Muebles y enseres

Although the effects of attention have been primarily studied at the level of individual neurons, several studies have begun looking at the effects of attention on the activity of whole populations of neurons. The most important finding for our purposes here is that attention has been found to increase thesynchronyof populations of neurons responding to the attended stimulus (Fries et al. 2001). Attention seems to cause the neurons responding to the attended stimulus to fire in unison. Interestingly, although this feature is readily ap- parent at the population level, it was obscured by the methods used to study the effects of attention on individual neurons. By averaging across trials in order to average out variabil- ity in neural response, they also averaged out information regarding the precise timing of each spike, which is necessary for noticing synchronization.

An interesting feature of these results is that attention does not result in the synchro- nization of a population of neurons at just any temporal interval. Instead, they preferentially synchronize their firing with gamma oscillations in the local field potential (LFP). What are these gamma oscillations? Oscillations at different frequencies are based in extracellular voltage fluctuations, which arise from summed electrical activity in a population of neu- rons. These oscillations can be measured on the scalp by EEG and in the brain by an electrode measuring the LFP. Both the EEG signal and the LFP can be decomposed into different frequency components, allowing a precise consideration of the role of oscillations

in each frequency band.3 When neural networks are activated, the power in higher frequen- cies increases, particularly in the gamma range (30-80 Hz). This has led some to argue that a prominent gamma rhythm is the signature of an engaged network. What this means is that the production of gamma oscillations in LFP are not necessarily intrinsically com- putationally relevant. They may merely be a byproduct of the activity in the network as a whole (see: Jia and Kohn 2011, for an overview of gamma oscillations). But, even though the gamma oscillations themselves may not have a particular functional role, they seem to have been co-opted for coordinating the firing of populations of neurons. However, it is im- portant to note that the synchronization of a population of neurons with gamma oscillations occurs in several different parts of the brain, completely independently of attention. Atten- tion does not uniquely induce gamma synchrony. There tends to be a degree of temporal coherence across the neural population even in the absence of attention. What attention does do is increasethe synchronization with gamma oscillations of the neurons encoding the attended stimulus anddecrease(but not eliminate) the synchronization of the neurons encoding the unattended stimuli (Fries et al. 2001; Bichot et al. 2005).

Interestingly, Womelsdorf et al. (2006) found that we can even predict the speed of change detection on the basis of gamma synchronization. This demonstrates that the pres- ence of one of the typical behavioral measures of attention (improved reaction time) can be predicted on the basis of the presence of one of the primary neural signatures of at- tention (gamma synchronization). This suggests, in turn, that gamma synchronization is facilitating neural communication in a way that ultimately results in improved reaction times. Given these findings, perhaps increased gamma synchrony can help ground atten- tion’s function of increasing the connectivity of attended representations. There is evidence that is can increase connectivity as it has been demonstrated that spikes arriving simulta- neously have a greater impact there than unsynchronized spikes (Usrey et al. 1998; Salinas

3_{These components are delta ¡ 4 Hz, theta 4-8 Hz, alpha 8-12 Hz, beta 12-30 Hz, gamma 30-80 Hz, and} high-gamma ¿ 80 Hz.

and Sejnowski 2001). Therefore, if sensory neurons carrying information about a stimulus synchronize their firing, the downstream areas are more likely to have a strong represen- tation of that stimulus. Thus, increased synchrony seems to be an ideal mechanism for enhancing the connectivity of a selected representation at the expense of others.

More recently it has been argued that the improved connectivity caused by gamma synchronization can be further enhanced by the presence of gamma synchrony at both the sourceand the target location. That is, instead of just having gamma synchronization on the input side, if you have gamma synchrony on the output side as well, particularly when they are in phase with each other, communication is even more effective. This hypothesis, called the ‘communication through coherence’ (CTC) hypothesis, has been defended re- cently by Pascal Fries (2005, 2009). As Bosman et al. (2012) put it, “rhythmic activity in a target group entails corresponding fluctuations in postsynaptic membrane potentials and postsynaptic shunting, which render input most effective if it is consistently timed to the peaks of depolarization, i.e., if it is synchronized with the target” (p.875). The idea be- hind the CTC is that when a population of neurons manage to selectively synchronize their firing with their downstream target, they ‘block’ the other competing representations from controlling the firing of that target population of neurons. If this proposal is correct, then it is specifically the synchronization of firing rates across areas that results in increased connectivity. I think that CTC is an interesting proposal with some strong, but not conclu- sive, evidence behind it. For now I will concentrate primarily on the role played by gamma synchronization on the input side, but it is important to keep in mind that this may only improve connectivity if it results in gamma synchronization on the output side.

6.3.3 Conclusion

There are two key features of the neural signature of attentional emphasis. First, there is an increase in firing rates for neurons encoding the attended stimulus and a decrease in firing rates for the other neurons. Second, there is increased synchronization amongst the

populations of neurons encoding the attended stimulus, particularly in the gamma band, and decreased synchronization amongst populations of neurons encoding the unattended stimuli. Both of these effects seem to have the function of increasing the strength of the neural representation of the attended stimulus and the promotion of its connectivity with distant regions.