ORIGINAL PAPER

Diminishing returns: the influence of experience and environmenton time-memory extinction in honey bee foragers

Darrell Moore • Byron N. Van Nest •

Edith Seier

Received: 30 August 2010 / Revised: 5 January 2011 / Accepted: 5 January 2011 / Published online: 20 January 2011

� Springer-Verlag 2011

Abstract Classical experiments demonstrated that honey

bee foragers trained to collect food at virtually any time of

day will return to that food source on subsequent days with

a remarkable degree of temporal accuracy. This versatile

time-memory, based on an endogenous circadian clock,

presumably enables foragers to schedule their reconnais-

sance flights to best take advantage of the daily rhythms of

nectar and pollen availability in different species of flow-

ers. It is commonly believed that the time-memory rapidly

extinguishes if not reinforced daily, thus enabling foragers

to switch quickly from relatively poor sources to more

productive ones. On the other hand, it is also commonly

thought that extinction of the time-memory is slow enough

to permit foragers to ‘remember’ the food source over a

day or two of bad weather. What exactly is the time-course

of time-memory extinction? In a series of field experi-

ments, we determined that the level of food-anticipatory

activity (FAA) directed at a food source is not rapidly

extinguished and, furthermore, the time-course of extinc-

tion is dependent upon the amount of experience accu-

mulated by the forager at that source. We also found that

FAA is prolonged in response to inclement weather, indi-

cating that time-memory extinction is not a simple decay

function but is responsive to environmental changes. These

results provide insights into the adaptability of FAA under

natural conditions.

Keywords Honey bees � Time-memory � Foraging �Circadian rhythms � Food-anticipatory activity

Introduction

The honey bee time-memory (Zeitgedachtnis) enables

individual forager honey bees (Apis mellifera) to return to a

food source at the same time each day. This behavioral

rhythm is driven by an endogenous circadian clock, as

shown by the presence of food-anticipatory behavior with a

period close to 24 h, under constant environmental condi-

tions (Renner 1955, 1957; Beier 1968; Beier and Lindauer

1970; Frisch and Aschoff 1987). The time-memory clock

also satisfies Pittendrigh’s (1958) definition of a ‘continu-

ously consulted’ oscillator because honey bees can be

trained to collect food at virtually any time of day and

return to that particular food source at the appropriate time

on subsequent days, even in the absence of food (Beling

1929; Wahl 1932; Moore and Rankin 1983; Moore et al.

1989; Moore and Doherty 2009). Continuously consulted

clocks also are required for the time-compensated sun

compass (von Frisch 1950; von Frisch and Lindauer 1954;

Meder 1958). In contrast, most behaviors under circadian

control (e.g., locomotor activity, mating, oviposition,

responsiveness to pheromone, stridulation, etc.) are pro-

grammed to be performed only at certain fixed times of the

day (Brady 1981).

It is assumed that the time-memory allows honey bees to

schedule their foraging flights in anticipation of species-

specific rhythms of nectar secretion, although a direct

connection between the time-memory and foraging

behavior has not been established under natural conditions.

It is known, for instance, that honey bees forage most

actively at the time of day that coincides with the highest

D. Moore (&) � B. N. Van Nest

Department of Biological Sciences, East Tennessee State

University, Box 70703, Johnson City, TN TN 37614, USA

e-mail: moored@etsu.edu

E. Seier

Department of Mathematics and Statistics, East Tennessee State

University, Box 70663, Johnson City, TN 37614, USA

123

J Comp Physiol A (2011) 197:641–651

DOI 10.1007/s00359-011-0624-y

nectar concentrations for each flower species (Kleber 1935;

Butler 1945; Corbet and Delfosse 1984) or with the highest

total available sugar (Giurfa and Nunez 1992; Rabinowitch

et al. 1993). However, the role of the time-memory in these

temporal foraging patterns has not been differentiated from

other factors that may bring foragers to the flower patch,

such as discovery of the patch by scouts, reactivation of

foragers with previous experience at the patch, or recruit-

ment of new foragers to the patch by those foragers already

exploiting it. On the other hand, it is well established, from

experiments with artificial feeders, that forager honey bees

preferentially ‘remember’ the time of day corresponding to

the highest offered sucrose concentration despite collecting

lower concentrations of sucrose at the feeder before and

after the elevated time on the previous day (Wahl 1933).

It is commonly believed that the time-memory of for-

ager bees is ‘‘fairly easily extinguished without positive

reinforcement’’ (Saunders 2002) or that, upon encountering

a previously productive food source that is now empty, the

forager ‘‘will rapidly erase this from her memory, and not

visit again’’ (Tautz 2008). Clearly, rapid extinction of the

time-memory would seem to be adaptive since most natural

food sources are ephemeral. Accordingly, the colony is

constantly exploiting new food sources and abandoning old

ones according to their profitabilities (Butler 1945; Vis-

scher and Seeley 1982). Individual foragers make the

decision to abandon or to continue with a particular food

source by assessing information such as nectar concentra-

tion and flight time (Seeley et al. 1991) and by the waiting

time upon returning to the hive required before it can

unload its nectar to food-receiver bees (Lindauer 1948;

Seeley 1986, 1989; Seeley et al. 1996), a function of the

current nutritional status of the colony. If a forager aban-

dons a food source, presumably the spatio-temporal

memory for that particular source should be rapidly

extinguished, enabling the forager to become ‘unem-

ployed’ (Seeley and Towne 1992) so that it can be recruited

to a different, more productive source.

In contrast to the assumption that the time-memory is

rapidly extinguished so that foragers can easily switch from

low quality to higher quality food sources, it is also rea-

sonable to assume that foragers should retain a robust time-

memory for high quality sources that will last at least for a

few days during inclement weather (Saunders 2002). Fur-

thermore, a strong persistent time-memory should con-

tribute to the efficiency of resource exploitation by the

colony, in that productive food sources would not need to

be rediscovered each day by scouts.

It appears, therefore, that there may be two competing

selection pressures acting on honey bee time-memory

behavior: one promotes a rapid extinction of food-antici-

patory activity (FAA) so that foragers can switch quickly to

more favorable resources whereas the other favors a slower

extinction so that the time-memory may be retained over a

stretch of bad weather. How are these two apparently

antagonistic requirements of the time-memory reconciled

to the benefit of the colony? To answer these questions, we

first need to know more about the time course of the

extinction process. For the purposes of the present study,

‘extinction process’ will refer to the diminution of the overt

expression of the time-memory, the performance of

reconnaissance flights (i.e., food-anticipatory activity)

directed toward a previously rewarded, time-restricted food

source. It is certainly possible that some foragers may

retain an internal, spatiotemporal ‘memory’ for a particular

food source but remain in the hive rather than expending

energy on reconnaissance trips.

To gain insights into the extinction process, we began by

examining the performance of FAA in forager honey bees

that had been trained to collect sucrose solution from

artificial feeding stations at certain fixed times of day. In

particular, we explored the pattern of abandonment of the

food source over a span of several days by foragers with

different degrees of experience (i.e., days of training) at

that source. Previous work (Moore and Doherty 2009)

examined reconnaissance flights of time-trained foragers to

a previously rewarded feeding station on the first day in

which no food was provided: the probability of a forager

expressing FAA was dependent on the amount of experi-

ence accumulated at that source. The present study was

designed to address several questions concerning the con-

tinued expression of FAA over several days. Does the time-

memory extinguish rapidly, as previously assumed? Once a

time-memory is established, is the time-course of FAA

extinction the same for all foragers? Alternatively, do

different experience levels yield different rates of extinc-

tion? Next, in a second series of experiments, we observed

patterns of FAA extinction to determine how foragers

retain a functional time-memory through 1 or 2 days of

inclement weather. Our findings revealed an unexpected

influence of weather conditions on the expression of FAA

on subsequent days.

Materials and methods

Fair-weather experiments

To determine the influence of experience on the extinction

of FAA associated with time-memory, we conducted five

field experiments (numbered 1–5 in Table 1) in which

forager honey bees (Apis mellifera) were given different

numbers of days of training to a sucrose solution at a food

source located 100 m distant from the colony. Each of the

experiments involved a different colony. All of the training

and subsequent testing occurred during fair weather: bees

642 J Comp Physiol A (2011) 197:641–651

123

from each colony were actively foraging from sunrise to

sunset on each of these days. Experiments 1–4 involved

colonies that were kept in 3- or 4-frame glass-sided

observation hives (6,000–8,000 bees/colony). Experiment 5

involved a standard, commercial field colony (approxi-

mately 20,000 bees). To shield the observation colonies

from direct sunlight, they were housed in a protective shed.

All of the experiments were conducted in a field containing

wildflowers at the Marine Corps Armory site in Johnson

City, Tennessee. This site may be characterized as a series

of meadows among clusters of trees.

Bees were time-trained according to established meth-

ods (von Frisch 1967; Moore and Rankin 1983; Moore and

Doherty 2009). First, strips of filter paper soaked with

sucrose solution were placed at the hive entrance. Bees

feeding from the paper were then transferred to a small

table containing a petri dish (9 cm diameter, trimmed to a

height of 4 mm) filled with 2 M sucrose, centered over a

filter paper disc (15 cm diameter). This process was repe-

ated until foragers began flying to the table to collect

sucrose. The filter paper disc was scented with four drops

of essential oil of lavender, anise, or gardenia. Previous

experiments demonstrated that different scents produced

no detectable differences in time-accuracy (Moore and

Rankin 1983; Moore and Doherty 2009). The training table

was then moved in steps until it was 100 m distant from the

hive: this process typically was accomplished within 1 or

2 days. Recruitment of new foragers occurred throughout

the step-wise moving of the training table. While feeding,

all of these foragers were given a paint mark on the thorax

to distinguish them from foragers that were recruited on

subsequent days. All contact with the sucrose solution was

restricted to a predetermined training time (Table 1). At the

conclusion of the training time, the petri dish and table

were washed with water to remove all traces of sucrose and

the filter paper disc was exchanged for a new one.

The training phase proper began on the day immediately

following establishment of the training table at 100 m.

Bees already experienced at this feeder appeared at the

training table and, after collecting sucrose and returning to

the colony, recruited naıve foragers to this training station.

It is important to note here that 2 M sucrose and scent were

present only during the previously established training

time, which varied between 1.0 and 1.25 h in duration

(Table 1). Upon their first arrival at the station, new

recruits (the focal bees for the experiments) were individ-

ually marked using combinations of colored paint dots

(Testors Enamel: The Testor Corporation, Rockford, Illi-

nois, USA) applied to the thorax and abdomen (von Frisch

1967). These color codes allowed us to track newly

recruited foragers as they visited the feeding station during

training as well as their reconnaissance visits to the empty

station during the testing phase. At the end of the training

time, all traces of sucrose were eliminated with water, and

the empty petri dish was placed on the table with a new

(unscented) circle of filter paper. In four experiments,

training lasted 5 days. This yielded cohorts of foragers with

different amounts of experience (from 1 to 5 days) at the

training station because new foragers were recruited on

each successive day of training. Foragers skipping a day of

training were not used in the analyses. In one experiment,

training lasted 3 days, thereby yielding forager cohorts

with 3, 2, or 1 day(s) of experience at the training station.

For the testing phase, the feeding station was monitored

from 0800 to 1700 hours beginning on the day immedi-

ately following the last day of training and continuing for 3

or 4 days (called ‘test days’), depending on the experiment

(Table 1). The arrival times of all individually marked

foragers approaching and/or landing on the table were

recorded. The petri dish remained empty and no scent was

applied to the filter paper disc. During the testing phase, a

list of all individually marked bees visible by scanning both

sides of the observation colony was compiled at least twice

daily, providing a census of test bees still alive on each

particular test day. In one case (experiment 5, Table 1), a

census was obtained by presenting the scent and 2 M

sucrose once again at the previous training time on the day

immediately following the last scheduled test day. Foragers

Table 1 Summary of the eight

fair-weather experiments

performed in this study,

including the test day dates,

number of training and test

days, training time, scent used,

and whether or not the trained

foragers were censused

Expt. Test dates # Training

days

# Test days Training

time

Scent Census

1 August 2003 5 4 11:00–12:00 Anise Yes

2 July–August 2005 5 4 11:00–13:00 Lavender Yes

3 July–August 2006 5 4 10:00–11:15 Anise Yes

4 September 2006 5 4 14:30–15:30 Anise Yes

5 July 2006 3 3 12:15–13:30 Lavender Yes

6 June 2003 3 3 11:15–13:15 Anise No

7 July 2003 3 3 11:45–13:45 Gardenia No

8 August 2003 4 3 11:15–12:45 Gardenia No

J Comp Physiol A (2011) 197:641–651 643

123

previously trained to this source were reactivated by sev-

eral persistent foragers and returned to the feeding station.

Most relevant to this study were the relative proportions of

each experience cohort that returned to the training station

on each successive, unrewarded test day.

An additional three experiments (numbered 6–8 in

Table 1) were performed using standard 10-frame com-

mercial field hives (approximately 20,000 bees/colony).

Bees were trained and individually marked as described

above for the observation hive experiments. Depending on

the particular experiment, there were either three or four

training days but only three test days (Table 1). Training

times varied between 1.75 and 2.0 h in duration. No census

of marked foragers was performed on these colonies.

Inclement-weather experiments

In four experiments (numbered 9–12 in Table 2), honey

bee foragers were time-trained as described for the fair-

weather experiments, but at least one of the test days was

interrupted by rain or the imminent threat of rain. Three of

these experiments (numbers 9–11) used observation hives,

thus enabling a census of individually marked bees to be

conducted for each test day. One experiment (number 12),

however, used a commercial field hive and no census was

taken. The proportions of each experience cohort that

arrived on subsequent test days were compared with those

exhibited by similar-experience cohorts in the fair-weather

experiments. The inclement-weather observation hive

experiments were compared only with fair-weather cens-

used hive experiments. Likewise, the one field hive

experiment that encountered inclement-weather was com-

pared only to the uncensused, fair-weather field hive

experiments. Fisher’s Exact Test was used for statistical

comparisons of FAA performance (i.e., the number of

trained foragers inspecting and not inspecting the training

station) of like-experience cohorts between fair-weather

and inclement-weather experiments.

A logistic regression model was created with the

GENMOD procedure (with unstructured correlation)

using SAS software (SAS Institute Inc., Cary, NC, USA)

to describe the performance of FAA over consecutive

unrewarded test days with respect to both experience at

the food source as well as weather conditions. Both fair-

weather and inclement-weather behaviors were repre-

sented in the model. Only those data obtained from

censused experiments were used in the model. The

observations for each experience cohort in each experi-

ment were treated as repeated measures. The descriptive

parameters were (1) number of days of experience at the

food source, (2) the number of the unrewarded test day

(number of days since the last day of food at the training

station), (3) the presence or absence of inclement

weather on the test day, and (4) the presence or absence

of inclement weather on the previous day. To provide an

indication of sample sizes, the numbers of foragers

belonging to each experience cohort on test day 1 for all

of the experiments in this study are compiled in Table 3.

These values show small declines over subsequent test

days; the actual numbers of trained foragers inspecting

and not inspecting the food source each test day are

incorporated into the model.

Table 2 Summary of the four inclement-weather experiments performed in this study

Expt. Test dates # Training days # Test days Training time Scent Census

9 September 2003 3 3 12:45–13:45 Almond Yes

10 October 2003 3 4 14:00–16:00 Gardenia Yes

11 June 2004 5 6 14:00–16:00 Gardenia Yes

12 September 2002 4 3 11:33–12:33 Gardenia No

Table 3 Number of foragers in each experience cohort on test day 1 for all fair-weather and inclement-weather experiments

# Days of experience Experiment #

Fair-weather experiments Inclement-weather experiments

1 2 3 4 5 6 7 8 9 10 11 12

5 15 10 16 21 8

4 28 11 18 14 10 18 13

3 5 17 7 17 27 18 29 12 26 26 16 17

2 19 22 13 8 39 45 63 29 17 28 8 14

1 41 27 30 38 48 49 44 31 20 33 14 13

Total 108 87 84 98 114 112 136 82 63 87 64 57

644 J Comp Physiol A (2011) 197:641–651

123

Results

Fair-weather experiments

In all time-training experiments, forager bees collected

sucrose solution that was available only at a fixed time of

day for several consecutive days. Different cohorts

received different amounts of experience (i.e., days of

training) at the feeding stations. Over a span of several

(3–5) days after the food was withdrawn, foragers contin-

ued to visit the feeding station at the previously rewarded

time of day, with the well described degree of anticipation

(Moore 2001; Moore and Doherty 2009). The results from

a typical experiment (Experiment 2, Table 1) are illustrated

in Fig. 1. Although the response, measured as the number

of arrivals, diminished progressively over several days (i.e.,

exhibited a process of ‘extinction’), these reconnaissance

visits maintained a remarkable degree of accuracy with

respect to the training time. However, the extinction

response differed among the experience cohorts. The rel-

ative proportion of visits to the training station made by

foragers with greater experience (4 or 5 days compared to

3 or fewer days of training) at the food source increased

with the number of unrewarded test days. Highly experi-

enced foragers accounted for 52.3, 56.3, 65.9, and 76.0% of

visits made to the training station on test days 1, 2, 3, and 4,

respectively, despite accounting for only 24.1, 24.7, 25.3,

and 26.0% of the total number of trained bees still living on

those days.

For all of the fair-weather experiments, the proportion of

foragers returning to the feeding station (showing food-

anticipatory activity, FAA) on the unrewarded test days

depended upon the level of experience accumulated by

each forager at the station—the greater the experience, the

higher the FAA. The results from a single experiment

(Experiment 1, Table 1) are shown in Fig. 2a. All of the

foragers with 5 days of experience and approximately 89%

of foragers with 4 days of experience returned to the

training station on test day 1. In comparison, about 80, 68,

and 51% of foragers with 3 days, 2 days, and 1 day of

experience, respectively, demonstrated FAA on test day 1.

All of the experience cohorts showed a decline in FAA

over several days. For example, about 33% of foragers with

4 days of experience continued to reconnoiter the feeding

station on test day 4. Only those foragers with just 1 day of

experience showed complete FAA extinction by test day 4.

Similar trends were seen in all of the other fair-weather

experiments. Pooled data from all of the experiments in

which a census was performed (Fig. 2b) revealed a con-

sistent relationship between experience level and the pro-

portion of foragers exhibiting FAA as well as a decline in

FAA over several days. The average rate of FAA decline

was similar among the experience cohorts: 18.7, 18.2, 19.4,

17.4, and 11.4% per day for foragers with 5, 4, 3, 2, and

1 day(s) of experience, respectively, at the feeding station.

Similar FAA levels and time-courses of extinction were

observed in the uncensused field hives (experiments 6–8,

Table 1): pooled data from these experiments showed that

78, 62, and 27% of foragers with 3, 2, and 1 day of

experience, respectively, at the food source showed FAA

on test day 1. On test day 3, these values had declined to

29, 7, and 2%, respectively. The field hive results showed

that the properties of time-memory extinction are a general

phenomenon and not a consequence of small colony size

such as those used in the observation hive experiments.

Inclement-weather experiments

Based upon the results from the fair-weather experiments,

the honey bee time-memory (as measured by FAA)

apparently decays over time with a pattern that appears

very much like extinction of an unreinforced learned

5

10

15

20

25

Num

ber

of A

rriv

als

at T

rain

ing

Stat

ion

5

10

15

20

5

10

15

Time of Day (h)7 9 11 13 15 17

5

10

Trainingtime

Day 1

Day 2

Day 3

Day 4

Fig. 1 Example (experiment 2, Table 1) showing that honey bee

time-memory response (FAA) diminishes over successive days but

retains accuracy with respect to training time. Arrivals at the training

station are plotted in 15 min intervals from 07:00 to 17:00 hours for

four consecutive unrewarded test days. Black arrivals made by

foragers with 4–5 days of experience at the food source. Gray arrivals

of foragers with three or fewer days of experience. With each

successive test day, higher-experience foragers contribute a greater

proportion of the total reconnaissance flights to the training station

relative to low-experience foragers

J Comp Physiol A (2011) 197:641–651 645

123

behavior. It has been assumed for many years that the time-

memory is retained over a day or two of inclement weather

(Saunders 2002). Does this ability occur because the time-

course of FAA extinction, as seen in the previous set of

experiments, is adequate to ensure that at least a few for-

agers will still reconnoiter the food source, or does some

factor or set of factors associated with inclement weather

prolong the time-memory? If FAA is governed by a simple,

invariable decay function, then the response levels on any

particular test day after inclement weather should be sim-

ilar to those levels typically observed on that test day in the

absence of bad weather.

We conducted four extinction experiments (Table 2) in

which one or more unrewarded test days were interrupted

by inclement weather. In the first of these experiments

(number 9, Table 2), foragers were time-trained from an

observation colony for 1, 2, or 3 days and then their

arrivals at the training station were monitored for three

consecutive unrewarded test days. Rain occurred through-

out test day 2; all other test days were fair-weather days

(Fig. 3a). The proportion of trained foragers returning to

the feeding station (i.e., exhibiting FAA) was depressed

during the rainy test day 2, as expected. However, FAA

was surprisingly elevated on test day 3 (the day immedi-

ately following the rain event) relative to levels observed in

observation hive experiments in fair weather. For bees with

3 days of training, FAA on test day 3 was elevated, though

not significantly (Fisher’s Exact Test, P = 0.053). The

proportion of foragers with 2 days of training that showed

FAA on test day 3 was significantly higher (P = 0.002)

than that seen in fair-weather experiments; for bees with

1 day of training, FAA on test day 3 was not significantly

elevated (P = 0.875).

In a second experiment involving inclement weather

(number 10, Table 2), the arrivals of foragers from an

observation hive with 1, 2, or 3 days of time-training were

monitored at the unrewarded feeding station for four con-

secutive test days. Test day 2 was dark and cloudy; the

other test days were bright and sunny (Fig. 3b). On test day

3, bees with 3 and 2 days of training exhibited FAA in

significantly higher proportions (Fisher’s Exact Test:

P \ 0.0001 in both cases) than those observed in fair-

weather experiments. As in the previous experiment, the

proportion of foragers with just 1 day of training that

returned on test day 3 did not differ significantly from the

proportions observed in fair-weather experiments

(P = 0.998). The proportions remained elevated 2 days

after the threat of rain: for bees with 3 and 2 days of

training, the levels observed on test day 4 were signifi-

cantly higher (P = 0.002 and 0.001, respectively) than test

day 3 levels in fair weather experiments. As before, for

bees with 1 day of training, FAA was unchanged relative to

fair-weather results (P = 0.970).

In a third inclement weather experiment (number 12,

Table 2), bees from a standard 10-frame field hive received

4, 3, 2, or 1 day(s) of time-training and their arrivals at the

training station were monitored for three consecutive

unrewarded test days (Fig. 3c). On test day 1, light rain

occurred both before and after, but not during the

11:33–12:33 training time. On test day 2, the conditions

were partly cloudy before, overcast during, and heavy rain

after the training time. Relative to the proportions of for-

agers showing FAA on test day 3 in fair-weather experi-

ments with field hives, the levels were significantly

elevated for bees with 4, 3, and 2 days of training (Fisher’s

Exact Test: P = 0.037, P \ 0.001, and P \ 0.0001,

respectively) but not for bees with only 1 day of training

(P = 0.98).

Fig. 2 Fair-weather experiments: time-course of FAA extinction

depends on level of experience. a Proportion of time-trained bees

returning to the training station from each experience cohort on

successive unrewarded test days for experiment #1 (Table 1).

b Pooled results from all five censused fair-weather experiments.

FAA diminishes at a similar rate for all cohorts. Note high levels of

persistence (expression of FAA) on test day 1 for foragers in high-

experience cohorts and relatively low persistence levels for low-

experience cohorts

646 J Comp Physiol A (2011) 197:641–651

123

A final inclement weather experiment (number 11,

Table 2) was performed in which foragers received 5, 4, 3, 2,

or 1 day(s) of time-training. Their arrivals at the unrewarded

feeding station (not shown) were monitored for six consec-

utive test days. Light rain occurred throughout test days 1, 3,

and 5, but test days 2, 4, and 6 were partly cloudy throughout

the day. On test day 4, FAA levels were significantly elevated

for bees with 4 and 3 days of training (Fisher’s Exact Test:

P = 0.003 and P = 0.013, respectively) relative to the

proportions of foragers returning on test day 4 in fair-weather

experiments with observation hives. Although the propor-

tions appeared to be elevated for bees with 5 and 2 days of

training, the differences were not statistically significant

(P = 0.061 and P = 0.998, respectively). As in all of the

previous inclement weather experiments, the proportion of

bees with 1 day of training that showed FAA on test day 4 did

not differ from the proportion seen in fair-weather experi-

ments using observation colonies (P = 0.901).

A composite view of the differences in expression of

FAA between fair-weather and inclement weather experi-

ments is shown in Fig. 3d. Depicted are the proportions of

time-trained foragers returning to the training station on

test day 3, with respect to the various experience cohorts,

for all five censused fair-weather experiments and two

inclement weather experiments. In the latter, inclement

weather occurred on test day 2. The general trends were an

elevation of FAA on the test day immediately following the

day of inclement weather with the exception being the

cohort with just 1 day of experience at the feeding station.

A statistical treatment of these trends is provided in the

following section.

Linear regression model of foraging behavior

The degree of food-anticipatory behavior with respect to

both experience level and weather conditions may be

described by a linear regression model:

lnp

1� p¼ 0:63E � 0:84T � 1:48Rþ 0:87P

where p is the estimated probability of return to the training

station and 1- p the estimated probability of not returning.

Fig. 3 Inclement weather prolongs the expression of FAA. Graphs

show the proportion of time-trained bees returning to the training

station (i.e., expressing FAA) on successive test days, by experience

cohort. Arrows days with inclement weather. Asterisks significant

differences between FAA levels and those observed in fair-weather

experiments. a Experiment in which the presence of rain on test day 2

was followed by significant elevation of FAA in the 2 days-

experience cohort on test day 3. b Experiment in which the threat

of rain on test day 2 was followed by significant elevation of FAA on

both test days 3 and 4 for both 2 days- and 3 days-experience cohorts.

c Experiment in which rain (before and after, but not during the

training time) on test day 1 and more rain (after, but not during the

training time) on test day 2 was followed by significant FAA

elevation for cohorts with 2, 3, and 4 days of experience. d Summary

of test day 3 levels of FAA for experiments in which inclement

weather occurred on test day 2 (black diamonds) and fair-weather

experiments (white diamonds), with respect to experience levels

(number of days of training). Dotted lines indicate mean FAA levels

J Comp Physiol A (2011) 197:641–651 647

123

E represents the number of days of experience at the

training station, T is the test day (number of days since the

last day of food at the training station), R is the presence or

absence of inclement weather on the test day (1 if present,

0 if absent), and P is the presence or absence of inclement

weather on the previous day (1 if present, 0 if absent). The

model (summarized in Table 4) illustrates several charac-

teristics of time-memory behavior. First, with all other

conditions held constant, each additional day of experience

at the food source nearly doubles the odds of expressing

FAA (P \ 0.0001, odds ratio 1.88). Second, the odds of

returning to the feeding station are reduced by more than

half (P \ 0.0001, odds ratio 0.43) for each consecutive

unrewarded test day. Third, unsurprisingly, the odds of

performing FAA are severely reduced during poor weather

(P \ 0.0001, odds ratio 0.23). Finally, the odds of

expressing FAA more than double when inclement weather

occurs on the previous day (P = 0.0083, odds ratio 2.38).

