Paleonet: Splitting of small samples

Jere H. Lipps jlipps at
Thu Oct 11 16:04:05 UTC 2012

It depends on what you want to do.   If, for 
example, you are only interested in diversity 
(number of species and what they are), you might 
do just as well by scanning through different 
size-fractions sorted by some kind of mesh.  Such 
mesh is cheap in stores.   You just need to 
figure out what sizes you want and can get.  We 
commonly make reference slides of a set of 
samples this way so that we know what we will 
encounter in more labor-intensive work.  We can 
have a lot of fun with forams without  a lot of work and time.

If you really want to do statistically valid 
sorts to 300-500 specimens for quantitative 
studies, then you should probably increase your 
number of specimens.  The usual 300 or so 
specimens was something that came out of 
sedimentary geology where people wanted to know 
the proportions of mineral and rock types in 
sand.  In sand there are seldom more than a few 
different types to look for and 300 specimens 
might be a valid sample size.  But if you have a 
100 different species in a foram sample for 
example, then you need far more specimens to get 
an valid sampling.   Some key species may be very 
rare (I've looked at samples wanting species that 
were 1 in 10,000 because of dominance of 
others).   The only way to deal with these is by 
rarefaction where you pick or look until you see 
no more new species and at the same time keep 
track of the numbers of specimens counted.   This 
is plotted on a graph of newly discovered species 
vs total number of specimens, and when the curve 
flattens out, you can assume you have sampled 
enough.   It is still an assumption and a lot of work.

You can split samples, but I doubt that is a 
truly statistically valid sampling of the 
original distribution of species.    A number of 
different kinds of splitters can be had, but each 
has its own problems.   The best ones are those 
that have multiple dividers that send a sample to 
two bins.   The fewer dividers, the more likely 
error will be introduced since the foram shapes 
and sizes will influence how they are arranged as 
introduced to the splitter and where they 
go.  Biases exist at every step from sampling 
through splitting to your selection of 
specimens.  All of that has to be considered in 
your work plan.  If your original sample is not 
good, then being rigorous later on is just a 
waste of time.   You might get, by splitting, a 
near-statistically valid sample of your original 
sample, if you are careful about how you 
processed the sample and how you introduced the 
sample to the splitter (I try to spread it more 
or less evenly on a or tray piece of  paper, then 
dump that in more or less evenly across the 
splitter from one side to the other), and then 
being sure that you examine every specimen.   Also a lot of work.

A far simpler way is just to start looking 
through a lot of sample to pick out the 
interesting (defined any way you want) forams, 
and then draw whatever conclusions are possible 
(not statistically based ones, obviously).  Also 
what should be obvious is that I don't like to 
pick samples, so I try to maximize my labor 
(sounds good like that, but perhaps you might 
think a better word would be "lazy").  A lot of 
good science or fun can be had without statistically valid samples.

There's a lot more to this, of course, than can 
be told on PaleoNet, and if you really want to 
deal with statistically valid samples, there's 
plenty of good books that would help.


At 06:25 AM 10/11/2012 Thursday, you wrote:
>Dear Paleonetters,
>how may I split a foraminifera-rich 
>sample-residue (>63µm) of about 20g into even portions of about 0,5g ?
>So far I mix such a rich residue of about 20g in 
>a small container and quickly dump it onto a 
>sheet of paper. Then I cut the residue carefully 
>with a knife into 4 parts and so forth until I 
>end up with about 0,5g. The method doesn't seem 
>very reliable for evenness and puts the whole 
>picking and counting into question. As I want to 
>process many samples rich in small forams 
>counting of 5g with 5000+ specimens is not an 
>option and I need to count 300-500 specimens per sample.
>Will a common micro-splitter do such a delicate 
>job ? So far I didn't want to spend $400 on a 
>micro-splitter as I am an amateur :)
>Thanks for your advice.
>Michael Hesemann
> Project
>Hamburg, Germany
>Paleonet mailing list
>Paleonet at
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the Paleonet mailing list