 

 

IIBRARY
Michigan State
University

....—-

This is to certify that the
thesis entitled

VISION AS A SENSORY MODALITY IN HONEY BEE (AP/S
MELLIFERA) DANCE COMMUNICATION

presented by

MERCEDES RAMIREZ

has been accepted towards fulﬁllment
of the requirements for the

MS de in ,1 Zoology

 

 

 

C/ﬂVlajor Professor's Si atur

lL/(Q/ﬁ

Date

MSU is an Affirmative Action/Equal Opportunity Employer

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Acc&Pres/ClRC/DateDue.indd

VISION AS A SENSORY MODALITY IN HONEY BEE (APIS' MELLIFERA)
DANCE COMMUNICATION

By

h’Iercedes Ramirez

A THESIS
Submitted to
Michigan State University

in partial fulﬁllment of the requirements
for the degree of

MASTER OF SCIENCE
Zoology

2009

ABSTRACT

VISION AS A SENSORY MODALITY IN HONEY BEE (APIS
MELLIFERA) DANCE COMMUNICATION

By

Mercedes Ramirez

Communication requires a sender, a receiver and a signal that transmits informa-
tion. In honey bees (Apis), foragers transmit distance and direction information to
other bees, but the sensory modality of the signal still remains unclear. There are
ﬁve candidate sensory modalities: substrate vibration, olfaction, mechanical antennal
sensation, audition and vision, but we do not yet know which are most important.
In A. mell'zlfem, dances are performed in dark cavities, potentially negating a role
for vision in dance communication. However, A. mell'ife'ra dances do sometimes take
place in bright light, e.g. on swarms, where vision might be useful. Also, the an-
cestral dance likely took place entirely on exposed nests, and extant open nesting
species have conspicuous visual displays during the dance. Thus, A. mellzifem may
have maintained an ancestral ability to use vision. We tested whether vision is used
as a sensory modality by measuring both dancer and follower behaviors under light
and dark conditions. The numbers of waggle runs performed by dancers, followers
that attended dances, waggle runs followed and follower persistence did not differ sig-
nificantly between the two light conditions. Despite evidence previously suggesting a
difference in follower attraction in light conditions, these results suggest that honey
bees do not rely on vision as a sensory modality. It is possible, however, that other
measures could reveal a role for vision. That A. mellz’fera does not employ vision in
dance communication raises questions about the evolution of the dance signals: the
phylogenetic patterns suggest either that the use of vision for dance communication
arose independently in open nesting bees, or that the ability to use vision was lost. in

the ancestors of A. mellifem, perhaps when they began using cavities as nests.

TABLE OF CONTENTS

List of Tables ................................. 1v
List of Figures ................................ V

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 The honey bee dance language ...................... 1
1.2 Sensory modalities for transmitting dance information ......... 3
1.3 Historical evidence of dance sensory modalities ............. 7
1.4 The focus of this study .......................... 11
Methods . . . . . . . . . . . . ..... . . . . . . . . . . . . 12
2.1 Experimental trials ............................ 13
2.2 Video analysis ............................... 14
2.3 Dancer behavior experiments ...................... 14
2.4 Follower behavior experiments ...................... 16

Results.............................17

3.1 Dancer behavior .............................. 17
3.2 Follower behavior ............................. 29
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.1 Summary and future considerations ................... 38

iii

2.1

3.1

3.2

3.3

3.4

3.5

LIST OF TABLES

Light levels in different natural and experimental environments mea-
sured in lux. LED: light emitting diode, CF/FS= compact ﬂuores-
cent / full spectrum bulbs ..........................

_ General linear model for the number of waggle runs performed by

dancers in 2007. Includes estimates of coefﬁcient (0), standard er-
ror (SE), 95 % conﬁdence intervals (CI), p-value and signiﬁcance level
(sig level). Stars indicate signifcance level: * = < 0.01, ** = < 0.05,
*** = < 0.001. ..............................

‘Best’ general linear model for the number of waggle runs performed by
dancers in 2007, selected using AIC. Includes estimates of coefficient
(,8), standard error (SE), 95 % conﬁdence intervals (CI), p—value and
signiﬁcance level (sig level)- Stars indicate signifcance level: * = <
0.01, ** = < 0.05, *** = < 0.001 .....................

General linear model for the number of followers attracted by dancers
per dancing bout in 2008. Includes estimates of coefﬁcient (0), stan-
dard error (SE), 95 % conﬁdence intervals (CI), p-value and signiﬁcance
level (sig level). Stars indicate signifcance level: * = < 0.01, ** = <
0.05, *** = < 0.001. ...........................

‘Best’ general linear model for the number of followers attracted by
dancers per dancing bout in 2008, selected using AIC. Includes esti-
mates of coefﬁcient (B), standard error (SE), 95 ‘70 conﬁdence intervals
(CI), p-value and signiﬁcance level (sig level). Stars indicate signif-
cance level: * = < 0.01, ** = < 0.05, *** = < 0.001. .........

General linear model for the number of waggle runs followed by follow-
ers in 2008. Includes estimates of coefﬁcient (0), standard error (SE),
95 % conﬁdence intervals (CI), p-value and signiﬁcance level (sig level).

13

21

22

28

Stars indicate signifcance level: * = < 0.01, ** = < 0.05, *** = < 0.001. 30

iv

1.1

1.2

1.3

1.4

3.1

3.2

3.3

3.4

LIST OF FIGURES

Waggle run angles correspond to changes in the angle between the food
source traveled to and the sun’s azimuth, which changes throughout
the day. Modiﬁed from Dyer (2002) ...................

Diagram of a waggle run: a) dancer, b followers, c) return loop. Mod-
iﬁed from Von Frisch (1967) .......................

Dance communication related traits mapped onto condensed A pz‘s phy-
logeny. Modiﬁed from Dyer (2002) and Raffiudin and Crozier (2007) .

Illustrations of a) lateral and dorsal/ ventral body waggling and wing
positions found in A. ﬂorea and A. dorsata, b) lateral body waggling
only and wing positions found in A. mellz’fera and A. cerana. Modiﬁed
from Towne (1985). ...........................

Histogram of the number of individuals vs. waggle runs performed in
a dancing bout in summer 2007 ......................

Box plot of number of waggle runs performed for each individual dancer
in summer 2007. Whiskers indicate highest and lowest points within
the 1. 5 interquartile range. Open circles indicate data points that lie

outside of the 1. 5 interquartile range. ..................

Box plots of number of waggle runs performed in a) light and dark
conditions, and b) experimental days in summer 2007. Whiskers indi-
cate highest and lowest points within the 1.5 interquartile range. Open

18

19

circles indicate data points that lie outside of the 1.5 interquartile range. 20

Box plot of number of waggle runs performed versus temperatures in
summer 2007. Whiskers indicate highest and lowest points within the
1.5 interquartile range. Open circles indicate data points that lie out-
side of the 1.5 interquartile range .....................

23

3.5

3.6

3.7

3.8

3.9

3.10

3.11

Histogram of the frequencies of the number of followers in dancing
groups in summer 2008. .........................

Box plots of number of followers in dancing groups versus a) individual
dancer identity, and b) experimental date in summer 2008. Whiskers
indicate highest and lowest points within the 1.5 interquartile range.
Open circles indicate data points that lie outside of the 1.5 interquartile
range. ...................................

Box plots of number of followers in dancing groups versus a) the number
of waggle runs performed by dancers, and b) light condition. Whiskers
indicate highest and lowest points within the 1.5 interquartile range.
Open circles indicate data points that lie outside of the 1.5 interquartile
range. ...................................

Histogram of the frequencies of followers that followed a certain number
of waggle runs in 2008. ..........................

Box plots of number of waggle runs followed versus a) the number
of waggle runs dancers performed, and b) the individual identity of
dancers in summer 2008. Whiskers indicate highest and lowest points
within the 1.5 interquartile range. Open circles indicate data points
that lie outside of the 1.5 interquartile range ...............

Box plots of number of waggle runs followed versus a) experimental
date and b) light condition in summer 2008. Whiskers indicate high-
est and lowest points within the 1.5 interquartile range. Open circles
indicate data points that lie outside of the 1.5 interquartile range. . .

Box plot of number of waggle runs performed for each individual dancer
in summer 2007. Whiskers indicate highest and lowest points within
the 1.5 interquartile range. Open circles indicate data points that lie
outside of the 1.5 interquartile range. ..................

vi

24

25

26

29

31

32

34

Chapter 1

Introduction

1.1 The honey bee dance language

Communication is generally characterized as the transmission of information from
a sender to a receiver via a signal. The honey bee dance language is a prominent
example of communication Among non-human animals, ﬁrst discovered by Karl von
Frisch in the early 20th century and most extensively studied in the western honey
bee, A pz's mellifem. The honey bee “waggle” dance is a complex series of stereotypic
movements that communicates distance and direction information of food and water
sources and nest sites (Von Frisch, 1967; Seeley, 1995). A forager gathers distance
and direction information on her trips to the resource, using the sun’s azimuth as a
reference point. When she returns to the hive, she performs a pantomime of her flight
on the hive comb surface, using gravity as a reference, instead of the sun’s azimuth
(see Figure 1.1).

The waggle dance itself consists of two distinct parts, the waggle run and the return
loop. During a waggle run, the dancer will take a single stride and rapidly “waggle”
her abdomen, which produces a unique and prominent multimodal signal for other

bees in the vicinity (Von Frisch, 1967; Tautz, 1996; Michelsen, 2003). The duration

 

 

Food 20. 40.

 

UP

20 400
x (I I

Figure 1.1: Waggle run angles correspond to changes in the angle between the food
source traveled to and the sun’s azimuth, which changes throughout the day. Modiﬁed
from Dyer (2002)

of this waggle run can be translated into the actual distance the dancer perceived
herself to travel to the source she is indicating, itself influenced by the amount of
optic flow feedback she recieves as she flies (Esch, 1967; Dyer, 2002; Von F risch, 1967).
After the waggle run, the dancer travels in an arc to the point where her previous
waggle run began, completing one dance circuit. The speed with which the dancer
performs the return loop is indicative of her assessment of the relative proﬁtability
of the source (Seeley et al., 2000). Dance circuits are often repeated numerous times
in succession, called a dance bout, before the dancer leaves to return to the source
herself (Von F risch, 1967). Bees that collect dance information are called followers, as
they will often move to follow the dancer as she performs her waggle runs (see Figure

1.2).

 

Figure 1.2: Diagram of a waggle run: a) dancer, b followers, c) return loop. Modiﬁed
from Von Frisch (1967)

1.2 Sensory modalities for transmitting dance in-

formation

Although we know much about the function of dance communication in A. mellifem,
the mechanisms by which waggle dances are performed, and especially how informa—
tion travels between dancer and follower bees, remain unclear. As a forager dances,
other bees in the hive have access to a wealth of information about foraging sites. How
are followers accessing this information? At least ﬁve candidate sensory modalities
have been proposed in A. mellz'fera, including substrate vibration, olfaction, mechani-
cal antennal sensation, audition and vision, but no evidence points deﬁnitively to any
of these (Esch, 1967; Michelsen et al., 1987; Tautz, 1996; Rohrseitz and Tautz, 1999;
Thom et al., 2007; Dyer, 2002). Additionally, we do not know whether the trans-
mission of information is mono- or multi-modal, whether there are any interactions
between or among sensory modailites or whether redundant information across mul-
tiple modalities allows for ﬂexibility in the face of variable environmental conditions

(Dyer, 2002; Griiter and Farina, 2009).

Substrate vibration

During waggle runs, dancers create a weak 200-300 Hz substrate vibration, and given
a choice, seem to prefer dancing on open comb, which transmits vibrations better
(Tautz, 1996; Nieh and Tautz, 2000). Additionally, dances performed on open comb
were more attractive to followers than those performed on sealed comb (Tautz and
Rohrseitz, 1998). The substrate vibrations are just within the detection threshold
of the subgenual organ (Sandeman et al., 1996). As the vibrations move though the
comb, they cause the cell walls to move in response. At a particular distance from the
origin of the vibrations, phase reversal of the cell walls occur, such that within a cell,
the walls further from the origin are displaced in reverse phase compared to those
walls closer to the origin. The result is that the vibration signal is ampliﬁed (Tautz
et al., 2001). Followers could easily detect this signal by straddling a cell and sensing
differences in the positions of pairs of legs as the vibrations move from one end of the
cell to another. Indeed, followers located in areas where phase reversal is greatest show
the most attraction (Tautz et al., 2001). However, substrate vibrations are clearly
not necessary for dance communication: dances for nest sites take place on swarms
and are performed on the back of other bees, the movement of which would negate
any substrate vibrations. Although perhaps not necessary, substrate vibrations may
play a role in attraction of followers to a dancer. Substrate vibrations, then, may not
be necessary to successfully transmit dance information, though these results suggest
a role in attraction of followers to a. dancer. Further experiments are required to

determine what role they may play in the transmission of dance information.

Olfaction

Foragers returning from food sources also carry with them floral odors, and this scent

influences bee behavior in a number of ways: it can reactivate foragers that have

ceased foraging at a food source, and increases the success of locating the proper food
source in new recruits. Odorants from the Nasanov gland also serve, in part, to mark
successful foraging sites for other foragers (Von Frisch, 1967). Additionally, dancing
bees release a unique scent that increases the number of bees that leave the hive
(Thom et al., 2007). However, given the rapidity with which dancers vibrate their
abdomens, and that this waggling carries in it speciﬁc information about distance,
it seems unlikely that a odor gradient could transmit this information with sufﬁcient
speciﬁcity to be useful as a dance communication sensory modality. It appears more
likely that dance Odorants may be important in attracting followers to dancing bees,

but more research is required.

Mechanical antennal sensation

Mechanical antennal sensation, separate from audition, includes tactile contact of
follower antennae with the body of the dancer, and using antennae to sense air jets
created around a dancer as she dances. Followers seem to prefer to position them-
selves within i 30 degrees of dancers’ midlines at the rear (Judd, 1995). Follower
antennae are buffeted by the rapidly waggling body of the dancer, and in this posi-
tion, followers may be able to accurately receive both duration of waggling and body
angle in relation to gravity, two major components of dance communication, though
more experimentation is required to determine how tactile sensations are used (Rohr-
seitz and Tautz, 1999). A dancing bee also produces a unique set of air jets distinct
from air movement caused by the 15 Hz waggling movement and 220 Hz vibration
of the wings (Michelsen, 2003). These air jets occur near the abdomen of dancers,
and may provide a cue, detectable by the followers’ antennae, as to the orientation of
the dancer’s body. Although air jets have been detected, no experiments have been
performed to determine if, or how, followers may perceive and use these potential

clues.

Audition

Audition is the favored modality for dance communication. A. mellz’fem dancers cre-
ate dance ‘sounds’ by vibrating their wings at around 280 Hz during waggle runs.
This causes movement of air particles behind the dancer, know as near-ﬁeld sounds.
The antennae of follower bees can detect these sounds (Towne, 1985; Towne and
Kirchner, 1989). A comparison of the dance sounds produced by regular and mutant,
small winged, honey bees found that the smaller wings vibrate at a higher frequency,
which produced a higher frequency sound, but also reduced the amplitude of the
sound produced. When successful follower recruitment rates were compared among
bees with normal wings, bees with experimentally clipped wings and mutant small
winged bees, recruitment was signiﬁcantly lower for bees with smaller (either mutant
or clipped) wings, suggesting that sounds play an important role in successful fol-
lower recruitment (Kirchner and Sommer, 1992). Finally, follower recruitment to an
artificial model honey bee dancer was improved when dance sounds were also emitted
during waggle runs (l\/Iichelsen et al., 1992). Because antennae are used for multiple
potential modalities, it may be difficult to disentangle the effects of one modality over

another.

