Skip to content

The problem of synchrony in multiunit recordings

October 19, 2011

Extracellular multiunit recordings are great for analyzing the relationships between multiple neurons, but I wonder if their ubiquity belies a significant lack of understanding of the techniques involved. Any multiunit analysis requires the use of spike sorting which attempts to figure out which events correspond to which neurons. My background is primarily with intracellular recording, so I am a novice on the topic. I was curious last year when an experimentalist told me he did most of his spike sorting by hand, as opposed to trusting the automated software that was available with his recording equipment. This is interesting to me because of a fascinating 2000 article on sorting accuracy with tetrodes by Harris et al. One of the conclusions of the article was that operators generally do worse than software. Obviously this depends on the experiment, the equipment, the operator, and the software. One can easily understand that a generic algorithm might require significant tuning to be useful, so perhaps there are many who do not bother with that. However, it makes me wonder what types of significant mistakes might be lurking in the spike analysis literature.

One topic raised by the article is the identification of synchronous events. That is, what happens when synchronous spikes from multiple neurons are superimposed in a recording? The article is more concerned with operator error vs. software error, but it does estimate a possible software error rate of up to 30%! Unfortunately for me, this problem does not receive much treatment in the article. Yet it seems extremely important because of two reasons. First, this study involved the use of tetrodes which offer significantly better spatial resolution than larger multielectrode arrays. Therefore the problem must certainly become more significant with larger arrays. Second, one of the most common measures of interest in multiunit recordings is correlation and the phenomenon of spike synchrony.

I find this striking, but I’m aware that these authors are leaders in this field. A look through later articles that cite this one did not turn up any subsequent interest in this issue, so perhaps it’s not as serious of a problem as it seems. I certainly need to learn more about spike sorting.


From → Uncategorized

One Comment
  1. josic permalink

    Chris – this is a huge issue, and most people know about it. For instance, just look at the cross-correlation functions you get from data after cells are identified by spike sorting. There is frequently a significant dip around 0 lag. This indicates that pairs of spikes have not been resolved when they occurred too close together. This is just one of the artifacts, and there are numerous others. Because of this it is probably very difficult to say much about synchrony from multi-unit recordings.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: