Neural plasticity is induced and amplified by increased neuronal activity (Kleim and Jones, 2008; Magee and Grienberger, 2020). Early studies of plasticity in the hippocampus showed that synaptic connectivity between neurons increased following trains of high-frequency electrical stimulation and that the connectivity decreased following trains of low-frequency stimulation (Hebb, 2005). Transcranial magnetic stimulation (TMS) has been used to non-invasively replicate these principles in humans (Thickbroom, 2007). TMS causes a brief magnetic pulse at the scalp that induces an electrical current in the targeted, underlying cortical brain region (Harris-Love, 2012). When done repetitively (rTMS), at high or low frequencies, stimulation results in increases or decreases in cortical excitability which affect behavior, mimicking the effects of long-term potentiation or depression (Rossini et al., 2015).

One form of rTMS, theta-burst stimulation (TBS) performs this repetitive stimulation in 50 Hz “bursts,” decreasing the needed time for stimulation and making it a more tolerable methodology for human studies (Huang et al., 2005). TBS uses a rapid train of single TMS pulses to change the state of the underlying cortical excitability (Bestmann and Krakauer, 2015; Suppa et al., 2016). In the motor system, continuous TBS (cTBS) depresses the amplitude of motor evoked potentials (MEPs), while intermittent TBS (iTBS) increases the amplitude of MEPs (Suppa et al., 2016). This broadly suggests that iTBS is excitatory and that cTBS is inhibitory (Suppa et al., 2016). The literature also suggests differing effects of timing between each type of TBS. On average, cTBS has decreased MEP amplitude for up to 50 min and iTBS has increased MEP amplitude for an average of 60 min (Wischnewski and Schutter, 2015). However, the greatest change in MEP amplitude occurs 18–33 min post-stimulation with iTBS or 5 min after cTBS (Hinder et al., 2014; Vernet et al., 2014). These principles of the time course of TBS have not been examined in more complex cognitive networks, such as the reading network.

The reading network is inherently more complex than the network required to produce an MEP. Despite this, the effects of TBS on the reading network have thus far been assumed to be on time scales that are similar to the motor system (Arrington et al., 2022). Stimulation of the motor system results in an approximate 1:1 relationship between stimulation of cortical neurons and production of an MEP. Two motor neurons (corticospinal tract and alpha motor neuron) are all that is needed to evoke the desired behavioral response (Brasil-Neto et al., 2009). Stimulation of a more complex cognitive system, like the reading network, is theoretically less straightforward. The reading network integrates visuospatial pattern-recognition, language, attention, memory, executive function, and articulatory motor networks. Therefore, the more complex neural architecture and number of networks recruited during reading may reflect a less linear relationship between stimulation of cortical neurons and reading output than the system needed to produce an isolated MEP (Krishnamurthy et al., 2019). The reading network is not one system, but the coordination of three distinct systems supporting underlying cognitive outcomes (Pugh et al., 1996). The coordination of these networks is more effortful and the spread of stimulation likely requires a longer time course, leading the effects of TBS to potentially peak later and last longer in the reading network than the time course reported in the motor literature. In fact, a review found that effects of TBS are much more variable when affecting complex cognitive systems than motor and somatosensory function (Demeter, 2016). Understanding the time course effects of TBS on complex cognitive systems is critical for evaluating the effects of stimulation on cognitive functions such as reading. It also has implications for the use of TBS in treatment paradigms.

According to conventional models of reading, the reading system is comprised of three primary subnetworks: dorsal, ventral, and anterior (Pugh et al., 1996). The dorsal stream is classically thought to be involved in phonological decoding, or mapping sounds to visual representation, with a primary node in the supramarginal gyrus (SMG) (Pugh et al., 2001). The ventral stream facilitates reading fluency and sight word recognition, with a primary node in middle temporal gyrus (Pugh et al., 2001). The anterior stream is involved in articulatory recoding with a primary node in pars triangularis (Pugh et al., 2001). Tasks that involve phonological discrimination have been shown to engage the dorsal stream while tasks that involve sight word recognition have been shown to engage the ventral stream (Pugh et al., 2001). The SMG sits within the dorsal stream and is expected to influence phonological decoding but not sight word recognition allowing these two distinct components of reading to be both functionally and structurally differentiated. Here, we use this model to examine the facilitation and timing principles of TBS that have yet to be fully understood in the complex reading system. Specifically, this study assessed change in phonological discrimination and sight word recognition reaction time post-stimulation to SMG. We aimed to address the following research questions within the context of well-studied reading networks:

The goal of this study was to provide preliminary evidence about the facilitation type and timing effects of TBS in the reading network. In 14 young adults, we found that there were no differences in performance (i.e., percent change in reaction time) between cTBS and iTBS for our active condition or our behavioral control tasks. Thus, the facilitation and inhibition effects of TBS are not seen in behavioral reading measures as has been previously reported in the motor system. Instead, within our sample, both types of stimulation were associated with near equal facilitation or speeding of phonological processing. In contrast, the effect of TBS was significantly influenced by time post-stimulation, with the greatest effect demonstrated 60–70 min, and exceeding 80 min, post-stimulation. We did not see this effect of time in our behavioral control condition. This indicates that, in this study, a timing effect exists outside of a repeated exposure learning effect or task habituation. When considering the neural bases of these findings, the SMG is situated in the dorsal stream and therefore stimulation of this area should have more of an effect on performance during PD than in OA, as is seen in our results.

Interestingly, there was no significant difference between facilitation of change in reaction time on the PD task following cTBS and iTBS. In the motor system, cTBS is thought to mimic long-term depression, leading to an inhibitory effect, while iTBS is thought to mimic long-term potentiation that causes an excitatory effect (Chung et al., 2016). There is evidence, however, that this effect is inconsistent. Previous studies have found that MEP amplitude was consistently affected by on-line, single-pulse stimulation. However, when TBS was performed prior to single-pulse stimulation, the MEP response was variable, and for individual participants, the elicited MEP responses could not be reliably reproduced across sessions (Wischnewski and Schutter, 2015). This previous study added to a growing body of evidence that cTBS can elicit an excitatory response within the motor system (Demeter, 2016; Ozdemir et al., 2021). In our sample, this excitatory response after cTBS was found in the reading system with both forms of TBS resulting in general facilitation of change in phonological processing speed in all participants. Inhibitory effects were not seen in any participant.

Several studies have suggested that the effects of TBS observed in the motor cortex might not necessarily apply to other cortical regions (Tupak et al., 2013; Tsang et al., 2014; Painter et al., 2015). There are limited studies available to contextualize the timing results found in this study within the cognitive system. One previous study found that trains of TBS applied during the retention phase of a working memory task increased response time more than trains of TBS applied during the retrieval phase, indicating that, at least for short bursts of stimulation, some time is required for TBS to have an effect (Luber et al., 2007). This difference, however, was over the course of seconds—not on the order of minutes as seen in our results. No other studies have reported the effects of time for TBS and cognition (Demeter, 2016). Several studies, however, have called for future work examining the facilitation and timing effects of TBS in cognitive tasks (Luber et al., 2007; Demeter, 2016).

There are several potential explanations for the discrepancy between response to iTBS and cTBS seen in this study and those seen in previous literature. The first is that the difference is dependent on the relative size and complexity of the networks (Pugh et al., 1996). The reading network is comprised of three streams working in parallel across all lobes of the brain while the network that is required for production of an MEP is relatively simple, in comparison (Pugh et al., 1996). In addition, cognitive tasks, such as reading, require the coordination of multiple brain regions which increases the computational likelihood of variability of neuronal response (Malins et al., 2018). Finally, cognitive tasks have more competition from neural noise which may also increase individual variability and play a role in the discrepancies seen in this study (Hancock et al., 2017). The competition of neural noise within the reading network and the individual variability has been demonstrated in imaging studies of reading impairment (Hancock et al., 2017; Malins et al., 2018; Arrington et al., 2019).

Many studies simply infer that a behavioral outcome is due to excitatory or inhibitory stimulation without neurophysiological measures to confirm that assertion (Ozdemir et al., 2021). Some studies, however, have gathered neurophysiological data along with cTBS outcomes. These studies have used cTBS to suppress neural activation or reduce functional connectivity and have had varying degrees of success in this effort (Ott et al., 2011; Valchev et al., 2015; Steel et al., 2016; Shang et al., 2020). This effect has been demonstrated in the semantic system, which is intertwined with the reading system (Jung and Lambon Ralph, 2016, 2021). It is worth noting that many of the studies that have more consistent results demonstrating inhibition after cTBS stimulation are performed in the motor system with stimulation of the primary motor cortex (Chung et al., 2016). Functional data analyses of the current sample is ongoing to provide neuroimaging data to add to these results.

The preliminary timing effects seen in this analysis have potential implications for future study design. As discussed previously, the available guidance on timing effects suggest windows of effectiveness up to 50 (cTBS) and 60 (iTBS) minutes post-stimulation with a peak effect after 18–33 min (iTBS) or 5 min (cTBS) post-stimulation (Hinder et al., 2014; Vernet et al., 2014; Wischnewski and Schutter, 2015). Our results suggest that in the more complex reading network, this window is much longer with a potentially much later peak. SMG is situated in the dorsal stream and therefore stimulation of this area should have more of an effect on performance during PD than in OA, as is seen in our results.

Understanding the timing of TBS effects is needed to design studies that target behavioral change or measure outcomes during the peak effect of stimulation. Inaccurate understanding of timing effects could decrease the effectiveness of TBS as an adjuvant to behavioral change techniques or lead to null or inaccurate results with outcome measures collected outside of an optimal range. This effect is particularly robust when you consider that these data were collected in a sample of typically developed readers. It is unusual to get a large effect from this population as there is often a ceiling effect for change in an otherwise well-established and automatized behavior. These results should be replicated in a larger sample size before firm conclusions about future study design are drawn. It would also be useful, in larger samples, to look at some of the known influences on response to TBS (genetic variability, developmental factors, age, cortical network activity, and neurotransmitter and receptor variation) to better understand what is driving the observed effects. We were unable to test for these potential causes of variability within this sample.

Despite promising implications, this study has several limitations that impact generalizability. Most importantly, the parent study was not initially designed to answer the question of timing and facilitation effects. As such, there is a range of time post-stimulation for each time point, and some overlap is observed in timepoints A to B and timepoints C to D. We have addressed this limitation by analyzing the data as a continuous-time variable via mixed-effects linear modeling instead of binned time points. Additionally, our control condition for learning effect was limited. We employed a behavioral control rather than sham TMS in this pilot sample. Though sham would have been preferred, our lab lacked sham capabilities during the period in which this data were collected. The effects of the behavioral control, however, are robust and show that the increased processing speed seen in the PD task after stimulation of the SMG is not solely a learning or habituation effect. In fact, it appears that the learning effect hovers between 10 and 15% while behavioral change occurs between 15 and 25% change in reaction time. Participants also performed the baseline and primary outcome measures while lying in the MRI scanner. Because this setting was consistent across baseline and outcome measure settings, we do not believe that this influenced the results. The sample size is relatively small, but repeated measures across participants help to address this limitation and Cohen’s d shows a very large effect of time in the PD condition. TBS is known to have wide inter-individual variability (Hinder et al., 2014). We controlled for this by including each participant in the mixed linear effects model as a random effect. We do not believe, however, that this variability was a contributing factor to our results, as demonstrated by individual spaghetti plots as shown in Figure 1.

For future analyses, we strive to further understand the neurobiology that supports TBS-induced changes by pairing the behavioral information reported in this study with gathered functional imaging data post-stimulation. We plan to extend this study design to a population of participants with aphasia, alexia, and dyslexia. These novel and exciting results from TBS in the reading system show that mechanisms derived from the motor system should not be assumed to be applicable to the more complex cognitive networks. If our results are replicated, it will be necessary to reevaluate how we have interpreted past results using cTBS as an inhibitory technique in the language and reading system, particularly as an intervention and rehabilitation technique, and reevaluate how we plan future intervention and rehabilitation studies.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://osf.io/bewxv/.

The studies involving humans were approved by the Georgia State University/Georgia Tech Joint Center for Advanced Brain Imaging Institutional Review Board. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

RH designed analysis, gathered data, performed analysis, and wrote manuscript. CA designed protocol, gathered data, designed analysis, performed analysis, and edited manuscript. LK designed analysis and edited manuscript. AO and AD gathered data, performed analysis, designed analysis, and edited manuscript. MJ gathered data, performed analysis, and edited manuscript. EB performed analysis and edited manuscript. AW contributed to analysis and edited manuscript. RM oversaw project, designed protocol, designed analysis, and edited manuscript. All authors contributed to the article and approved the submitted version.