ApiaryActiveLive
Try: pause · settings · learn · wipe
← Community / Reading Room
CT
nature · 24 min read

Count the Bees on Two Willows, Then Count the Flowers

By Austin Little

By Austin Little

AI disclosure: Drafted with AI assistance for Apiary. The two-shrub comparison is an invented teaching example; scientific results are attributed to the relevant studies.

You might make it past a willow twice before recognizing that it has flowers. There’s no ring of showy petals and sometimes not much leaf to frame the view. The flowers are grouped into catkins, the little cylinders or tufts you notice along the branches. On a warm spring day, a willow in bloom may have enough prospecting insects to make the plant audible. On another visit it can look nearly unoccupied. Neither snapshot tells you what kind of food the insects found, whether they transferred pollen, or how many bees were in the neighborhood.

A study in New Brunswick makes the uncertainty useful. Researchers compared which flower clusters bees visited on male and female willows. In one of the controlled garden comparisons, most recorded Andrena visits on each of two willow species were to male plants. The percentages differed between species, and the strength of the preference varied through the day. The headline case is not that a yellow catkin always wins. It is that a question as simple as “which tree got more bees” can have more than one defensible answer, depending on the plants, timing and unit of comparison. Mosseler and colleagues: bee foraging on three willows

What a catkin can tell you if you get a good look

A catkin is a cluster of individual flowers, not a single flower with lots of long petals. The form is common enough on some woody plants that you might assume all catkins work the same way. They do not. On the willows in the New Brunswick study, male and female flowers occur on separate plants. The botanical shorthand for this is dioecious. The distinction is useful because a branch with pollen-bearing flowers is not expected to be the branch carrying developing seed capsules on those plants. In different willow species, flowers may become evident before or around the arrival of leaves. US Forest Service: yellow willow · US Forest Service: black willow

Male plants in the New Brunswick work had flowers providing pollen and nectar. Female catkins could offer nectar but are not the source of male pollen. This gives a bee a reason to behave differently between the two, depending on what it is collecting and what else is available. It also gives the plant a reason to matter to insects even on the side of the pair that does not provide pollen. If you are looking from a footpath, you may see some details clearly and need to leave others unconfirmed. An angle that shows individual flower parts is more informative than one that shows only the tip of a fuzzy catkin.

Some flowers with catkins are associated with wind pollination, but you cannot determine the mechanism just from that shared shape. In willows, insects can be important to pollen transfer; accounts of black willow also describe wind carrying pollen. The useful correction is to avoid assuming a single pollination route from the shape of the display. Which mechanism contributes most in a particular place is a separate question. US Forest Service: black willow reproduction

If you are trying to identify the host species, the plant’s full growth form and leaf details may help later. Photo angles that work for insect ID are often too tight for tree ID. Save the whole-plant view, the branch view and the close view. Those pictures answer different questions. If you cannot tell which cluster is pollen-bearing, leave the sex unclassified in your count. You can still report that both sites had visitors and revisit when the floral features are clearer. A table with an “unknown” category makes the observations usable without guessing.

There is a difference between a favorite plant and a favorite food

The word “specialist” is easy to overread. It can refer to the pollen a bee typically collects for its offspring, not every place that an adult bee visits. Minnesota’s Department of Natural Resources describes the eastern willow miner bee, Andrena bisalicis, as a Salix specialist in early spring. That is information about a named bee. It does not establish that any miner bee you can photograph during a willow visit is that species. Nor does it tell you what a bee in a different genus is doing on the next branch. Minnesota DNR: eastern willow miner bee

The reverse caution is also useful. A bee known to many people by association with a different plant may use willows. James Watson and colleagues reported evidence of willow pollen collection by a bee usually regarded as associated with blueberries and other plants in the heath family. The setting was southeastern Manitoba. The finding does not show that every population of that bee uses willow equally or establish how much food comes from it over a whole season. It does show why the common name and broad reputation should not prevent you from recognizing a new documented observation. Watson and colleagues: willow pollen collection by a blueberry specialist bee

It is easy for a garden visitor to conflate what they can see with what they want to know. You can often see an individual land, move across a cluster or leave. A good image might show a load that appears to be pollen. You cannot necessarily identify its plant source without other evidence. The insect could have come from another flowering plant before the part of the trip you watched. That is why a study that checks pollen is asking a more specific question than one that counts visits.

Think carefully about what a useful caption should say. “A small bee moving across a willow catkin in March” is a description if that’s what the image shows. “Willow specialist pollinates this shrub” compresses insect identity, behavior and outcome into a sentence that one photo may not support. A caption can be short without dropping the distinction. If all you can tell is that an insect was there, give that encounter its own place in the record and let another piece of evidence do the rest of the work.

The New Brunswick findings, in their proper scale

In the garden test, Mosseler and colleagues compared Andrena visitation on male and female Salix eriocephala and S. interior. They reported that about 87% and 71% of visits, respectively, were to the male plants within those comparisons. Counts in another part of the project, at a reclaimed mine site, involved an additional willow species and a broader set of bees. The researchers also found that the number of plants available in flower could change the way a summary looked. That’s a useful field habit to borrow even if you have no intention of replicating their study. Mosseler and colleagues: bee foraging on three willow species

An important limit: these are results for the species, places, sample and methods of that research. They are not percentages to use as rules for every backyard or every willow. The paper also discusses earlier observations in which some willow species had different patterns. Nor do the reported shares show that female plants are unnecessary. A male plant can supply pollen and nectar; a female plant is a different part of the reproductive system. A plant that gets fewer visitors can still receive pollen and set seed. Visit frequency by itself settles neither nectar output nor the success of the next generation.

The time-of-day result deserves the same care. Activity on male flowers increased and decreased over the observing day, but the researchers did not interpret that as proof the bees collectively switched which sex they preferred midday. If you only count on your walk to work, that might tell you which visible plants got insects along your route at that hour. It can’t establish what happened during the rest of the day. When comparing observations, say what parts of the clock were covered.

A separate research team working with black willow assessed the kinds of insects visiting flowers and the activity of particular native Andrena species. That work used its own methods, sites and questions. From an article reader’s perspective, this is a reason to compare the design of studies before comparing the apparent size of the results. One set of counts can be useful for describing a floral insect community, and another for testing preferences under a defined comparison. Neither is made stronger by borrowing the denominator from the other. Research study: black willow floral insects and native Andrena bees

A worked observation: a louder bush versus a more visited cluster

Suppose a volunteer watches two nearby willows from a public path. Over a pair of equal-length observation windows, Willow A receives 20 recorded landings and Willow B receives 10. A has twice as many observed landings. If the volunteer wants to know which shrub drew more activity overall in those windows, that is a reasonable answer. It does not matter that the plants are different sizes for that limited question.

But suppose Willow A has about 100 visible flowering clusters in the observed area and Willow B has about 20. The comparison now has a second reading. A’s tally is 0.20 recorded landings per visible flowering cluster for the window, and B’s is 0.50. These are rough exposure rates, not probabilities that any particular cluster received a visit. Many visits may have occurred on the same cluster or by the same insect. The two answers are compatible because they answer different questions: where was more total action, and how did the tally compare to the scale of the display?

Now add a small complication. The volunteer counted all of the visible catkins on B, but leaves and other branches hid much of A. This does not imply the arithmetic is wrong. It means “visible” is doing important work. The counts and flowering estimate should cover the same defined area. If most landings were visible on the near face but the denominator included a guess about the far face, the comparison mixes observation and imagination. For a second visit, the volunteer can choose one clearly visible branch on each plant, use the same length of time, and record exactly which area was watched.

Then ask what would change if the purpose were seed production rather than attracting insects. Information about the visits is still relevant, but the evidence needed now changes. You would want to know where pollen arrived and whether a later reproductive outcome occurred. A photograph showing twenty landings on the busier male shrub would tell you little by itself about seed set on the quieter female shrub. The denominator cannot rescue a measurement that addresses the wrong question.

How to watch without making your records harder to compare

A quick observation is not a field experiment, but you can make it useful to yourself:

  1. Choose a safe, permitted viewpoint. Do not enter a closed riparian area or climb out over water to follow a bee.
  2. Decide what an event is. Counting a clear landing on a cluster is easier to repeat than counting every swing past a plant. Record uncertain events separately.
  3. Use a short, equal time window for comparisons. Let a few minutes finish even when it seems nothing is happening. Keep the zero, since it is tied to a real observation window.
  4. Note the conditions. Wind, sun, time and current bloom stage can help explain why successive visits differ. Count roughly comparable visible clusters if you plan to compare activity per cluster.
  5. Write down the identification level you actually have. Willow uncertain, bee-like insect or visible pollen-looking load are legitimate intermediate descriptions. Photograph from a safe position when the insects are already there.
  6. Repeat in another window when possible. A warm midday watch and a cooler return visit may provide different evidence. Keep the separate rows before combining them.

There is no need to trap a bee or pick a willow branch to practice this. It can be as simple as two five-minute notes from an established trail. To reduce your own bias, choose the area before the count instead of choosing the most active branch halfway through. You may still learn interesting things outside the selected area; put them in a separate note rather than recategorizing them as if they occurred in the defined window. That preserves the usefulness of both kinds of record.

Consider a stormy morning when no insect lands during your watch. If you write “no bees here,” the result will be harder to interpret later. If you write “no confirmed landings in five minutes at this cluster in gusty weather,” you have given the observation a scale. When someone comes back on a mild afternoon and watches several visits, that doesn’t make the morning record false. It is a reminder that insect activity and the conditions under which we detect it both change.

What this means for a yard, a streambank or a school project

A willow can help fill an early season flowering window where the species and site fit. It is not a universal substitution for planning flowers across a season, and the results of bee visits are not a permit to disturb a managed creek or install woody plants wherever there is open ground. For an actual planting, start with the site’s locally appropriate plants, space, access and hydrology. Ask local restoration staff or cooperative extension which approach fits. A plot designed to restore native riparian vegetation may have different goals from a small yard’s bee observation spot.

The amount of visible bloom is worth noting when discussing plant value. Suppose a park’s willow patch gets very busy for one week, while the nearby area appears quiet before and after. That can be a valuable brief resource. It can also show which timing gaps deserve more looking. A story can give credit to the busy week and still set a separate question about continuity. Counting a bloom as useful does not require it to cover the whole period.

For a school or community group, have different observers compare notes on the same plant without forcing agreement first. One person may record tiny hovering insects. Another may use a stricter landing definition and tally only a few. Both can explain what they were counting. If the group then repeats the watch with a shared definition, it will encounter a central rule of field work: agreement starts with describing the measure, not with averaging incompatible numbers. A surprising difference can give the next watch a better design.

The best part of this little study is that it changes what you notice. A shrub you thought of as an ordinary bank-side plant turns out to have two kinds of flower displays across different individuals, a schedule of visits that can change across one day and questions you can’t answer from the busier branch alone. You can leave your field notes small and still make a solid contribution. Say what landed, on which measurable part of the display, and under what conditions. Then look again when the plant changes. A useful count makes the next observation easier to ask for.

Decide whether you are measuring a display, a choice or an encounter

A landing count can serve at least three questions. A visit-total question concerns the activity people can observe around a defined display. An exposure question asks about that activity relative to the amount of flowering display in view. A choice question asks whether animals favor one option when there are meaningful alternatives they can actually encounter. A public observation can produce evidence for the first two in a narrow window without establishing the third. Writing the desired question in advance prevents a striking afternoon tally from silently changing jobs.

A fictional plant near a sunny path illustrates the difference. Its flowers show a lot of activity; a shaded plant down the trail shows less. A sign explaining overall visitor activity could report that contrast as observed. A claim of choice, such as “these bees favor this willow type,” must also take seriously the light, available plants, flower stage, and paths available to moving insects. The animals may never have had comparable access to the two flowers in the time you watched. A preference experiment defines a comparison those animals could make. Your field card does not need to rival the experiment, but it should not adopt the same verb unless its evidence does the same work.

Set down an actual unit before entering any counts. “Observed display” could mean one small branch framed between two forks, or the part of a shrub visible from a path. “Active flowering cluster” must mean something a repeat viewer can see without taking samples. Own the definition; it will not represent the same practical scale at every site. To improve comparability across people, write one example that counts and one that doesn't. A cluster whose details can be seen well enough to judge flowering can enter some studies. A fuzzy pale form behind other branches may enter a separate unknown tally. Definition disagreements are often the real differences between observers, and they can be addressed in the notebook instead of through post hoc exclusions.

A reasonable field card can hold the intended question, target shape, view status, rough time, time actually watched, and stage definition. If you then decide that the plant was too large to score reliably, retain the original intention and define an amended trial as such. Put “partial front branch; back side not covered” in plain language rather than leaving room for someone to cite the number as plant-wide. Compare results only once they have shared enough definitions to discuss a common question.

Rotate the observer and the clock together

Time can act like a hidden advantage when one observer always starts with the same plant. Consider four fictional opportunities: the sunlight first reaches plant A in the morning, clouds arrive during the return watch, both branches are exposed on a still noon outing, and the park closes before the late slot. On a rushed field card, the line for plant B can end up containing two cold observations while A receives the fair weather. Equal minute totals will not by themselves make the conditions comparable. Before the visits, decide which differences you intend to compare and which you intend to note honestly as uncontrolled.

A simple paired design for an informal group can alternate the starting plant across opportunities, or use two observers watching at the same clock time from comparably safe stations. If two people are available, rotate their assignments on later comparable days. Treat this as an attempt to expose order and observer effects, not as a claim that the experiment is exhaustive or controlled. Record all starts and gaps, including the time it took to walk to the next plant. Sudden weather changes can happen between two theoretically adjacent windows. When that happens, discussion may reveal why preserving separate time blocks is more informative than using one pooled row.

This matters if your audience wants to compare the apparent flower values of the two plants across a whole spring. A visitor activity count depends both on the animals that use a place and on the period a person spent looking. If a single unusually quiet interval provides most of the evidence for a plant, say so. Differences between willows might be robust across stories and windows, or a striking comparison may be nearly all from one interval. Show the raw event/time lines as the starting point. A return visit is a chance to ask the same limited question in another sufficiently comparable setting, not permission to drop the earlier visit from the record.

Be careful about conditional rules that sound efficient but alter the question. “Start counting when the first bee arrives” is not the same as a randomly chosen or scheduled visit. It discards an initial quiet interval by construction and therefore can't fairly quantify when activity was not visible. “End when things settle down” gives the viewer a vague choice about what qualifies as an end. Set an easily followed period before looking for action, then let quiet time count. If you must pause to move clear of an approaching group, record the gap as time not watched; don't mark the lost minutes all quiet.

Independent tallies do not always count independent animals

A field observer may see three landings by the same individual, two swift arrivals that cannot be separated confidently, and one arrival whose start was missed. Adding six events as a shorthand for bee abundance would be misleading. A count of landings is perfectly legitimate when repeated behavior is part of the question. The shift happens when a reader turns the same tally into “six bees visited” or “the shrub supported six individual pollinators.” Without an appropriate identification design, the two answers can differ by more than the count suggests.

There are useful ways to preserve context without attempting to recognize each bee. Say whether activity appeared clustered in short bursts, dispersed across the watch, or dominated by repeated movements near one side of the defined branch. Note unambiguously visible repeated moves when a specific individual stayed continuously in frame. Don't attach an identity merely because one insect looks about the same size or has a similar color later in the sequence. When in doubt, your event unit can remain valid even after individual identity stops being knowable.

Now imagine one observer entering a tally each time contacts with a catkin are visible, while another tallies only arrivals from outside the branch frame. A bee flies between two clusters on that branch without leaving the frame. Both observers may be entirely faithful to their own rules and still return different numbers. Rather than resolving the difference by averaging, teach the readout with an example. “Within-frame moves not counted as new arrival” tells the second observer where the events start; “each clear new cluster contact counts” tells the first. Preserve the phrase defining the count wherever that number travels.

Note late or ambiguous events explicitly. Did contact happen just outside the five-minute period, or did a rapid movement make it unclear that the targeted cluster was touched? A useful tally sheet has a small separate place for uncertain events and enough scratch notes to explain why they were uncertain. If the same ambiguity repeats across several windows, you may need to change the definition for future data. A record of when the rule changed lets later readers understand whether older and newer watches can actually be pooled. Good consistency can mean resisting a comparison that becomes easier to calculate than to defend.

Join the display snapshot to the watch without inventing precision

Cluster counts offer a better description of the display if they correspond closely to what visitors and observers actually used. But a single estimate can be unreliable where the whole display is large, partly hidden or progressing across the watch. State whether you counted exact visible units in a tightly bounded frame, estimated a larger figure, or bracketed an approximate number. Avoid attaching a decimal place to an approximate flower count solely because a spreadsheet displays one. It is more honest to report that one plausible category spans fifteen to twenty clusters than to promote a best photographic guess to a physically measured number.

Take a synthetic example with eight contacts in a five-minute watch and an indistinct display estimated at ten to sixteen visibly available clusters. The simple ratio ranges from 0.50 to 0.80 observed contacts per visible cluster in the window across that denominator bracket. The range is a sensitivity illustration, not an error bar for a population estimate or a claim about how any given cluster was used. For another plant with nine contacts and eighteen clusters visible, the ratio of 0.50 overlaps one end of the first case's bracket. This may be enough to avoid declaring a strong difference on the back of a dubious flower estimate. Clearly state what counts as a cluster and what part of the canopy each number covers before doing any ratio.

A photograph can help document a denominator, but don't assume one per visit is always equivalent to the watched view. A person rushing to get a close image of insects might record a much smaller frame than the person actually scanning the branch. A reference picture from the watching position and a later detail image answer different questions. The first helps readers locate the area for the numerical comparison; the second offers visual evidence about a particular plant surface or insect. Make those roles obvious in filenames and notes rather than asking later readers to reconstruct them from a set of similar thumbnails.

There is a counterintuitive case when a denominator becomes zero. If no flowering clusters are observable, “zero clusters observed under the view rules” can be a valid description of the display. “Contacts divided by zero is zero” is not a valid per-cluster comparison. Write “not applicable” for the rate and keep whatever source observations support the call. If the flowers are behind an obstruction, that is “unknown,” not a confirmed zero. Treat no actual watched time the same way: a missing rate does not have to be patched with a value. A table gains credibility when empty cells document meaningful limitations rather than hiding them.

A case study in what two observers can learn from a difference

Imagine three pairs of made-up windows. In morning A and B, Bright Bend has eight confirmed contacts in six minutes on roughly thirty visible catkins, while Quiet Bend has four in six minutes on roughly twelve. At noon, Bright Bend has twelve in six minutes on roughly thirty catkins; Quiet Bend has nine in the same period on roughly twelve. A final pair is scheduled for the afternoon. A group of hikers blocks the view across Quiet Bend for much of its window; that period is genuinely incomplete and no comparable count should be manufactured. The totals make Bright Bend look busier in both complete pairs, while rough exposure ratios remain consistent with higher observed contact activity per visible cluster at Quiet Bend in these windows.

A second observer notes that they counted two of Bright Bend's noon events as too obscured and couldn't tell one rapid contact from the insect's prior position. The group now has a methodological opportunity. Examine a usable image or compare the notes and justify any changes transparently. If you agree on stricter rules for the future, present the first tally as original and the second as a reclassification rather than simply deleting events without trace. If the ambiguity cannot be resolved, say so and explain whether the central contrast depends on those events. Often the substantive finding is not that one estimate wins. It's that the apparent ranking depends on both the question and data handling.

Now read the same fictional record as a project curator preparing a public sign about the area's willows. The best sign may say only that small flowering branches supported visible contacts in several watched intervals and that cluster counts change how visitors discuss the activity. It can also include a small illustration asking visitors to count the clusters before ranking the two branches. It does not have to give a particular bee a confirmed name from the toy data or tell land managers where to plant next. A school activity can teach a sophisticated measurement distinction without collecting pollen, manipulating insects, or implying a site recommendation.

