Introduction

The idea of a globally interconnected functional network in the neocortex has in recent years gained increased recognition. Studies have shown that tactile information is decoded in several different cortical areas and that information from visual input in awake mice can be found across the cortex[1](#ref-CR1 “Enander, J. M. D. & Jorntell, H. Somatosensory cortical neurons Decode tactile input patterns and location from both dominant and Non-dominant digits. Cell. Rep. 26, 3551–3560e3554. https://doi.org/10.1016/j.celrep.2019.02.099

(2019).“),[2](#ref-CR2 “Enander, J. M. D. et al. Ubiquitous neocortical decoding of tactile input patterns. Front. Cell. Neurosci. 13, 140. https://doi.org/10.3389/fncel.2019.00140

(2019).“),[3](https://www.nature.com/articles/s41598-025-26688-5#ref-CR3 “Findling, C. et al. Brain-wide representations of prior information in mouse decision-making. 2023.2007.2004.547684 https://doi.org/10.1101/2023.07.04.547684

%JbioRxiv (2023).“). Recent wide-field calcium imaging studies also indicate that there is a global activation of the cortex across a variety of behavioral actions[4](https://www.nature.com/articles/s41598-025-26688-5#ref-CR4 “Nietz, A. K. et al. Wide-Field calcium imaging of neuronal network dynamics in vivo. Biology (Basel) 11, 1601. https://doi.org/10.3390/biology11111601

(2022).“). Motivated by a need for more non-invasive methods to explore global activity distributions, we recently showed that a method designed to analyze global changes in cortical activity distributions can be used to detect even weak tactile inputs using the less invasive recording techniques of electroencephalogram (EEG) or electrocorticogram (ECoG)[5](https://www.nature.com/articles/s41598-025-26688-5#ref-CR5 “Mellbin, A., Rongala, U., Jörntell, H. & Bengtsson, F. ECoG activity distribution patterns detects global cortical responses following weak tactile inputs. iScience 27 https://doi.org/10.1016/j.isci.2024.109338

(2024).“).

If the cortex operates as a globally interconnected network, any damage or disruption to the cortex would be expected to alter the activity and the processing of the whole network. For example, a stroke in a remote cortical area decreases the ability of the neurons in the primary somatosensory cortex (S1) to process tactile information[6](https://www.nature.com/articles/s41598-025-26688-5#ref-CR6 “Wahlbom, A., Enander, J. M. D., Bengtsson, F. & Jörntell, H. Focal neocortical lesions impair distant neuronal information processing. J. Physiol. 597, 4357–4371. https://doi.org/10.1113/jp277717

(2019).“). It is reasonable to assume that if a small, localized lesion can impact cortical processing at such a distance, then more widespread disruptions could potentially have bigger effects on the cortical network and its processing capabilities. Amphetamine is a drug known to have a widespread effect on the brain, affecting the neurotransmitters noradrenaline, dopamine, acetylcholine and serotonin which through axonal projections from the brainstem impact essentially all parts of the cortex and thalamus[7](https://www.nature.com/articles/s41598-025-26688-5#ref-CR7 “Stratmann, P., Albu-Schäffer, A. & Jörntell, H. Scaling our world view: how monoamines can put context into brain circuitry. Front. Cell. Neurosci. 12, 506. https://doi.org/10.3389/fncel.2018.00506

(2018).“),8. Amphetamine can be used to treat ADHD and narcolepsy[9](https://www.nature.com/articles/s41598-025-26688-5#ref-CR9 “Heal, D. J., Smith, S. L., Gosden, J. & Nutt, D. J. Amphetamine, past and present–a Pharmacological and clinical perspective. J. Psychopharmacol. (Oxford, England). 27, 479–496. https://doi.org/10.1177/0269881113482532

(2013).“) and has therefore been examined for some of its effects on the cortex. D-amphetamine has been found to impact the frequency content of EEG in a way that suggested an activation of D1-receptors, with a switch to activation of D2-receptors with repeated administration[10](https://www.nature.com/articles/s41598-025-26688-5#ref-CR10 “Stahl, D., Ferger, B. & Kuschinsky, K. Sensitization to d-amphetamine after its repeated administration: evidence in EEG and behaviour. Naunyn. Schmiedebergs Arch. Pharmacol. 356, 335–340. https://doi.org/10.1007/PL00005059

(1997).“). Another study found evidence of amphetamine causing forebrain arousal by acting on noradrenergic β-receptors[11](https://www.nature.com/articles/s41598-025-26688-5#ref-CR11 “Berridge, C. W. & Morris, M. F. Amphetamine-induced activation of forebrain EEG is prevented by noradrenergic β-receptor Blockade in the halothane-anesthetized rat. Psychopharmacology 148, 307–313. https://doi.org/10.1007/s002130050055

(2000).“). Amphetamine also increases the release of acetylcholine in the cortex, by a mechanism that appears to depend on more than just an activation of D1- and D2-receptors[12](https://www.nature.com/articles/s41598-025-26688-5#ref-CR12 “Arnold, H. M., Fadel, J., Sarter, M. & Bruno, J. P. Amphetamine-stimulated cortical acetylcholine release: role of the basal forebrain. Brain Res. 894, 74–87. https://doi.org/10.1016/S0006-8993(00)03328-X

(2001).“). When amphetamine is misused as a drug, users can exhibit symptoms from a wide range of modalities, such as extreme moods, ataxia, stereotypical mouth movements, increased sympathetic stimulation and paranoia[13](https://www.nature.com/articles/s41598-025-26688-5#ref-CR13 “Connell, P. H. Clinical manifestations and treatment of amphetamine type of dependence. JAMA 196, 718–723. https://doi.org/10.1001/jama.1966.03100210088024

%J JAMA (1966).“).

Other studies focused on the effect of amphetamine on the functional connectivity using fMRI, a method that reports the spatial features of the brain activity integrated over time. A reduction in functional connectivity was found in the cortico-striato-thalamic network, as well as in the default mode networks and the salience-executive networks[14](https://www.nature.com/articles/s41598-025-26688-5#ref-CR14 “Schrantee, A. et al. Effects of dexamphetamine-induced dopamine release on resting-state network connectivity in recreational amphetamine users and healthy controls. Brain Imaging Behav. 10, 548–558. https://doi.org/10.1007/s11682-015-9419-z

(2016).“). Other studies have reported a reduced functional connectivity between nucleus accumbens and the basal ganglia, medial prefrontal cortex, temporal cortex, and the anterior cingulate cortex, with an increase in functional connectivity between nucleus accumbens and medial frontal regions as well as between putamen and the left inferior frontal gyrus[15](https://www.nature.com/articles/s41598-025-26688-5#ref-CR15 “Weafer, J., Van Hedger, K., Keedy, S. K., Nwaokolo, N. & de Wit, H. Methamphetamine acutely alters frontostriatal resting state functional connectivity in healthy young adults. Addict. Biol. 25, e12775. https://doi.org/10.1111/adb.12775

(2020).“),[16](https://www.nature.com/articles/s41598-025-26688-5#ref-CR16 “Ramaekers, J. G. et al. Methylphenidate reduces functional connectivity of nucleus accumbens in brain reward circuit. Psychopharmacology 229, 219–226. https://doi.org/10.1007/s00213-013-3105-x

(2013).“). Furthermore, D-amphetamine induces an auditory-sensorimotor-thalamic functional hyperconnectivity measured with fMRI[17](https://www.nature.com/articles/s41598-025-26688-5#ref-CR17 “Avram, M. et al. Characterizing thalamocortical (Dys)connectivity following D-Amphetamine. Biol. Psychiatry: Cogn. Neurosci. Neuroimaging. 7, 885–894. https://doi.org/10.1016/j.bpsc.2022.04.003

(2022). https://doi.org/https://doi.

MDMA Administration.“). In the cortex, amphetamine has been shown to reduce both REM and non-REM sleep times in rodents, while also reducing low frequency EEG activity[18](https://www.nature.com/articles/s41598-025-26688-5#ref-CR18 “Authier, S. et al. Effects of amphetamine, diazepam and caffeine on polysomnography (EEG, EMG, EOG)-derived variables measured using telemetry in cynomolgus monkeys. J. Pharmacol. Toxicol. Methods. 70, 86–93. https://doi.org/10.1016/j.vascn.2014.05.003

(2014).“). Amphetamine was also reported to modulate synaptic plasticity in the motor cortex, allowing better task specific recovery after brain lesions, and speeding up the learning of motor tasks[19](https://www.nature.com/articles/s41598-025-26688-5#ref-CR19 “Gilmour, G. et al. Amphetamine promotes task-dependent recovery following focal cortical ischaemic lesions in the rat. Behav. Brain. Res. 165, 98–109. https://doi.org/10.1016/j.bbr.2005.06.027

(2005). https://doi.org/https://doi.org/

“),[20](https://www.nature.com/articles/s41598-025-26688-5#ref-CR20 “Bütefisch, C. M. et al. Modulation of use-dependent plasticity by d-amphetamine. Ann. Neurol. 51, 59–68. https://doi.org/10.1002/ana.10056

(2002). https://doi.org/https://

“). However, another study found that while amphetamine increased short lasting neuronal excitability, it suppressed long lasting plasticity induced by stimulation[21](https://www.nature.com/articles/s41598-025-26688-5#ref-CR21 “Ziemann, U., Tam, A., Bütefisch, C. & Cohen*, L. G. Dual modulating effects of amphetamine on neuronal excitability and stimulation-induced plasticity in human motor cortex. Clin. Neurophysiol. 113, 1308–1315. https://doi.org/10.1016/S1388-2457(02)00171-2

(2002).“).

However, as the dynamic global collaboration between the neurons appears to be a critical aspect of brain operation, it follows that methods designed to quantify the dynamically changing distributions of global activity may potentially provide for sensitive indicators also of more subtle, but systematic, changes in brain activity. Since a high temporal resolution is a key to address activity dynamics, electrophysiological recording methods remain advantageous in this regard[22](https://www.nature.com/articles/s41598-025-26688-5#ref-CR22 “Kristensen, S. S., Kesgin, K. & Jörntell, H. High-dimensional cortical signals reveal rich bimodal and working memory-like representations among S1 neuron populations. Commun. Biology. 7, 1043. https://doi.org/10.1038/s42003-024-06743-z

(2024).“). But given that it is impossible to record the electrical activity of every single neuron in the brain, mass recordings such as electrocorticogram (ECoG) across multiple electrodes may be useful to provide insights into systematic shifts in the dynamic activity distribution across the global network[5](https://www.nature.com/articles/s41598-025-26688-5#ref-CR5 “Mellbin, A., Rongala, U., Jörntell, H. & Bengtsson, F. ECoG activity distribution patterns detects global cortical responses following weak tactile inputs. iScience 27 https://doi.org/10.1016/j.isci.2024.109338

(2024).“) and provide results that are potentially translatable to humans. Here we used the same methodology to examine how D-amphetamine affects the global ECoG activity distribution patterns[5](https://www.nature.com/articles/s41598-025-26688-5#ref-CR5 “Mellbin, A., Rongala, U., Jörntell, H. & Bengtsson, F. ECoG activity distribution patterns detects global cortical responses following weak tactile inputs. iScience 27 https://doi.org/10.1016/j.isci.2024.109338

(2024).“). Using PCA and kNN analysis, we find that D-amphetamine significantly alters the activity distribution patterns both for spontaneous activity and for activity recorded during ongoing tactile inputs. We also find that it reduces the difference between spontaneous activity patterns and activity patterns during ongoing sensory inputs, suggesting a general disorganization of the dynamic structure of the brain network activity.

Results

D-amphetamine induces changes in the brain activity distribution patterns

We used a set of eight ECoG electrodes distributed globally across the cortex as shown in Fig. 1A to record cortical activity from anesthetized rats. The ECoG activity was continually recorded during the stimulation protocol (Fig. 1B-D), which contained periods of spontaneous activity alternated with periods with ongoing tactile stimulation to the second digit of the forepaw or the hind paw (Fig. 1B). An evoked response to the stimulation was only observed in S1 and this remained after D-amphetamine administration (Supplementary Fig. S1) and Fig. 2. To investigate the impact of D-amphetamine on the brain activity distribution patterns, we compared the brain activity data both for the spontaneous activity and for the activity with ongoing stimulation (Fig. 3A). Using principal component analysis (PCA) of the ECoG activity distribution patterns across the eight ECoG electrodes, each time step of the recorded activity was mapped to the PC-space (Fig. 3B, Supplementary Fig. S2), in which the clustering of the data for the two different conditions in the comparison was quantified using the k-Nearest Neighbor (kNN) method. Across the 7 experiments, each with 7 stimulation episodes at different frequencies for two different stimulation sites, a total of 98 different comparisons were made (both for spontaneous activity and for activity during ongoing tactile stimulation). We first quantified the changes in the spontaneous activity induced by D-amphetamine (Fig. 3C). As a control, we also compared different half-segments of spontaneous activity under each condition (with and without D-amphetamine, respectively) with each other (Fig. 3A, C). The kNN accuracies of the difference in activity induced by D-amphetamine were significantly higher than those obtained for the control data (Friedmans < 0.01, post hoc sign test < 0.01 against both sets of control data). Similar results were obtained for the changes D-amphetamine induced in the activity with ongoing tactile stimulation (Fig. 3D, data from forepaw and hind paw stimulations combined) (Friedmans < 0.01, post hoc sign test < 0.01 against both sets of control data). The results are summarized in Table 1.

Fig. 1

Recording setup and stimulation protocol. A The location of craniotomies and recording electrodes in relation to different cortical areas. M1, primary motor cortex; S1, primary somatosensory cortex; A1, primary auditory cortex; V1, primary visual cortex. B The location of the tactile stimulation electrodes on the distal left forepaw and right hind paw. C Visualization of the stimulation protocol that was repeated before and after amphetamine administration. D ECoG traces before and after amphetamine administration, recorded from the right S1 area, with artifacts removed and Savitzky-Golay filter applied. The traces were recorded one minute before (top) and 15 min after the administration of D-amphetamine (bottom).

Fig. 2

Overview of the basic principles of the principal component analysis. A Example data from two ECoG channels. Red and blue parts of the trace represent two types of fictional activity. B Plot visualizing how the data from the traces in A might be positioned in a two dimensional space. The colors of each point denotes which part of the trace it is a part of. Note that the coordinates of each point is fictional and given to maximize clarity. In this plot both the x- and y- coordinates are needed to determine the color of a point. C The gray vectors added to the plot shown in B to visualize where the principal component analysis might create new vectors, capturing as much of the information as possible. D Plot showing how the data from B would be positioned in the new coordinate system created by the principal component analysis. Please note that in this new coordinate system the y-coordinate would be enough to determine the color of a point, visualizing how the PCA increases the amount of information about the data contained in a dimension, without changing the positioning of the points in relation to each other.

Fig. 3

Both spontaneous and stimulated activities were altered by D-amphetamine. A Illustration of the comparisons made in the kNN analysis. To make comparisons between matching periods of the protocol, each period of spontaneous and stimulated activity was divided into two halves (dark and light nuances of the same color). The two halves were compared against each other, both for the spontaneous and the evoked activity, to obtain control values (pre-control’ and ‘post-control’). These control values were then compared to the activity differences obtained after D-amphetamine (‘pre-post 1’ & ‘pre-post 2’, which were combined into one ‘Pre-post’ value, to ensure similar size of the data sets used in kNNs that are directly compared). B Distribution of the spontaneous activity in principal component space (the subspace defined by PCs #4–6) before and after the administration of D-amphetamine in a sample experiment. C The kNN accuracy of the comparison between the spontaneous activities before and after D-amphetamine administration (‘Pre-post’). Also shown are the kNN accuracies from the control comparisons before and after D-amphetamine administration (‘Pre-control’ and ‘Post-control’). Asterisks indicate significantly different distributions at p < 0.01 (Sign test). Each box with outliers shows all 98 kNNs for the group. D Similar to C but for activity recorded during stimulation periods. Data from forepaw and hind paw stimulations are combined. E Results from the data shuffling. Illustration shows the average kNN accuracy from one of the 98 kNNs, chosen for being the accuracy closest to the median of all 98 kNN accuracies. The red distribution curve represents the results for the shuffled data (the shuffling was repeated 100 times), and the black line shows the kNN accuracy in the test data (pre-post’). F Similar to E but for stimulated activity.

As an additional control, we used data from 5 experiments, which had the same total recording durations, but without the D-amphetamine administration. The kNN accuracy of the difference between two time segments of spontaneous activity separated by two hours were significantly worse than for the two segments of spontaneous data with and without D-amphetamine (Wilcoxon rank sum test p < 0.01), indicating that D-amphetamine significantly altered the spontaneous activity patterns also relative to this control.

Using data shuffling, we found that the actual chance level for the kNN analysis was centered around the theoretical chance level of 50%, i.e. that the kNN could not detect any difference between the normal activity and the activity under the D-amphetamine regime when the conditions (labels) of the data points were shuffled (Fig. 3E-F). Moreover, the shuffled data had a very narrow distribution and the actual data was located many standard deviations away from the shuffled data (Fig. 3E-F).

D-amphetamine reduced the difference between spontaneous and stimulated activity

As we have previously reported, the analysis of the global ECoG activity distributions can be used to detect differences between spontaneous activity and the activity when there is a weak ongoing tactile stimulation (Mellbin et al., 2024). We repeated the same analysis here but in addition compared the difference between spontaneous and stimulated activity after D-amphetamine administration (Fig. 4A). Under both conditions (with and without D-amphetamine) the kNN accuracy of the actual data was substantially different from distribution of the shuffled data (Fig. 4B, C). Across the whole data set, including when the forelimb and the hindlimb stimulations were considered separately, the results were equivalent (Fig. 4D, left) also under D-amphetamine (Fig. 4D, right). However, when compared against the shuffled data, the difference between spontaneous and stimulated activity was smaller after D-amphetamine (Fig. 4B, C). Indeed, the kNN accuracy was significantly higher before the administration of D-amphetamine compared to after, for both forepaw and hind paw stimulation (p < 0.01 for either stimulation) (Table 2). This indicates that D-amphetamine reduced the difference between spontaneous and stimulated activity. We also quantified whether the difference between the spontaneous and stimulated activity was significantly different depending on whether forepaw or hind paw stimulation was used. We found that this was not the case, neither before (p = 0.25) nor after (p = 1) D-amphetamine. No consistent relationship was observed between stimulation frequency and the pre-post D-amphetamine difference. This may reflect the smaller sample size available for each frequency tested.

Fig. 4

The difference between spontaneous and stimulated activity was reduced by D-amphetamine. A Illustration of the comparisons made in the kNN analysis. Spontaneous and stimulated activity was compared to each other, before and after administration of D-amphetamine, respectively. B kNN results compared to the shuffled data. The result (black line) represents the average accuracy for one of the 98 kNN analyses made, chosen for being the comparison with an accuracy closest to the median accuracy of all the 98 kNN accuracies. The red curve represents the kNN results for the shuffled data, the shuffling being repeated 100 times. C Similar to B but for data after D-amphetamine administration. D The kNN accuracy of the difference between spontaneous and stimulated activity for the activity recorded before and after D-amphetamine administration (left and right diagram, respectively. Comparisons were made both for activity recorded during stimulation of the forelimb and for the activity recorded during stimulation of the hindlimb, as well as the activity recorded under either stimulation (‘Both’). In “Both” the box with outliers represent the result of all 98 kNN analyses, whereas “Forepaw” and “Hind paw” data included 49 kNN analyses each. Asterisk denotes groups which were found to be significantly different by the sign test.

The dimensionality of the brain activity was not affected by D-amphetamine

Our basic analysis approach was to use PCA to map the brain activity distribution recorded at each time step to a location in the high-dimensional space defined by the PCs and then to calculate the Euclidean distances between the data points to obtain a kNN value. To explore if D-amphetamine affected the dimensionality of the data, the kNN accuracy was iteratively calculated based on subsets of the PCs. We first examined the contribution of each PC to explain the alteration induced by D-amphetamine in the spontaneous activity and in the stimulated activity (Fig. 5A, B). The accuracy for separating the control condition from the D-amphetamine condition increased for each added PC, for both spontaneous and stimulated activity (Fig. 5A). Likewise, removing any PC decreased the kNN accuracy for both types of activity (Fig. 5B). Hence, in this case the effect of D-amphetamine appeared to be broadly distributed across all dimensions of the brain activity data. This indicates that D-amphetamine did not alter the dimensionality of the brain activity distribution patterns or bias the activity to any specific such dimension.

Fig. 5

Each PC contributed to the measured difference induced by D-amphetamine in both the spontaneous and the stimulated activity. A The median kNN accuracy (from 98 kNNs) of the difference in the activity data induced by D-amphetamine as a function of the number of included PCs in the kNN analysis. Red data points show the median kNN accuracy for the data recorded during spontaneous activity (before and after D-amphetamine) and blue data points show the median kNN accuracy for data recorded during stimulated activity (before and after D-amphetamine). The median cumulative variance (from seven rats) explained as a function of the number of PCs is shown by the black data points. B Dashed lines indicate the median kNN accuracy (from 98 s kNNs) for separating the activity data with and without D-amphetamine for both spontaneous and stimulated activity. Solid lines and data points indicate the medina kNN accuracy per each PC removed from the analysis. Black data points indicate the median variance explained (from seven rats) by each PC.

We next analyzed how the difference between spontaneous and stimulated activity was altered by D-amphetamine. Figure 6A shows that each PC added to the kNN accuracy also after D-amphetamine, again suggesting that D-amphetamine did not impact any specific aspect of the brain activity distribution patterns. However, the effect of each added PC was reduced after the administration of D-amphetamine (sign test p < 0.01, all 98 kNN accuracies included). This suggests that the structure of the ECoG activity data was impacted by the drug, making it harder for the kNN analysis to separate the spontaneous and stimulated activity after D-amphetamine. Figure 6B instead analyzed if any specific PC had a particularly large importance for explaining the difference between the spontaneous and stimulated activity. But like Fig. 5B, we found that any PC removed affected the resulting accuracy, both before and after D-amphetamine administration. Moreover, in neither case did the magnitude of the reduction in accuracy correlate with how much of the variance the PC explained (Pearson’s correlation coefficient − 0.22 pre administration, −0.06 post administration).

Fig. 6

Each PC contributed to the difference between spontaneous and stimulated activity, both before and after D-amphetamine. (A) The median kNN accuracy (from 98 kNNs) of the difference between the spontaneous and the stimulated activity data as a function of the number of included PCs. Red data points show the median kNN accuracy for data recorded before D-amphetamine and blue data points show the median kNN accuracy for data recorded after. The median cumulative variance explained (from seven rats) per PC for the entire data set is shown by the black data points. (B) Dashed lines indicate the median kNN accuracy (from 98 kNNs) of the difference between the spontaneous and the stimulated activity data. Solid lines and data points indicate the median kNN accuracy per each PC removed from the kNN analysis. Black data points indicate the median variance explained (from seven rats) by each PC.

D-amphetamine caused a general decrease in power across frequency bands

To allow for a more concrete measure of the cortical activity changes, we also analyzed the power across the frequency bands of the ECoG signal before and after the administration of D-amphetamine. Comparing changes in the five frequency bands across all 7 animals, a significant, but small, decrease in power was found for all frequency bands after the D-amphetamine administration (Sign test, p < 0.01). This was true also when the data was split into spontaneous and stimulated activity, with the exception of the Beta frequency band (12–30 Hz) for the stimulated data, where no significant difference was observed (Fig. 7A, C). When examining the difference in frequency content before and after D-amphetamine in each individual channel and frequency band, there was a significant decrease in power after D-amphetamine administration in 50 out of the 80 comparisons (sign test, p < 0.05). There was no systematic difference between the areas in terms of the changes in frequency content after D-amphetamine administration (Fig. 7B, D).

Fig. 7

D-amphetamine reduced the power across all frequency bands. A The median power in each frequency band during all spontaneous activity for all 7 experiments, divided into before (black bars) and after (opaque, red bars) the administration of D-amphetamine. Green error bars show the upper and lower quartile before D-amphetamine administration and blue error bars show the upper and lower quartile after. Asterisks signify a significant difference in the power for the frequency band. B The median power in each frequency band for each channel, during all spontaneous activity for all 7 experiments. Black bars represent the amplitude before D-amphetamine administration and red, opaque, bars the amplitude after. Green error bars show the upper and lower quartile before D-amphetamine administration and blue error bars show the upper and lower quartile after. Asterisks signify a significant difference in the power for the frequency band. C Similar to A but for the stimulated activity. D Similar to B but for the stimulated activity.

Discussion

D-amphetamine significantly altered the activity distribution patterns recorded by multi-channel ECoG activity. This was true for both spontaneous activity and for brain activity in the presence of ongoing sensory input. We also found that the difference between the spontaneous and the stimulated activity was decreased by D-amphetamine. As discussed in greater detail below, our analysis indicates that D-amphetamine altered the cortical activity distribution patterns, equivalent to that D-amphetamine pushed the cortical network activity towards new locations in its state space. This is a new angle of how to interpret the effects of D-amphetamine on cortical activity dynamics, extending the more traditional descriptions that it impacts dopamine, serotonin, noradrenaline and acetylcholine transmitter systems or specific aspects of averaged functional connectivity.