Vision

Finally, vision has been generally unstudied as a sensory modality in A. mellz'fem, as
they generally nest in dark cavities where vision would be limited. However, dances
for nest sites take place in daylight, and the openings of natural cavities vary in size,
thus creating relatively lighted situations. In an experiment investigating the im-
portance of tactile contact in follower attraction to dancers, dances performed under
artiﬁcial lights were more attractive than those performed with infrared illumination

(Tautz and Rohrseitz, 1998). Followers appeared to require more tactile contact to

become attracted to dances performed in dark conditions, and less tactile contact
in light conditions. From this, the authors suggest that vision may have a role in
dance communcation, speciﬁcally with attraction to dancers. However, to date, no
experiments have investigated changes in dancer or follower behaviors in light versus
dark environmental conditions, which may indicate whether vision can be used by A.

melltfera.

1.3 Historical evidence of dance sensory modali-
ties

Because all honey bees use a symbolic dance language to communicate, but partic-
ular features and requirements of dancing differs among species, information about
dance language mechanisms in other Apis species could provide clues on how the
A. mellifem dance works. Fortunately, the Apts genus is small (~10 species), and
an older proposed A pis phylogeny has been recently conﬁrmed using molecular tech-
niques (Lindauer, 1956; Rafﬁudin and Crozier, 2007). We can look at trends and
correlations related to dance communication across species, and to infer plausible
transitions that produced currently observed diversity of species (see Figure 1.3).

A. ﬂorea, the dwarf bees, likely represent the extant species that most closely
resembles an ancestral A pis species. This species forages during the day, and builds
nests by wrapping wax comb around a tree branch, creating a nearly horizontal comb
surface at the top of the branch. To orient dances, dwarf bees use landmarks or a
view of either the sky or the sun as a reference point (reviewed in Dyer 2002). Unlike
other Apis species, A. ﬂorea are unable to use gravity to orient their dances 011 the
comb. Waggle runs in this species consist of both dorsal-vental and lateral abdomen
movements, and outspread wings, which do not produce near-ﬁeld sounds (Towne

1985; see Figure 1.4a). The giant bees, A. dorsata, are also an open nesting species,

€83 eaeo eat ceased

was Amoomv Emma 80.: voczcog .cﬁaowovEg miﬁ comzeccoU 3:0 EASE: 365 page?“ cosmoEsﬁﬁoQ 8:an ”ma 2:.wE

AeambzuE Ax .wdv
mama 9:: 98.595
Rosa?
morn
met».

8%

333

Eco 1383

3.238 1.x .wdv

moon oz: :umumwm

EBB?

we»

:Bocxcs
338

3:0 1832

683.8% ex .wdv
meme acmmw
Bose?
mo»
we»
caoqus
coao

Essa?» HM 3593

Aeoaomx ex .wdv
803 “3.26

.EEONCOF :OEmuceEo 8st
o: momma 33qu
o: ways—Sm 8me

we». pew: mv—uwsvawq
some 0.5.898 ammz

EBB? aw 183.3 £85958 mﬁom

a)

 

Figure 1.4: Illustrations of a) lateral and dorsal/ventral body waggling and wing
positions found in A. ﬂorea and A. dorsata, b) lateral body waggling only and wing
positions found in A. mellz'fem and A. cemna. Modiﬁed from Towne (1985).

but build their comb dangling from tree branches or rock overhangs, creating only
vertical comb surfaces. Consistent with comb direction, A. dorsata perform dances
in the vertical plane, likely using gravity to anchor their waggle dances. No studies
have investigated whether giant bees can use landmarks to orient dances. Their
waggle runs are similar to those found in A. ﬂorea; however, unlike A. ﬁorea, A.
dorsata do produce dance sounds, particularly when dancing after foraging flights on
moonlit nights (Kirchner and Dreller, 1993). In contrast, A. mellz‘fera and the closely
related eastern hive bee, A. cerana, build vertical comb in dark cavities, and use
gravity to orient their dances on the comb. Dancers in these species do not perform
a dorsal/vental waggle, but only a lateral one during their waggle runs, and produce
signiﬁcant dance sounds by vibrating their closed wings (see Figure 1.4b).

This comparative analysis suggests two trends related to sensory modalities through
the evolution of dance communication in Apz’s. The ﬁrst trend includes light avail-
ablilty in the hive environment, using prominent visual signals, and having silent
dances, as seen in A. ﬂorea. The second, as seen in A. dorsata, A. mellifem and A.
cerana, is a correlation between dark hive conditions (either at night or from cavity
nesting), less prominent visual displays, and sound production during dances. What
do these two trends suggest about the sensory modalities used by these species for

dance communication? For A. florea, silent dances exclude audition, while the gen-

erally high environmental light levels, along with visually prominent body and wing
displays most strongly support vision as the principle sensory modality. In contrast,
a consistently dark nest environment and relatively subdued wing and body displays
would seem to exclude vision as a dance sensory modality for A. melltfem and A.
cemna, while the prominent sounds produced during dances in both species supports
audition as the most important sensory modality for dance communication. A. dor-
sata seems to hold a unique intermediate position between A. flared and A. mellifera
and A. cerana in terms of sensory modalities. This species is not cavity nesting, with
light hive conditions, and also has the prominent visual displays seen in A. ﬂorea
dances, both traits suggesting vision. However, unlike other species, A. dorsata can
forage on moonlit evenings, and during these times have been found to produce more
robust dance sounds like A. mellife'ra, which lends support to the position of A. dor-
sate using audition as a sensory modality. No experiments have been performed to
determine which sensory modality is used during A. dorsata dance communication,
but it appears as though A. domain may be able to use both sensory modalities,
depending on environmental conditions.

Thus, a comparative analysis suggests that both A. ﬂorea and A. dorsata use
vision as a sensory modality in dance communication (Towne, 1985). We can then
ask whether A. mellifera have retained this ancestral ability, and can use vision when
light is available in the hive environment. Besides evidence from the Apis phylogeny,
A. mellsze-ra dances for nest sites often take place on swarms in daylight conditions,
where vision would be a useful modality (Seeley et al., 2000; Dyer, 2002). Finally,
some preliminary evidence suggests that light increases the ease of attraction of bees
to a dancer, suggesting that follower bees may be able to use vision to ﬁnd dancers
(Tautz and Rohrseitz, 1998). Is vision is a potential sensory modality for transmission

of information in the honey bee dance language?

10

1.4 The focus of this study

Our hypothesis is that A. mellz'fera can indeed use vision as a sensory modality for
dance communication in light conditions. Finding evidence of vision use in Amellifem
would conﬁrm our assumptions about the ancestral use of vision based on our current
understanding of dance language evolution within the Apis phylogeny, namely that
the Apis species ancestral to the whole modern genus relied on visual signals to
transmit dance information, and that other sensory modalities evolved later. While
adding to our knowledge of dance language evolution within the A pas genus, support
for vision use in A. melltfera would at same time complicate our understanding of
the speciﬁc dance communication mechanisms in this species by adding evidence for
yet another potentially important sensory modality. On the other hand, ﬁnding that
vision does not seem to be used by A. mell'ifem would not contribute signiﬁcantly to
our knowledge about vision in ancestral A pis, but would allow us to rule out vision as
a potential sensory modality in A. mellifera speciﬁcally and focus on others in future
research.

To test whether A. mellz‘fera can use vision as a sensory modality for dance com-
munication, we applied two global conditions, light and dark, to an observation hive,
and measured both dancer and follower behaviors in each condition. We looked for
differences between the two light conditions in the number of waggle runs performed
by dancers, the number of followers a dancer contacted, the number of waggle runs
followed by followers and the following persistence of followers. Differences between
light conditions would suggest that vision was being used in some capacity by either

dancers or followers.

