A dual-pathway neural architecture for specific temporal prediction

Neuroscience and Biobehavioral Reviews 37 (2013) 2587–2596

Contents lists available at ScienceDirect

Neuroscience and Biobehavioral Reviews journal homepage: www.elsevier.com/locate/neubiorev

A dual-pathway neural architecture for specific temporal prediction Michael Schwartze a,b,∗ , Sonja A. Kotz a,b a b

Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany School of Psychological Sciences, University of Manchester, Manchester, United Kingdom

a r t i c l e

i n f o

Article history: Received 7 January 2013 Received in revised form 19 July 2013 Accepted 15 August 2013 Keywords: Prediction Temporal processing Cerebellum Thalamus Oscillation

a b s t r a c t Efficient behavior depends in part on the ability to predict the type and the timing of events in the environment. Specific temporal predictions require an internal representation of the temporal structure of events. Here we propose that temporal prediction recruits adaptive and non-adaptive oscillatory mechanisms involved in establishing such an internal representation. Partial structural and functional convergence of the underlying mechanisms allows speculation about an extended subcortico-cortical network. This network develops around a dual-pathway architecture, which establishes the basis for preparing the organism for perceptual integration, for the generation of specific temporal predictions, and for optimizing the brain’s allocation of its limited resources. Key to these functions is rapid cerebellar transmission of an adaptively-filtered, event-based representation of temporal structure. Rapid cerebellar transmission engages a pathway comprising connections from early sensory processing stages to the cerebellum and from there to the thalamus, effectively bypassing more central stages of classical sensory pathways. © 2013 Elsevier Ltd. All rights reserved.

Contents 1. 2. 3. 4. 5. 6.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Representations of temporal structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A dual-pathway architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rapid cerebellar transmission of auditory input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cerebellar connections to early stages of auditory processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Introduction ‘Ready, set, go!’ In the appropriate context, such a simple sequence of acoustic events evokes complex behavior. It prepares the organism to perform a specific type of action at a specific time. Ideally, this action is adequate to the goal and triggered in time, i.e. it is neither performed too early nor too late. These processes ultimately require predictive adaptation defined as the adjustment of overt and covert behavior to future events. On the one hand, predictive adaptation may pertain to the form, or formal structure, of events (see Glossary; Fig. 1), e.g. ‘set’ is associated with ‘ready’. On the other hand, it may pertain to the temporal structure of events. The latter is twofold: (i) ‘set’ will follow ‘ready’,

∗ Corresponding author. Tel.: +44 161 3060443. E-mail address: [email protected] (M. Schwartze). 0149-7634/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.neubiorev.2013.08.005

2587 2588 2589 2590 2592 2593 2594 2594

or (ii) the temporal relation of ‘ready’ and ‘set’ will approximate the temporal relation of ‘set’ and ‘go’. Both render predictive adaptation a powerful mechanism in optimizing an individual’s behavior beyond mere reaction. Accordingly, a predictive bias is increasingly recognized as fundamental to brain function (Bar, 2007; Bubic et al., 2010; Friston, 2005). From this perspective, the brain is a constructive organ that predicts environmental demands in order to efficiently deploy its limited resources (Engel et al., 2001; Friston, 2012; Raichle, 2010). The above example suggests that formal and temporal structure are conceptually independent, but may interact to optimize adaptive behavior. To this end, the predictive mechanism has to adjust to novel situations and events in real time and it has to be able to infer future events from perceived regularity in formal and temporal structure. Predictive adaptation is probably most efficient when it pertains to a specific point in time. Such specific temporal prediction needs to be distinguished from temporal order, which does

2588

M. Schwartze, S.A. Kotz / Neuroscience and Biobehavioral Reviews 37 (2013) 2587–2596

Fig. 1. Type and timing of events. (A), (B) Dynamic signals imply interplay of persistence and change, expressed in the concepts of succession (two events are perceived as different) and duration (the interval between two events; Fraisse, 1984). Events (e) can be described in terms of their formal structure (consisting of a set of features (f) which allow for their identification). Changes (c) in formal structure generate the temporal structure of successive events. For example, in morse-code, the alternation of gap events with short and long events with a temporal relation of 1:1:3 is used to encode messages. Successive events may develop on different time-scales in parallel, e.g., ‘ready, set, go’ but also ‘set’ instantiate three successive events, albeit with differing temporal structure. (C) Predictions concerning successive events can be temporally specific if the temporal relation between events is known (t), or non-specific, if the temporal relation between events is unknown, with several variants of uncertainty implied in neuropsychological contexts (Kahneman and Tversky, 1982). Specific temporal prediction may be more efficient on particular time-scales, while it may be irrelevant on others. In order to obtain knowledge about temporal structure and to use this information to optimize behavior, it is necessary to generate some internal representation of temporal structure. (D) Dynamic attending theory (Large and Jones, 1999) suggests that the allocation of attention can be modeled as an adaptive oscillatory process. Phase- and period adjustment of this oscillation based on the temporal relation between two events provides the opportunity to predictively focus attention. The striatal beat frequency model (Matell and Meck, 2004) proposes that the activity of nonadaptive oscillations (following an initial phase reset) provides a ‘timestamp’ for a specific interval (coincidence detection). Such an explicit encoding of intervals allows for evaluation and storage of temporal structure as a potential basis for predictive adaptation. (E) Linear representations (reflecting more continuous sampling of a dynamic signal) may be necessary to unambiguously identify an event, whereas a non-linear representation may provide an unambiguous representation of temporal structure.

not automatically suggest temporal specificity. Crucially, specific temporal prediction implies some form of an adequate neural representation of temporal structure and temporal relations. However, it is still unclear how precisely the brain deals with this task. The issue is further aggravated by the fact that different mechanisms may contribute to one and the same goal: the ability to exploit temporal structure to generate temporal predictions to optimize adaptive behavior. This ability most likely arises from the interaction of different functions and different brain areas, which may demand a revision of classical concepts starting with primary sensory processing and extending to higher level cognitive processes in order to explain the phenomenon.

2. Representations of temporal structure One possibility to represent temporal structure is via synchronization, defined as coupling between different oscillations that start to oscillate with a common frequency (Pikovsky et al., 2001; Fig. 1). Oscillations occur naturally in neurons and in neuronal populations. Since oscillations imply repetitive behavior, they are useful for temporal prediction (the ‘when’ aspect of events)—in

principle independent of formal prediction (the ‘what’ aspect of events; Buzsáki and Draguhn, 2004). Limited cognitive resources such as attention and memory are associated with neural oscillations across different frequency bands (Jensen et al., 2007). These constructs interact and overlap (Awh et al., 2006; Gazzaley and Nobre, 2011) and partly determine the quality of other cognitive operations. For example, attention enforces rhythmic shifting of neuronal excitability, thereby amplifying responses to events in a stimulus stream (Lakatos et al., 2008; Schroeder et al., 2008). In other words, enhanced responses to attended events result from the alignment of high-excitability oscillation phases that are phase-locked to temporal structure (Schroeder and Lakatos, 2008). This view is also central to the entrainment hypothesis put forward in Dynamic Attending Theory (DAT; Jones, 1976; Large and Jones, 1999). According to DAT, the allocation of attention proceeds in a stimulus-driven, oscillatory fashion. Adjusting the phase and the period of an adaptive oscillation (McAuley, 1995) establishes a synchronized, future-oriented attending mode if the temporal structure of the environment is ‘coherent’ (non-arbitrary). This stands in contrast to an analytic attending mode that is employed if temporal structure is incoherent (Jones and Boltz, 1989). DAT thus provides a framework for

M. Schwartze, S.A. Kotz / Neuroscience and Biobehavioral Reviews 37 (2013) 2587–2596

temporal prediction (‘when’) shaping efficient resource allocation and formal processing (‘what’) based on synchronization via unidirectional coupling of oscillations (Barnes and Jones, 2000; Large and Kolen, 1995; McAuley and Jones, 2003). In line with this proposal, coherent temporal structure has been shown to enhance the attention-dependent response to changes in formal structure embedded in verbal and non-verbal auditory sequences (Schwartze et al., 2011; Otterbein et al., 2012). A fundamentally different way to represent temporal structure is via a dedicated temporal processing system (Ivry and Schlerf, 2008). Models in this category share the notion of an internal clock mechanism that represents temporal structure in an explicit manner (Buhusi and Meck, 2005; Ivry and Schlerf, 2008). It is important to note that, while entrainment and dedicated temporal processing take different perspectives, they are not necessarily exclusive, because analytic attending may well entail reliance on explicit internal or external dedicated clock mechanisms (Grondin, 2001; Jones and Boltz, 1989). However, primarily based on selective impairments found in cerebellar patients, the cerebellar timing hypothesis posits that the cerebellum instantiates such a dedicated system, that is engaged in precise automatic, salient-feature, or event-based (i.e. ‘discrete-event’ (Buhusi and Meck, 2005)) temporal processing in the sensory- and sensorimotor domain (Ivry and Schlerf, 2008; Ivry and Spencer, 2004; Spencer et al., 2003). Related proposals such as the Striatal Beat Frequency (SBF) model bring the basal ganglia and several neocortical areas, including dorsolateral prefrontal cortex and supplementary motor area into the equation (Buhusi and Meck, 2005; Coull et al., 2011). The SBF model deals with attention-dependent, interval-based (i.e. ‘continuous-event’ (Buhusi and Meck, 2005)) temporal processing. It suggests that, following an initial phase reset, the oscillatory activity of ensembles of non-adaptive cortical oscillators with different periods is integrated by the basal ganglia, and more specifically striatal spiny neurons, to generate a ‘timestamp’ for a specific interval (coincidence detection; Matell and Meck, 2004; Meck et al., 2008; Miall, 1989). However, while dedicated cerebellar and cortico-striatal temporal processing systems may work in parallel, structural and functional connections between the cerebellum, the basal ganglia, the thalamus, and the supplementary motor area may allow for interaction within an integrative temporal processing network (Schwartze et al., 2012). This network bridges the spatial segregation of functionally related processes by means of a dual-pathway architecture. 3. A dual-pathway architecture One crucial aspect of the respective network is the dissociation of temporal and formal structure also at the neurofunctional level. This proposal rests on four basic assumptions: (i) transmission of events via the classic ascending sensory pathways preserves a detailed representation of formal structure, (ii) rapid cerebellar transmission reduces input to a discrete, event-based representation of temporal structure that implicitly encodes the temporal relation between successive events, (iii) rapid cerebellar transmission drives adaptive and/or non-adaptive cortical oscillations, and (iv) these oscillations explicitly represent temporal relations, thus providing a basis for specific temporal prediction. In the following, we will trace these assumptions across different processing stages, starting with an integral structural component of the network, the thalamus. Just considering the size and the strategic position of the thalamus suggests that it is more than a relay for sensory information (Bartlett and Wang, 2007; Crick, 1984; Sherman, 2007). As Sherman and Guillery (2006, p. 403) emphasized, ‘cortex without thalamus is rather like a great church organ without an organist: fascinating, but useless’. Consequently, an understanding of temporal

2589

prediction depends critically on knowledge about what kind of information the thalamus receives, how it acts upon it, and what kind of information the thalamus sends to its targets. In general, the thalamo-cortical system is well fitted for temporal binding, dynamic filtering of temporal structure, and the synchronization of auditory and attentional processes (Alitto and Usrey, 2005; Llinás and Steriade, 2006; Winer and Lee, 2007). The thalamus mediates information flow between subcortical and cortical areas by transmitting non-linear and linear signals (Akkal et al., 2007; Sherman and Guillery, 2006). Accordingly, the proposed differentiation of an event-based, non-linear representation of temporal structure and a more continuous, linear representation of the same input rests on a functional interpretation of electrophysiological characteristics of thalamic cells (Fig. 2). From the perspective of the dual-pathway architecture, the primarily tonic-encoded linear representation is conveyed to primary sensory cortex, while a primarily burst-encoded, non-linear representation of temporal structure is transmitted to frontal cortices. These representations of an input are considered primarily burstand tonic-encoded, because, in principle, each firing mode may contribute to both representations. However, functionally, the primarily tonic-encoded representation preserves a greater level of linear detail in a combination of lower and higher frequency portions, whereas the primarily burst-encoded representation affords temporally precise and rapid transmission of temporal structure. As indicated, the linear representation reaches its neocortical targets via classic ascending connections comprising the thalamus. The combination of lower and higher frequency portions within this representation is most likely reflected in oscillatory activity across different frequency bands, which is in turn modulated by attention (Besle et al., 2011). The prime example for a highly complex dynamic input, the speech signal, conveys temporally structured energy fluctuations associated with syllabic and segmental aspects (in the delta/theta frequency bands), as well as rapid spectro-temporal transitions (in the gamma band: >25 Hz). It has been shown that these properties map onto neural oscillatory activity (Giraud et al., 2007; Ghitza and Greenberg, 2009) and that phase-locked cortical oscillations across these frequency bands are instrumental during memory access (Ghitza, 2011). Thalamic bandpass-filtering and selective amplification of salient events in burst-mode may drive the oscillatory responses in the delta and theta range, entrain thalamo-cortical loops, and thereby facilitate sensory integration in general (Llinás and Steriade, 2006; Luo et al., 2010; Schroeder et al., 2008). However, there are many open questions regarding the functional significance of modulations across different frequency bands, e.g. interactions between lower-frequency delta-to-theta and higher-frequency alpha, beta, and gamma responses (Arnal and Giraud, 2012). A sufficiently detailed representation of formal structure that is acquired via low-pass-filtering in tonic-mode deems necessary for stable memory processing, as reflected, for example, in high intelligibility, and correspondingly in better comprehension, in speech processing (Luo and Poeppel, 2007; Ghitza, 2011; Peelle and Davis, 2012; but see Cummins, 2012). Low-pass filtering of the speech signal can be quite substantial and still allow for decent recognition, especially for vowels (Shannon et al., 1995). This suggests that the rhythmic succession of increasing acoustic energy that is commonly associated with vowel onsets, especially stressed ones, may be part and parcel of the interaction of formal and temporal structure during speech processing (Kotz and Schwartze, 2010), probably reflecting the fact that perceived temporal regularity in speech processing does not conform to onset regularity (Morton et al., 1976; Scott, 1998). However, the example of speech processing also illustrates that higher-level perceptual integration of an input requires more than lower-level sensory integration comprising spectro-temporal analysis and some simple mapping

2590

M. Schwartze, S.A. Kotz / Neuroscience and Biobehavioral Reviews 37 (2013) 2587–2596

Fig. 2. Thalamic burst- and tonic-firing-modes. The thalamus is part of ascending sensory pathways (here the primary auditory pathway (A)). If stimulated in an oscillatory fashion (in this example from the visual domain it is a sinusoidal grating), thalamic cells respond in two distinct modes ((B) adapted from Sherman, 2001; Sherman and Guillery, 2002). Tonic-firing affords greater linear summation and encodes a more faithful stimulus representation (acting as a lowpass filter constrained by the resolution limit; Sherman and Guillery, 2006). Burst-firing occurs within frequency limits of 1–10 Hz (acting as a band-pass filter), and is most efficient at about 4 Hz. It occurs approximately at phase zero of the underlying oscillation and encodes a non-linear stimulus representation associated with the efficient detection and signaling of sudden novel, interesting, or threatening events in the environment. Burst-firing exerts a powerful “wake-up call” to neocortical targets, potentially guiding the deployment of attentional resources (Sherman, 2001; Swadlow and Gusev, 2001; Lesica et al., 2006). Since these characteristics are the same across thalamic relay cells, the dual-mode principle is most likely applicable to all sensory transmission involving the thalamus.

from signal to meaning. As all adaptive behavior, it depends on the circular exchange of information between the organism and its environment. This process engages the dorsolateral prefrontal cortex and its ability to perform perceptual integration, i.e. the bridging of temporally separate elements into a behavioral gestalt, as well as the function of the dorsolateral prefrontal cortex in generating moment-to-moment models of patterns of events (Fuster, 1995; Huettel et al., 2002). As a matter of consequence, the processing of formal structure and sensory integration hinge mostly on temporal order and require accumulation of information over time, whereas specific temporal prediction may be more relevant in perceptual integration. Processing of formal structure and specific temporal prediction may hence evolve on different time-scales, as the processing of formal structure may demand more detail and higher temporal resolution. Applied to the initial example, this suggests that the identification of ‘ready’ requires rapid accumulation of numerous features in the correct temporal order relative to the temporal bridging of ‘ready’ and ‘set’. The dorsolateral prefrontal cortex copes with the temporal syntax of behavior, including the expectation of, and the preparation for, anticipated events, which in turn involves working memory and selective attention that is directed towards dynamic sensory input (Fuster, 2001). In this specific context, memory may be defined as an adaptive pattern-recognition process that allows detection of predictive cues to guide behavior, serving the human disposition to infer causality and predictability even when it does not exist (Ivry and Knight, 2002). Dynamic allocation of attention, as proposed by DAT, and dedicated temporal processing, as suggested by SBF, may support prefrontal cortex by adding temporal specificity to the predictive perceptual integration of dynamic input. Yet, frontal cortex is not expected to provide the elements of sensory- or long-term memory itself. It is hence necessary to consider both, the initiation and the time-course of the processes that are responsible for the retrieval and the bridging of these elements. If frontal cortex implements specific temporal prediction in perceptual integration, it is reasonable to assume that it is activated simultaneously or even in advance, i.e. before it receives elements retrieved from

memory. Ongoing perceptual integration of these elements most likely recruits connections between sensory and frontal areas. From the perspective of the proposed dual-pathway architecture, the memory-related function is based on the primarily linear representation conveyed to sensory cortices, which holds a sufficient amount of detail in order to identify specific elements. The activating function relies on the primarily non-linear representation conveyed to frontal cortex, which sacrifices detail in favor of temporal discreetness. Additionally, this connection may contribute to fronto-central oscillatory responses in the delta range (Stefanics et al., 2010). However, these pathways may not only differ in their functional role and in their cortical targets, but also in the origin of input. Here, we propose that the non-linear representation arises from an event-based representation of temporal structure, established by non-motor parts of the cerebellum and cerebellum-like structures (Fig. 3). 4. Rapid cerebellar transmission of auditory input It is evident that the interaction between the frontal cortex and the cerebellum is at least as significant as cerebellar interaction with motor cortex (Ramnani, 2006; Strick et al., 2009). In addition to its role in motor control, there is increasing evidence on the role of the cerebellum in prediction and in particular in temporal prediction (Courchesne and Allen, 1997; Nixon, 2003; O’Reilly et al., 2008). Attentional orienting and shifting have also been related to cerebellar function, including the precisely timed dynamic adjustment or ‘movement’ of attention (Akshoomoff and Courchesne, 1992). However, in general, the picture is mixed and a unified theory of cerebellar function remains elusive (Baillieux et al., 2008; Timmann and Daum, 2007). The question arises as to whether there is a common denominator among some of the proposed cerebellar functions. One candidate may be a broad interpretation of the term ‘tracking behavior’, i.e. the continuous orienting and maintenance of attention towards a dynamic input, following an initial cerebellothalamo-cortical ‘wake-up-call’ for attention (Huang and Liu, 1990;

M. Schwartze, S.A. Kotz / Neuroscience and Biobehavioral Reviews 37 (2013) 2587–2596

2591

Fig. 3. A dual-pathway architecture. Transmission of formal and temporal structure recruits different ascending pathways (A). In this auditory example, a primarily linear stimulus representation, which preserves detailed formal structure, passes several processing stages, including inferior colliculus (IC) and thalamus (THAL) on its way to temporal cortex (TC, blue). A primarily non-linear stimulus representation reaches frontal cortex (FC) via rapid cerebellar transmission, using a pathway from dorsal cochlear nucleus (DCN) to the cerebellum (CE) and thalamus (red). Whereas the former plays a role in establishing sensory-memory representations, in accessing long-termmemory, and in formal prediction (e.g. in terms of a “segment prediction, error”; Gagnepain et al., 2012), the latter prepares frontal cortex for perceptual integration and drives oscillations capable of explicitly representing the temporal relation between events (B). This is accomplished by means of phase- and period-adjustment of adaptive oscillations and/or via coincidence detection of non-adaptive oscillations providing input to basal ganglia (BG) and striato-thalamo-cortical circuits, respectively (green). Temporal relations are not bound to events (E) arising from absolute changes (e.g. stimulus onset), because also relative change (e.g. rising spectral energy) may be signaled as a discrete event to the cortex (lower panel). Rapid cerebellar transmission exerts a “wake-up-call” to frontal cortical areas that is substantial to the initial, orienting of attention, quick responding, and continuous maintenance of attention to a dynamic input. Rapid cerebellar transmission thereby prepares frontal cortex for the perceptual integration of memory representations (black arrow). Crucially, preparation based on rapid cerebellar transmission is beneficial also with incoherent temporal structure (temporally non-specific preparation). However, it is expected to be most efficient with temporal structure that is perceived as coherent (temporally specific preparation). Temporal predictability in the latter case allows for the establishment of a synchronized processing mode. Accordingly, the system may per default operate in a preparatory mode, but switch to a temporally specific predictive mode whenever it encounters (physical or perceived) regularity in temporal structure. Thus, ‘prediction’ is inherent to the dual-pathway architecture, while specific temporal prediction also depends on stimulus characteristics and cognitive disposition. Regularity may be used to predictively modulate processing via reciprocal, direct, and descending connections, e.g., between different levels of the central auditory system, including different cortical areas, but also between cortex and thalamus (Lee and Sherman, 2011), as well as between the cerebellum, the thalamus, and the basal ganglia (Hoshi et al., 2005; Bostan and Strick, 2010), going as far as to adjust cerebellar internal models via cortico-pontine connections, or to tune adaptive filtering of sensory input in the DCN. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Schwartze et al., 2012a). Efficient tracking behavior depends on rapid transmission of sensory input, source localization and identification, suppression of unnecessary input, and adequate adaptation. While the identification of an event builds upon formal structure, it may be sufficient to transmit the mere presence of successive events and their temporal relation in order to focus attention in time. These aspects may be part of ‘a more encompassing theory’ relating the cerebellum to preparatory signaling and

preparatory actions, with attention conceived of as an ‘act of preparation and selective modulation’ (Courchesne and Allen, 1997). Specific temporal prediction may help to tune these preparatory and modulatory processes. However, this requires that the cerebellum has access to temporally precise information, ideally provided via connections to early stages of sensory processing. Such a cerebellar pathway may engage in the rapid transmission of sensory input to targets outside of primary sensory cortices.

2592

M. Schwartze, S.A. Kotz / Neuroscience and Biobehavioral Reviews 37 (2013) 2587–2596

It is well known that the cerebellum engages in purely sensory processing, including auditory processing (for reviews see Callan et al., 2007; Petacchi et al., 2005; Sens and de Almeida, 2007). However, a more fine-grained look at the characteristics of cerebellar processing is required to further delineate the potential cerebellar contribution to temporal prediction. Data come from a number of invasive studies in various animal species. While direct mapping from animal to human function is not feasible, a comparative approach provides the required level of detail and underlines some more general aspects of predictive adaptation. For example, in monkeys, the production of salient periodic sounds serves at least two functions: It attracts attention and ensures that the signal is recognized over the background noise of an animal colony (Remedios et al., 2009). This nicely summarizes the functional aspects discussed so far. Following the observation of short-latency evoked responses after click stimulation, Snider and Stowell (1944) confirmed the existence of a cerebellar auditory area. Bilateral damage to the inferior colliculus abolished the responses, while other studies report, albeit attenuated, responses after intercollicular transection (Aitkin and Boyd, 1975; Lorenzo et al., 1977). The cerebellar area is polysensory, insofar as auditory and visual receiving areas are overlapping, with enhanced responses obtained for auditory and visual stimuli presented in a time-window of 10-25 ms, and attenuated responses outside this window (Deura and Snider, 1964; Levy et al., 1961; Snider and Stowell, 1944). Such an integrative amplification may be useful to filter coherent temporal structure across different modalities into a single event. The short latency of the responses deems critical as about 10 ms suggest few synaptic passages in a pathway for rapid auditory transmission of soundprovoked impulses (Aitkin and Boyd, 1975; Highstein and Coleman, 1968). After conditioning, click-evoked responses can be observed in the cerebellar dentate of cats after 4–6 ms (Wang et al., 1991; Xi et al., 1994). Cerebellar responses may precede neocortical responses (Lorenzo et al., 1977; Misrahy et al., 1961). Misrahy et al. (1961) and Shofer and Nahvi (1969) discuss a triphasic pattern comprising early cerebellar responses followed by responses in auditory cortex, followed by secondary cerebellar responses, with only the latter being suppressed by anesthesia, cooling, or destruction of temporal cortex. The most common response is a burst occurring at the onset of an auditory stimulus, with variations in latency from about 6 to 30 ms (Aitkin and Boyd, 1975; Altman et al., 1976; Freeman, 1970; Huang and Burkard, 1986). Reminiscent of thalamic burst-firing, the amplitude of the cerebellar response decays with stimulation rates above 10 Hz (Lorenzo et al., 1977). Cells in different locations of the cerebellar auditory area display broad frequency tuning and rather poor selectivity to sound intensity and duration (Altman et al., 1976; Freeman, 1970; Huang and Burkard, 1986), potentially pointing towards a function in signaling just the presence of an input, as well as in localizing its source. In other words, while the cerebellum seems unfit to process detailed formal structure, it seems to be sensitive to relative changes in sensory input occurring in a timewindow of roughly 10 to 30 ms. If the relative change evokes a response, the cerebellum signals this as an event to the thalamus. In turn, the thalamus transmits this information to cortical targets, where it is reflected in oscillatory responses, most likely in the delta-to-theta range. This mechanism may be advantageous for encoding a temporally precise event-based representation of the temporal structure of an input, thus providing distinct temporal markers also for a complex input such as ‘ready, set, go!’. However, analogous to thalamo-cortical interaction, it is necessary to specify the actions of the brainstem branches of the cerebellothalamic drivers to entertain this possibility (Sherman and Guillery, 2006).

5. Cerebellar connections to early stages of auditory processing Input to the cerebellum occurs via climbing- or mossy-fiberactivation. If one considers the cerebellum as a relay for rapid transmission, a direct sensory afferent pathway to the cerebellum is required. Indeed, there is some evidence that such a pathway exists between the cerebellum and the dorsal cochlear nucleus (DCN). Using retrograde degeneration, Niemer and Cheng (1949) identified fibers that reach the cerebellar auditory area from the DCN. Retrograde transport after tracer injections into the cerebellar auditory area was found to label the cochlear nuclei bilaterally, but neither the superior olive nor the inferior colliculus (Huang et al., 1982). Labeled cells in the cochlear nucleus were also found after injection into the cerebellar vermis and hemispheres (Zhang et al., 1990), as well as after injections into the cerebellar dentate (Wang et al., 1991). Wang et al. (1991) suggest that a primary auditory transmission pathway exists between dorsal and ventral cochlear nuclei, the cerebellar dentate, and rostral thalamus. Certainly not exhaustive, these results, nevertheless, allow speculation about cerebellar afferents originating in the DCN, while they do by no means rule out functional significance of cerebellar afferents from other, progressively later, processing stages, including climbing-fiber-activation (Wu et al., 2011). These later inputs may stem from both peripheral and neocortical sites of origin. However, what renders the DCN interesting with respect to rapid cerebellar transmission, and the processing of temporal structure in particular, is not only its anatomical location close to the periphery, but also its ability for ‘event-coding’. The DCN is considered a cerebellum-like structure as it parallels cerebellar organization. It receives input via the auditory nerve and from several brain areas, including the inferior colliculus and the auditory cortex (Bell et al., 2008; Oertel and Young, 2004). Any functional interpretation of the DCN must hence consider interaction between input from the sensory periphery and input from more central sites (Bell et al., 2008). Cells in the DCN are fitted to signal important acoustic events such as spectral maxima or minima to the auditory system, especially steeply rising spectral edges (Nelken and Young, 1996; Reiss and Young, 2005). Edgesensitivity may be involved in signaling relative increases in the energy-level of a stimulus as a discrete event to the cerebellar auditory area and from there to thalamic and neocortical targets, thereby transmitting a representation of temporal structure. The DCN performs adaptive filtering of sensory input that serves to predict and to effectively cancel out sensory consequences of selfgenerated action (Reiss and Young, 2005; Requarth and Sawtell, 2011; Roberts and Portfors, 2008). Downward modulation of this process based on specific temporal predictions gives rise to the possibility of predictive event-coding already at the level of the DCN (Nelken, 2012; Roberts et al., 2006; Schwartze et al., 2012a). The superior temporal precision granted by this pathway could be beneficial to the automatic adjustment of overt behavior to the temporal structure of the environment. For example, phasecorrection in adaptive sensorimotor synchronization (e.g. by means of finger-tapping to an auditory pacing signal) is considered an automatic process that rests on reference ‘time points’ (events), i.e. the times a tap or a tone occurs (Repp and Keller, 2004, 2008). Rapid cerebellar transmission via the DCN may provide the cerebellum with sensory feedback and a temporal reference according to which the actual motor behavior is adjusted, thus supporting the role of the cerebellum in subliminal error-correction (Bijsterbosch et al., 2010). However, it stands to reason how such a reference relates to the actual optimization of behavior, e.g. in terms of the periodic modulation of activity in the beta band, which may support the initiation of movement (Fujioka et al., 2012). A second process, period-correction, is considered to be

M. Schwartze, S.A. Kotz / Neuroscience and Biobehavioral Reviews 37 (2013) 2587–2596

attention-dependent and to rest on the interval between events (Repp and Keller, 2004, 2008). While such supraliminal error correction also engages the cerebellum, it recruits additional parietal and frontal areas (Bijsterbosch et al., 2010). Moreover, patients with basal ganglia lesions demonstrate impaired period-correction, potentially reflecting dysfunctional allocation of attention in time (DAT) and/or interval-based temporal processing (SBF; Schwartze et al., 2011a), substantiating the role of the basal ganglia in the processing of temporal relations. In principle, sensorimotor synchronization is possible at rates of about 10 Hz if biomechanical constraints are circumvented, i.e. by employing faster pacing than movement rates (1:n synchronization; Repp, 2005). Convergence of cerebellar and thalamic responses and sensorimotor performance around this rate may hint at a fundamental constraint in synchronization via rapid cerebellar transmission and thus also in the predictive adaptation of behavior. Sensorimotor synchronization is of course a simplistic form of such a behavior, as it typically involves periodic pacing and movement (Repp, 2005). However, more generally speaking, connections from the DCN to the cerebellum may initiate processing and provide direct sensory feedback to large-scale cortico-cerebellar networks, which generate and adjust internal models (Ito, 2008). This mechanism could afford event-based small-scale automatic adjustments of behavior, thereby sparing cortical neural and cognitive resources, including attention-dependent mechanisms. Hence, if such a pathway exists in humans, it may be critically involved in optimizing adaptive predictive behavior involving an auditory component. This optimizing function needs to be formally dissociated from basic and also from pathological function. As indicated, rapid cerebellar transmission is supposed to drive cortical oscillations engaged in establishing an internal representation of temporal structure. It thereby adds the opportunity for specific temporal prediction to an otherwise functional system to make maximal use of both, formal and temporal structure. Rapid cerebellar transmission and specific temporal prediction are, however, not a prerequisite for sensory processing.

6. Summary and conclusions The dual-pathway architecture outlined in the preceding sections proposes that temporal prediction recruits a network of brain areas, in which sensory processing interfaces with temporal processing in order to optimize adaptive behavior. Since oscillatory dynamics are inherent to the network, they inevitably serve different functions, including the allocation of attention, memory access, temporal processing, temporal binding, and the development of cortical circuits (Engel et al., 2001; Uhlhaas et al., 2009). Here, we focused on the role of oscillatory processes in specific temporal prediction, as well as processes that precede cortical processing stages. Furthermore, we used auditory information as an example for a dynamic input, acknowledging the fact that some aspects may be specific to this modality, while the overarching concept of specific temporal prediction based on rapid cerebellar transmission is expected to hold true beyond audition. However, further differentiation of thalamo-cortical and cortico-cortical mechanisms would have to be considered, for example in the case of cross-modal sensory input (Arnal and Giraud, 2012; Schroeder et al., 2008). Starting at the level of the thalamus, two ascending pathways are discussed: one that conveys a primarily tonic-encoded linear representation of an input to sensory cortical areas; the other transmits a primarily burst-encoded non-linear representation of the same input to frontal cortical areas. Sensory integration of the former establishes sensory objects, sensory-memory representations, and a basis for long-term-memory access. The latter prepares frontal cortex for the perceptual integration of sensory

2593

objects and elements retrieved from memory. To this end, the linear representation retains detailed formal structure, whereas the non-linear representation involves rapid transmission of temporal structure. The non-linear representation may drive oscillatory mechanisms capable of representing temporal structure and establishes a functional link between dedicated temporal processing systems such as the cerebellum, the supplementary motor area, and the basal ganglia. These aspects may also be reflected in the shortand longer-latency, non-feature-specific, non-selective responses to sensory input obtained from frontal cortical areas, including the supplementary motor area (Giard et al., 1994; Kurata and Tanji, 1985; Shalgi and Deouell, 2007). In addition, the proposed role of the cerebellum in the precise encoding of temporal structure and the role of the basal ganglia in the processing of temporal relations establish a conceptual link between the dual-pathway architecture and formal models of predictive coding. According to this theory, recipients invert a hierarchical model to recover information embedded across different time-scales of a dynamic signal (Kiebel et al., 2009; Friston, 2012). From this perspective, the dual-pathway architecture may be instrumental in isolating slower dynamics, i.e., high-level patterns, which guide the predictive integration of faster dynamics (Friston, 2012; Schwartze et al., 2012a). Perceptual integration is expected to be most efficient with optimal allocation of limited cognitive resources such as attention. With dynamic input, optimal allocation implies allocation of attention in time. Crucially, this requires predictive adaptation to the temporal structure of the signal. While predictions concerning the future course of events may be unspecific with respect to the precise temporal locus of an event, they may be temporally specific if the brain encodes an adequately precise internal representation of external temporal structure. Here we ascribe this function to the non-linear stimulus representation, which drives adaptive and/or non-adaptive cortical oscillations engaged in the dynamic allocation of attention in time (DAT; Large and Jones, 1999) and in interval-based temporal processing (SBF; Matell and Meck, 2004), respectively. If temporal structure gives rise to perceived regularity, these mechanisms allow the system to switch from a temporally non-specific preparatory mode to a temporally specific, potentially synchronized, predictive mode. This process may relate to the switching between a rhythmic and a continuous processing mode (Schroeder and Lakatos, 2009). Obviously, dynamic signals differ in the degree with which they may give rise to perceived regularity. However, even with respect to speech processing, it stands to reason whether specific temporal prediction may not contribute differently to the development (e.g. motherese, nursery rhymes), the proficient use (e.g. under adverse listening conditions, rhetoric), or the compensation for dysfunction and decline (e.g. cochlear-implants, ageing). In turn, any such differences may interact with ontogenetic factors which constrain the ability to use temporal structure to optimize behavior (McAuley et al., 2006; Huss et al., 2011; Lustig and Meck, 2011; Trainor, 2012). The non-linear representation makes use of rapid cerebellar transmission and thalamic burst-firing to transmit saliency and in order to evoke reliable postsynaptic responses (Izhikevich, 2004, 2007). With successive events, feedback from the cortex to the thalamus may lead to further amplification of thalamo-cortical oscillations in response to a relatively weak input (Buzsáki and Draguhn, 2004). Similar feedback mechanisms may operate at earlier processing stages, which adaptively adjust their function on the basis of modulatory input derived from perceived regularity among events. This could lead, for example, to facilitatory convergence of cerebrocortical and peripheral inputs already at the level of the cerebellum (Freeman, 1970) and, even earlier, to the filtering of specific aspects of an input in cerebellum-like structures (Bell et al., 2008). In combination with cross-modal integrative amplification of cerebellar responses, such mechanisms may be used to enhance

2594

M. Schwartze, S.A. Kotz / Neuroscience and Biobehavioral Reviews 37 (2013) 2587–2596

the contrast between an event and its vicinity. The enhancement of contrast arising from physical signal change is a fundamental operation of sensory systems (Kluender et al., 2003). The cerebellum and cerebellum-like structures may perform this operation to guarantee the fast and stable transmission of temporal structure, which is mandatory for efficient predictive adaptation of behavior. However, the specific contributions of filtering and selective amplification to this process, as well as their interplay, remain to be determined. While the consequences of predictive adaptation are manifold and may be seen in both, simplistic and complex behavior, it builds upon some straightforward principles. Perhaps the most important one is the notion of optimized behavior with adequate timing. The fundamental nature of this task is reflected in the mechanisms that underlie specific temporal prediction. As indicated, the role of the cerebellum in this context may trace back to tracking behavior in general, and ‘sound-evoked orientation’ in particular (Aitkin, 1986), whereas the role of the basal ganglia in the encoding of the temporal relation between events may be rooted in learning, e.g. of the time delay between an event and a reward (Jin et al., 2009). On the one hand, these primitives may have remained virtually the same, while increasingly more complex cognitive operations developed around them, including the ability to efficiently adjust overt and covert behavior to future events. On the other hand, damage to core components of the underlying temporal processing network should result in suboptimal use of temporal structure, which in turn should affect a wide range of functions, not only in production but also in purely perceptual tasks. As a matter of consequence, the dual pathway architecture is expected to serve temporal prediction irrespective of whether it manifests in the perception or the production of music and speech, in sensorimotor synchronization via finger tapping, or in the timely action following a simple sequence of acoustic events such as ‘ready, set, go!’. Acknowledgments This work is supported by DFG KO 2268/6-1 granted to S. A. K. References Aitkin, L.M., Boyd, J., 1975. Responses of single units in cerebellar vermis of the cat to monaural and binaural stimuli. J. Neurophysiol. 38, 418–429. Aitkin, L.M., 1986. The Auditory Midbrain: Structure and Function in the Auditory Pathway. The Humana Press, Clifton. Akkal, D., Dum, R.P., Strick, P.L., 2007. Supplementary motor area and presupplementary motor area: targets of basal ganglia and cerebellar output. J. Neurosci. 27, 10659–10673. Akshoomoff, N.A., Courchesne, E., 1992. A new role for the cerebellum in cognitive operations. Behav. Neurosci. 106, 731–738. Altman, J.A., Bechterev, N.N., Radionova, E.A., Shmigidina, G.N., Syka, J., 1976. Electrical responses of the auditory area oft he cerebellar cortex to acoustic stimulation. Exp. Brain Res. 26, 285–298. Alitto, H.J., Usrey, W.M., 2005. Dynamic properties of thalamic neurons for vision. Prog. Brain Res. 149, 83–90. Arnal, L.H., Giraud, A.L., 2012. Cortical oscillations and sensory predictions. Trends Cogn. Sci. 16, 390–398. Awh, E., Vogel, E.K., Oh, S.H., 2006. Interactions between attention and working memory. Neuroscience 139, 201–208. Baillieux, H., De Smet, H.J., Paquier, P.F., De Deyn, P.P., Mariën, P., 2008. Cerebellar neurocognition: insights into the bottom of the brain. Clin. Neurol. Neurosurg. 110, 763–773. Bar, M., 2007. The proactive brain: using analogies and associations to generate predictions. Trends Cogn. Sci. 11, 280–289. Barnes, R., Jones, M.R., 2000. Expectancy, attention, and time. Cogn. Psychol. 41, 254–311. Bartlett, E.L., Wang, X., 2007. Neural representations of temporally modulated signals in the auditory thalamus of awake primates. J. Neurophysiol. 97, 1005–1017. Bell, C.C., Han, V., Sawtell, N.B., 2008. Cerebellum-like structures and their implications for cerebellar function. Annu. Rev. Neurosci. 31, 1–24. Besle, J., Schevon, C.A., Metha, A.D., Lakatos, P., Goodman, R.R., McKhann, G.M., Emerson, R.G., Schroeder, C.E., 2011. Tuning of the human neocortex to the temporal dynamics of attended events. J. Neurosci. 31, 3176–3185. Bijsterbosch, J.D., Lee, K., Hunter, M.D., Tsoi, D.T., Lankappa, S., Wilkinson, I.D., Barker, A.T., Woodruff, P.W.R., 2010. The role of the cerebellum in sub- and supraliminal

error correction during sensorimotor synchronization: evidence from fMRI and TMS. J. Cogn. Neurosci. 23, 1100–1112. Bostan, A.C., Strick, P.L., 2010. The cerebellum and the basal ganglia are interconnected. Neuropsychol. Rev. 20, 261–270. Bubic, A., von Cramon, D.Y., Schubotz, R.I., 2010. Prediction, cognition and the brain. Front. Hum. Neurosci. 4, 1–15. Buhusi, C.V., Meck, W.H., 2005. What makes us tick?: functional and neural mechanisms of interval timing. Nat. Rev. Neurosci. 6, 755–765. Buzsáki, G., Draguhn, A., 2004. Neuronal oscillations in cortical networks. Science 304, 1926–1929. Callan, D.E., Kawato, M., Parsons, L., Turner, R., 2007. Speech and song: the role of the cerebellum. Cerebellum 6, 321–327. Coull, J.T., Cheng, R., Meck, W.H., 2011. Neuroanatomical and neurochemical substrates of timing. Neuropsychopharmacology 36, 3–25. Courchesne, E., Allen, G., 1997. Prediction and preparation, fundamental functions of the cerebellum. Learn. Mem. 4, 1–35. Crick, F., 1984. Function of the thalamic reticular complex: the searchlight hypothesis. PNAS 81, 4586–4590. Cummins, F., 2012. Oscillators and syllables: a cautionary note. Front. Psychol. 3, 364. Deura, S., Snider, R.S., 1964. The interaction of various impulses in the cerebellum. J. Neurol. Sci. 1, 78–196. Engel, A.K., Fries, P., Singer, W., 2001. Dynamic predictions: oscillations and synchrony in top-down processing. Nat. Rev. Neurosci. 2, 704–716. Fraisse, P., 1984. Perception and estimation of time. Ann. Rev. Psychol. 35, 1–36. Freeman, J.A., 1970. Responses of cat cerebellar Purkinje cells to convergent inputs from cerebral cortex and peripheral sensory systems. J. Neurophysiol. 33, 697–712. Friston, K., 2005. A theory of cortical responses. Philos. Trans. R. Soc. B 360, 815–836. Friston, K., 2012. Prediction, perception and agency. Int. J. Psychophysiol. 83, 248–252. Fujioka, T., Trainor, L.J., Large, E.W., Ross, B., 2012. Internalized timing of isochronous sounds is represented in neuromagnetic beta oscillations. J. Neurosci. 32, 1791–1802. Fuster, J.M., 1995. Temporal processing. In: Grafman, J., Holyoak, K.J., Boller, F. (Eds.), Structure and Function of the Human Prefrontal Cortex. N.Y. Academy of Sciences, New York, pp. 173–181. Fuster, J.M., 2001. The prefrontal cortex—an update: time is of the essence. Neuron 30, 319–333. Gagnepain, P., Henson, R.N., Davis, M.H., 2012. Temporal predictive codes for spoken words in auditory cortex. Curr. Biol. 22, 615–621. Gazzaley, A., Nobre, A.C., 2011. Top-down modulation: bridging selective attention and working memory. Trends Cogn. Sci. 16, 129–135. Ghitza, O., Greenberg, S., 2009. On the possible role of brain rhythms in speech perception: intelligibility of time-compressed speech with periodic and aperiodic insertions of silence. Phonetica 66, 113–126. Ghitza, O., 2011. Linking speech perception and neurophysiology: speech decoding guided by cascaded oscillators locked into the input rhythm. Front. Psychol. 2, 1–13. Giard, M.H., Perrin, F., Echallier, J.F., Thévenet, M., Froment, J.C., Pernier, J., 1994. Dissociation of temporal and frontal components in the human auditory N1 wave: a scalp current density and dipole model analysis. Electroencephalogr. Clin. Neurophysiol. 92, 238–252. Giraud, A.L., Kleinschmidt, A., Poeppel, D., Lund, T.E., Frackowiak, R.S.J., Laufs, H., 2007. Endogenous cortical rhythms determine cerebral specialization for speech perception and production. Neuron 56, 1127–1134. Grondin, S., 2001. From physical time to the first and second moments of psychological time. Psychol. Bull. 127, 22–44. Highstein, S., Coleman, P.D., 1968. Responses of the cerebellar vermis to binaural auditory stimulation. Brain Res. 10, 470–473. Hoshi, E., Tremblay, L., Féger, J., Carras, P.L., Strick, P.L., 2005. The cerebellum communicates with the basal ganglia. Nat. Neurosci. 8, 1491–1493. Huang, C., Liu, G., Huang, R., 1982. Projections from the cochlear nucleus to the cerebellum. Brain Res. 244, 1–8. Huang, C., Burkard, R., 1986. Frequency sensitivities of auditory neurons in the cerebellum of the cat. Brain Res. 371, 101–108. Huang, C., Liu, G., 1990. Organization of the auditory area in the posterior cerebellar vermis of the cat. Exp. Brain Res. 81, 377–383. Huettel, S.A., Mack, P.B., McCarthy, G., 2002. Perceiving patterns in random series: dynamic processing of sequences in prefrontal cortex. Nat. Neurosci. 5, 485–490. Huss, M., Verney, J.P., Fosker, T., Mead, N., Goswami, U., 2011. Music, rhythm, rise time perception and developmental dyslexia: perception of musical meter predicts reading and phonology. Cortex 47, 674–689. Ito, M., 2008. Control of mental activities by internal models in the cerebellum. Nat. Rev. Neurosci. 9, 304–313. Ivry, R.B., Knight, R.T., 2002. Making order from chaos: the misguided frontal lobe. Nat. Neurosci. 5, 394–396. Ivry, R.B., Spencer, R.M.C., 2004. The neural representation of time. Curr. Opin. Neurobiol. 14, 225–232. Ivry, R.B., Schlerf, J.E., 2008. Dedicated and intrinsic models of time perception. Trends Cogn. Sci. 12, 273–280. Izhikevich, E.M., 2004. Which model to use for cortical spiking neurons? IEEE Trans. Neural Netw. 15, 1063–1070. Izhikevich, E.M., 2007. Dynamical Systems in Neuroscience: The Geometry of Excitability and Bursting. MIT Press, Cambridge.