D-amphetamine impacts the preferred state space locations of the network

The impact of D-amphetamine was quantified using a previously published approach[5](https://www.nature.com/articles/s41598-025-26688-5#ref-CR5 “Mellbin, A., Rongala, U., Jörntell, H. & Bengtsson, F. ECoG activity distribution patterns detects global cortical responses following weak tactile inputs. iScience 27 https://doi.org/10.1016/j.isci.2024.109338

(2024).“), where the activity distribution patterns across multiple ECoG recording electrodes are used as an estimation of changes in the global activity distribution patterns across sets of neuron populations. The changes in the neural activity distributions, which occurred for every 1 ms-time step of the recording data, can be regarded as a proxy of the changes in the global state of the network. As previously described, the ‘realm’ of possible activity distributions in the global neuron population will form a high-dimensional state space (see for example[23](https://www.nature.com/articles/s41598-025-26688-5#ref-CR23 “Luczak, A., Barthó, P. & Harris, K. D. Spontaneous events outline the realm of possible sensory responses in neocortical populations. Neuron 62, 413–425. https://doi.org/10.1016/j.neuron.2009.03.014

(2009).“). Whereas the unperturbed network activity may normally have preferred and non-preferred locations in that state space[23](https://www.nature.com/articles/s41598-025-26688-5#ref-CR23 “Luczak, A., Barthó, P. & Harris, K. D. Spontaneous events outline the realm of possible sensory responses in neocortical populations. Neuron 62, 413–425. https://doi.org/10.1016/j.neuron.2009.03.014

(2009).“), here we wanted to quantify if D-amphetamine impacted the preferred state space locations of the brain activity, similar to what has been observed for tactile and visual inputs[5](https://www.nature.com/articles/s41598-025-26688-5#ref-CR5 “Mellbin, A., Rongala, U., Jörntell, H. & Bengtsson, F. ECoG activity distribution patterns detects global cortical responses following weak tactile inputs. iScience 27 https://doi.org/10.1016/j.isci.2024.109338

(2024).“),[22](https://www.nature.com/articles/s41598-025-26688-5#ref-CR22 “Kristensen, S. S., Kesgin, K. & Jörntell, H. High-dimensional cortical signals reveal rich bimodal and working memory-like representations among S1 neuron populations. Commun. Biology. 7, 1043. https://doi.org/10.1038/s42003-024-06743-z

(2024).“) and for auditory inputs[23](https://www.nature.com/articles/s41598-025-26688-5#ref-CR23 “Luczak, A., Barthó, P. & Harris, K. D. Spontaneous events outline the realm of possible sensory responses in neocortical populations. Neuron 62, 413–425. https://doi.org/10.1016/j.neuron.2009.03.014

(2009).“). Since we have previously shown that ongoing, weak tactile stimulations can induce systematic changes in state space locations, we quantified the impact of D-amphetamine on both spontaneous activity and on activity recorded during ongoing tactile stimulation. The findings that D-amphetamine perturbed both types of activity (Fig. 3) and that the difference in the state space locations between the spontaneous and the stimulated activity was reduced (Fig. 4), both effects of which occurred across all dimensions of the activity state space (Figs. 5 and 6), indicate that D-amphetamine also resulted in a loss of information about real-world inputs in the brain circuitry activity. Altogether, this suggests that one effect of D-amphetamine was a reduced structure in the brain activity, such that the stimulated activity was less separated from the spontaneous activity in each dimension. Less structure in each dimension of the brain activity data would imply a more chaotic temporal evolution of the brain activity, potentially implying a less predictable or controlled behavior which could potentially explain some of the symptoms exhibited by people using amphetamine as a drug[13](https://www.nature.com/articles/s41598-025-26688-5#ref-CR13 “Connell, P. H. Clinical manifestations and treatment of amphetamine type of dependence. JAMA 196, 718–723. https://doi.org/10.1001/jama.1966.03100210088024

%J JAMA (1966).“). Changes in the functional connectivity between the thalamus and the cortex observed after administration of D-amphetamine in previous studies could also be part of the explanation for the reduction in the ability to separate spontaneous and stimulated activity after D-amphetamine administration[17](https://www.nature.com/articles/s41598-025-26688-5#ref-CR17 “Avram, M. et al. Characterizing thalamocortical (Dys)connectivity following D-Amphetamine. Biol. Psychiatry: Cogn. Neurosci. Neuroimaging. 7, 885–894. https://doi.org/10.1016/j.bpsc.2022.04.003

(2022). https://doi.org/https://doi.

MDMA Administration.“).

It should be noted that the observation that each dimension of the data, i.e. each feature in the ECoG activity distribution patterns, carried similar weight in explaining the differences induced by D-amphetamine (Figs. 5 and 6) argues against any simple interpretation such as an amplification of the thalamocortical loop activity. If D-amphetamine’s effect would simply have been an impact on the general excitability of the thalamocortical loop, then it would have been expected to result in more prominent effects in one or a few dimensions of the data. Rather, this observation supports that the alterations induced by D-amphetamine impacts the physiological network structure in multiple ways, such that new pathways of activity spread across the network opens up[24](https://www.nature.com/articles/s41598-025-26688-5#ref-CR24 “Szeier, S. & Jörntell, H. Neuronal networks quantified as vector fields. bioRxiv, 2024.2006.2029.601314 https://doi.org/10.1101/2024.06.29.601314

(2024).“), and that this is what caused the network to find new preferred locations in its activity state space. Given that amphetamine affects neurotransmitters that are present throughout the cortex and thalamus, it would be surprising if the effect of D-amphetamine was not widely distributed in the brain[7](https://www.nature.com/articles/s41598-025-26688-5#ref-CR7 “Stratmann, P., Albu-Schäffer, A. & Jörntell, H. Scaling our world view: how monoamines can put context into brain circuitry. Front. Cell. Neurosci. 12, 506. https://doi.org/10.3389/fncel.2018.00506

(2018).“),8.

However, since our analysis method is very sensitive to fine nuanced changes in brain activity, we used comparatively low doses of D-amphetamine. It is conceivable that at higher doses there may exist dose-dependent reductions in the dimensionality of the brain activity. Other studies of the effects of D-amphetamine on EEG recordings detected an effect only after repeated administrations of higher doses (0.6 mg/kg I.P. as opposed to 0.25 mg/kg I.V. used here) and these effects may possibly involve additional factors compared to those recorded in the present study[10](https://www.nature.com/articles/s41598-025-26688-5#ref-CR10 “Stahl, D., Ferger, B. & Kuschinsky, K. Sensitization to d-amphetamine after its repeated administration: evidence in EEG and behaviour. Naunyn. Schmiedebergs Arch. Pharmacol. 356, 335–340. https://doi.org/10.1007/PL00005059

(1997).“).

On the probability of obtaining clustered data with the kNN method

Because the pattern of the recorded multichannel ECoG signal will be constantly changing due to ongoing internal processes in the brain networks, even comparing two separate segments of spontaneous activity under the same condition will have a certain probability to be reported to be different by the PCA-kNN approach we used. This can be seen in Fig. 3C, D, where different half-segments of recorded ECoG data recorded under the same condition (i.e., with or without D-amphetamine) were found to be different by the kNN analysis. This effect would be expected to be time-dependent, i.e., if we had recorded spontaneous activity for 48 h and just compared the first 24 h to the second, the kNN accuracy would have been more likely to be closer to chance (50%). Conversely, the accuracy of such a comparison would have increased if we instead compared 10-minute segments and increased even more if we compared just one-minute segments to each other. This is not surprising but merely indicates how sensitive the method is and how complex the brain activity patterns normally are[25](#ref-CR25 “Stringer, C. et al. Spontaneous behaviors drive multidimensional, brainwide activity. Sci. (New York N Y). 364, eaav7893. https://doi.org/10.1126/science.aav7893

(2019).“),[26](#ref-CR26 “Norrlid, J., Enander, J. M. D., Mogensen, H. & Jörntell, H. Multi-structure cortical States deduced from intracellular representations of fixed tactile input patterns. 15 https://doi.org/10.3389/fncel.2021.677568

(2021).“),[27](https://www.nature.com/articles/s41598-025-26688-5#ref-CR27 “Etemadi, L., Enander, J. M. D. & Jörntell, H. Remote cortical perturbation dynamically changes the network solutions to given tactile inputs in neocortical neurons. iScience 25, 103557. https://doi.org/10.1016/j.isci.2021.103557

(2022).“).

The effect of D-amphetamine on the distribution of ECoG frequency power

A potential explanation for the differentiation of activity before and after D-amphetamine could be the altered frequency content of our ECoG signal. The ECoG primarily signals local field potentials, which in turn primarily reflects synaptic activity. Field potentials thereby indicate changes in the activity in the underlying neuron population[28](https://www.nature.com/articles/s41598-025-26688-5#ref-CR28 “Gallego-Carracedo, C., Perich, M. G., Chowdhury, R. H., Miller, L. E. & Gallego, J. Á. Local field potentials reflect cortical population dynamics in a region-specific and frequency-dependent manner. eLife 11, e73155 https://doi.org/10.7554/eLife.73155

(2022).“). The broad-band loss of power in all frequencies would suggest that there were fewer or less synchronized activity changes in the neuron population after amphetamine. However, the difference was relatively small, possibly reflecting temporal shifts in frequency content. This can be seen as an increase in higher frequencies at certain times after D-amphetamine administration in some traces (Fig. 1.D). The only exception was the Beta band for the activity recorded during ongoing tactile stimulation, where no significant difference was found. While frequency analysis can provide a broad overview of major cortical activity changes between the two conditions, it does not replace PCA and kNN, which can detect differences beyond frequency content. And at the same time, PCA and kNN can not replace frequency analysis and other traditional methods of analysis which give concrete information on what changes in the cortical activity.

Previous studies of the effect of amphetamine across EEG frequency bands have reported a reduction in slow wave activity after administration of amphetamine using a lower dose than here (0.15 mg/kg, I.V.)[11](https://www.nature.com/articles/s41598-025-26688-5#ref-CR11 “Berridge, C. W. & Morris, M. F. Amphetamine-induced activation of forebrain EEG is prevented by noradrenergic β-receptor Blockade in the halothane-anesthetized rat. Psychopharmacology 148, 307–313. https://doi.org/10.1007/s002130050055

(2000).“). A similar reduction in slow wave activity, together with an increase in alpha band frequency was seen after using a higher dose of D-amphetamine than used in this study (0.6 mg/kg, I.P.)[10](https://www.nature.com/articles/s41598-025-26688-5#ref-CR10 “Stahl, D., Ferger, B. & Kuschinsky, K. Sensitization to d-amphetamine after its repeated administration: evidence in EEG and behaviour. Naunyn. Schmiedebergs Arch. Pharmacol. 356, 335–340. https://doi.org/10.1007/PL00005059

(1997).“). Similarly, another study found that a low dose of D-amphetamine (0.4 mg/kg I.P.) causes a desynchronization with general lowering of power in all frequency bands, while a high dose (4 mg/kg I.P.) increased power at the 7 to 9.5 Hz range (alpha-1 band), likely due to different receptors being affected at different dosages[29](https://www.nature.com/articles/s41598-025-26688-5#ref-CR29 “Ferger, B., Kropf, W. & Kuschinsky, K. Studies on electroencephalogram (EEG) in rats suggest that moderate doses of cocaine ord-amphetamine activate D1 rather than D2 receptors. Psychopharmacology 114, 297–308. https://doi.org/10.1007/BF02244852

(1994).“). This decrease in power in all frequency bands is in agreement with our findings. The lack of significant effect of D-amphetamine on the stimulated activity in the Beta band (Fig. 7) could potentially be a sign of our stimulation having an amphetamine-dependent effect on those specific frequencies, though the stimulation frequencies were disjunct from this band.

Previous studies have reported a decrease in slow-wave activity during wakefulness following oral D-amphetamine administration, accompanied by a reduction in both non-REM and REM sleep duration[18](https://www.nature.com/articles/s41598-025-26688-5#ref-CR18 “Authier, S. et al. Effects of amphetamine, diazepam and caffeine on polysomnography (EEG, EMG, EOG)-derived variables measured using telemetry in cynomolgus monkeys. J. Pharmacol. Toxicol. Methods. 70, 86–93. https://doi.org/10.1016/j.vascn.2014.05.003

(2014).“). I.V. administration of D-amphetamine at doses of 0.3–3.3 mg/kg has been shown in rats to induce a dose-dependent arousal from sevoflurane anesthesia, reduce the time to emergence from propofol anesthesia, and accelerate the recovery of consciousness and respiratory drive following fentanyl administration. Its effects during ketamine anesthesia appear to differ, however. One study found that while I.V. D-amphetamine at a dose of 1 mg/kg did reduce the time to emergence after dexmedetomidine administration, a higher dose of 3 mg/kg did not significantly reduce the time to emergence from ketamine anesthesia[30](#ref-CR30 “Kenny, J. D., Taylor, N. E., Brown, E. N. & Solt, K. Dextroamphetamine (but not Atomoxetine) induces reanimation from general anesthesia: implications for the roles of dopamine and norepinephrine in active emergence. PloS One. 10, e0131914. https://doi.org/10.1371/journal.pone.0131914

(2015).“),[31](#ref-CR31 “Kato, R. et al. D-Amphetamine rapidly reverses Dexmedetomidine-Induced unconsciousness in rats. Front. Pharmacol. 12, 668285. https://doi.org/10.3389/fphar.2021.668285

(2021).“),[32](https://www.nature.com/articles/s41598-025-26688-5#ref-CR32 “Moody, O. A. et al. D-Amphetamine accelerates recovery of consciousness and respiratory drive after High-Dose Fentanyl in rats. Front. Pharmacol. 11 https://doi.org/10.3389/fphar.2020.585356

(2020).“). This means that while D-amphetamine have a general arousal effect, this effect seems to be reduced or removed in some way during ketamine anesthesia, potentially due to the fact that ketamine can increase the release of dopamine in the prefrontal cortex[33](https://www.nature.com/articles/s41598-025-26688-5#ref-CR33 “Lorrain, D. S., Baccei, C. S., Bristow, L. J., Anderson, J. J. & Varney, M. A. Effects of ketamine and < em > n-methyl-d-aspartate on glutamate and dopamine release in the rat prefrontal cortex: modulation by a group II selective metabotropic glutamate receptor agonist LY379268. Neuroscience 117, 697–706 https://doi.org/10.1016/S0306-4522(02)00652-8

(2003).“),[34](https://www.nature.com/articles/s41598-025-26688-5#ref-CR34 “Moghaddam, B., Adams, B., Verma, A. & Daly, D. Activation of glutamatergic neurotransmission by ketamine: a novel step in the pathway from NMDA receptor Blockade to dopaminergic and cognitive disruptions associated with the prefrontal cortex. J. Neuroscience: Official J. Soc. Neurosci. 17, 2921–2927. https://doi.org/10.1523/jneurosci.17-08-02921.1997

(1997).“) and thereby interfere with some of the effects that amphetamine is expected to have on this transmitter system.

Limitations of the field potential approach to analyze brain activity dynamics

Could coarse mass electrode recordings from some perspectives offer advantages compared to multi-neuron recordings? Neuron recordings naturally have a higher resolution, but the ECoG recordings are naturally more globally distributed and more easily conducted. It is not theoretically possible to record from every single neuron in the brain, in fact the tissue-destructive effects of inserting electrodes into the brain tissue limits the single neuron approach to record from extreme subsets of the entire neuron population. If cortical operation is the effect of globally integrated signals, the more distributed signal pickup may provide at least some advantages compared to the details of local signals. With the major disadvantage of course being that the recorded transitions in neuron population activities become extreme under-representations of which specific neurons alter activity in which direction. However, recent work using multichannel ECoG recordings to drive a diversified speech synthesizer indicate that this type of recording can indeed contain a lot of information about the underlying brain processing[35](https://www.nature.com/articles/s41598-025-26688-5#ref-CR35 “Littlejohn, K. T. et al. A streaming brain-to-voice neuroprosthesis to restore naturalistic communication. Nat. Neurosci. 28, 902–912. https://doi.org/10.1038/s41593-025-01905-6

(2025).“). Moreover, ECoG is closely related to non-invasive EEG, and the results we obtained here are likely to b