish developing with group protection as a
result of shoaling with adults, resulting in reduced antipredator responses when reared with adults. This study underscores how
social context during development can be critical in shaping how individuals perceive and respond to potential threats in their
environment.
1 | Introduction
Despite the costs of social living, such as competition for food
and mates, it confers profound advantages to social animals, and
in particular improved foraging and predator evasion (Ward and
Webster2 016). To fully reap the benefits of social living, individ-
uals in groups must effectively coordinate their activities with
one another, as some degree of group coordination ultimately
provides these benefits, and improved coordination of activities
between individuals generally improves individual outcomes
within a group (Sirot and Touzalin 2009). Flocks of pigeons
exhibit highly coordinated responses to predators that utilize
self- organization rules to reduce the amount of attention and
cognitive effort individuals must invest in predator vigilance
(Papadopoulou etal.2022). In zebra finches, synchronizing re-
production with groupmates is associated with higher reproduc-
tive success (Brandl etal.2021). Across taxa and social contexts,
modulating group coordination is essential for the evolution of
social complexity (Griesemer and Shavit2023).
Two essential aspects of effective group coordination are co-
hesion and communication. Cohesion is when the average dis-
tance between group members is low, and maintaining group
cohesion following a threat is a powerful factor in group pred-
ator evasion in many species (Miller et al. 2013; Riley, Gillie,
Savage, etal.2019). Cohesive groups generally coordinate group
This is a n open access ar ticle under the terms of t he Creative Commons Attr ibution License, which p ermits use, dis tribution and repro duction in any medium, p rovided the orig inal work is
properly cited.
© 2024 T he Author(s). Ecology and Evolution publis hed by John Wiley & Sons L td.
2 of 12 Ecology and Evolution, 2024
movements more effectively, an effect demonstrated in multiple
species including many fishes (Herbert- Read et al. 2011). High
cohesion generally provides antipredator benefits, with cohe-
sive groups more likely to successfully evade predators (Sogard
and Olla 1997; Chivers et al. 1995; Viscido and Wethey 2002).
Communication between group members also facilitates group
coordination and enhances group antipredator response. Some
forms of group communication, such as alarm calls, involve
the clear transfer of information from one individual, or a sub-
set of a group, to the rest of the group (Schel etal. 2010), often
containing specific information about the potential predator
(Templeton 2005). Group communication before and during
a predation threat can be influenced by a prey species' fear of
predation, and in the Trinidadian guppy Poecilia reticulata, the
perceived risk of predation caused fish to develop stronger social
ties and higher group cohesion (Heathcote etal.2 017).
While communication and coordination are crucial to survival
and reproduction in many species, individuals generally require
developmental inputs in order to learn and express effective so-
cial coordination behaviors. In the cichlid P. taeniatus, group co-
ordination is a socia lly developed behavior, and sub- adults raise d
in isolation showed increased aggression, a delayed response to
external stimuli, less shoal cohesion, and lowered growth com-
pared to their age- matched peers raised in groups (Hesse and
Thünken 2014). In the highly social cichlid Neolamprologus
pulcher, being raised with or without adults also affects social
behavior, albeit in concert with ecological experience: fish that
grew up in mixed- age family groups in a seemingly safe envi-
ronment showed the most pro- social behavior and best integra-
tion into their social group, whereas fish that only experienced
same- age full siblings during development were less socially
competent (with predator experience altering these effects;
Fischer etal.2017). The predatory mite Phytoseiulus persimilis
exhibits a similar effect of social isolation, with mites reared in
social isolation showing lower social competence and poorer re-
productive outcomes (Schausberger, Gratzer, and Strodl 2017).
Additionally, the absence of adult fish during development
likely makes individuals worse at coordination under stress, due
to fear of predation (Kelley and Magurran2003). These studies
underscore the importance of social experiences in the develop-
ment of group coordination skills in individuals. Social rearing
generally promotes behaviors that are beneficial for group liv-
ing, such as reduced aggression, quicker responses to stimuli,
and stronger group cohesion; it also suggests that isolation can
lead to significant behavioral and growth deficits (Hayes and
Solomon2004; Chapman, Ward, and Krause2008; Liedtke and
Schneider2 017).
Our study species, Corydoras aeneus, exhibits a clear develop-
mental effect of isolation: larvae raised in isolation exhibit lower
social competence than larvae reared socially (Riley etal.2020).
In this study, we reared larvae either socially or in isolation, and
then placed larvae in groups and assessed their social interac-
tions. We found that larvae reared in isolation initiated fewer
nudges with groupmates, and furthermore, were more likely to
c- start (an involuntary antipredator response) when they phys-
ically touched others. This underscores that nudging has an in-
nate component, but nonetheless requires social experience to
properly develop. This raises interesting questions about how
different kinds of social experience impact social development.
Corydoras aeneus is a promising system for investigating ques-
tions related to social development, in part because Corydoras
catfish exhibit tactile interactions as a form of communication.
These discrete tactile interactions, termed “nudges,” have been
shown to facilitate coordination and cohesion in groups of fish
(Riley, Gillie, Savage, etal.2019). They are also deployed during
group threat responses, in which nudges serve as a form of com-
munication to coordinate group responses to a threat (Riley,
Gillie, Savage, etal.2019). In the wild, these fish have been ob-
served in group sizes as small as three (Riley 2012) and live in
mixed- age, mixed- size groups (Lambourne 1995; Riley personal
observation) that form due to their extremely low levels of ag-
gression and defensive anatomy (in contrast to many other fish
systems; Hoare etal.2000). Having a variety of sizes may make
the shoal less effective in confusing predators (as confusion ef-
fects depend on group homogeneity [Krakauer 1995]), so there
are likely benefits to overcome that disadvantage (Lambourne
1995; Riley 2012; Hoare etal. 2000). It is also likely that very
young juveniles that hatch as a group must live in juvenile- only
shoals for some time before encountering and joining a mixed-
age shoal, since C . aeneus eggs are typically laid in clusters and
parents do not directly provide parental care after eggs have
been laid (Lambourne 1995).
This paper examines the role of social rearing condition in the
environmental exploration and threat evasion effectiveness of
C. aeneus sub- adults, comparing those who are housed with
(and thus have observed and interacted with) adult fish to those
who have never had social experience with an adult and have
lived entirely with age- matched peers. The discrete, quantifiable
nature of nudging behavior allows us to quantify individual and
group social responses to threats, and this study will shed light
on how social exposure to mature individuals impacts the devel-
opment of the ecologically critical behavior of predator evasion.