M. Schwartze, S.A. Kotz / Neuroscience and Biobehavioral Reviews 37 (2013) 2587–2596 Jensen, O., Kaiser, J., Lachaux, J., 2007. Human gamma-frequency oscillations associated with attention and memory. Trends Neurosci. 30, 317–324. Jin, D.Z., Fuji, N., Graybiel, A.M., 2009. Neural representation of time in cortico-basal ganglia circuits. PNAS 106, 19156–19161. Jones, M.R., 1976. Time, our lost dimension: toward a new theory of perception, attention, and memory. Psychol. Rev. 83, 323–355. Jones, M.R., Boltz, M., 1989. Dynamic attending and responses to time. Psychol. Rev. 96, 459–491. Kahneman, D., Tversky, A., 1982. Variants of uncertainty. Cognition 11, 143–157. Kiebel, S.J., Daunizeau, J., Friston, K.J., 2009. Perception and hierarchical dynamics. Front. Neuroinform. 3, 20. Kotz, S.A., Schwartze, M., 2010. Cortical speech processing unplugged: a timely subcortico-cortical framework. Trends Cogn. Sci. 14, 392–399. Kurata, K., Tanji, J., 1985. Contrasting neuronal activity in supplementary and precentral motor cortex of monkeys. II. Responses to movement triggering vs. nontriggering sensory signals. J. Neurophysiol. 53, 142–152. Kluender, K.R., Coady, J.A., Kiefte, M., 2003. Sensitivity to change in perception of speech. Speech Commun. 41, 59–69. Lakatos, P., Karmos, G., Mehta, A.D., Ulbert, I., Schroeder, C.E., 2008. Entrainment of neuronal oscillations as a mechanism of attentional selection. Science 320, 110–113. Large, E.W., Kolen, J.F., 1995. Resonance and the perception of musical meter. Connect. Sci. 6, 177–208. Large, E.W., Jones, M.R., 1999. The dynamics of attending: how people track timevarying events. Psychol. Rev. 106, 119–159. Lee, C.C., Sherman, S.M., 2011. On the classification of pathways in the auditory midbrain, thalamus, and cortex. Hear. Res. 276, 79–87. Lesica, N.A., Weng, C., Jin, J., Yeg, C., Alonso, J., Stanley, G.B., 2006. Dynamic encoding of natural luminance sequences by LGN bursts. PLoS Biol. 4, e209. Levy, C.K., Loeser, J.D., Koella, W.P., 1961. The cerebellar acoustic response and its interaction with optic responses. Electroencephalogr. Clin. Neurophysiol. 13, 235–242. Llinás, R.R., Steriade, M., 2006. Bursting of thalamic neurons and states of vigilance. J. Neurophysiol. 95, 3297–3308. Luo, H., Poeppel, D., 2007. Phase patterns of neuronal responses reliably discriminate speech in human auditory cortex. Neuron 54, 1001–1010. Luo, H., Liu, Z., Poeppel, D., 2010. Auditory cortex tracks both auditory and visual stimulus dynamics using low-frequency neuronal phase modulation. PLoS Biol. 8, e1000445. Lustig, C., Meck, W.H., 2011. Modality differences in timing and temporal memory throughout the lifespan. Brain Cogn. 77, 298–303. Lorenzo, D., Velluti, J.C., Crispino, L., Velluti, R., 1977. Cerebellar sensory functions: rat evoked potentials. Exp. Neurol. 55, 629–636. Matell, M.S., Meck, W.H., 2004. Cortico-striatal circuits and interval timing: coincidence detection of oscillatory processes. Cogn. Brain Res. 21, 139–170. McAuley, J.D., 1995. Perception of Time as Phase: Toward an Adaptive-oscillator Model of Rhythmic Pattern Processing. In: Indiana University Dissertation. McAuley, J.D., Jones, M.R., 2003. Modeling effects of rhythmic context on perceived duration: a comparison of interval and entrainment approaches to short-interval timing. J. Exp. Psychol. Hum. Percept. Perform. 29, 1102–1125. McAuley, J.D., Jones, M.R., Holub, S., Johnston, H.M., Miller, N.S., 2006. The time of our lives: live span development of timing and event tracking. J. Exp. Psychol. Gen. 135, 348–367. Meck, W.H., Penney, T.B., Pouthas, V., 2008. Cortico-striatal representation of time in animals and humans. Curr. Opin. Neuroniol. 18, 145–152. Miall, C., 1989. The storage of time intervals using oscillating neurons. Neural Comput. 1, 359–371. Morton, J., Marcus, S.M., Frankish, C., 1976. Perceptual centres (P-centres). Psychol. Rev. 83, 405–408. Misrahy, G.A., Spradley, J.F., Beran, A.V., Garwood, V.P., 1961. Acoustic cerebellar pathways in cats. J. Neurophysiol. 24, 159–166. Nelken, I., Young, E.D., 1996. Why do cats need a dorsal cochlear nucleus? J Basic Clin. Physiol. Pharmacol. 7, 199–220. Nelken, I., 2012. Predictive information processing in the brain: the neural perspective. Int. J. Psychophysiol. 83, 253–255. Niemer, W.T., Cheng, S.K., 1949. The ascending auditory system, a study of retrograde degeneration. Anat. Rec. 103, 116. Nixon, P.D., 2003. The role of the cerebellum in preparing responses to predictable sensory events. Cerebellum 2, 114–122. Oertel, D., Young, E.D., 2004. What’s a cerebellar circuit doing in the auditory system? Trends Neurosci. 27, 104–110. O’Reilly, J.X., Mesulam, M.M., Nobre, A.C., 2008. The cerebellum predicts the timing of perceptual events. J. Neurosci. 28, 2252–2260. Otterbein, S., Abel, C., Heinemann, L.V., Kaiser, J., Schmidt-Kassow, M., 2012. P3b reflects periodicity in linguistic sequences. PLoS One 7, e51419. Peelle, J.E., Davis, M.H., 2012. Neural oscillations carry speech rhythm through to comprehension. Front. Psychol. 3, 320. Petacchi, A., Laird, A.E., Fox, P.T., Bower, J.M., 2005. Cerebellum and auditory function: an ALE meta-analysis of functional neuroimaging studies. Hum. Brain Mapp. 25, 118–128. Pikovsky, A., Rosenblum, M., Kurths, J., 2001. Synchronization: A Universal Concept in Nonlinear Sciences. Cambridge University Press, Cambridge. Raichle, M.E., 2010. Two views of brain function. Trends Cogn. Sci. 14, 180–190. Ramnani, N., 2006. The primate cortico-cerebellar system: anatomy and function. Nat. Rev. Neurosci. 7, 511–522.

2595

Reiss, L.A.J., Young, E.D., 2005. Spectral edge sensitivity in neural circuits of the dorsal cochlear nucleus. J. Neurosci. 25, 3680–3691. Remedios, R., Logothetis, N.K., Kayser, C., 2009. Monkey drumming reveals common networks for perceiving vocal and nonvocal communication sounds. PNAS 42, 18010–18015. Repp, B.H., Keller, P.E., 2004. Adaptation to tempo changes in sensorimotor synchronization: effects of intention, attention, and awareness. Q. J. Exp. Psychol. 57A, 499–521. Repp, B.H., 2005. Sensorimotor synchronization: a review of the tapping literature. Psychon. B. Rev. 12, 969–992. Repp, B.H., Keller, P.E., 2008. Sensorimotor synchronization with adaptively timed sequences. Hum. Mov. Sci. 27, 423–456. Requarth, T., Sawtell, N.B., 2011. Neural mechanisms for filtering self-generated sensory signals in cerebellum-like circuits. Curr. Opin. Neurobiol. 21, 602–608. Roberts, P.D., Portfors, C.V., Sawtell, N., Felix, R., 2006. Model of auditory prediction in the dorsal cochlear nucleus via spike-timing dependent plasticity. Neurocomputing 69, 1191–1194. Roberts, P.D., Portfors, C.V., 2008. Design principles of sensory processing in cerebellum-like structures: early stage processing of electrosensory and auditory objects. Biol. Cybern. 98, 491–507. Schroeder, C.E., Lakatos, P., 2008. Low-frequency neuronal oscillations as instruments of sensory selection. Trends Neurosci. 32, 9–18. Schroeder, C.E., Lakatos, P., Kajikawa, Y., Partan, S., Puce, A., 2008. Neuronal oscillations and visual amplification of speech. Trends Cogn. Sci. 12, 106–113. Schroeder, C.E., Lakatos, P., 2009. The gamma oscillation: master or slave? Brain Topogr. 22, 24–26. Schwartze, M., Rothermich, K., Schmidt-Kassow, M., Kotz, S.A., 2011. Temporal regularity effects on pre-attentive and attentive processing of deviance. Biol. Psychol. 87, 146–151. Schwartze, M., Keller, P.E., Patel, A.D., Kotz, S.A., 2011a. The impact of basal ganglia lesions on sensorimotor synchronization, spontaneous motor tempo, and the detection of tempo changes. Behav. Brain Res. 216, 685–691. Schwartze, M., Rothermich, K., Kotz, S.A., 2012. Functional dissociation of pre-SMA and SMA-proper in temporal processing. NeuroImage 60, 290–298. Schwartze, M., Tavano, A., Schröger, E., Kotz, S.A., 2012a. Temporal aspects of prediction in audition: cortical and subcortical neural mechanisms. Int. J. Psychophysiol. 83, 200–207. Scott, S.K., 1998. The point of P-centres. Psychol. Res. 61, 4–11. Sens, P.M., de Almeida, C.I.R., 2007. Participation of the cerebellum in auditory processing. Braz. J. Otorhinolaryngol. 73, 266–270. Shalgi, S., Deouell, L.Y., 2007. Direct evidence for differential roles of temporal and frontal components of auditory change detection. Neuropsychologia 45, 1878–1888. Shannon, R.V., Zeng, F., Kamath, V., Wygonski, V., Ekelid, M., 1995. Speech recognition with primarily temporal cues. Science 270, 303–304. Sherman, S.M., 2001. A wake-up call from the thalamus. Nat. Neurosci. 4, 344–346. Sherman, S.M., Guillery, R.W., 2002. The role of the thalamus in the flow of information to the cortex. Phil. Trans. R. Soc. Lond. B 357, 1695–1708. Sherman, S.M., Guillery, R.W., 2006. Exploring the Role of the Thalamus and Its Role in Cortical Function. MIT Press, Cambridge. Sherman, S.M., 2007. The thalamus is more than just a relay. Curr. Opin. Neurobiol. 17, 417–422. Shofer, R.J., Nahvi, M.J., 1969. Firing patterns induced by sound in single units of the cerebellar cortex. Exp. Brain Res. 8, 327–345. Snider, R.S., Stowell, A., 1944. Receiving areas of the tactile, auditory and visual systems in the cerebellum. J. Neurophysiol. 7, 331–357. Spencer, R.M.C., Zelaznik, H.N., Diedrichsen, J., Ivry, R.B., 2003. Disrupted timing of discontinuous but not continuous movements by cerebellar lesions. Science 300, 1437–1439. Stefanics, G., Hangya, B., Hernádi, I., Winkler, I., Lakatos, P., Ulbert, I., 2010. Phase entrainment of human delta oscillations can mediate the effects of expectation on reaction speed. J. Neurosci. 30, 13578–13585. Strick, P.L., Dum, R.P., Fiez, J.A., 2009. Cerebellum and nonmotor function. Ann. Rev. Neurosci. 32, 413–434. Swadlow, H.A., Gusev, A.G., 2001. The impact of ‘bursting’ thalamic impulses at a neocortical synapse. Nat. Neurosci. 4, 402–408. Timmann, D., Daum, I., 2007. Cerebellar contributions to cognitive functions: a progress report after two decades of research. Cerebellum 6, 159–162. Trainor, L., 2012. Predictive information processing is a fundamental learning mechanism present in early development: evidence from infants. Int. J. Psychophysiol. 83, 256–258. Uhlhaas, P.J., Roux, F., Rodriguez, E., Rotarska-Jagiela, A., Singer, W., 2009. Neural synchrony and the development of cortical networks. Trends Cogn. Sci. 14, 72–80. Wang, X.F., Woody, C.D., Chizhevsky, V., Gruen, E., Landeira-Fernande, J., 1991. The dentate nucleus is a short-latency relay of a primary auditory transmission pathway. Neuroreport 2, 361–364. Winer, J.A., Lee, C.C., 2007. The distributed auditory cortex. Hear. Res. 229, 3–13. Wu, X., Ashe, J., Bushara, K.O., 2011. Role of olivocerebellar system in timing without awareness. PNAS 108, 13818–13822. Xi, M., Woody, C.D., Gruen, E., 1994. Identification of short latency auditory responsive neurons in the cat dentate nucleus. Neuroreport 5, 1567–1570. Zhang, S., Sun, X., Jen, P.H.S., 1990. HRP study of neural projections of acoustic signal to the cerebellum of echolocating bats. Acta Theriol. Sin. 10, 110–113.

2596

M. Schwartze, S.A. Kotz / Neuroscience and Biobehavioral Reviews 37 (2013) 2587–2596

Glossary Formal vs. temporal structure: Formal structure stands for a combined set of multidimensional features which allows for the identification of an event, e.g. the color, shape, position, or sound of a sensory object. Temporal structure is the result of salient changes in this set, giving rise to the concepts of succession and duration (Fraisse, 1984). Specific vs. non-specific temporal predictions: Temporal structure allows for two types of temporal predictions. Specific temporal predictions define temporal relations explicitly (event1 follows event2 after time T), whereas non-specific temporal predictions do not (event1 follows event2 ), i.e. non-specific temporal predictions are associated with temporal order.

Adaptive vs. non-adaptive oscillations: Adaptive oscillations adjust their phase and period in response to perturbations, i.e. adaptive oscillations are capable of internalizing the temporal structure of an input (McAuley, 1995). Non-adaptive oscillations may react to perturbations, e.g. via phase resetting, but continue to oscillate at a given frequency. Linear vs. non-linear stimulus representations: Linear stimulus representations encode sensory input over time, establishing an adequate mode for the processing of formal structure. Non-linear stimulus representations encode salient changes in time, potentially establishing the basis for explicit definitions of temporal relations.