The human brain is a system composed of multiple levels organized into integrative network configurations. At a gross topological scale, spatially distributed, interconnected brain areas interact to perform circumscribed functions—communicating via patterns of synchronization presumably supported by long-range white matter fiber tracts (for reviews see Buzsáki and Draguhn, 2004; Buzsáki, 2006; Bullmore and Sporns, 2009; Bressler and Menon, 2010; Sporns, 2010; Breakspear and McIntosh, 2011). Through the evaluation of coherence in spontaneous low-frequency fluctuations (0.01–0.1 Hz) of the blood-oxygenation-level-dependent (BOLD) signal, resting-state fMRI (RS-fMRI) has proven to be a valuable tool for characterizing these functional relationships. While the term “state” implies a well-controlled period of investigation, the “resting-state” is simply a period of recording in the absence of any explicit task paradigm. Investigations are not time-locked to specific task-stimuli and subjects are free to cycle through normal cognitive processes while passively lying awake in the scanner with eyes open and fixating or closed. Owing to the non-invasive nature of the technique and its ease of implementation, the vast majority of an ever-growing field of studies examining intrinsic brain activity and resting-state functional connectivity has been primarily conducted on human subjects. Great strides have been made in furthering our understanding of large-scale brain topology, including the identification of multiple, anatomically distributed resting-state networks (RSNs) (Biswal et al., 1995; Beckmann et al., 2005; Damoiseaux et al., 2006) that are involved in processes that cover a broad range of lower- and higher- order processes. Further motivation for the use of RS-fMRI has come from the discovery of RSN alterations across a spectrum of disease states (Auer, 2008; Greicius, 2008; van den Heuvel and Hulshoff Pol, 2010).

Given the notable successes of this relatively new area of neuroimaging (Biswal et al., 1995), the field is rapidly advancing toward more complex characterizations of brain architecture and potential clinical applications. The goal of this review, however, is to caution against premature conclusions being drawn when there are a number of crucial open questions concerning both the technique and interpretation of RS-fMRI. For example, the evolution of network topology, the neural origins underlying spontaneous BOLD activity, the potential function of such activity, the effectors of disease-related changes of the signal correlations, and the correspondence between functional and structural connectivity all remain unresolved albeit essential elements in making inferences from resting-state findings. While there are human studies attempting to examine each of these issues, they are limited by ethical or practical restrictions that preclude the invasive experimentation necessary to directly assess these questions. Therefore, we contend that the examination and manipulation of the functional brain organization in animals is a requisite for forthcoming resting-state research as they allow these necessary experimental manipulations to be carried out. While rodent models will serve an integral role in these studies, we advocate for the use of non-human primates as the most suitable model of human brain networks, discussing ongoing work establishing homologies between the species, outlining relevant contributions of macaque resting-state investigations to elucidating the aforementioned open questions, and discussing critical future experiments.

Beyond the intrinsic motivation to explore and classify species, comparative biology can significantly enhance our understanding of mammalian brain organization and evolution through cross-species evaluation of homologies. Determining the relationships among cortical areas and networks between species can be difficult because they differ in both brain and body size, the relationship of which is nonlinear (Van Dongen, 1998). Over the course of evolution, brain regions can duplicate, fuse, reorganize, or expand, changing the proportions of different regions as well as its microstructure and connectivity (Sereno and Tootell, 2005; Hill et al., 2010). The task of comparison is made more difficult because the measurement techniques are often different between species. RS-fMRI circumvents this final limitation and provides a more direct tool for cross-species comparisons as the same methodology can be used to directly compare species because there is no task requirement. By determining which features are conserved across species it can indicate brain regions and patterns that have a basic functional and/or developmental role. The results can also help validate previously extrapolated findings derived from animal models (e.g., disease or knockout models) that afford a greater range of experimental manipulations not practical or possible in humans. However, in comparison to the exponential growth of resting-state publications in humans (Snyder and Raichle, 2012), RS-fMRI investigations of other species is considerably lacking. To date, published reports have included the mouse (Jonckers et al., 2011), rat (Lu et al., 2007; Pawela et al., 2008; Hutchison et al., 2010; Liang et al., 2011), macaque (Vincent et al., 2007; Margulies et al., 2009; Hutchison et al., 2011), and chimpanzee (Rilling et al., 2007). While these studies have begun to expand our knowledge of unique and preserved network properties, a greater range of animals within and between evolutionary branches needs to be examined, capitalizing on the strength of RS-fMRI.

The BOLD signal is not a direct measure of neural activity as would be recorded in electrophysiological experiments. Instead, it represents a surrogate signal reflecting local variations in the deoxyhemoglobin concentration that reflects fluctuations in blood flow, blood volume, and oxygen metabolism that are then partially coupled to the underlying neural activity (Raichle and Mintun, 2006). Task-evoked BOLD responses have been best linked to the induced changes in local field potentials (LFPs; Logothetis et al., 2001) and mechanisms of neurovascular coupling have been proposed (Attwell and Iadecola, 2002; Uludağ et al., 2004; Raichle and Mintun, 2006; Iadecola and Nedergaard, 2007). The same correlation has not been confirmed for spontaneous hemodynamic fluctuations and it is yet to be determined if the coupling mechanisms are the same across spontaneous and evoked states.

There are a number of studies employing simultaneous electroencephalography (EEG)-fMRI in humans that have explored the relation of spatial and temporal hemodynamic patterns to bands of electrophysiological activity (Goldman et al., 2002; Laufs et al., 2003; Mantini et al., 2007, 2011; Sammer et al., 2007; Laufs, 2008, 2010; He et al., 2008; Olbrich et al., 2009; Britz et al., 2010; Michels et al., 2010; Wu et al., 2010; Musso et al., 2010). Many of these studies have shown that specific frequency bandwidths (or power fluctuations of the bandwidth) are in fact correlated with the ongoing hemodynamic fluctuations. Though these results do much to support a neural origin of the RS-fMRI signal, drawing overall conclusions from the studies is difficult as they each implicate different bands across the frequency spectrum including delta, theta, alpha, beta, and gamma. Other evidence has begun to reconcile these findings by ascribing unique electrophysiological signatures to different brain states and RSNs that are represented by power variations across the EEG range (Laufs et al., 2003; Mantini et al., 2007; Britz et al., 2010; Musso et al., 2010; for review see Laufs, 2008).

First proposed by Fox and Raichle (2007), infraslow oscillations represent another potential candidate. Using direct current-coupled EEG, which circumvents the limited recording bandwidth of most EEG systems (>0.5 Hz), large-scale infraslow oscillations (0.02–0.2 Hz) can be recorded across widespread regions in the human cortex (Vanhatalo et al., 2004). The oscillations are themselves correlated with changes in the power of higher frequency bands including gamma, leading to the notion of a causal role between the two processes in which the infraslow oscillations modulate the power of higher frequency activity. In this model, the rapid fluctuations coordinate the neuronal activity at small spatial scales, whereas the much slower power fluctuations allow for long-range coordination. This is supported by empirical evidence (Buzsáki and Draguhn, 2004), however its direct relationship to BOLD fluctuations remains to be determined.

The failure to record the complete range of physiologically relevant electrophysiological signals when using standard EEG highlights one of a number of its disadvantages that prevent adequate exploration and understanding of the neural activity underlying hemodynamic fluctuations and their organization into complex networks. The most prominent is that EEG, despite its excellent temporal resolution, is unable to accurately discern or record all cortical and subcortical activity (Gloor, 1985). The limited source localization is due to biophysical challenges related to convolution of the cortical surface, distortion from cerebral spinal fluid, neuron orientation, synchronization, skull conduction, and other sources of attenuation or loss (Ritter and Villringer, 2006) that further exacerbate the inverse problem. The problem refers to the infinite number of possible locations and magnitudes of the intracranial current sources that make reconstructing a unique mathematical solution, and by consequence a precise spatial mapping, impossible (Niedermeyer and da Silva, 2004). Electrocorticography (ECoG) can circumvent several of these limitations and allow better source localization, but it is extremely invasive and only suitable for a small group of patient populations. He et al. (2008) have reported that a similar “correlation structure” of the sensorimotor network recorded by ECoG and RS-fMRI in patients undergoing surgical treatment for intractable epilepsy. Slow cortical potentials (<0.5 Hz) and gamma frequency power were found to best correspond with the RS-BOLD fluctuation profiles across multiple states (He et al., 2008).

Beyond the electrophysiological correlates, a number of other hemodynamic and metabolic variables have been put forth by Fox and Raichle (2007) and require consideration. Oxygen availability, nicotinamide adenine dinucleotide levels, spontaneous neurotransmitter release, cytochrome oxidase activity, blood volume, and blood flow demonstrate spontaneous low-frequency fluctuations that can have 1/f distributions and similar spatial patterns to those seen with BOLD (Fox and Raichle, 2007). Therefore, precise spatial localization and characterization of temporal patterns must come from recording electrodes and optical probes placed within the brain. Ideally, these will be recorded simultaneously with RS-fMRI. Nir and colleagues (2008) have reported slow (<0.1 Hz) spontaneous fluctuations of neuronal activity (LFP gamma power modulations) in the auditory cortex of a small group of awake patients as well as significant interhemispheric correlations between the homologous areas using bilateral single-unit and LFP recordings (Nir et al., 2008). However, owing to the same restrictions requiring the examination of non-human primates to elucidate the task-based neurovascular coupling, non-human primates will be critical in establishing the correlate of the low frequency oscillations as distributed areas can be directly assessed with depth electrodes and optical probes, high-field studies, and pharmacological investigations.

The growing body of animal studies does support the notion that resting-state BOLD fluctuations of cortical and sub-cortical regions originate from the coupling of spontaneous neuronal activity to a hemodynamic response function. Early multi-modal evidence in anesthetized rats demonstrated tight coupling between spontaneous cerebrovascular fluctuations and bursts of electrocortical activity (Golanov et al., 1994). Pioneering work in the macaque has further established slow fluctuations of power in the gamma frequency range as the leading candidate for the neural correlate of spontaneous BOLD fluctuations (Leopold et al., 2003; Shmuel and Leopold, 2008). Leopold and colleagues (2003) were able to show that the power of gamma at a particular moment fluctuates, albeit at a much slower rate (<0.1 Hz) than its LFP activity. Besides sharing a similar frequency as the hemodynamic changes, the slow power fluctuations exhibited 1/f behavior and was correlated across large regions of cortex (Leopold et al., 2003). Later, it was shown using simultaneous fMRI and LFP recordings again in non-human primates, that the power fluctuations are in fact directly correlated with spontaneous BOLD fluctuations (Shmuel and Leopold, 2008). Macaques have also been used to support a neural origin of the global component of RS-fMRI (Schölvinck et al., 2010) and investigate differences of electrically induced and spontaneous activity (Matsui et al., 2011).

Non-human primates also afford the opportunity to invasively explore the generalizability of the neurovascular mechanisms contributing to the spontaneous BOLD fluctuations across cortical and subcortical areas. Implicit in most hypothesis-driven (seed-region analysis) and exploratory (ICA, principal component analysis, clustering) approaches is an assumption of regional homogeneity in the mechanisms underlying neurovascular coupling. However, recent task-based fMRI investigations of the rat (Sloan et al., 2010) and humans (Martuzzi et al., 2009; Conner et al., 2011; Gonzalez-Castillo et al., 2012) examining distributed response patterns have revealed variations in regional coupling between neural activity and metabolism. Whereas task-based studies may constrain their analysis to the task paradigm, increase averaging, and utilize more flexible or dynamic hemodynamic response functions (Gonzalez-Castillo et al., 2012), RS-fMRI does not afford such time-locked events and are dependent upon comparing the timecourses of voxels containing considerable differences in cell type, laminar organization, connectivity, in addition to variations in vascular, neuron, synapse, and astrocyte density (Logothetis and Wandell, 2004). Detailed optical imaging in primate studies will help elucidate the mechanisms and scale of the variability and could suggest that measures beyond region correlations will be necessary to truly characterize ongoing functional relationships that may be captured using RS-fMRI.

A large proportion of the resting-state investigations of animals have utilized anesthesia as a method to eliminate motion effects, physiological stress, and training requirements. As outlined throughout this review, RSNs have been found in rodents and macaques despite the use of various types of anesthesia. In rodents, these have included alpha-chloralose (Lu et al., 2007; Majeed et al., 2009), medotomidine (Pawela et al., 2008, 2010; Zhao et al., 2008; Magnuson et al., 2010), ketamine/xylazine (Hutchison et al., 2010), and isoflurane (Kannurpatti et al., 2008; Hutchison et al., 2010; van Meer et al., 2010; Liu et al., 2011; Wang et al., 2011). Isoflurane represents the most commonly used anesthetic in monkeys (Vincent et al., 2007; Shmuel and Leopold, 2008; Teichert et al., 2010; Hutchison et al., 2011; Mars et al., 2011) though propofol (Matsui et al., 2011; Adachi et al., 2012) and a combination of ketamine and medetomidine (Moeller et al., 2009) have been used successfully. Despite convergence of results across anesthesia types and even with awake animals (Zhang et al., 2010b; Liang et al., 2011; Mantini et al., 2011) it is incorrect to say that quantifiable differences do not exist between the different states (Lu et al., 2007; Moeller et al., 2009; Williams et al., 2010; Wang et al., 2011). The mechanisms of action of many anesthetics remain poorly understood, but certainly modulate neural activity and consequently influence the cerebral blood flow if not effecting blood flow directly. Isoflurane in particular has been shown to disrupt functional thalamocortical connectivity (Alkire et al., 2000; Steriade, 2001; Arhem et al., 2003) in addition to causing vasodilation that can alter cerebrovascular activity (Farber et al., 1997; White and Alkire, 2003; Schlünzen et al., 2006). The effects could be manifested as changes in the correlation strength, localization, or inclusion of distributed nodes within specific networks (Lu et al., 2007; Vincent et al., 2007; Liu et al., 2011; Wang et al., 2011). Anesthesia can also limit longitudinal studies in cases such as alpha-chloralose that require animal sacrifice or recoverable anesthetics that require a necessary interval between scanning sessions. Finally, there are concerns of possible drug interactions when studying pharmacological interventions.

Despite the disadvantages and limitations, the utility of anesthesia should not be understated. Beyond allowing extended, motion-free acquisition in naive animals, anesthesia experimentation can serve to explore the fundamental physiological relationships underlying spontaneous fluctuations and functional connectivity by exploiting their unique mechanisms of action and effect on neural activity, neurovascular coupling, and vascular reactivity. Further, anesthesia can eliminate conscious processes such as passive mind wandering, active monitoring, memory formation, or changes in attention and arousal during image acquisition that may confound certain experiments (Hutchison et al., 2012b). Reciprocally, there is the potential to explore the mechanisms that account for the anesthesia's diverse effects on memory, pain, and consciousness. While the use of anesthesia in human research is possible (Greicius et al., 2008) it is severely limited and provides an additional need for animal investigations.

There are a number of fundamental questions regarding the potential relationship between functional connectivity, measured as the temporal relationships between brain regions, and the underlying structural connectivity that represents the anatomical white matter fiber tracts that still remain unanswered. By definition, RSNs are composed of anatomically separated brain regions. These can include contralateral homologs or more distributed patterns within and between hemispheres. Given the growing evidence supporting a neural origin of resting-state fluctuations and their synchronization, many have hypothesized that the functional connectivity is supported by direct structural connections (Damoiseaux and Greicius, 2009). While direct evidence is limited in human subjects, simulated and empirical investigations of humans have suggested an overall good correspondence between the two (Damoiseaux and Greicius, 2009; Greicius et al., 2009), although clear discrepancies have emerged.

The majority of studies examining the relationship between spontaneous BOLD correlations and anatomical connectivity have used diffusion imaging techniques such as diffusion tensor imaging (DTI) or diffusion spectrum imaging (DSI). At a local level, Koch and coworkers found that regions on either side of a sulcus showing high functionally connectivity were also structurally connected by short-range fibers (Koch et al., 2002). This has since been shown at the whole brain level, in which regions with a higher level of structural connectivity showed higher levels of functional connectivity (Honey et al., 2007, 2009; Hagmann et al., 2008). Indeed, almost all functionally linked regions of the most commonly reported RSNs appear to be roughly constrained by known white matter tracts (Honey et al., 2007; Vincent et al., 2007; Greicius et al., 2009; van den Heuvel et al., 2009). The relationship between structural and functional connections is not, however, one-to-one and there are a number of discrepancies. Studies have reported areas that share no direct connections (Habas, 2009; Honey et al., 2009; Krienen and Buckner, 2009). For example, primary visual cortex has been found to be robustly connected to its contralateral homolog, though no direct connections exist (Van Essen et al., 1982). This implies that some of the reported functional connectivity is driven by polysynaptic pathways. The opposite pattern has also been reported in which areas known to have structural connections do not show functional connectivity (Adachi et al., 2012; Hutchison et al., 2012). Evaluation of the correspondence between connectivity types is made difficult by methodological limitations of diffusion techniques that do not allow precise delineation of the origins, crossings, and terminations of pathways, thereby restricting the interpretation of results. The few case studies examining congenital or surgical alteration of the callosal fiber connections have produced mixed results (Quigley et al., 2003; Johnston et al., 2008; Uddin et al., 2008; Tyszka et al., 2011). Further insight into the relationship between the two connectivity types is needed and will likely come from an experimental system in which anatomical connectivity can be more easily assessed and manipulated—that is, in an animal model. Histological dissection and staining, degeneration methods, and axonal tracing carried out in non-human primates remains the gold standard for uncovering the precise information concerning the origins and terminations of white matter connections.

The first qualitative comparisons of functional resting-state connectivity maps and structural connectivity maps derived from experimental tracer studies in the macaque demonstrated remarkable consistency between the patterns (Vincent et al., 2007; Margulies et al., 2009; Mars et al., 2011, Figure 6). In fact, macaque tracing findings can be accurately extrapolated to predict human functional connectivity patterns (Margulies et al., 2009; Kelly et al., 2010).

To allow for a more quantitative assessment of functional/structural relationships it is necessary to examine the correspondence beyond single areas. Recent work has compared macaque RS-fMRI connectivity to structural connectivity derived from macaque axonal tract tracing studies contained within the CoCoMac database (Adachi et al., 2012). CoCoMac is a systematic record of the known anatomical connectivity of the primate brain containing details of hundreds of tracing studies in their original descriptions (Stephan et al., 2001). The primary finding of this work was that functional connectivity between areas with no direct structural connection is driven by common afferents and common efferents (as opposed to serial relays). The work however, did not attempt to explore the overall correspondence between the connectivity types and was limited to unilateral visual and sensorimotor areas in two monkeys. Current investigations are building on this work, establishing similarity measures between functional and structural connectivity across the complete extent of both cortical hemispheres (Shen et al., in preparation; Figure 6C). The results revealed a relatively high degree of overlap between connectivity measures. Further, the strength of functional connectivity was proportional to the strength of the anatomical connection. These findings confirm the general notion that white matter constrains functional architecture, however, as has been previously reported, the relationship between the two is not perfect. Future investigations directly comparing diffusion tractography and functional connectivity results with neuroanatomical tracing data of the same monkey will be necessary to most accurately determine the precise relationship of functional and structural connectivity in the primate.

The macaque model has also allowed for the controlled evaluation of fiber pathway contribution to functional connectivity through surgical manipulation. As stated, there are inconsistent results when examining patients with agenesis or resection of the corpus callosum. A case study of a 6-year-old child following resection of the corpus callosum (Johnston et al., 2008) and of a small group (N = 3) of patients with agenesis (Quigley et al., 2003) found significantly decreased functional connectivity between the neocortices. However, contradictory reports from a patient (age = 73) who underwent complete forebrain commissurotomy (Uddin et al., 2008) and a sample (N = 8) of patients with complete agenesis of the corpus callosum (Tyszka et al., 2011) found preserved bilateral connectivity. In response to these discrepancies, that may be possibly related to compensatory mechanisms occurring over time, Croxson et al. (2011) scanned monkeys (N = 4) before and after surgical transection of the forebrain commissures (including the body and genu of the corpus callosum, anterior commissure, and splenium including the hippocampal commissure). The authors found significantly decreased interhemispheric correlations between pre- and post- operative scans across multiple bilateral homologs.

Overall, the strong positive similarity values between functional and structural connectivity matrices and the loss of interhemispheric correlations following surgical resection of the tracts suggest that the two measures are intricately related. The relationship, however, is not one-to-one as evinced by a less than perfect correlation and presence of correlations in the absence of direct connections. The results are not necessarily surprising given that the structural brain network needs to facilitate a vast array of functional configurations to achieve different states (van den Heuvel and Hulshoff Pol, 2010). It is here that the non-human primate will continue to serve an invaluable role by allowing the exploration of polysynaptic connectivity and regulation by common inputs.

Networks facilitate efficient information transfer and allow the emergence of properties not possible when the nodes are in isolation. These could include, for example, increased processing capabilities, stability, or resource sharing. However, alteration or breakdown of the network, especially of central (hub) nodes, can create detrimental dynamics and catastrophic failure across the entire system. The brain is especially sensitive to manipulations that alter its functional and structural organization. A growing and promising avenue of research is exploring the use of RS-fMRI measures in assessing clinical disorders; the overarching hypothesis across many of these studies being that alteration of brain networks are the cause or consequence of the abnormal manifestations characteristic of the disease. The technique is particularly well suited for investigations of non-normal populations, such as subjects with severe cognitive or physical impairments compared to other methodologies, including task-based fMRI. This is because resting-state investigations require minimal task compliance and therefore allow for accurate comparisons of brain connectivity and dynamics. For example, a task requiring memory encoding can be of particular concern when evaluating patients suffering from neurodegenerative diseases. Caution, however, must be taken when comparing groups that may exhibit systematic differences unrelated to paradigm such as motion that can induce systematic, but spurious correlation structures throughout the brain (Power et al., 2012).

Although their meaning is not fully understood, changes in functional RSNs have been reported in multiple psychiatric and neurologic disorders including depression (Greicius et al., 2007; Kühn and Gallinat, 2011; Lui et al., 2011), attention deficit-hyperactivity disorder (Castellanos et al., 2008; Fair et al., 2010), schizophrenia (Whitfield-Gabrieli et al., 2009; Kühn and Gallinat, 2011; Bassett et al., 2012), Alzheimer's disease (Greicius et al., 2004; Chen et al., 2011), epilepsy (Waites et al., 2006; Zhang et al., 2011), disorders of consciousness (Soddu et al., 2011) including coma (Norton et al., in press), multiple sclerosis (Lowe et al., 2002, 2008), and amyotrophic lateral sclerosis (Mohammadi et al., 2009) (for reviews see Auer, 2008; Greicius, 2008; van den Heuvel and Hulshoff Pol, 2010). Many early studies focused on the default-mode RSN, as the network seems particularly sensitive to disruption in disease states, but more recent work has now started to examine other networks as well as changes in the overall organization of functional brain network using graph analysis techniques (Jafri et al., 2008). For example, through graph analysis of resting-state data it was revealed that the locations of high concentrations of amyloid deposits in Alzheimer's disease patients were highly correlated with the location of highly connected hub-regions in the human brain suggesting that disruption of integrative hubs may result in the decreased functional brain efficiency in these patients (Buckner et al., 2008). Taken together, the extensive documentation of altered RSN topology suggests that brain diseases are targeting interconnected cortical networks, rather than a single region and may help explain some of the complex manifestations seen in these patient populations.

Given that the examination of spatiotemporal properties of RSNs studied with RS-fMRI can delineate abnormal neural functional architecture, the natural extension of the methodology would be using RSN-related metrics as potential screening devices for disease. However, many of the robust changes across the range of aforementioned disorders have been derived and significance-tested for “proof of concept” at the group level. These represent valuable contributions toward understanding abnormal brain activity and connectivity, but characterizing patterns of functional variability between normal and patient groups is far from providing clinical diagnostics at the single-subject level. The correlative results also present a directionality problem, in that the relationship between the disease and altered connectivity are unclear. The functional disruptions could represent a consequence of the disease or be the underlying cause and this could vary across disease types. Given the extraordinary potential for RSNs as possible diagnostic or prognostic markers, it is crucial to understand the mechanisms by which they are altered. While rodent models do share features of brain organization and afford examination of genetically altered models, the homologous spatial and temporal brain properties establish the macaque monkey as the most suitable animal model of human brain organization. The macaque provides the flexibility to explore causal effects by allowing pharmacological, lesion, or even optogenetic interventions that extend well beyond limited case studies and transcranial magnetic stimulation in human subjects (Diester et al., 2011). Additionally, many rodent models and few non-human primate models exist across the spectrum of neurological and psychiatric diseases shown to be accompanied by abnormal functional disruptions. While these models must be interpreted with caution, exploring large-scale topology changes in these models combined with an enhanced understanding of the physiological mechanisms of fluctuation, regulation, and entrainment of LFFs gained from wild-type animals will further RS-fMRI's clinical potential. Following the determination of causal relationships between the disease and altered connectivity patterns, the animal models will also allow the assessment of early diagnostic biomarkers and the development of better drug treatments.

Resting-state investigations in humans are being conducted at an astonishing pace whereas the same technique is still in its infancy in animal models. Nevertheless, the available reports have presented a promising assessment of preserved attributes upon which multiple research directions can now be developed. It is with the knowledge of normal brain topology and dynamics in the rodent, and even more the macaque that experimental manipulations can be properly interpreted providing new and valuable insights into how the human brain operates across multiple states. In addition to those outlined in the individual sections, below we present additional potential avenues that present an exciting and potentially revealing future for non-human primate (as well as other animals model) investigations.

Ongoing work using RS-fMRI has demonstrated that brain network topology and its reciprocal temporal features are ubiquitous across mammalian species. Greater brain complexity can also be demonstrated from more distributed network organization and lateralization of functional connectivity patterns. From an applied research perspective, the findings support the use of animal models, particularly non-human primates for the study of network disruptions in disease and exploration of the true underpinnings of the RS-fMRI signal. The applications and future avenues of research extending from current work are broad and will aid in our understanding of normal and abnormal brain function.