This is my first question here, so I apologize for all mistakes I could have possibly made.

I'm a high school student in East-Central Europe and I need to complete some research for a biology contest (asking for advice is accepted, so I'm not cheating). My task is to analyze the influence of certain environmental factors (temperature etc., it's not that important for my question) on the activity of bees. The method is pretty simple: once a week I record the entrance to the bee hive (I do it on four hives), play it in slow-mo (It's impossible to count them properly without doing so) and simply count the number of bees entering or leaving.

Problem is, I don't know how long the observation should take. I play it in like X1/8, you need to play it twice (entering/leaving), so it takes a lot of time to gather one piece of information for a certain day. Till now I've been doing it for one minute - and there seems to be some kind of pattern to their activity analyzed that way. Yet, I'm not sure if it's actually eligible. I can't do the observation for hours - I still need to learn and have other duties.

So, what should I do? Could anyone give me some advice? Is one minute (several, like 6 times a day per hive) legitimate enough?

Thank you in advance.

  • $\begingroup$ I know of a person who used to do these experiments. They mentioned that they also used to mark each bee to track exactly. $\endgroup$
    Commented May 21, 2014 at 8:05
  • $\begingroup$ One minute a day sounds far too short. Several one minute measurements throughout the day would be one way, or measuring a day in full, and then using a representative time-point would also be neat. I think it depends on the question you're answering. "hot bees are more productive" and "Hot bees hives have more bees" could look like the same result in your setup. @WYSIWYG I heard a similar story - painting moving bees sounds a bit precarious to me! $\endgroup$
    – James
    Commented Sep 14, 2016 at 9:48

3 Answers 3


One thing that's certain is that the activity of bees varies according to time of day, so more important than how long you record for is probably at what time you record. If you always record at the same time of day, this should allow a reasonable comparison between results of different days. For example, recording from time 14:30 to 14:35 every day could be fine. You could then record conditions at this time each day (temperature, weather, etc.) along with this.

As for the exact length of your recordings: Of course, the longer they are, the better. More samples usually increases reliability and representativeness of your results. However, more importantly you want a practically useful time as well. This really depends on the actual number of bees entering into the hive within different time frames, as well as the baseline variability between days irrespective of the conditions you are investigating. I don't know anything about either of these :)

You want the number of entrances in your recording to be at a level where you will notice if there is a meaningful difference between days, and the only way of going about this is usually to start with a guess and then change if you notice that you can't really make sense of your results.

  • $\begingroup$ To further reduce the time it takes you, you could start out counting both entrances and exits from the hive - but then check if you get the same results if you only consider entrances. As a measure of bee activity, I would expect entrances and exits to be equivalent and thus it shouldn't matter whether you count either one or both. $\endgroup$
    – Armatus
    Commented May 19, 2014 at 22:06

Ideally, you would gather preliminary statistics to design the experiment. If you have a high variability, and your observation window is short, then small effects will be swamped by the variability. However, if what you are looking for has a big effect in the number of bees, then counting for a shorter time is going to be fine.

A rule of thumb would be to try recording 5 to 10 segments of 1 minute analysis and see what the variability is. If the standard deviation is small compared to the effects you are seeing, then 1 minute is fine.

If you want a more theoretical justification, Poisson distribution could be used as a first order approximation of the distribution. I don't know much about statistics specific for this kind of insect behaviors though.


I'm no expert at this but these are some of the things that I would consider thinking about. How practical is it for you to count the total number of bees in the hive, based on lets say hive weight or the number of bees on one rack in the hive (then multiply by the total number of racks to give you an estimate of the total bee numbers in a hive) before you start collecting video data? This will at least put your influx/outflow number into some sort of prospective if you are counting in different times of days or if there are different wheather conditions in different days. This is easily achieved if you divide your influx and outflow numbers by the total (estimated) number of bees. This gets you two numbers (ratios) which you plot on the Y-axis. On the x-axis you can plot the passage of time and even colour code different days based on average (lets say) temperature to see if temperature had an obvious effects on your observations. But obviously all your instances of recording and time intervals between recordings needs to be constant!

So in many fly behavioural or locomotion function assays, scientists record fly movements at certain fps and analyse the videos. Now most of this video data analysis is done using scripting in Matlab which is (I know to be) the tool choice for motion analysis. Now since you say you are at high school, this would probably be quite a challenge since the program is costly and scripting is not that easy (at least for me), so see if you can utilise the help of someone to get over this barrier. Once you do then you can analyse much longer videos at much higher frequencies, which will give you a better resolution on your data. If you could't then consider looking through web sites using search engines such as google for Matlab codes for motion tracking analysis scripts. I did this some time ago and there are wealth of information out there and once you do that you can get the communities help in websites such as stack overflow to help you refine your codes or any problems you may have with it, so long as you show evidence of your work. This is a great option since you can add variety of data (or different data aspects) to your project such as what is the average speed/velocity which your bees travel at in different times when entering/exiting the hive, is there a pattern to their travelling (e.g. straight line, zig-zag etc) and much more. I'm aware that this is beginning to sound like a PhD project and not a high school project but these are some things to consider. If programming is totally out of the question then you can look on internet to see what free programs are out there to aid you with your data collection. I briefly looked and saw this (http://www.youtube.com/watch?v=jWD8CHW06Ks) and this (http://www.researchgate.net/post/Does_anyone_know_of_FREE_software_for_motion_capture) just by simply typing into google "video motion tracking software" so look around and see if you can find anything which you can tailor for your own need.

Now regarding video length time, as correctly mentioned in Armatus response (if I'm not wrong and correct me if I am) it shouldn't matter for how long or how frequent you are recording so long as you keep the variables (i.e. time of recording, frequency of recording etc) to the minimum (i.e. not change them), or in other words keep your observation methodology constant through time, and try to come up with a practical schedule and if you explain your thought rational (experimental design) in your report, even a professor (i dare to say) wouldn't be able to argue with it but you always have to talk about the limitations of your study and that is key and that's where peoples project falls apart when an expert reviews it and finds holes and criticises claims made, which can not be backed up by the data/types of analysis gathered/performed!!

EDIT: Just to address some statistics issues, I would make your recordings from minimum of three hives per recording time-interval/point since in science n (number of observations) should be >= 3.

Anyways I hope my ramblings have been somewhat useful.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .