**1. Introduction**

Owing to its dynamicity, the brain responds to an intense, novel stimulus with enhanced, transient neural activity. This rapid response, referred to as a startle, is considered to play a critical function in promoting survival [1]. However, exposure to a higher number of identical sensory stimuli yields attenuation of neural activity in the responding network, a phenomenon known as repetition suppression (RS) [2]. RS has been well-characterized across several brain regions, employing various stimulus categories and modalities [3–6]. In the motor system, RS has been demonstrated as a decrement in the amplitude of motor-evoked potentials (MEPs) when transcranial magnetic stimulation (TMS) is applied to an optimal motor cortex location [7,8]. Although it has been suggested that the attenuation observed in RS may serve to provide an energy-efficient neuronal information processing [9], the exact mechanisms underlying RS have remained elusive. RS was initially portrayed merely as an expression of bottom-up mechanisms [2,3,10,11]. However, more recent theories have emphasized the role of top–down mechanisms within a predictive coding scheme, relying on iterative

comparison between prior expectations and sensory inputs [12]. Interestingly, RS of MEPs have been demonstrated to be closely associated with neuroplasticity [13].

Neuroplasticity is considered one of the key mechanisms that grants living organisms the ability to adapt and respond flexibly in the face of changing environmental demands [14]. Depending on the speed of these changes, neuroplasticity can take different forms and occur at different timescales. Neuroplasticity is considered the keystone of learning, memory, and recovery from (mild) brain injuries [15–17]. Aberrant neuroplasticity has been put forth as the pathophysiological basis of several neuropsychiatric disorders, such as schizophrenia, depression, and chronic pain [18–20]. Long-term potentiation (LTP) consists of persistent synaptic activity, which is often considered as the cellular basis in the mediation of these functions [21].

Currently, TMS provides the opportunity to study neuroplasticity at the system level, ranging from synaptic plasticity to network-level plasticity [22]. The shifts towards either elevated excitation or diminished inhibition have been proposed as potential underlying mechanisms of neuroplasticity, with the short-term plasticity most likely mediated by the reduction of GABAergic inputs onto excitatory synapses [23].

A well-established and widely used TMS paradigm to induce short-term, topographically specific plasticity in the motor cortex is paired associative stimulation (PAS), in which electrical peripheral nerve stimulation is paired with cortical stimulation [24,25]. If the peripheral input precedes the cortical stimulation, PAS can lead to elevated cortical excitability that manifests itself via an increase in the MEP amplitude (LTP-like plasticity) [26,27]. By contrast, if the order of the arrival of inputs is reversed, depression of cortical excitability is likely to occur (long-term depression (LTD)-like plasticity) [24]. Due to its dependency on timing, PAS has been suggested to induce spike-timing dependent plasticity [28].

In the present paper, to investigate neuroplastic effects induced with PAS, RS is hypothesized to represent the interplay of two states: (1) one reflecting the efficient processing of a novel input, "dynamic RS", indexed by the initial decrement from the first amplitude to the second one, and (2) one reflecting the overall cortical excitability free from the ongoing dynamics, "stable RS". Stable RS, described here as the suppressed amplitude level of the second to the fourth MEPs within the RS trials, might potentially display the capacity of the brain to maintain the processed input as an initial "memory trace". We investigated the dynamic and stable RS prior to and following a common PAS-LTP protocol [24]. We hypothesized that the brain would demonstrate a trend towards a state with low variation in MEP amplitude, which we consider the target level of neuronal network excitability as it is independent from reactive dynamics within the network. As an implication, for long-term neuroplastic effects, the modulation of this stable level could potentially be targeted by neuromodulation, and to create optimal conditions for adaptive neural changes.