11

Chapter 2

Methods

Experiments took place between the summer of 2007 and fall 2008 at Michigan State
University, East Lansing, hr‘lichigan. A small two frame observation hive of mixed
stock A. mell'zlfem was housed in a trailer in 2007, and at an MSU bee facility in
2008. Both sites were surrounded mostly by ﬁelds; the 2008 site also had numerous
larger hives in the area. The hive entrances at both sites faced roughly WNW. Bees
were trained to a scented (lemon in 2007, anise in 2008; see Wray et a1. 2008) sucrose
solution feeder ~500m NNW of the hive entrance, which standardized dances for
distance and relative proﬁtability. Trained foragers were individually tagged at the
feeder. Because of the presence of other local hives at the 2008 site, all bees that
exited the hive were dusted with colored powder to identify and exclude non-dusted
individuals from other hives that might have arrived at the feeder by chance. Local
weather conditions (temperature, wind speed and wind direction) were measured in
2007 and included as predictors in a general linear model of dancer waggle runs, but

did not affect results, and were not collected in 2008.

12

Table 2.1: Light levels in different natural and experimental environments measured
in lux. LED: light emitting diode, CF/FS: compact fluorescent/full spectrum bulbs.

 

Environment irradiance (lux)
Dark 0
LED array (2007) 5
CF/FS array (2008) 8
Open nesting conditions 40+
Full sunlight 900+

2.1 Experimental trials

To prepare the hive for experimental trials, a sucrose solution feeder was set out, and
marked bees were allowed to forage ad libitum. Unmarked individuals were removed
from the feeder and destroyed to minimize recruitment and encourage dancing from
marked individuals. Each trial consisted of two 30 minute periods of behavior record-
ing, one period for each light condition. To create a light environment for dancers and
followers, we used LED and incandescent lights in 2007, and full-spectrum compact
fluorescent lights in 2008 to more closely approximate both the intensity and spec-
trum of sunlight, measured using a lux meter (see Table 2.1). Dance floor illumination

should have been sufficient to allow bees to see one another (see Menzel 1981).

The starting condition was selected randomly at the beginning of the day. Trials
began once at least 15 marked bees were actively foraging from the feeder. Digital
video of dancers and followers was collected in both light and dark conditions; record-
in g was made possible by an infrared (IR) source on the camera which provided enough
illumination for the IR—sensitive camera (Sony Handycam, DCR—HC52). Dancing and
following behaviors were captured by aiming the camera at the portion of comb near
the hive entrance, where the majority of dances take place. Because forager moti-
vation to continue foraging and dancing was influenced by many variables, including

weather, sucrose concentration and time of day, the number of successful trials varied

13

on a day to day basis.

2.2 Video analysis

Prior to video editing and analysis, digital video was imported onto a laptop computer
(IV'lacbookPro, Apple, 2006) from digital tapes using Media Edit Pro (Machare,
2003). Dancers were selected based on the following criteria: dancing bout lasted for
at least ﬁve waggle runs, dancers performed dancing bouts at least once in each light
condition on a. given day, and dancers and followers were clearly visible throughout
the dancing bout, and not blocked by other bees. Once a dancer was chosen, the
number of waggle runs she performed was counted and recorded, as well as the start
time of her dancing bout. Followers were chosen using the deﬁnition from Wray et a1.
(2008): bees facing a dancer that were within one bee-length of the dancer, that
moved to keep themselves near the dancer, and that followed at least two consecutive
waggle runs. These criteria elin'iinated bees that happened to be near, or passing by a
dancer, but were not actually part of the dancing group. To keep of track of followers
as they joined and left a dancing group, we marked them digitally. However, because
of this, it was impossible to continually track followers once they left dancing groups.
Thus, it is likely that some followers are counted multiple times over a number of days.
However, this recounting should occur randomly, and so we treated each follower as
a different individual. All video was coded using JWatcher (National Institute of

Health, 2009).

2.3 Dancer behavior experiments

We predicted that dancers would perform more waggle runs in the dark than the

light, attempting to increase the efﬁcacy of transmitting dance information, since it

14

may be harder for followers to locate a dancer in the dark. To test for differences
in the mean number of waggle runs performed by individuals in both light and dark
conditions, we measured the number of waggle runs individual dancers performed. We
used a general linear model to assess the signiﬁcance of date, weather data (includes
outside temperature, wind speed, wind direction), individual dancer identity, light
condition and a dancer identity by light interaction on the number of waggle runs
performed by dancers in 2007. We log—transformed waggle run values in the model
to meet the assumptions of normality and homoscedasticity. We performed a step-
wise algorithm to create models from our full model, and compared AIC (Akaike
Information Criteria.) values to determine the ‘best possible’ model from variable
measured. To specifically determine whether light was a signiﬁcant influence, we
compared the ’best‘ model, which excluded the light variable, to the same model
with the addition of the light variable using AN OVA. We compared waggle run values
between summers 2007 and 2008 using ANOVA and found they were not significantly
different, so we combined the datasets. We used a paired t-test (n: 31) to compare
the mean numbers of waggle runs performed by dancers in light and dark conditions.

We also predicted that dancers would be able to attract more followers in light
conditions, as followers would be able to use vision to ﬁnd dancing groups. We created
a general linear model to access the signiﬁcance of date, dancer identity, number of
waggle runs performed by dancers and light condition on the number of followers that
were part of a dancing group during individual dancing bouts in 2008. This model
met the assumptions of homoscedasticity and normality. We performed a step-wise
algorithm to create simpler models from our full model and compared them using
AIC values to determine the ‘best possible’ model. To determine whether light was
an important predictor of dance follower number, we used ANOVA to compare the

‘best possible’ model with one that included light condition.

15

2.4 Follower behavior experiments

We predicted that followers would be able to follow more waggle runs in the light,
since they would be able to use vision as well as other senses to ﬁnd and follow
dancers.We used a general linear model to access the signiﬁcance of date, dancer
identity, the number of waggle runs performed by dancers and light condition on the
numbers of waggle runs followers followed in 2008. We log transformed the follower
waggle run number to meet the assumptions of homoscedasticity and normality.
Finally, we predicted that followers in light conditions would be able to persist
longer in following once they joined a dancing group, again because they would be
able to use vision in light conditions. We used only uncensored data, where the
start and stop times of following were observed, and analysed the number of waggle
runs individual followers followed in both light and dark conditions using a Cox
proportional hazard regression model, which gave us following persistence statistics
in both light conditions. All statistics were performed in R (R Development Core

Team, 2008).

16

Chapter 3

Results

3.1 Dancer behavior

Do dancers perform more waggle runs in light conditions?
Exploratory statistics

In the summer of 2007, dancers performed an average of 19 (i 12) waggle runs per
dancing bout. The range and direction of skew of waggle run data varies between
individual dancers (see Figure 3.2). Given this variation between dancers, we expect
dancer identity to be an important predictor of the number of waggle runs performed.
The most frequent number of waggle runs per dancing bout is skewed towards the
left, with the most dancers performing between 10 and 20 waggle runs per dancing
bout (see Figure 3.1).

The ranges and skew of waggle run values between Julian calendar days are fairly
consistent, but there appears to be a slight decline in the median waggle run value
from days later in the season (see Figure 3.3a). Waggle run values plotted against
temperature do not reveal any consistent trend; within a temperature waggle runs

values tend to be slightly skewed towards lower numbers, but median values hover

17

4O 60 80 100 120

Number of individuals

20

 

 

 

 

| a
1 I
’ I
H1
i1 2b 4b 60 8'0

Number of waggle runs per dance bout

 

Figure 3.1: Histogram of the number of individuals vs. waggle runs performed in a
dancing bout in summer 2007.

around 20 waggle runs (see Figure 3.4). Data for the number of waggle runs performed
in light and dark environments appear to have a relatively normal distribution (see
Figure 3.3b). There is a slightly wider range of values in the dark condition, and the
upper quartile extends further, however, this is likely the result of a one extremely
high value for waggle run number. There is no immediately obvious difference between
the two distributions, suggesting that light condition will not be a good predictor of

waggle run number.

Model selection

We wanted to see if weather variables inﬂuenced the number of waggle runs that
dancers performed, in addition to the influence of the experimental light variable.
From the full model, only dancer identity was a signiﬁcant predictor of the number
of waggle runs that dancers performed (5 out of 28 dancers in 2007, see 3.1). This

model accounted for approximately 30% of the variation in dancer waggle run number

18

 

80

60

° 1»

t LATDEE as tv,t.as as

1 2 3 5 6 7 8 9 111214151'81921023242526272830313231334
Individual dancer identity

40

20

 

 

Number of waggle runs performed by dancers
1
2|

Figure 3.2: Box plot of number of waggle runs performed for each individual dancer in
summer 2007. Whiskers indicate highest and lowest points within the 1.5 interquartile
range. Open circles indicate data points that lie outside of the 1.5 interquartile range.

(F28,282=5-541 R2=0.29, p= <0.001). The ﬁnal ‘best possible’ model created from
the stepwise algorithm included dancer identity, Julian calendar day, wind speed
and headwind (F28,282=5-541 R2=0.29, p: <0.001, see 3.2). Dancer identity and
wind speed are signiﬁcant predictors of waggle run number (AN OVA, p D :< 0.001,

pW =< 0.03).

Summary: number of waggle runs performed

The main goal of this study was to assess the influence of light on the number of
waggle runs dancers performed. There was no signiﬁcant difference between a model
that included light condition and one without it, suggesting that light condition was
not a signiﬁcant predictor of waggle run number in 2007 (ANOVA, p= 0.58), and the
means of waggle runs performed in light and dark conditions in both 2007 and 2008

were not signiﬁcantly different (paired t—test, t = 0.3568, df = 33, p = 0.72).

19

 

80

60

40

 

20

 

 

 

 

‘

dark tht
Hive light condition

0

 

 

 

Number of waggle runs performed by dancers m

 

 

 

 

U)
L.
OJ
(c) o
C
see.
>
.D
"U
Q)
E°~
(.0 o
o _V"
t o l
‘1’ _l_
O.
V) 1 0
Co- I 0
3V ‘ 0 FT.
1. ‘ o ‘
co 7 .9— ' : ‘T‘
‘0: Tn ? : : 1
o) . ,_ .‘
§°~ ( T ;
I,” E .
._ _1_ : L
O C I l
‘D I "— I .
E04 __ O _1.__ + 4- .a_
3
Z

318 321 327 323 329 335 336
Julian calendar date
Figure 3.3: Box plots of number of waggle runs performed in a) light and dark
conditions, and b) experimental days in summer 2007. Whiskers indicate highest and

lowest points within the 1.5 interquartile range. Open circles indicate data points
that lie outside of the 1.5 interquartile range.

20

Table 3.1: General linear model for the number of waggle runs performed by dancers
in 2007. Includes estimates of coefficient (8), standard error (SE), 95 ‘70 conﬁdence
intervals (CI), p-value and signiﬁcance level (sig level). Stars indicate signifcance
level: * = < 0.01, ** = < 0.05, *** = < 0.001.

 

 

Response
WR Predictors 13 SE 95% CI P-value Sig Level
Intercept 9.90 6.92 -3.73—23.53 0.15
Dancer identity

 

2 -0.02 0.82 -1.82-1.41 0.80
3 0.31 0.78 -1.22-1.84 0.69
5 0.43 0.76 -1.07—1.93 0.57
6 -0.36 0.72 -1.77-1.05 0.62
7 0.38 0.80 -1.19-1.97 0.63
8 0.05 0.75 -1.43-1.54 0.94
9 0.59 0.70 -0.79-1.97 0.94

11 0.72 0.72 -0.70—2.15 0.32
12 0.93 0.70 -0.44-2.30 0.18
14 1.15 0.71 —O.25-2.55 0.11
15 1.41 0.84 -0.24-3.07 0.09
18 1.27 0.69 —0.10-2.64 0.07
19 1.32 0.71 -0.70-2.71 0.06
20 1.49 0.70 0.11-2.86 0.03 *
23 0.38 0.70 -l.00-1.77 0.60
24 1.29 0.69 —0.07—2.65 0.06
25 1.23 0.69 -0.13-2.59 0.08
26 1.07 0.79 —0.48-2.63 0.17
27 1.32 0.83 -0.31-2.96 0.11
28 1.55 0.76 0.06-3.04 0.42
30 1.77 0.79 0.22-3.32 0.03 *
31 1.09 0.77 —0.42-2.60 0.16
32 1.74 0.69 0.38—3.10 0.01
33 1.65 0.70 0.26-3.03 0.02
34 1.46 0.76 -0.31-2.95 0.06
Light condition 0.25 0.77 -1.27-1.77 0.75
Julian day -0.02 0.02 —0.07-0.02 0.31

V’Vind direction 0.001 0.002 0.002-0.006 0.51
Wind speed 0.36 0.20 -0.05—0.76 0.09
Temperature -0.04 0.05 -0.14-0.05 0.37

Headwind -0.28 0.16 -0.59-0.03 0.08
Crosswind -0.20 0.16 -0.51-0.10 0.19
Dancer identity x

light condition*

2 x light 0.46 0.98 -1.47-2.38 0.64

21

Table 3.2: ‘Best’ general linear model for the number of waggle runs performed by
dancers in 2007, selected using AIC. Includes estimates of coefﬁcient (,6), standard
error (SE), 95 % conﬁdence intervals (CI), p-value and signiﬁcance level (sig level).
Stars indicate signifcance level: * = < 0.01, ** = < 0.05, *** = < 0.001.

 

 

Response
VVR Predictors ,13 SE 95% CI P-value Sig Level
Intercept 12.16 4.88 255-2177 0.13
Dancer identity

2 0.16 0.38 -0.58-0.91 0.66
3 0.44 0.41 —0.37—1.25 0.28
5 0.58 0.41 -0.23-1.38 0.16
6 0.13 0.35 -0.56-0.81 0.71
7 0.68 0.40 -0.11-1.46 0.09
8
9

 

0.55 0.37 -0.19-1.29 0.14
0.67 0.36 -0.04-1.39 0.07

11 0.84 0.34 0.18-1.51 0.01 *
12 1.10 0.35 0.40-1.79 0.002 **
14 1.05 0.38 0.30-1.80 0.006 **
15 1.44 0.38 0.70-2.18 < 0.001 ***
18 1.18 0.37 0.45—1.92 0.002 **
19 1.28 0.38 0.52-2.03 < 0.001 ***
20 1.46 0.39 0.68-2.23 < 0.001 ***
23 0.86 0.39 0.10-1.62 0.03 *
24 1.32 0.34 0.65-1.99 < 0.001 ***
25 1.29 0.37 0.57-2.01 < 0.001 ***
26 1.48 0.36 0.78-2.18 < 0.001 ***
27 1.24 0.35 0.54-1.93 < 0.001 ***
28 1.26 0.41 0.45-2.07 0.002 **
30 1.90 0.39 1.12-2.67 < 0.001 ***
31 1.40 0.35 0.72-2.09 < 0.001 ***
32 1.59 0.37 0.86-2.32 < 0.001 ***
33 1.60 0.40 0.82-2.39 < 0.001 ***
34 1.32 0.37 0.59-2.06 < 0.001 ***
Julian day -0.03 0.01 0.06—0.004 0.03 *
Wind Speed 0.13 0.05 0.03-0.22 0.01 *

Headwind -0.07 0.04 -0.16—0.02 0.13

22

 

80

60

Egg qBHHE MN: EHBHHB E

24’ 226231236 23.9 24.1 24.3 24.5 24.7 24.9 25.1 25.3 25.5 26.3 29.4 29. 633.10
Outside temperature

40

1.
j

I~4ﬂ

20

l .1

.1.

.50

l_u

 

 

0

 

Number of waggle runs performed by dancers

Figure 3.4: Box plot of number of waggle runs performed versus temperatures in
summer 2007. Whiskers indicate highest and lowest points within the 1.5 interquartile
range. Open circles indicate data points that lie outside of the 1.5 interquartile range.

Do dancers attract more followers in light conditions?

Exploratory statistics

A dancer attracts an average of 13 (:t 8) followers during a dancing bout, although
the variation between individual dancers is high (see Figure 3.6a). The number of
followers is skewed to the left, suggesting that lower followers numbers were more
prevalent.(see Figure 3.5). There was no clear trend in the number of followers based
on experimental date. The median values differed between days, but the ranges for
each day largely overlapped (see Figure 3.6b). Perhaps unsurprisingly, there appears
to be a positive correlation betwen the number of waggle runs performed by dancers,
and the number of followers they attracted (see Figure 3.7a). The values for the
number of dance followers attracted in light and dark are very similar in median value
and range, suggesting no signiﬁcant difference in the number of followers attracted

across conditions (see Figure 3.7b).

23

15

10

Number of individuals

 

 

 

 

 

o .3 .__ ,,,,, __ -a. ___-_._ .._, _ i._7.___,-_. ___ __.l

l l r

5 10 15 20 25 30
Number of followers in a dancing group

 

Figure 3.5: Histogram of the frequencies of the number of followers in dancing groups
in summer 2008.

Model selection

We created a model for dance follower number that. included date, dancer identity,
the number of waggle runs performed by dancers, and light condition. Of these
variables, date and waggle runs performed by dancers were signiﬁcant predictors of
follower numbers (see Table 3.3). Approximately 55% of the variation in the number
of followers attracted to dancing groups was explained by this model (F15,13=3.28,
R2=0.55, p = 0.02). The signiﬁcant predictors from the full model, date, dancer
identity and waggle runs performed remained in the ‘best possible’ model created
through stepwise regression (see Table 3.4). The model explained approximately 58%

of the variation in follower attraction numbers (F14,14=3.78, R2=0.58, p = 0.009).

Summary: number of followers attracted

Light condition was not a significant predictor of the number of followers attracted

to dancers was, as there was no signiﬁcant difference in the amount of variation

24

DJ

40

30

Number of followers in a dancing group
20

 

 

i23 4515;59101112
Individual dancer identity

0'

 

 

 

 

 

 

 

Number of followers in a dancing group

 

 

 

05t3 05:9 0ci11 Oct14
Date

Figure 3.6: Box plots of number of followers in dancing groups versus a) individual
dancer identity, and b) experimental date in summer 2008. Whiskers indicate highest
and lowest points within the 1.5 interquartile range. Open circles indicate data points
that lie outside of the 1.5 interquartile range.

25

Q)

40

l

30

 

0

Number of followers in a dancrng group
20
1:1.
I
ill
:1:
A |
L J
|
|
l
l
|

10
I
E
I
|

 

"6 3 1'1 1'3 14 1'5 1'6 1'7 1'8 1'9 2'1 22 2'3 2'4 272's 2'9 3'3 3'5 37 47
Number of waggle runs performed by dancer

0'

40

0

-|

20

”E

 

 

 

 

 

 

 

Number of followers in a dancing group

0

 

dark light
Hive light condition

Figure 3.7: Box plots of number of followers in dancing groups versus a) the number of
waggle runs performed by dancers, and b) light condition. Whiskers indicate highest
and lowest points within the 1.5 interquartile range. Open circles indicate data points
that lie outside of the 1.5 interquartile range.

26

Table 3.3: General linear model for the number of followers attracted by dancers per
dancing bout in 2008. Includes estimates of coefﬁcient (,8), standard error (SE), 95
‘70 conﬁdence intervals (CI), p-value and signiﬁcance level (sig level). Stars indicate
signifcance level: * = < 0.01, ** = < 0.05, *** = < 0.001.

 

 

 

Response
WR Predictors [3 SE 95% CI P-value Sig Level
Intercept 3.94 6.05 -9.13-17.00 0.53
\Vaggle run number 0.44 0.13 0.16-0.73 0.005 **
Light condition 0.17 2.01 —4.19-4.51 0.94
Date
10/09/2008 —1.67 4.19 ~10.70-7.38 0.70
10/11/2008 8.68 3.37 1.40-1597 0.02 *
10/14/2008 3.50 3.49 -4.04-11.04 0.33
Dancer identity
2 -5.77 6.08 -18.91—7.37 0.36
3 —1.15 5.67 -13.40-11.10 0.84
4 -0.66 6.42 ~14.53-13.21 0.92
5 -5.88 5.28 -17.29-5.52 0.29
6 2.05 5.45 -9.73—13.83 0.71
7 -8.65 6.43 ~22.54-5.24 0.20
8 -3.29 5.74 -15.70-9.12 0.58
9 -2.97 6.81 -17.68-11.74 0.67

-6.69 5.90 -19.44-6.06 0.28
-l.11 8.12 48.65-16.44 0.89

p-AH
r—‘O

27

Table 3.4: ‘Best’ general linear model for the number of followers attracted by dancers
per dancing bout in 2008, selected using AIC. Includes estimates of coefﬁcient (,8),
standard error (SE), 95 % conﬁdence intervals (CI), p-value and signiﬁcance level (sig
level). Stars indicate signifcance level: * = < 0.01, ** = < 0.05, *** = < 0.001.

 

 

 

Response
WR Predictors [5 SE 95% CI P-value Sig Level
Intercept 4.13 5.38 0.46
Waggle run number 0.44 0.13 0.003 **
Date
10/09/2008 —1.64 4.02 0.69
10/11/2008 8.60 3.09 0.015 *
10/14/2008 3.46 3.34 0.32
Dancer identity
2 -5.79 5.86 I 0.34
3 -1.19 5.45 0.83
4 -0.68 6.18 0.91
5 -5.88 5.09 0.27
6 2.04 5.26 0.70
7 —8.66 6.20 0.18
8 -3.30 5.54 0.56
9 —3.03 6.53 0.65
10 -6.74 5.66 0.25
11 -1.26 7.63 0.87

28

so so 00
,3 _J

I

Number of individuals
40

20

l

 

 

 

 

T l I

0 5 10 15
Number of waggle runs followed

Figure 3.8: Histogram of the frequencies of followers that followed a certain number
of waggle runs in 2008.

explained by the model that included light condition (ANOVA, p = 0.87). There was
not a signiﬁcant difference in the mean number of followers attracted in light and

dark conditions (t-test, t = -1.17, df = 15.58, p = 0.26).

3.2 Follower behavior

Do followers follow more waggle runs in light conditions?
Exploratory statistics

Follower bees followed an average of 5 (:i: 3) waggle runs before disengaging from
a dancing group. The number of waggle runs followed is left skewed, with most
followers following 5 or fewer waggle runs, and frequencies of each number of waggle
runs followed tapers off in a regular fashion (see Figure 3.8).

There was not a noticeable trend in the number of waggle runs followed by fol-

29

lowers compared to the number of waggle runs performed by dancers, or individual
dancer identity. Although the median values differ, they do so seemingly randomly,
and the ranges overlap to a great degree (see Figure 3.9). Waggle runs followed do
not seem to change signiﬁcantly based on experimental date (see Figure 3.10a). Both
the median values and ranges for the number of waggle runs followed in light and

dark are very similar between conditions (see Figure 3.10b).

Table 3.5: General linear model for the number of waggle runs followed by followers
in 2008. Includes estimates of coefﬁcient (,0), standard error (SE), 95 ‘70 conﬁdence
intervals (CI), p-value and signiﬁcance level (sig level). Stars indicate signifcance
level: * = < 0.01, ** 2:: < 0.05, *** = < 0.001.

 

 

 

Response
WR Predictors [3 SE 95% CI P-value Sig Level

Intercept 3.94 6.05 -9.13-17.00 0.53
Waggle run # 0.44 0.13 0.16-0.73 0.005 **

Light condition 017 2.01 —4.19-4.51 0.94

Date

10/09/2008 -1.67 4.19 -10.70-7.38 0.70

10/11/2008 8.68 3.37 140-1597 0.02 *

10/14/2008 3.50 3.49 -4.04—11.04 0.33
Dancer identity
2 -5.77 6.08 -18.91-7.37 0.36

3 -1.15 5.67 -13.40-11.10 0.84
4 -0.66 6.42 -14.53—13.21 0.92
5 -5.88 5.28 -17.29-5.52 0.29
6 2.05 5.45 ~9.73—13.83 0.71
7 -8.65 6.43 -22.54-5.24 0.20
8 -3.29 5.74 -15.70-9.12 0.58
9 -2.97 6.81 -17.68-11.74 0.67
10 -6.69 5.90 -19.44-6.06 0.28
11 -1.11 8.12 -18.65-16.44 0.89

Model selection

Our full model for the number of waggle runs followers followed included the num-

ber of waggle runs dancers performed in the dancing bout, the dancer identity, the

30

 

15

1

10

5.
D

L .

o

 

Number of waggle runs followed

0

 

 

6'3 8 1'1 13 1'4 1'5 1'6 1'7 1'8 1'9 2'1 2'2 2'3 2'4 27 2'8 2'9 3'3 3'5 3'7 4'7
Number of waggle runs performed by dancers

b

 

15

1
O u -r- 0
.

111.444

1234587501'01'11'2
Individual dancer identity

10
o
o
4
i

0

5

+11

0

Number of waggle runs followed

0

 

 

 

Figure 3.9: Box plots of number of waggle runs followed versus a) the number of
waggle runs dancers performed, and b) the individual identity of dancers in summer
2008. Whiskers indicate highest and lowest points within the 1.5 interquartile range.
Open circles indicate data points that lie outside of the 1.5 interquartile range.

31

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

'0 7
U ‘—
0) o
3
—° ° ° ° w.—
E o g .
391 a ; ; l
9 : 3 5 l
m : : : '
a ' = 1 ——
a : z a *m l
:0: ‘0 '1 " '
OJ
.0
E ' i
3 ' i
z 5 —4— , 1
ca l
Oc't 3 0c'19 0cf11 Oct' 14
Date
b
‘0 _
s F .
5 o o
— 0 0
'3 o a l
C 0 q '—r_
a '- —*— ; l
2 ' l
8 l
g 1
.. l
S "’1 l l 1
é ; 1 1 l ‘
3 2 .
z I a
O i
dark light

Hive light condition

Figure 3.10: Box plots of number of waggle runs followed versus a) experimental date
and b) light condition in summer 2008. Whiskers indicate highest and lowest points
within the 1.5 interquartile range. Open circles indicate data points that lie outside
of the 1.5 interquartile range.

32

experiment date, and light condition. This model is not very good at explaining vari—
ation in the number of waggle runs followed, and none of the variables are signiﬁcant

predictors (see Table 3.5, 1713;359:127, 11220.02, p = 0.10).

Summary: number of waggle runs followers follow

In regards to the effect of light condition on the number of waggle runs followed, there
was no significant difference in the amount of variation explained by a model that
included light condition and one without it (ANOVA, p=0.29). There was also no
signiﬁcant difference in the mean number of waggle runs performed in light and dark

(t-test, t=0.36, df=240.8, p=0.72).

Does following persistence improve for followers in light con-
ditions?

Following persistence, or the probability of a bee continuing to follow once she begins
following, remained constant between light conditions and number of waggle runs

performed by dancers (see Figure 3.11, Cox proportional hazard, likelihood ratio,

test: 37.7, df=31, p=0.l9).

33

 

0.4 0.6 0.8 1.0

0.2

 

 

0.0

 

0 5 1'0 1'5
Number of waggle runs followed

Probability of continuing to follow waggle runs

 

Figure 3.11: Box plot of number of waggle runs performed for each individual dancer
in summer 2007. Whiskers indicate highest and lowest points within the 1.5 interquar-
tile range. Open circles indicate data points that lie outside of the 1.5 interquartile
range.

34

Chapter 4

Discussion

In 1967, Karl Von Frisch published ”The Dance Language and Orientation of Bees,”
a collection of the work he began in the 1940's where he proposed that honey bees
communicated the location of food sources using a ‘dance’ language (Von Frisch,
1967). Since that time, intensive research has determined when dance communication
is used, how foragers translate and produce dances, what dance communication is
used for (Seeley, 1995; Dyer, 2002). The honey bee dance language has become a
prominent example of animal communication. In general, communication requires a
sender, a receiver and a signal; in the case of honey bees, we still have only a murky
understanding of the signal and how it is received and used by the receiver, and thus
we have an incomplete picture of the dance language and the workings of honey bee
colonies. In this study, we focused on sensory modalities by which honey bee receivers

may collect information contained in dance communication.

Based on our knowledge of the dance language in other A p23 species and known
evolutionary relationships among these species, we hypothesized that A. 77iellz'fem has
maintained the ancestral ability to use vision in the dance language, and speciﬁcally,
that visual signals are used to communicate dance information. To test this hypoth-

esis, we used measures of both dancer and follower behaviors that are important to

35

dance communication, such as the number of waggle runs performed, the number of
followers in a dancing group, the number of waggle runs followed, and the persistence

of dance followers throughout a dancing bout.

There is modest evidence that followers require more antennal contacts with a
dancer to become attracted to and join a dancing group in dark versus light conditions
(Tautz and Rohrseitz, 1998). we also know that the number of waggle runs performed
by dancing foragers has been previously shown to be influenced by both environmental
and individual factors (Von Frisch, 1967; Waddington and Kirchner, 1992; Seeley
et al., 2000). Because it may be harder to attract and/or maintain followers in dark
conditions, we wondered if dancers altered the number of waggle runs they performed
between light conditions to improve the attractiveness of the dance. We measured
the number of waggle runs dancers performed in both light and dark conditions and
found that they did not respond to changes in light conditions by altering the number
of waggle runs they performed within a dancing bout. Thus, dancers do not seem to
vary the number of waggle runs as a way to increase follower attraction to dances in
dark conditions.

As more waggle runs are performed, dancers contact more potential followers, and
followers have increased opportunities to follow dance bouts. Dancers that performed
more waggle runs (had longer dancing bouts) had greater numbers of followers than
those with shorter dancing bouts. As dancers increase the number of waggle runs
they perform, they also increase the opportunities to contact followers that then
join the dancing group. Indeed, waggle run number was a. signiﬁcant predictor of the
number of dance followers. Because there was no difference in the number of followers
in a dancer’s group between light conditions, there appears to be no difference in
attractiveness of the dance to followers.

This result contrasts with that of Tautz and Rohrseitz (1998), who found that

followers in the dark needed more antennal contacts to become attracted to dancers.

36

The present study is different in two key ways that may explain the difference in
results from the Tautz and Rohrseitz (1998) study. First, Tautz and Rohrseitz (1998)
deﬁned a follower as a bee who followed at least one waggle run and one return phase,
whereas followers in our study needed to follow at least two waggle runs to be counted
as followers, as deﬁned by W’ray et a1. (2008). Thus, some followers that would have
been included in earlier study were excluded in this one. We chose to use the Wray
et al. (2008) criterion because it was more stringent, and excluded bees that. may have
passed by a dancer and been momentarily part of the group, but that then moved
on. The second key difference is the number of followers used for analysis. Tautz and
Rohrseitz (1998) sampled 40 followers, while this study looked at over 300. These

methodological differences have likely led to opposing results between the two studies.

This study conﬁrmed earlier reports that followers typically follow ﬁve waggle runs
consecutively (Von Frisch, 1967). Unexpectedly, we found no correlation between the
length of dancing bouts and the number of waggle runs followed by a follower. This re-
sult may suggest that followers gain sufﬁcient information to guide their flights within
following ﬁve waggle runs, and do not need to ‘take advantage’ of the availability of
more waggle runs to increase the accuracy of their interpretation of dance informa-
tion. We also found no difference in the number of waggle runs followed between
light and dark conditions, suggesting that followers are able to follow equally well,
regardless of light condition. Thus, vision does not seem necessary to follow waggle
runs, and therefore A. mell'zifem followers must use a different sensory modality to

receive information from dancers.

Another perspective on follower behaviors in light versus dark is seen in our anal-
ysis of follower persistence, which measures how quickly bees drop out of the popula-
tion following a dancer with each successive waggle run. If followers were using vision
to follow waggle runs, we might expect that it would be more difﬁcult to maintain

contact with a dancer in darkness. However, we found no difference in follower per-

37

sistence between the two conditions, indicating that it is not any more difﬁcult for

followers to follow waggle runs in the dark.

4.1 Summary and future considerations

None of the behavioral measures used in this study indicate that A. mellifem use
vision as a sensory modality for dance communication. It remains possible that
vision does play a role in the dance language, either in other aspects of behavior not
measured in this study, or in follower attraction to dancing groups, as suggested by
Tautz and Rohrseitz (1998), although further experiments are necessary. One major
concern is whether the light condition was sufficiently bright to allow bees to see each
other. The light levels were dark compared to full sunlight in the open, but. likely
con‘iparable to levels experienced by A. florea nests, which are usually shaded by the
tree canopy. Brighter conditions also appeared to confuse bees trying to ﬁnd the
hive exit, and so the experimental levels were a. compromise between brightness and
non-interference in hive activities. That said, we believe the lights were sufﬁcient,
and the negative results reported here valid.

These results may' also be surprising in light of the evolutionary history of dance
communication in the Apis genus. Vision is thought to be important in most other
Apia species, and so a lack of evidence for continued use of vision by A. mellifem
suggests that they differ from other species not only in their nesting habits, but also
in the primary sensory modality that they use in dance communication.

At best, vision is not an important sensory modality for A. mellz’fem dance com-
munication, and may prove to not be used for communication at all. If so, we should
look at other sensory modalities as the critical means for dance communication in
this species. Just as there are stark differences in how dark the nest environment is

across Apis species, there are also great differences in the ‘noisiness’ of dances, with

38

A. florea having totally silent dances, A. dorsata. dances being perhaps facultatively
noisy , and A. mellife'ra (and A. cerana) producing prominent dance sounds. This
seems to be the likeliest successful avenue of future investigations, although because
of the unique sensory role of follower antennae as mechanosensors, it may be difficult
distinguish between audition and tactile sensation.

Understanding the mechanism of information transmission in the honey bee waggle
dance is important for a number of reasons. Primarily, knowledge of dance commu-
nication mechanisms will shed light on the evolution of the dance language through
the A 1125 genus. Based on our current knowledge, the ancestral sensory modality for
dance communication is likely vision, much like modern day A. ﬂor‘ea. If the results
of this study are supported by further research, there are two evolutionary possibil-
ities for the lack of vision as a sensory modality in A. mellz’fe'ra. The ﬁrst is that
vision was the ancestral sensory modality, and was subsequently lost by A. mell'ifera,
perhaps due to the evolution of cavity nesting. The second possibility is that vision
as a sensory modality was not used by ancestral A piss, and evolved independently in
open nesting species like A. ﬁerce and A. dorsata. Until we more fully understand
information transfer in A. rriellzlfe’ra, and indeed the other species within the genus,
we will not be able to resolve this ambiguity. Additionally, the honey bee dance
language is one of the most sophisticated examples of communication in non-human
animals. Understanding the mechanisms of this system is vital to determining how,
and why, evolution has produced such an advanced system so rarely throughout the
animal kingdom. We will be able to compare features of dance language evolution
with other advanced systems, like our own, to gain deeper understanding of language

and communication more generally.

39

Bibliography

Dyer, F. C., 2002. The biology of the dance language. Annual Review of Entomology
47, 917—49.

Esch, H., 1967. The sounds produced by swarming honey bees. Journal of Compar-
ative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology
56 (4), 408—411.

Griiter, C., Farina, W., 2009. The honeybee waggle dance: can we follow the steps?
Trends in Ecology & Evolution 24 (5), 242—247.

Judd, T., Jan 1995. The waggle dance of the honey bee: Which bees following a dancer
successfully acquire the information. Journal of Insect Behavior 8 (3), 343—354.

Kirchner, W. H., Dreller, C., 1993. Acoustical signals in the dance language of the
giant honeybee, ap'z's dorsata. Behavioral Ecology and Sociobiology 33 (2), 67—72.

Kirchner, W. H., Sommer, K., 1992. The dance language of the honeybee mutant
diminutive wings. Behavioral Ecology and Sociobiology 20, 181—184.

Lindauer, M., 1956. iiber die verst an digung bei indischen bienenn lo. Z. Vgl. Physiol
38, 521—57.

l\l-‘Ienzel, R., Jan 1981. Achromatic vision in the honeybee at low light intensities.
Journal of Comparative Physiology A: Neuroethology.

hi'lichelsen, A., 2003. Signals and ﬂexibility in the dance communication of honey-
bees. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and
Behavioral Physiology 189 (3), 165—174.

l\=lichelsen, A., Andersen, B. B., Storm, J., Kirchner, W. H., Lindauer, M., 1992.
How honeybees perceive communication dances, studied by means of a mechanical
model. Behavioral Ecology and Sociobiology 30 (3), 143—150.

lVIichelsen, A., Towne, W ., Kirchner, W., Kryger, P., 1987. The acoustic near ﬁeld of

a dancing honeybee. Journal of Comparative Physiology A: Sensory, Neural, and
Behavioral Physiology 161 (5), 633-643.

40

Nieh, J ., Tautz, J., Jan 2000. Behaviour-locked signal analysis reveals weak 200-300
hz comb vibrations during the honeybee waggle dance. Journal of Experimental
Biology 203, 1573—1579.

Rafﬁudin, R, Crozier, R. H., May 2007. Phylogenetic analysis of honey bee behavioral
evolution. Molecular Phylogenetics and Evolution 43 (2), 543—552.

Rohrseitz, K., Tautz, J ., 1999. Honey bee dance communication: waggle run direction
coded in antennal contacts? Journal of Comparative Physiology A: Neuroethology,
Sensory, Neural, and Behavioral Physiology 184, 463—470.

Sandeman, D., Tautz, J ., Lindauer, M., 1996. Transmission of vibration across honey-
combs and its detection by bee leg receptors. The Journal of Experimental Biology
199, 2585—2594.

Seeley, T., 1995. The Wisdom of the Hive: The Social Physiology of Honey Bee
Colonies. Harvard University Press.

Seeley, T., l\rlikheyev, A., Pagano, G., Jan 2000. Dancing bees tune both duration and
rate of waggle-run production in relation to nectar-source proﬁtability. Journal of
Comparative Physiology A: Neuroethology.

Tautz, J ., 1996. Honeybee waggle dance: recruitment success depends on the dance
floor. The Journal of Experimental Biology 199, 1375—1381.

Tautz, J ., Casas, J ., Sandeman, D., 2001. Phase reversal of vibratory signals in hon-
eycomb may assist dancing honeybees to attract their audience. Journal of Exper-

imental Biology 204, 3737—3746.

Tautz, J., Rohrseitz, K., Nov 1998. What attracts honeybees to a waggle dancer?
Journal of Comparative Physiology A: Neuroethology 183, 661—667.

Thom, C., Gilley, D., Hooper, J., Esch, H., 2007. The scent of the waggle dance.
PLoS Biology 5 (9), 1862—1867.

Towne, W. F., 1985. Acoustic and visual cues in the dances of four honey bee species.
Behavioral Ecology and Sociobiology 16 (2), 185—187.

Towne, W. F., Kirchner, W. H., 1989. Hearing in honey bees: Detection of air—particle
oscillations. Science 244 (4905), 686—688.

Von Frisch, K., 1967. The Dance Language and Orientation of Bees. Belknap Press
of the Harvard University Press.

Waddington, K., Kirchner, W., 1992. Acoustical and behavioral correlates of prof-
itability of food sources in honey bee round dances. Ethology 92, 1—6.

Wray, M., Klein, B., Mattila, H., Seeley, T., Jan 2008. Honeybees do not reject dances
for ‘implausible’ locations: reconsidering the evidence for cognitive maps in insects.
Animal Behaviour.

41

||l.l
H
l“
Ill"
ll"
I“
H
“
E91
1“
T"
I

1

11111111

1293 03063 3733

3