We evaluated flight responses by applying a threat stimulus to
groups of subadult fish and qualitatively and quantitatively eval-
uating their response. The first category is composed of subadult
fish raised in a mixed- age social housing condition (which we
term MAC [mixed- age condition]), and the second category is
composed of subadults raised without adults and only with their
peers in a same- age social housing condition (which we term
SAC [same- age condition]). The MAC category emulates the
same social environment these fish have in the wild, while the
SAC category is socially deprived without the presence of adults.
We define a subadult C. aeneus as a fish that has attained adult
coloration, but is not yet sexually mature. Our previous study
of C . aeneus larvae showed that habituation to tactile stimulus,
and thus to receiving nudges, occurs starting in the larval stage,
which results in a gradual decrease in flight response from tac-
tile interactions as the fish becomes an adult (Riley etal.2020).
Accordingly, the subadult fish in this study had fully developed
nudging behavior. We evaluated threat responses in subadult
C. aeneus in two contexts: a context that is unfamiliar to the fish
but is not directly threatening (termed baseline in this study),
and then a second, threatening context in which fish were faced
with a direct potential threat (in the form of a startle test).
We have two hypotheses about the effect of social rearing con-
ditions on the subadult fish. The first is that the absence of adult
C. aeneus during development makes subadult fish less socially
3 of 12
coordinated. We would thus predict differences between the
baseline behavior (i.e., behavior observed before any startle
tests) in SAC and MAC groups. In particular, we expect low-
ered group cohesion and nudging in SAC fish, which have not
been exposed to socially active adults. Second, the absence of
adult C. aeneu s during development may lead to subadult fish
being unable to adequately respond to a threat. We predict that
this should result in differences in group responses to threats,
with SAC fish showing reduced group cohesion and increased
individual- level fear responses. Through this study, we aim to
elucidate the role of mixed- age social grouping during the devel-
opment of antipredator behavior and group coordination.
2 | Methods
2.1 | General Husbandry
All juvenile/subadult fish in this experiment were bred from
our adult stock population, housed in the Biology- Psychology
Building at the University of Maryland in College Park,
Maryland, USA. Our stock population consists of wild C. aeneu s
(CW097 designation) catfish caught in the Madre de Dios region
of the southern Peruvian Amazon region in October 2018. We
obtained these fish in January 2021 and they have been main-
tained in our lab since. Our adult stock population was kept
in two standard 109.78 L tan ks (76.2 0 × 30.48 × 45.72 cm). The
stocking density of these tanks was 14–15 adult fish per ~109 L
aquarium. Adult fish were fed daily with Tetra brand tropi-
cal fish granules as well as twice a week with thawed frozen
bloodworms (Chironomidae larvae), and water changes were
performed once a week from a central RO supply remineralized
with Tropic Marin Remineral Tropic to 95–100 ppm total dis-
solved solids (as in Riley, Gillie, Savage, etal.2019; Riley, Gillie,
Cat, et al. 2019). The stock tanks were kept in a temperature-
controlled room with a set 12:12 day/night cycle. Each tank was
also equipped with a power filter (Tetra brand) and two air-
driven sponge filters, as well as sand substrate, artificial plants,
and large PVC pipes for cover.
Eggs were collected from adult stock tanks in two batches, one
in January 2022 and one in March 2022; both batches were col-
lected from the same two adult stock tanks. Multiple clutches
of eggs were laid by multiple sets of parents in each adult stock
tank over 2 days per spawning event. Because of the unique
sperm- drinking copulation mechanism in these fish (Kohda
etal. 1995), we can be confident that eggs in the same clutch
(visibly distinguishable as clusters of eggs) are full siblings. On
each day of the 2- day spawning event for both batches, we put all
of the eggs laid in both adult stock tanks in a separate enclosure,
thereby mixing full sibling groups. This ensures that the degree
of relatedness did not differ when fish were later allocated to
experimental social- rearing tanks. This also simulates natural
spawning conditions, when many pairs of adults spawn around
the same time and lay eggs in the same areas.
Eggs were placed i n smal l ~22.7 L (22.23 × 50.4 8 × 20.32 cm)
aquaria and stayed in those aquaria as larvae. Larvae were fed
daily with crushed TetraColor tropical fish granules and thawed
frozen baby brine shrimp, and water changes were performed
every other day with the same procedure described above for the
adult stock fish. After 8–9 weeks in the small aquaria (9 weeks
for the first batch, 8 weeks for the second batch), larvae had de-
veloped to the juvenile stage of their life cycle and had reached
a minimum standard length of 1 cm. They were then randomly
placed in one of the four 56.78 L social rearing enclosures
(30.23 × 46.4 8 × 68.58 cm), described below.
2.2 | Experimental Husbandry
Four social rearing enclosures were used during each batch of
this experiment, two SAC enclosures and two MAC enclosures,
for eight total enclosures across both experimental batches. In
each batch, two enclosures were randomly designated as SAC
housing, and two were randomly designated as MAC housing.
Juveniles were randomly allocated to enclosures to eliminate
differences in the degree of relatedness between SAC and MAC
enclosures. The two SAC enclosures in each batch contained
15–18 juvenile fish from the same larval- rearing aquaria. The
two MAC enclosures in each batch contained 13 juvenile fish
and 5 adult C. aeneu s. Fish were kept in the enclosures for
1 month, until they reached the subadult stage of development
90 days after hatching. The presence or absence of adult C. ae-
neus during development distinguished the two treatment cate-
gories. Fish were fed daily with TetraColor tropical fish granules
and thawed frozen bloodworms twice a week. Water changes
were performed twice a week from a central RO supply rem-
ineralized with Tropic Marin Remineral Tropic to 95–100 ppm
total dissolved solids. The experimental enclosures were kept in
a temperature- controlled room with a set day/night cycle (the
same room and parameters as the adult social- rearing tanks de-
scribed above). Each tank was also equipped with two sponge
filters, and contained cover in the form of two terracotta pots
to minimize stress. Additionally, a lid composed of plastic egg
crate was placed on top of each aquarium to prevent fish from
jumping out while allowing light in.
2.3 | Experimental Housing
Across both batches of the experiment, we collected data from
28 groups of three juvenile fish randomly chosen from the same
enclosures (so that all individuals in each group were familiar
with one another) were placed in 37.85 L test tanks initially com-
posed of half water from adult stock tanks and half remineral-
ized RO water. Three test tanks were used simultaneously, each
with the same water level and each containing one small sponge
filter. The sponge filter was turned off during filming, but was
left in the test tank to provide cover. Cameras were mounted
above each tank at a uniform height for filming fish behavior.
Each of the three tanks also had one vertically oriented lamp
8–12 in. away from the tank to ensure even lighting.
Each test tank was filmed for 2 h. Starting 30 min after filming
began, a stimulus, in the form of a single tap on the glass of
the test tank with one finger, was applied to each tank in 15-
min intervals. Tap tests like the one we employ in this study
have been used to assess antipredator responses and responses
to potential danger in other fish systems (Gotanda etal.2012;
Chanin etal.2012). We refer to these taps as “startles” hereaf-
ter. After 2 h of filming, the fish finished the experiment. The
4 of 12 Ecology and Evolution, 2024
three fish from each test tank were then placed into a social
housing tank designated for fish who had completed the exper-
iment, thus avoiding multiple testing of the same individuals.
When all fish were transferred, a 50% water change in each
test tank was performed using remineralized RO water. This
process was then repeated for each experimental group, result-
ing in 28 total trials (NS AC = 13, NMAC = 15) that were analyzed
across both batches.
We excluded groups (four groups, three SAC groups, and one
MAC group) that exhibited no movement during the baseline
period (described below) or experienced camera errors that pre-
vented filming (one MAC group); so 33 groups of three fish were
filmed initially.
2.4 | Data Collection and Video Analysis
We analyzed two aspects of the videos: (1) a baseline period,
defined as the 5 min before the first startle event (starting after
~25 min of filming) and (2) the behavior of the fish 3 s before and
3 s after each startle event.
For the baseline period, we scored the videos for nudges, and we
collected information on the total number of nudges performed,
following the protocol in Riley, Gillie, Savage, etal.(2019). This
analysis includes nudges that take place off of the bottom, that
is, during short periods of free swimming in the water column.
The fish had 30 min to acclimate before we began the startle
stimuli, and the last 5 min of this pre- startle acclimation was
scored for a baseline rate of nudges.
In addition to scoring nudges, we recorded some basic activity
and sociality level information during the baseline period. We
recorded the cumulative amount of time (in seconds) of all of
the possible combinations of activity among the fish (no move-
ment, one individual moving, two individuals moving, all mov-
ing), the cohesion of the fish (whether all three were apart, all
three were together, or two fish were in a group separate from
the third fish (Appendix1)), and the amount of time in seconds
that “wall- surfing” behavior was observed. Cohesion was de-
fined as being within two body lengths of another fish (Riley,
Gillie, Savage, etal.2019). Wall surfing behavior occurs when
the fish move off the bottom of the tank and swim up and down
the sides, and is considered a non- standard behavior and a po-
tential indicator of stress (Martins etal. 2012). As part of re-
cording this information, we assumed that, as a default, all fish
were actively moving, all fish were grouped together, and that
wall surfing was not present. This default coordination state is
their most commonly observed state in the wild (Riley, personal
observation). Activity and wall surfing was measured with a 5 s
threshold, where a non- default behavior (i.e., fish were still/not
together or wall surfing behavior) was only recorded if it oc-
curred for 5 or more consecutive seconds.
To measure the cohesion of the fish, we calculated a “cohesion
fish seconds” metric. To do this, we multiplied the amount of
time each group spent all together (in a group size of 3) by three
and the amount of time each group spent with two fish together
and one apart by 2. We then added these products t ogether for the
following overal l equat ion: (3 × all three f ish together) + (2 × two
fish together, one apart) = cohesion fish seconds.
To measure the activity of the fish, we calculated an “active fish
seconds” metric, derived from summing the products of each
possible combination times the number of fish that were moving
in that combination. For example, when all three fish are mov-
ing, we multiply the amount of time they spend in that combi-
nation by 3, giving us the total number of time spent by all three
fish in that configuration. We multiplied the amount of time
two fish were moving by 2, and the amount of time just one fish
was moving by 1. We then added these products together, for
this overall equation: ([3 × all three fish moving] + [2 × two fish
moving] + [one fish moving]) = active fish seconds. This takes
into account the movement of every single fish, rather than just
analyzing the average of each group of 3.
To measure the wall surfing, we calculated a “wall surfing fish
seconds” metric in a manner similar to “active fish seconds.”
We calculated this by multiplying the amount of time all three
fish were wall surfing by 3, multiplying the amount of time two
fish were wall surfing by 2, and multiplying the amount of time
just one fish was wall surfing by 1, and then adding the products
together. This takes into account the movement of every single
fish to get an overa ll measure of wall surf ing for the entire group.
To score the behavioral responses to startles, we scored the same
basic activity information as described above at 3 s before each
startle and at 3 s after the startle. In addition, we recorded the
number of active fish and the number of fish that were in cover
3 s before and after the startle, as well as how many fish moved,
froze, or performed a c- start maneuver during each startle. The
c- start is an involuntary response where the fish adopts a c
shape with its body, then straightens out to travel in a direction
where the fish can escape their perceived threat (Domenici and
Blake1997; see Figure1).
2.5 | Statistics
All data analysis was conducted in R (2022.07.2, R Core Team
2021). For continuous data, we assessed data distributions vi-
sually and used nonparametric tests when data were not nor-
mally distributed. To compare total nudges between SAC and
MAC subadults, we used a Wilcoxon signed rank test. We used
a two- sample t- test to compare cohesion levels between SAC
and MAC fish. To compare activity levels between SAC and
MAC subadults, we used a Wilcoxon signed rank test. To com-
pare wall surfing tendencies, we also used a Wilcoxon signed
rank test.
We then analyzed the group behavior associated with startle
events. Because we performed four startle events, we checked
for an effect of habituation on cohesion status before and after
the startle, the number of c- starts following a startle, the number
of active fish before and after the startle, and the number of fish
that froze in response to the startle. We used a Spearman rank
correlation to assess if there was any association between each
of these measures and the startle event. We performed these
tests for the whole dataset (SAC and MAC fish combined), as
5 of 12
well as for each of the two rearing conditions separately. These
additional steps were included to ensure that we did not over-
look potential differences between SAC and MAC fish with re-
gard to habituation effects.
