EEGLAB is a great free place to start http://sccn.ucsd.edu/eeglab/

EEGlab is a good tool

Eeglab, I agree.

However, you should be aware that methods can only be applied if the data set satisfies certain conditions (e.g. Linear or non-linear coupling etc.).

Check carefully the validity criteria of the methods you are using.

EEGLAB is a fine choice. Another very good MATLAB-based choice would be Fieldtrip. http://fieldtrip.fcdonders.nl/ Both are interoperable to some degree (i.e., each can import/export to the other format). I don't think it would be possible to state how many trials minimum. An important factor in phase measurement is signal to noise ratio, which of course will be higher with more trials simply by noise reduction. However, some effects of interest have higher signal than others and will require fewer trials, boosting SNR on the signal side.

You can also try BESA, which however is neither open source nor free. I would also go with EEGlab or fieldtrip.

I agree with Christian that BESA provides a good alternative (although it is quite expensive). It has an excellent inter-areal coherence function in source space. The obvious caveat is that you need to have a sufficiently dense sensor array to realistically separate sources; you need to be aware that your source model is just that - a model, and you need to be ever-sensitive to the issue of simple volume conduction.

I think EEGLAB does not offer great variety in terms of algorithms for evaluating phase synchrony, and does not include a tool for the computation of cross-frequency coupling. Additionally, coherence measurements have the known limitations in terms of volume conduction of electrical signals on the scalp. Recently, I have tried to implement a method proposed by Guido Nolte and Petra Ritter (for reference see Nolte et al. 2004, Clin Neurophys, and Freyer et al. 2012, Frontiers in Hum Neurosc) that introduced the concept of Imaginary Coherency.

Also, for cross-frequency coupling I would suggest this reference: Cohen, MX 2008, J Neurosc Methods. The author is very responsive, and in the past he shared with me part of his algorithms that contributed to one of my recent publications (see Perfetti, Moisello et al. 2011, J Neurosc).

I think the best software to measure EEG parameters is Neuroguide http://www.appliedneuroscience.com/. It utilizes a well developed standardized data base and algorithims, though it is expensive, it can yield themost robust meaurements. A minimum of 1 minute of artifact free data is required, though more is more reliable.

To overcome the problem of volume conduction, the eLORETA phase synchronization algorithm uses a "lagged" phase synchronization measure that does not take into account the contribution of the instantaneous phase synchronization.

EEG-Lab has a a good filter function (eegfilt()) ..but sigTOOL is very good for PSD measurements and event related cross-frequency coupling and you can also import data from other recording tools (like Spike2 or NeuroExplorer etc)...best is too take them several tools and just use their functions in matlab ;)

This may be premature, but i have been hearing good things about Cartool. I have not looked into it myself yet though.

FieldTrip indeed has a lot of options for calculating connectivity measures, see here: http://fieldtrip.fcdonders.nl/development/connectivity. @carla, looks like imaginary coherence is implemented there as well.

Have you tried with this tool?: http://www.stat.physik.uni-potsdam.de/~mros/damoco.html

EEG signal is highly non linear and evidencing high complexity. Standard methoodlogies employ linear methods that are usefull but cannot see in the inner structure of the signal. I suggest to employ non linear emethods as example the recurrence quantification analysis or other advanced methdos that are discussed in the list of my publication. I am at your disposal if someone needs help since they are methods thata re not so simple to be used. Exploration as example of chaotic dynamics is not everyday matter buut of course we cannot continue every time to use also old methods that , I reopeat, have standardized advantages but cannot cover fully approrpiately the question

Not filter use .. not resampling.. not averaging. Certainly I know that they are curerntly used but they should not be taken in consideration when we have so complex . non linear signals as the EEG. We lose basic information. Consequently .. non linear methodlogies must be used

There is a special issue of Computational Intelligence and Neuroscience on "Academic Software Applications for Electromagnetic Brain Mapping Using MEG and EEG". You will find papers on the main general EEG/MEG tools (SPM, EEGLAB, FieldTrip) and also there are a few that are specific for brain connectivity. See here: http://www.hindawi.com/journals/cin/si/138743/

You can calculate phase synchronization using the free and open-source Neurophysiological Biomarker Toolbox (NBT); see http://www.nbtwiki.net/doku.php?id=tutorial:phase_locking_value#.Unun442-OCo

Depends on the EEG spectrum you are interested in.For instance if your synchronization is measure related to beta, I would recommend you take a time window of two cycles of the center frequency of your interest and perform simple cross correlation of the the sites of interest. The strength of the crosscorrelation is directly related to phase coherence of the frequency in question. If you then perform the calculation parametrically over several center frequencies you will get a "tuning" for the coupling. These steps are available on MATLAB.

If your interest is synchrony between different frequencies then I there are several methods available , but I am not familiar with any of them being commercially avilable. I know of research on phase-amplitude coupling... but no commercial programs the last time I looked.

This is a complex question with no simple answer. There are many manifestations of connectivity in the brain, and many methods to assess them. Each method has its advantages and limitations. I can recommend a new resource for learning about these methods and their implementations: http://mitpress.mit.edu/9780262019873