Discussion

Our findings provide insights into the workings of a cir-

cadian pacemaker system within an ecological context, an

area of chronobiology that, to date, has received scant

attention. Specifically, the results indicate that foraging

behavior driven by the honey bee time-memory, based on a

continuously consulted timing system, is adaptable. The

fair-weather experiments indicate that the probability of a

forager honey bee returning to investigate a previously

profitable food source (i.e., exhibit food-anticipatory

activity, FAA) depends upon the amount of experience the

forager has accumulated at that source. Contrary to com-

mon belief (Saunders 2002; Tautz 2008), the expression of

FAA does not extinguish easily without reinforcement.

Instead, FAA levels decline (i.e., undergo a process of

extinction) over several days at rates that are approximately

the same regardless of the amount of prior experience

(Fig. 2b). Consequently, cohorts of foragers with relatively

low levels of experience show complete FAA extinction

sooner than high-experience cohorts because their FAA

levels are already lower when the source becomes empty

(e.g., test day 1 of our experiments). The inclement weather

experiments reveal that the expression of FAA is not

governed by a simple, inflexible decay function. Instead,

the time-course of FAA extinction apparently is adjusted in

response to bad weather. Levels of FAA on days imme-

diately following inclement weather are significantly

higher than those expected in the absence of bad weather

(Fig. 3a–d): the end result is a prolongation of FAA at the

previously productive food source. Interestingly, foragers

with only 1 day of experience at the food source do not

show the effect. Prolonging FAA presumably ensures that a

sufficient number of foragers will resume foraging at a

familiar site after a relatively long stretch of bad weather.

Alternatively, FAA prolongation may serve a different

purpose. The environment surrounding the honey bee col-

ony is constantly changing: on any given day, some plants

come into flower while others become depleted. After a

period of rain, some depleted plants may recover their

ability to yield nectar. Rather than depending on the flower

patch being discovered again by scouts, the prolonged FAA

(at least in some individuals) would ensure that the source

receives significant reconnaissance.

The mechanisms underlying the inclement weather

effect remain to be determined. One possibility is that a

reduction in honey stores may lead to increased FAA.

Another possibility simply is that a bee’s internal foraging

motivation (Hogan 1997), linked to a previously reinforced

circadian phase, if not dissipated by the performance of

foraging activity, is added to the drive expressed on the

following day. A similar circadian-gated process has been

proposed as a model for human sleep (Daan et al. 1984).

Yet another alternative is that, during periods of bad

weather, overall hive activity is reduced. Such reductions

in general hive activity may lessen potential memory-

interfering interactions within the colony (e.g., recruitment

dances, dance following, nectar and pollen receiving,

Table 4 Logistic regression statistics describing the behavior of foragers in response to experience and weather conditions

Parameter Estimate SE of estimate 95% Confidence limits Z P Odds ratio

Upper Lower

Constant -0.0001 0.2745 -0.5381 0.5378 0 0.9997 –

E 0.6294 0.1007 0.4320 0.8269 6.25 \0.0001 1.8765

T -0.8409 0.0668 -0.9719 – -12.58 \0.0001 0.4313

R -1.4843 0.2682 -2.0100 -0.9586 -5.53 \0.0001 0.2267

P 0.8687 0.3290 0.2239 1.5134 2.64 0.0083 2.3838

Model parameters E represents the number of days experience accumulated by the foragers

T = number of days since the last day of food at the training station, R = the presence (1) or absence (0) of inclement weather on the test day;

P = the presence (1) or absence (0) of inclement weather on the previous day

648 J Comp Physiol A (2011) 197:641–651

123

foraging departures and arrivals, etc.). Minimizing inter-

ference from events that occurred post-learning has been

successful at prolonging memory retention in animals as

diverse as cockroaches (Minami and Dallenbach 1946),

pond snails (Sangha et al. 2003), and humans (Jenkins and

Dallenbach 1924). However, in one experiment (Fig. 3c),

rain occurred on test days 1 and 2, but outside of the for-

agers’ training time, allowing foragers to investigate the

food source even on those otherwise rainy days. This

window of opportunity presumably also allowed other

foraging groups to exploit their own flower patches. The

fact that FAA levels were still high on test day 3 suggests

that inclement weather itself may be the relevant signal,

rather than some indirect consequence of the inability to fly

out of the hive on the previous day. The finding that for-

agers with 1 day of experience at the food source do not

show FAA prolongation suggests that the relevant signal is

effective only if the internal state of the time-memory

reaches some minimal excitatory level.

Although FAA in the honey bee is known as the ‘time-

memory’, it may or may not be founded upon a learning

process. The honey bee FAA simply may be the behavioral

expression of a circadian oscillation that is entrained by

periodic food presentations (Frisch and Aschoff 1987). In

contrast to this non-learning explanation, Gallistel (1990)

proposed a computational model in which the time-mem-

ory is established when a forager encounters the food on

several successive days such that the onset times of

occurrence of this food are significantly clustered, thereby

indicating one particular circadian phase (Gallistel 1990).

In this scenario, the forager has been conditioned to asso-

ciate the time of day (read from the continuously consulted

circadian clock) with the presence of food. Two findings

from the present study are consistent with the honey bee

time-memory incorporating a learning component. First,

the proportion of foragers returning to the food source

increased with the number of days of training, suggesting

stronger associations between time of day and the presence

of food. Second, in contrast to the other experience cohorts,

foragers with just 1 day of training failed to show FAA

prolongation in response to inclement weather. This phe-

nomenon perhaps may reflect differences in responsiveness

to stimuli in the transition between the early and late stages

of long-term memory (Menzel 1999).

Another intriguing possibility, built on the assumption

that the honey bee time-memory is, indeed, an associative

learning phenomenon, is that prolongation of FAA in

response to inclement weather may represent spontaneous

recovery, an effect often seen in extinction studies (Bouton

2007; Terry 2003). For example, in the honey bee, after

repeated pairings of food (the unconditioned stimulus, US)

and odor (the conditioned stimulus, CS), proboscis exten-

sion can be elicited by presentation of the CS alone

(Takeda 1961). Prior to the CS-US association, odor does

not elicit the response. After such learning has occurred,

repeated presentations of the CS alone, however, yield a

steady decline in response performance (i.e., extinction of

the response) and, if there is a pause in the delivery of the

unreinforced CS, the response shows a so-called sponta-

neous recovery, reassuming much of its initial magnitude

(Takeda 1961; Bitterman et al. 1983). The level of spon-

taneous recovery is higher with increasing numbers of

conditioning trials but diminishes as the number of

extinction trials is increased (Sandoz and Pham-Delegue

2004). Applying learning and extinction theories to the

honey bee time-memory behavior, sucrose is the US and

time of day (phase of the circadian cycle) serves as the CS.

Repeated CS-US associations lead to FAA, the conditioned

response. Successive test days, in which there is no sucrose

offered at the training station, are analogous to extinction

trials: FAA is highest on the first test day and declines

steadily with each subsequent test day. According to this

scenario, days of inclement weather interspersed among

fair-weather days function as pauses within extinction tri-

als, thus enabling spontaneous recovery (expressed as

elevated levels of FAA). Our finding that foragers with just

1 day of training show little or no elevated FAA after a day

of inclement weather is in accordance with the fact that the

degree of spontaneous recovery depends on the amount of

prior conditioning (Sandoz and Pham-Delegue 2004).

However, the spontaneous recovery interpretation is not

consistent with the results of the experiment shown in

Fig. 3c in which rain occurred on test days 1 and 2, but

outside of the training time. Despite the fact that a high

proportion of trained foragers showed FAA on both test

days 1 and 2 (i.e., there was no interruption in the sequence

of extinction trials), FAA nevertheless was elevated sig-

nificantly on test day 3, except for those bees with just

1 day of training.

One experiment that may begin to differentiate among

the various hypotheses proposed here to account for the

prolongation of FAA will be to impede foraging (such as

by blocking the hive exit) during one sunny day in the

middle of a series of extinction days and measure FAA on

the following day, also sunny. If recovery of FAA still

occurs, then inclement weather will be shown to have

nothing to do with the pattern of responses observed.

Further experimentation will be necessary to discriminate

among the remaining hypotheses that rely on foragers

being impeded from leaving the hive—deterioration of

honey stores, elevation in foraging motivation, decrease in

memory interference, and spontaneous recovery.

Our empirical findings demonstrate that food-anticipa-

tory behavior in forager honey bees exhibits a considerable

degree of inter-individual variability. Experience matters:

the proportion of time-trained foragers reconnoitering a

J Comp Physiol A (2011) 197:641–651 649

123

food source (i.e., exhibiting FAA) is a function of the

amount of experience accumulated at that source by the

individual foragers. As detailed above, the probability of

exhibiting FAA is further subject to modification depend-

ing upon weather conditions, but (at least in our experi-

ments) only if the forager has accumulated at least 2 days

of experience at that food source. The foraging group,

therefore, is not homogeneous. Rather, each forager is a

unique agent and its behavior is a product of its own dis-

tinct set of experiences and influences. There are many

open questions regarding possible interactions between the

time-memory rhythm and the environment. For example,

does FAA or its decay rate depend upon the time of day at

which it became established? How do different reward

profitabilities influence time-memory extinction? One

obvious next stage of inquiry into the honey bee time-

memory is testing exactly how this level of complexity and

adaptability translates into benefits for the colony, espe-

cially under a variety of resource conditions.

Acknowledgments We thank Trevor England, Neha Barakam, Brad

Barker, Arianna Bruno, Stefani Coleman, Christopher Cronan, Patrick

Doherty, Erica Edmund, Alison Gagan, Forrest Harrison, Jenny

Hoekstra, Jonathan Humberd, Nathan Humphrey, Jennifer Johnson,

Tannille King, Guy Kramer, Adam Lewis, Christopher Litchfield, T.J.

Metcalf, Jaime McManus, Charles Miller, Samara Miller, Somer

Miller, Mary Ann Moore, Matt Otto, Caleb Paquette, Lia Pun-Chuen,

Ryan Rice, Lindsay Slemp, Will Smith, Kim Stroup, Jack Whitaker,

Jeremey Whitaker, Anthony Whitted, Ashley Williams, and Dmitri

Yampolsky for help with the field experiments. We also thank two

anonymous reviewers for valuable suggestions that strengthened the

manuscript. This work was supported by funds from the National

Research Initiative of the USDA Cooperative State Research, Edu-

cation, and Extension Service, Grant No. 2006-35302-17278 (D.M.)

and the Department of Biological Sciences, East Tennessee State

University, Denise I. Pav Research Award (B.V.N.). The present

study complies with the current laws of the country in which the

experiments were performed, including the ‘Principles of Animal

Care’, Publication No. 86-23, revised 1985 of the National Institutes

of Health.

References

Beier W (1968) Beeinflussung der inneren Uhr der Bienen durch

Phasenverschiebung des Licht-Dunkel-Zeitgebers. Z Bienenfors-

chung 9:356–378

Beier W, Lindauer M (1970) Der Sonnenstand als Zeitgeber fur die

Biene. Apidologie 1:5–28

Beling I (1929) Uber das Zeitgedachtnis der Bienen. Z vergl Physiol

9:259–388

Bitterman ME, Menzel R, Fietz A, Schafer S (1983) Classical

conditioning of proboscis extension in honeybees (Apis melli-fera). J Comp Psychol 97:107–119

Bouton ME (2007) Learning and behavior: a contemporary synthesis.

Sinauer, Sunderland, Massachusetts

Brady J (1981) Behavioral rhythms in invertebrates. In: Aschoff J (ed)

Handbook of behavioral neurobiology, vol 4. Plenum Press, New

York, pp 125–144

Butler CG (1945) The influence of various physical and biological

factors of the environment on honeybee activity. An examination

of the relationship between activity and nectar concentration and

abundance. J Exp Biol 21:5–12

Corbet SA, Delfosse ES (1984) Honeybees and the nectar of Echiumplantagineum L. in southeastern Australia. Aust J Ecol

9:125–139

Daan S, Beersma DGM, Borbely AA (1984) Timing of human sleep:

recovery process gated by a circadian pacemaker. Am J Physiol

246:R161–R183

Frisch B, Aschoff J (1987) Circadian rhythms in honeybees:

entrainment by feeding cycles. Physiol Entomol 12:41–49

Gallistel CR (1990) The organization of learning. The MIT Press,

Cambridge

Giurfa M, Nunez JA (1992) Foraging by honeybees on Carduusacanthoides: pattern and efficiency. Ecol Entomol 17:326–330

Hogan JA (1997) Energy models of motivation: a reconsideration.

Appl Anim Behav Sci 53:89–105

Jenkins J, Dallenbach K (1924) Oblivescence during sleep and

walking. Am J Psych 35:605–612

Kleber E (1935) Hat das Zeitgedachtnis der Bienen biologische

Bedeutung? Z vergl Physiol 22:221–262

Lindauer M (1948) Uber die Einwirkung von Duft- und Ges-

chmacksstoffen sowie anderer Faktoren auf die Tanze der

Bienen. Z vergl Physiol 31:348–412

Meder E (1958) Uber die Einberechtung der Sonnenwanderung bei

der Orientierung der Honigbiene. Z vergl Physiol 40:610–641

Menzel R (1999) Memory dynamics in the honeybee. J Comp Physiol

A 185:323–340

Minami H, Dallenbach K (1946) the effect of activity upon learning

and retention in the cockroach, Periplaneta americana. Am J

Psych 59:1–58

Moore D (2001) Honey bee circadian clocks: behavioral control from

individual workers to whole-colony rhythms. J Insect Physiol

47:843–857

Moore D, Doherty P (2009) Acquisition of a time-memory in forager

honey bees. J Comp Physiol A 195:741–751

Moore D, Rankin MA (1983) Diurnal changes in the accuracy of the

honeybee foraging rhythm. Biol Bull 164:471–482

Moore D, Siegfried D, Wilson R, Rankin MA (1989) The influence of

time of day on the foraging behavior of the honeybee, Apismellifera. J Biol Rhythms 4:305–325

Pittendrigh CS (1958) Perspectives in the study of biological clocks.

In: Buzzati-Traverso AA (ed) Perspectives in marine biology.

University of California Press, Berkeley, pp 239–268

Rabinowitch HD, Fahn A, Meir T, Lensky Y (1993) Flower and

nectar attributes of pepper (Capsicum annuum L.) plants in

relation to their attractiveness to honeybees (Apis mellifera L.).

Ann Appl Biol 123:221–232

Renner M (1955) Uber die Haltung von Bienen in geschlossenen,

kunstlich beleuchteten Raumen. Naturwissenschaften

42:539–540

Renner M (1957) Neue Versuche uber den Zietsinn der Honigbiene. Z

vergl Physiol 40:85–118

Sandoz J-C, Pham-Delegue M-H (2004) Spontaneous recovery after

extinction of the conditioned proboscis extension response in the

honeybee. Learn Mem 11:586–597

Sangha S, McComb C, Lukowiak K (2003) Forgetting and the

extension of memory in Lymnaea. J Exp Biol 206:71–77

Saunders DS (2002) Insect clocks. Elsevier Press, Boston

Seeley TD (1986) Social foraging by honeybees: how colonies

allocate foragers among patches of flowers. Behav Ecol Socio-

biol 19:343–354

Seeley TD (1989) Social foraging in honeybees: how nectar foragers

assess their colony’s nutritional status. Behav Ecol Sociobiol

24:181–199

Seeley TD, Towne WF (1992) Tactics of dance choice in honey bees:

do foragers compare dances? Behav Ecol Sociobiol 30:59–69

650 J Comp Physiol A (2011) 197:641–651

123

Seeley TD, Camazine S, Sneyd J (1991) Collective decision-making

in honey bees: how colonies choose among nectar sources.

Behav Ecol Sociobiol 28:277–290

Seeley TD, Kuhnholz S, Weidenmuller A (1996) The honey bee’s

tremble dance stimulates additional bees to function as nectar

receivers. Behav Ecol Sociobiol 39:419–427

Takeda K (1961) Classical conditioned responses in the honey bee.

J Insect Physiol 6:168–179

Tautz J (2008) The buzz about bees. Springer, Berlin

Terry WS (2003) Learning and memory: basic principles, processes,

and procedures. Allyn and Bacon, Boston

Visscher PK, Seeley TD (1982) Foraging strategy of honeybee

colonies in a temperate deciduous forest. Ecology 63:1790–1801

von Frisch K (1950) Die Sonne als Kompass im Leben der Bienen.

Experientia 6:65–69

von Frisch K (1967) The dance language and orientation of bees.

Harvard University Press, Cambridge

von Frisch K, Lindauer M (1954) Himmel und Erde in Konkurrenz

bei der Orientierung der Bienen. Naturwissenschaften

41:245–253

Wahl O (1932) Neue Untersuchungen uber das Zeitgedachtnis der

Bienen. Z vergl Physiol 16:529–589

Wahl O (1933) Beitrag zur Frage der biologischen Bedeutung des

Zeitgedachtnisses der Bienen. Z vergl Physiol 18:709–717

J Comp Physiol A (2011) 197:641–651 651

123