We first analyzed cohesion tendencies testing for the effects of
habituation on cohesion status. We compared the number of
fish in each possible cohesion status (all together, two together
and one apart, all apart) three 3 s before the startle event and 3 s
after the startle event in SAC and MAC fish. To assess whether
MAC and SAC fish differed in their cohesion tendencies fol-
lowing a startle event, we used a multinomial regression using
the nnet package in R; the response variables were the cohesion
state after a startle event (i.e., from all fish apart, all fish to-
gether, etc.) and the explanatory variable was the cohesion state
before and the social rearing condition (MAC or SAC). Startle
events are expected to be experienced frequently in nature, and
the time between events in our experiment is such that we ex-
pect no cumulative effect of the startle number (and found no
effect of habituation). Therefore, in this analysis we treat each
startle event as independent (rather than analyzing them as re-
peated trials). We compared models with and without the social
rearing condition using the Akaike information criterion (AIC)
to assess whether these transition tendencies differed between
the social rearing conditions (MAC and SAC). For the full
model including the social rearing condition, we found the fol-
lowing coefficients in a multinomial logistic regression model
(where we take cohesion before = 1 as our baseline).
Final cohesion 2 3
Cohesion before 0.31 9.90
House cond. 0.01 26.10
Cohesion before × House
cond.
0.28 −7.98
Because we had multiple startle events for each group, we used
a mixed- effect approach to analyze the number of fish that c-
started following a scare, the number of fish that were active
before and after a startle event, the number of fish in cover be-
fore and after a startle event, and the number of fish that froze
before and after a startle event. Linear mixed- effects models
(LME) and generalized linear mixed- effects models (GLMM)
were fitted using the lme4 package (Bates etal.2015). Data dis-
tributions were initially assessed visually, and model diagnos-
tics were subsequently checked to assure appropriate fits and
then verified using the DHARMa package in R (Hartig2022).
We fitted LMEs for normally distributed data (arcsin(sqrt) trans-
formed proportion of c- starts in each startle event). For propor-
tion data that could not be transformed to meet the assumptions
of an LME, we fitted GLMMs with a binomial distribution. We
used the optimx package (Nash and Varadhan2011) to modify
the optimizer function in binomial GLMMs when necessary to
ensure model convergence.
We used a linear mixed- effect model to compare the proportion
of fish that exhibited a c- start in each startle in MAC versus SAC
fish. We used the arcsin(sqrt) transformation on the proportion
of fish in a group that c- started following a startle (number of
fish that c- started divided by 3) as the response variable. The
fixed effects were the social rearing condition (MAC or SAC)
and startle event (1–4). The random effect was the group ID.
We used a binomial generalized linear mixed- effect model to
compare the proportion of active fish in each group before and
after a startle in MAC versus SAC fish. The proportion of fish
that were active was the response variable. The fixed effects
were the interaction between the social rearing condition and
timepoint (before or after the startle), and startle event. The ran-
dom effect was group ID. We used the optimx “L- BFGS- B” opti-
mizer to ensure model convergence.
We used a binomial generalized linear mixed- effect model to
compare the proportion of fish in cover before and after a startle
in MAC versus SAC fish. The proportion of fish in cover was the
response variable. The fixed effects were social rearing condi-
tion, timepoint (before or after the startle), and startle event. The
random effect was group ID.
We used a binomial generalized linear mixed- effect model to
compare the proportion of fish that exhibited a c- start in each
startle. We used the proportion of fish in a group that c- started
following a startle (number of fish that c- started divided by 3) as
the response variable. The fixed effects were the social rearing
condition (MAC or SAC) and startle event (1–4). The random
effect was the group ID.
FIGUR E | Diagram of a c- start threat response in Corydoras aeneus larvae.
6 of 12 Ecology and Evolution, 2024
We used the emmeans package in R (Lenth 2020) to conduct
posthoc comparisons whenever an interaction between fixed ef-
fects or startle event were significant predictors in our models.
3 | Results
3.1 | During Baseline
3.1.1 | Nudges
SAC groups exhibited significantly higher rates of nudging than
MAC groups (Wilcox test, W = 46. 5, p = 0 .020, Figure2a).
3.1.2 | Coordination
3.1.2.1 | Cohesion. MAC fish were less cohesive than SAC
fish, although this difference failed to be statistically significant
(two- sample t- test, t = −1.9, df = 26, p- value = 0.073, Figure2b).
3.1.2.2 | Activity Levels. SAC groups exhibited signif icantly
lower levels of activity than MAC groups (Wilcox test, W = 157,
p = 0.005, Fig ure2c).
3.1.2.3 | Wall Surf ing. There was no difference in wall
surfing tendencies based on the rearing conditions (two- sample
t- test, t = 0.6, p = 0.524, Figu re2d).
3.2 | Response to Startles
We found no association between any of the variables of interest
and startle event (spearman rank correlation, all p > 0 .2 0).
In Figure 3, we plot the probability of transitioning between
each cohesion state after a scare event. The probability is cal-
culated as the number of those in each final cohesion state,
compared to the total number that started in the cohesion state
before, for example, Prob cohesion after = 2 = (Number in cohe-
sion after = 2)/(“Number in cohesion after = 1” + “Number in
cohesion after = 2” + “Number in cohesion aft er = 3”). T hese are
compared for each social rearing condition.
Including social rearing condition (2 level factor: SAC or MAC)
in the model of changes to group cohesion in response to startles
improved model fit (AIC = 202 vs. AIC = 215). This difference in
AIC of 13 provides evidence that a model without housing condi-
tion has essentially no support (Burnham and Anderson2004).
SAC fish generally tended to improve cohesion in response to a
startle (Table1; 19/52 instances of increasing cohesion), whereas
MAC fish tended to reduce it (Table2; 15/60 instances of reduc-
ing cohesion). The same number of MAC groups increased co-
hesion as decreased (15 and 15), but more SAC groups increased
cohesion than decreased (19 vs. 8). Table entries labeled colored
blue represent groups that exhibited greater cohesion after the
startle than before, while entries colored red represent groups
that exhibited lesser cohesion after the startle than before.
FIGUR E | Data from the baseline period (exploration of the unfamiliar environment). (a) The number of front nudges in MAC versus SAC groups.
(b) The cohesion fish seconds in MAC versus SAC groups. (c) Active fish seconds in MAC groups versus SAC groups. (d) Fish seconds wall surfing in
MAC versus SAC groups. For all boxplots, the box hinges represent the interquartile range, IQR (first to third quartiles) and whiskers represent 1.5IQR.
7 of 12
Neither social rearing condition (F = 1.2 , p = 0.278) nor startle
event (F = 0.9 7, p = 0.327) were significant predictors of the pro-
portion of fish in a group that exhibited a c- start response to a
threat event (see Figure4).
Both startle event (LRT = 8.4, p = 0.038) and the interaction be-
tween the social rearing condition and timepoint (LRT = 5.2,
p = 0.022) were significant predictors of the number of active
fish per group. Posthoc analysis shows that MAC fish exhib-
ited a low number of fish that were active both before and after
a scare, whereas SAC fish were significantly more active be-
fore a startle than after a startle (see Figure5a,b). The number
of active fish per group per startle event significantly differed
between startles 2 and 3, but not any other combination.
Social r earing (LRT = 10, p = 0. 002), timepoint (LRT = 31.8,
p < 0.00 01), and startle ev ent (LRT = 5.4, p = 0.021) were
all significant predictors of the number of fish in cover (see
Figure5c,d). The number of fish in cover per startle event sig-
nificantly differed between startles 1 and 2 and 1 and 3, but
not any other combination.
Startle event was not a significant predictor of the proportion
of fish that exhibited a freeze response (LRT = 2.3, p = 0.130).
Social rearing condition, however, was a significant predictor
(LRT = 4.2 , p = 0 .041, Figure6).
4 | Discussion
Our results show that at a baseline (i.e., when fish were ex-
ploring an unfamiliar environment prior to any startle events),
SAC subadults have increased nudging behavior compared to
their MAC peers, but are less active. In terms of startle events,
we saw no effect of habituation across the four startle events
in this experiment: all correlations between our response
variables and startle event were not significant, and in models
where startle event was a significant predictor of the response
variable, there was no consistent trend that indicates habit-
uation (i.e., the differences between startle events seemed to
be random, and not a systematic change as startle events pro-
gressed) and SAC fish were more likely to freeze after a startle
than their MAC peers. Additionally, group cohesion before
and after four startle events was significantly different, with
SAC fish increasing cohesion more than MAC fish when being
startled. SAC fish also exhibited a higher degree of antipreda-
tor behavior in response to startle events, were more likely to
be in cover both before and after a startle, as well as to freeze
in response to startles, which is a common antipredator be-
havior in these fish (Riley, Gillie, Savage, etal.2019). These
results ran counter to our original hypotheses, with the SAC
individuals being more interactive and cautious than MAC in-
dividuals, despite being socially deprived of interactions with
adult fish. These unexpected findings highlight the diverse
ways developmental experience modifies behavior, with im-
portant ecological consequences.
Our hypothesis that MAC fish would exhibit heightened anti-
predator responses stems from the fact that young animals liv-
ing in mixed- age groups often acquire critical life skills from
observing adults (van Schaik 2010). The locality that our study
population originates does contain several predator species
(Riley, 2012), and this population of fish has a variety of behav-
ioral responses to predation pressure. Accordingly, we hypothe-
sized that juvenile C. aeneus would learn antipredator behaviors
(including surveillance) from observing adult groupmates. Our
results that MAC fish exhibited a lower degree of caution and
responses to startle seem to run counter to this, but nonethe-
less, both phases of our study indicate that the behavioral
FIGUR E | The probability of transitioning between cohesion states
based on the initial cohesion state and social rearing condition.
TABLE | Cohesion transitions of groups in SAC fish 3 s before and
after startle events.
After the Startle
All
apart
Two together,
one apart
All
together
Before the Startle
All apart 5 8 2
Two together,
one apart
614 9
All together 0 2 6
Note: Purple cells indicate the number of groups that maintained a certain
level of cohesion before/after startles, red cells indicate groups that decreased
cohesion levels, and blue cells represent groups that increased cohesion levels.
TABLE | Cohesion transitions of groups in MAC fish 3 s before and
after startle events.
After the Startle
All
apart
Two together,
one apart
All
together
Before the Startle
All apart 13 15 0
Two together,
one apart
11 15 0
All together 1 3 2
Note: Purple cells indicate the number of groups that maintained a certain
level of cohesion before/after startles, red cells indicate groups that decreased
cohesion levels, and blue cells represent groups that increased cohesion levels.
8 of 12 Ecology and Evolution, 2024
development of young fish is influenced by the presence of adult
conspecifics. The effect we observe is that, instead of acquiring
enhanced antipredator skills, adult groupmates modify the be-
havior of young fish by seemingly relaxing their antipredator
responses.
The behavior of fish during baseline indicates that SAC fish
are more cautious than their MAC peers. SAC individuals ex-
hibited heightened nudging behavior compared to their MAC
peers, despite markedly lower activity levels: this implies that
SAC individuals devot e much more of their active time to nudg-
ing. As nudging facilitates social coordination in these fish
(Riley, Gillie, Cat, etal.2019), our results support the idea that
SAC fish devote more energy to social coordination and less to
exploration. In addition, the lowered activity of SAC fish sug-
gests heightened attention to minimizing conspicuousness:
FIGUR E | The number of c- start responses at each startle for MAC and SAC groups.
FIGUR E | (a, b) Number of fish per group, either SAC or MAC condition, that were active 3 s before (a) and after (b) a startle. (c, d) The number
of fish per group, either SAC or MAC condition, that were in cover both before (c) and after (d) a startle event.
FIGUR E | The number of fish per group, either SAC or MAC
condition, that exhibited a freeze response after a startle event.
9 of 12
C. aeneus often deploys a freeze response to take advantage
of their cryptic coloration to escape detection (Riley, Gillie,
Savage, et al. 2019), and freezing and lowered activity as an
antipredator response strategy has been shown to be present
in other systems such as black carp Mylophar yngodon piceus
(Tang etal.2017). An additional reason for the decreased ac-
tivity among SAC fish may be attributed to predation pres-
sure leading prey species to conserve energy for predator
detection, an effect seen in other species, including Brown
Trout (Bachman1984). MAC fish may not exhibit this energy-
conserving pattern of behavior due to the group security that
they experienced during development: adults maintaining
vigilance within a shoal setting may allow young fish in that
shoal to forage without the need to reduce activity to devote
energy to antipredator vigilance, a phenomenon that has been
observed in meerkats (Santema and Clutton- Brock2013). This
is consistent with previous studies that show that small indi-
viduals benefit from shoaling with larger conspecifics in many
species, including zebrafish (Aslanzadeh etal.2019) and prai-
rie voles (Solomon1993). In addition to their larger size, adult
C. aeneus have fully developed armor and locking, venomous
spines, which may further deter predators and offer protec-
tion to more vulnerable juveniles (Riley, 2012). The Corydoras
genus also exhibits Mullerian co- mimicry between related
species, which suggests that predators may learn to recog-
nize that Corydoras catfish are generally poor prey items,
and juveniles benefit from associating with larger Corydoras,
which are more easily perceived and avoided (Alexandrou
etal. 2011). Given that Corydoras species do not exhibit ag-
gressive behavior (Riley, 2012), the costs to juveniles of asso-
ciating with older, larger fish are limited, further emphasizing
the benefits to young fish of shoaling with more visible adults,
whose appearance may deter predators and provide protection
to juveniles and subadults that can then forage more freely.
These benefits also shed light on the higher risk- taking of
MAC fish and higher caution of SAC fish during the startle
events. MAC fish were less likely than SAC fish to be in cover
both before and after startle events, and SAC fish were simi-
larly less likely to be active before the startle and more likely to
freeze in response to the startle. Given that freezing is a com-
mon antipredator response in these fish (Riley, Gillie, Savage,
etal.2019), we conclude that the SAC fish are both more cau-
tious during exploration and more likely to adopt behaviors
to be less conspicuous to predators in response to a startle.
This is also supported by the result that SAC fish tended to
remain in cover before a startle. Without the direct perception
of a threat, the tendency of SAC fish to remain in the cover is
suggestive of a higher degree of cautious behavior, as is their
tendency to increase their cohesion, given that increasing co-
hesion is a common response to potential threats across taxa
(Brown etal.2001; Viscido and Wethey2002). For SAC fish,
upregulating nudging and downregulating activity may be a
more advantageous way of balancing the risks of predation
versus the benefits of active social coordination. For MAC
subadults, the benefits of larger, more experienced conspecif-
ics may lead to the upregulating of exploration and foraging
behaviors, instead of increased social transmission of anti-
predator behaviors. Our previous work showed that nudging
is an innate behavior in these fish, but is strongly inf luenced
by social conditions during development (Riley et al. 2020);
accordingly, the SAC fish's increased nudging during baseline
and cohesion during startle events suggests that antipredator
behaviors have a substantial innate component, but individual
behavioral patterns are inf luenced by social development.
While the effect of the social rearing condition is apparent in
both the baseline and startle events, our results demonstrate
that the social rearing condition did not cause pathological
levels of stress or aberrant behaviors in either condition during
either phase of our study. The lack of a significant difference
in wall- surfing behavior during baseline between rearing
conditions supports that neither MAC nor SAC fish exhibited
higher levels of stress in response to a novel environment, but
rather that SAC fish were responding with increased caution.
Furthermore, the lack of difference in c- start frequency during
startle events between MAC and SAC fish implies that SAC
fish are not inherently more reactive to potential threats. These
commonalities in wall- surfing and c- start frequency show that
the observed differences between SAC and MAC fish reflect a
difference in caution and antipredator responses, and not that
SAC fish were adversely affected by the experimental design.
The results discussed above shed light on how a species' social
behavior can profoundly affect its ecology. For Corydoras cat-
fish, the combination of extremely low levels of aggression and
defensive anatomy in adults apparently leads to strong benefits
from mixed- age, mixed- size shoals (in contrast to many other
fish systems; Hoare etal. 2000), which likely accrue in direct
(to juveniles, who receive protection and improved foraging)
and indirect (an adult's offspring benefit from the presence of
adults) ways. By understanding how these fish adjust their be-
havior based on the relative vulnerability of their shoals, we can
gain insights into their survival strategies and their ability to
adapt to changing environmental conditions. While fish live in
mixed- age shoals in the wild, it is very much in the realm of
possibility that very young juveniles hatch as a group and live
in a juvenile- only shoal for some time before encountering and
joining a mixed- age shoal, since C. aeneu s eggs are typically
laid in clusters and parents do not directly provide parental care
(Lambourne 1995). Consequently, adjusting behavior based
on the relative vulnerability of juvenile- only or subadult- only
groups compared to mixed- age groups, where the experience
and physical presence of larger adults confers protection to
younger fish, is certainly adaptive. Continued assessment of the
role of social learning on the development of nudge coordina-
tion and group shoaling behavior in C. aeneus would add insight
into the ecodevelopmental effects of social behavior. Further
work focusing on the behavior of socially deprived individuals
(i.e., individuals who have had highly constrained social expe-
rience) can help us understand the effects of social deprivation
on behavior and survival. This is particularly relevant in the
context of habitat fragmentation or population declines, where
individuals may be more likely to experience social isolation.
Further work on how adults and juveniles interact would shed
light on the behavioral mechanisms that drive the increased
exploration and lower response to potential threats in MAC
fish. Corydoras is an extremely promising system for inves-
tigating the developmental effects of the social rearing en-
vironment, and furthermore, how social ecology can drive
behavioral evolution.
10 of 12 Ecology and Evolution, 2024
Author Contributions
Munir Siddiqui: conceptualization (lead), data curation (equal), for-
mal analysis (supporting), investigation (lead), methodology (lead),
supervision (supporting), visualization (supporting), writing – original
draft (equal), w riting – review and editing (equal). Austin Chiang:
conceptualization (supporting), investigation (equal), methodology
(equa l). Ethan Lac: data curation (equal), methodology (supporting),
writing – original draft (supporting), writing – review and editing
(sup port in g). Jesse Kern: conceptualization (supporting), investiga-
tion (equal), methodology (supporting). Gerald Wilkinson: funding
acquisition (supporting), supervision (supporting), writing – original
draft (supporting), writing – review and editing (supporting). Arne
Jungwirth: conceptualization (supporting), data curation (support-
ing), formal analysis (supporting), methodology (supporting), writing –
original draft (equal), writing – review and editing (supporting). James
Allen: data curation (supporting), formal analysis (lead), writing –
original draft (supporting), writing – review and editing (supporting).