Recognize the work a pollen claim would really need

A landing in a traceable watch is much more directly observed than successful pollen transport or reproductive benefit. Visible loads can be interesting, but they don't carry their identity in a photograph. A study can independently examine pollen material, or compare plant sex and later reproductive outcomes with a suitable design. A viewer often has no way to tell from the bee and bloom alone where material originated, what was deposited and whether it mattered to the plant. Good captions keep the action at the resolution provided by the image. Motions and visible contact are not inferior stories simply because their more ambitious outcomes remain unmeasured.

Published evidence also doesn't make one unexamined local sample conclusive. A research article on pollen uses evidence about particular people, plants, animals and places; its measurements apply in the scope of that work. The study on willow pollen collection by a bee often associated with blueberry is valuable in part because it cautions against using a host nickname as a categorical rule. It doesn't identify a blurred bee in a different place or predict the composition of a load you did not sample. Quoting the measurement accurately helps your local record stay open to surprise.

In projects where an identification is not attainable, use a hierarchy of observational confidence. “Insect in flight near the shrub” is low taxonomic resolution but may be visible. “Beelike visitor made a clear landing” requires more information from the scene. An identified species needs evidence appropriate to that name. For many outdoor records, preserving the first two without inventing the third is the best possible work. You can add corroboration later without altering what the original window actually showed.

Share a result somebody can extend rather than merely admire

Close with a simple reproducibility check. A reader should be able to find the branch in a lawful context view, understand how you counted flowering clusters, understand which actions counted, distinguish inactive time from missing time, and find the weather and viewpoint note that shapes interpretation. They should also be able to recalculate any ratio by looking at the event count and time or cluster denominator beneath it. If any step requires asking what was in the original observer's head, write down the interpretation that was actually used or mark the field unknowable.

For a group returning next year, preserve one blank version of the field card and one filled example with the safety notes attached. Ask a new observer to use both before picking a favorite bush. Compare their note to the original definition not to a desired numerical result. If they cannot locate the exact original frame, let their new position start a visibly separate series. One useful goal is to improve access and continuity, not to pretend all views are equal. There can be real ecological change between one spring and another, but there can also be changed plants, viewing conditions, and observer expectations. The purpose of keeping these factors separate in the record is to give actual changes a better chance of being recognized.

The payoff is not just a better bee count. The practice clarifies what kind of answer the people at the path actually found and what question still belongs to the next outing. A busy willow can be a good discovery. So can a scarcely visible shrub whose flowers receive clear attention during a quiet, well-observed interval. If your records preserve how you learned the difference, the next visitor won't have to choose between trust in the numbers and trust in their own eyes.

Sources and field references

Frequently asked
What is Count the Bees on Two Willows, Then Count the Flowers about?
By Austin Little
What should you know about what a catkin can tell you if you get a good look?
A catkin is a cluster of individual flowers, not a single flower with lots of long petals. The form is common enough on some woody plants that you might assume all catkins work the same way. They do not. On the willows in the New Brunswick study, male and female flowers occur on separate plants. The botanical…
What should you know about there is a difference between a favorite plant and a favorite food?
The word “specialist” is easy to overread. It can refer to the pollen a bee typically collects for its offspring, not every place that an adult bee visits. Minnesota’s Department of Natural Resources describes the eastern willow miner bee, Andrena bisalicis , as a Salix specialist in early spring. That is information…
What should you know about the New Brunswick findings, in their proper scale?
In the garden test, Mosseler and colleagues compared Andrena visitation on male and female Salix eriocephala and S. interior . They reported that about 87% and 71% of visits, respectively, were to the male plants within those comparisons. Counts in another part of the project, at a reclaimed mine site, involved an…
What should you know about a worked observation: a louder bush versus a more visited cluster?
Suppose a volunteer watches two nearby willows from a public path. Over a pair of equal-length observation windows, Willow A receives 20 recorded landings and Willow B receives 10. A has twice as many observed landings. If the volunteer wants to know which shrub drew more activity overall in those windows, that is a…
References & sources
  1. Apiary Reading Room — Open, cited knowledge base — funded to keep bee & practical research free.
From the Apiary Reading Room. Opinion & editorial — not financial advice. We don't overclaim.
More from the Reading Room