Riva J. R iley: conceptualization (equal), data curation (lead), formal
analysis (lead), funding acquisition (lead), investigation (equal), meth-
odology (equal), project administration (lead), supervision (lead), visu-
alization (lead), writing – original draft (equal), writing – review and
editing (equal).
Acknowledgments
The authors wish to thank Dr. Danielle Adams and Dr. Kimberly
Paczolt for their advice about experimental design and analysis, as well
as Mr. John Riley for assistance in husbandr y considerations.
Conflicts of Interest
The authors declare no conflicts of interest.
Data Availability Statement
All data, as well as the R code we used in our multinomial model, are
stored at Dryad and can be found here: https:// datad ryad. org/ stash/
share/ gKyMx MASkI CCiiq qcwYv titCK nja- QPulN mUR8y ECP0.
References
Alexandrou, M. A., C. Oliveira, M. Maillard, etal. 2011. “Competition
and Phylogeny Determine Community Structure in Müllerian Co-
Mimics.” Nature 469, no. 7328: 84–88.
Aslanzadeh, M., K. Ariyasiri, O.- H. Kim, et al. 2019. “The Body Size
of Stimulus Conspecifics A ffects Social Preference in a Binary Choice
Task in Wild- Type, but Not in Dyrk1aa Mutant, Zebrafish.” Zebrafish
16, no. 3: 262–267.
Bachman, R. A. 1984. “Foraging Behavior of Free- Ranging Wild and
Hatchery Brow n Trout in a Stream.” Transactions of the American
Fisheries Society 113, no. 1: 1–32.
Bates, D., M. Maechler, B. Bolker, et al. 2015. “Package ‘lme4’.”
Convergence 12, no. 1: 2.
Brandl, H. B., S. C. Griffith, D. R. Farine, and W. Schuett. 2021. “Wild
Zebra Finches That Nest Synchronously Have Long- Term Stable Social
Tie s.” Journal of Animal Ecology 90, no. 1: 76–86.
Brown, G. E., J. C. Adrian, T. P. Jr, and D. P. Chivers. 2001. “Fathead
Minnows Learn to Recognize Predator Odour When Exposed
to Concentrations of Artificial Alarm Pheromone Below Their
Behavioural- Response T hreshold.” Canadian Journal of Zoolog y 79, no.
12: 2239–2245.
Burnham, K. P., and D. R. Anderson. 2004. “Multimodel Inference:
Understanding AIC and BIC in Model Selection.” Sociological Methods
& Research 33: 261–304.
Chanin, S., C. Fryar, D. Varga, etal. 2012. “Assessing Startle Responses
and Their Habituation in Adult Zebrafish.” Zebrafish Protocols for
Neurobehavioral Research 66: 287–300.
Chapman, B. B., A. J. W. Ward, and J. Krause. 2008. “Schooling
and Learning: Early Social Environment Predicts Social Learning
Ability in the Guppy, Poecilia reticulata.” Animal Behaviour 76, no.
3: 923–929.
Chivers, D. P., G. E. Brown, R. Jan, and F. Smith. 1995. “Familiarity
and Shoal Cohesion in Fathead Minnows (Pimephales promelas):
Implications for Antipredator Behaviour.” Canadian Journal of Zoology
73, no. 5: 955–960.
Domenici, P., and R. W. Blake. 1997. “The Kinematics and Performance
of Fish Fast- Start Swimming.” Journal of Experimental Biology 200, no.
8: 1165–1178.
Fischer, S., L. Bohn, E. Oberhummer, C. Nyman, and B. Taborsky. 2017.
“Divergence of Developmental Trajectories Is Triggered Interactively
by Early Social and Ecological Experience in a Cooperative Breeder.”
Proceedings of the National Academy of Sciences of the United States of
America 114, no. 44: E9300 –E9307.
Gotanda , K. M., E. E . Reardon, S. M. C. Mur phy, and L. J. Chapman. 2 012.
“Critical Swim Speed and Fast- Start Response in the African Cichlid
Pseudocrenilabrus multicolor Victoriae: Convergent Performance in
Divergent Oxygen Regimes.” Canadian Journal of Zoology 90, no. 5:
545–554.
Griesemer, J., and A. Shavit. 2023. “Scaffolding Individuality:
Coordination, Cooperation, Collaboration and Community.”
Philosophical Transactions of the Royal Society, B: Biological Sciences
378, no. 1872: 20210398.
Hartig, F. 2022. “DHARMa: Residual Diagnostics for Hierarchical
(Multi- Level/Mixed) Regression Models.” R Package Version 0.4.6.1.
https:// github. com/ flori anhar tig/ dharma.
Hayes, L . D., and N. G. Solomon. 200 4. “Costs and Benefit s of Communal
Rearing to Female Prairie Voles (Microtus ochrogaster).” Behavioral
Ecology and Sociobiology 56: 585–593.
Heathcote, R. J. P., S. K. Darden, D. W. Franks, I. W. Ramnarine, and D.
P. Croft. 2017. “Fear of Predation Drives Stable and Dif ferentiated Social
Relationships in Guppies.” Scientific Reports 7, no. 1: 41679.
Herbert- Re ad, J. E., A. Perna , R. P. Mann, T. M. Schaer f, D. J. T. Sumpter,
and A. J. W. Ward. 2011. “Inferring the Rules of Interaction of Shoaling
Fish.” Proceedings of the National Academy of Sciences of the United
States of America 108, no. 46: 18726–18731.
Hesse, S., and T. Thünken. 2014. “Growth and Social Behavior in a
Cichlid Fish A re Affected b y Social Reari ng Environment and Kin ship.”
Naturwissenschaften 101: 273–283.
Hoare, D. J., J. Krause, N. Peuhkuri, and J.- G. J. Godin. 20 00. “Body
Size and Shoaling in Fish.” Journal of Fish Biology 57, no. 6: 1351–1366.
Kelley, J. L., and A. E. Magurran. 2003. “Learned Predator Recognition
and Antipredator Responses in Fishes.” Fish and Fisheries 4, no. 3:
216–226.
Kohda, M., M. Tan imura, M. Kikue- Nakamura, and S. Yamagishi. 1995.
“Sperm Dr inking by Female Catf ishes: A Novel Mode of Insemination.”
Environmental Biology of Fishes 42: 1– 6.
Krakauer, D. C. 1995. “Groups Confuse Predators by Exploiting
Perceptual Bottlenecks: A Connectionist Model of the Confusion
Ef fect.” Behavioral Ecology and Sociobiology 36: 421–429. https:// doi.
org/ 10. 10 07/ BF001 77338 .
Lambourne, D. 1995. Corydoras and Catf ish: An Aquarist's Handbook.
Blandford Press.
Lenth, R. 2020. “Emmeans: Estimated Marginal Means, Aka Least-
Squares Means.” R Package Version 1.4.7. https:// CRA N. R- proje ct. org/
packa ge= emmeans.
11 of 12
Liedtke, J., and J. M. Schneider. 2017. “Social Makes Smart: Rearing
Conditions Affect Learning and Social Behaviour in Jumping Spiders.”
Animal Cognition 20, no. 6: 1093–1106.
Martins, C. I. M., L . Galhardo, C. Noble, et al. 2012. “Behavioural
Indicators of Welfa re in Farmed Fish.” Fish Ph ysiology and Bi ochemistr y
38: 17–41.
Miller, N., S. Garnier, A. T. Hartnett, and I. D. Couzin. 2013. “Both
Information and Social Cohesion Determine Collective Decisions in
Animal Groups.” Proceedings of the National Academy of Sciences of the
United States of America 110, no. 13: 5263–5268.
Nash, J. C., and R . Varadhan. 2011. “Unif ying Optimization Algorithms
to Aid Software System Users: Optimx for R.” Journal of Statistical
Software 43, no. 9: 1–14.
Papadopoulou, M., H. Hildenbrandt, D. W. E. Sankey, S. J. Portugal,
and C. K. Hemelrijk. 2022. “Self- Organization of Collective Escape in
Pigeon Flocks.” PLoS Computational Biology 18, no. 1: e1009772.
R Core Team. 2021. R: A Language and Environment for Statistical
Computing. Vienna, Austria: R Foundation for Statistical Computing.
https:// ww w. R- proje ct. org/ .
Riley, R. J. 2012. Watching the Watchmen: Evidence for Mixed- Species
Associations and a Novel Social Behavior in Corydora s. Cambridge, MA:
B.A., Harvard University.
Riley, R. J., E. R. Gillie, J. L. Savage, N. J. Boogert, A. Manica, and A.
Jungwirth. 2019. “The Role of Tactile Interactions in Flight Responses
in the Bronze Cory Catfish (Corydoras aeneus).” Ethology 125: 810–820.
Riley, R. J., E. R. Gillie, H. Cat, R. A. Johnstone, N. J. Boogert, and M.
Andrea. 2019. “Coping With Strangers: How Familiarity and Active
Interactions Shape Group Coordination in Corydoras aeneus.” Royal
Society Open Science 6: 1 90587.
Riley, R. J., T. P. Roe, E. R. Gil lie, and A. Man ica. 2020 . “The Development
of Tactile Social Interactions in Corydoras aeneus L ar vae.” Behaviour
157, no. 6: 515–539.
Santema, P., and T. Clutton- Brock. 2013. “Meerkat Helpers Increase
Sentinel Behaviour and Bipedal Vigilance in the Presence of Pups.”
Animal Behaviour 85, no. 3: 655–661.
Schausberger, P., M. Gratzer, and M. A. Strodl. 2017. “Early Social
Isolation Impairs Development, Mate Choice and Grouping Behaviour
of Predatory Mites.” Animal Behaviour 127: 15–21.
Schel, A. M., A. Candiotti, and K. Zuberbühler. 2010. “Predator-
Deterring Alarm Call Sequences in Guereza Colobus Monkeys are
Meaningful to Conspecifics.” Animal Behaviour 80, no. 5: 799–808.
Sirot, E., and F. Touzalin. 2009. “Coordination and Synchronization
of Vigilance in Groups of Prey: The Role of Collective Detection and
Predators' Preference for Stragglers.” American Naturalist 173, no. 1:
47–59.
Sogard, S. M., and B. L. Olla. 1997. “The Influence of Hunger and
Predation Risk on Group Cohesion in a Pelagic Fish, Walleye Pollock
Theragra chalcogramma.” Environmental Biology of Fishes 50, no. 4:
405– 413.
Solomon, N. G. 1993. “Body Size and Social Preferences of Male and
Female Prairie Voles, Microtus ochrogaster.” Animal Behaviour 45:
1031–1033.
Tang, Z.- H., Q. Huang, W. Hui, L. Kuang, and F. Shi- Jian. 2017. “The
Behavioral Response of Prey Fish to Predators: The Role of Predator
Size.” PeerJ 5: e3222.
Templeton, C. N., E. Greene, and K. Davis. 2005. “Allometry of Alarm
Calls: Black- capped Chickadees Encode Information About Predator
Size.” Science 308, no. 5730: 1934–1937.
van Schaik, C. P. 2010. “Social Learning and Culture in Animals.” In
Animal Beh aviour: Evolut ion and Mechanisms , edited by Peter Kappeler,
623–653. Berlin, Heidelberg, Germany: Springer Berlin Heidelberg.
Viscido, S. V., and D. S. Wethey. 2002. “Quantitative Analysis of Fiddler
Crab Flock Movement: Evidence for ‘Selfish Herd’ Behaviour.” Animal
Behaviour 63, no. 4: 735–741.
Ward, A., and M. Webster. 2016. Sociality: The Behaviour of Group-
Living Animals, 4 07. Berlin, Germany: Springer.
12 of 12 Ecology and Evolution, 2024
Appendix 1
Our video scoring of cohesion involves three different cohesion configurations possible for a group of three individuals. Namely, the group can have
all individuals together (all together), have two individuals together and one individual apart (2–1 configuration), or all individuals apart (all apart).
These configurations are mutually exclusive. When analyzing all three configurations separately, we see no significant differences in the amount of
time spent in each configuration (two- sample t- tests, all p > 0.05, Figu reA1).
FIGUR E A | Time in all possible cohesion configurations in MAC versus SAC fish.
