Abstract

The mediodorsal nucleus of the thalamus (MD), with its extensive connections to the lateral pFC, has been implicated in human working memory and executive functions. However, this understanding is based solely on indirect evidence from human lesion and imaging studies and animal studies. Direct, causal evidence from humans is missing. To obtain direct evidence for MD's role in humans, we studied patients treated with deep brain stimulation (DBS) for refractory epilepsy. This treatment is thought to prevent the generalization of a seizure by disrupting the functioning of the patient's anterior nuclei of the thalamus (ANT) with high-frequency electric stimulation. This structure is located superior and anterior to MD, and when the DBS lead is implanted in ANT, tip contacts of the lead typically penetrate through ANT into the adjoining MD. To study the role of MD in human executive functions and working memory, we periodically disrupted and recovered MD's function with high-frequency electric stimulation using DBS contacts reaching MD while participants performed a cognitive task engaging several aspects of executive functions. We hypothesized that the efficacy of executive functions, specifically working memory, is impaired when the functioning of MD is perturbed by high-frequency stimulation. Eight participants treated with ANT-DBS for refractory epilepsy performed a computer-based test of executive functions while DBS was repeatedly switched ON and OFF at MD and at the control location (ANT). In comparison to stimulation of the control location, when MD was stimulated, participants committed 2.26 times more errors in general (total errors; OR = 2.26, 95% CI [1.69, 3.01]) and 2.86 times more working memory-related errors specifically (incorrect button presses; OR = 2.88, CI [1.95, 4.24]). Similarly, participants committed 1.81 more errors in general (OR = 1.81, CI [1.45, 2.24]) and 2.08 times more working memory-related errors (OR = 2.08, CI [1.57, 2.75]) in comparison to no stimulation condition. “Total errors” is a composite score consisting of basic error types and was mostly driven by working memory-related errors. The facts that MD and a control location, ANT, are only few millimeters away from each other and that their stimulation produces very different results highlight the location-specific effect of DBS rather than regionally unspecific general effect. In conclusion, disrupting and recovering MD's function with high-frequency electric stimulation modulated participants' online working memory performance providing causal, in vivo evidence from humans for the role of MD in human working memory.

INTRODUCTION

A key component of human executive functions is working memory, responsible for short-term storage and online manipulation of information. It has traditionally been associated with pFC and cortical networks (D'Esposito & Postle, 2014; Gazzaley, Rissman, & D'Esposito, 2004; Goldman-Rakic, 1995), but recent studies point toward the involvement of corticothalamic networks and subcortical structures, such as mediodorsal nucleus of the thalamus (MD). Human lesion (Van der Werf et al., 2003; Zoppelt, Koch, Schwarz, & Daum, 2003) and imaging studies (Piras, Caltagirone, & Spalletta, 2010; Monchi, Petrides, Petre, Worsley, & Dagher, 2001), as well as animal studies (Browning, Chakraborty, & Mitchell, 2015; Parnaudeau et al., 2015; Funahashi, 2013; Watanabe & Funahashi, 2012; Mitchell, Browning, & Baxter, 2007; Aggleton & Mishkin, 1983; Isseroff, Rosvold, Galkin, & Goldman-Rakic, 1982), suggest that MD plays a role in human executive functions and working memory. However, previous approaches have some limitations, which impact the conclusions about MD's role in human working memory.

Studies on patients with vascular lesion are limited in their neuroanatomical accuracy due to vascular lesions being rarely restricted to a specific structure, for example, MD, but extending to neighboring nuclei or white matter structures in an unspecific way (Rorden & Karnath, 2004). Such equivocal neuroanatomical accuracy obscures the conclusions drawn about MD's role in working memory. Furthermore, the use of a control group in lesion studies introduces a number of confounding factors, such as individual cognitive capabilities, which cannot be reliably controlled for. Also, comparison of a study group with a control group calls for a large sample size typically not available in studies depending on vascular lesion to a specific focal location. Human imaging studies, on the other hand, have their spatial-, temporal-, and contrast-related limitations. Both techniques typically provide associative rather than causal evidence for the role of a certain brain region in a specific cognitive process. Animal studies overcome some of the aforementioned limitations and allow for anatomically accurate and causal evidence, but one must be cautious when extrapolating results from animal cognition studies to human cognition (Preuss, 1995). In summary, the current understanding of MD's role in humans is based on evidence extrapolated from animal studies and indirect human lesion and imaging studies. Causal, in vivo evidence from humans is missing.

Deep brain stimulation (DBS) is an emerging treatment for psychiatric and neurological disorders, for example, refractory epilepsy (Kocabicak, Temel, Höllig, Falkenburger, & Tan, 2015; Salanova et al., 2015; Fisher et al., 2010). High-frequency DBS is thought to disrupt the functioning of the target nucleus by creating a reversible lesion (Benazzouz et al., 2000). Thus, DBS can be seen as a “lesion on demand” procedure, with the lesion emerging and disappearing at the target site as a function of electric stimulation. An alternative view of how the stimulation may work is that high-frequency, suprathreshold stimulation suppresses the firing of the soma while simultaneously inducing an efferent output in the axons at the stimulation frequency (McIntyre, Grill, Sherman, & Thakor, 2004). Common to both views is that normal neuronal activity around the active contacts is disrupted by stimulation. The DBS target for treatment of refractory epilepsy, anterior nucleus of the thalamus (ANT), is located superior and anterior to MD. When ANT-DBS lead is implanted transventricularly to the thalamus, one or two tip contacts of the lead penetrate through ANT into the superior-anterior section of MD.

The current study extends previous research by investigating the role of MD in executive functions and working memory in humans using DBS contacts reaching MD. We periodically disrupted and recovered MD's functioning using high-frequency electric stimulation while participants performed an ongoing, online cognitive test engaging several aspects of executive functions, that is, the Executive RT Test. Previous studies using this method to study the role of deep brain structures in human emotion and cognition found ANT to play a role in emotion–attention interaction (Sun et al., 2015; Hartikainen et al., 2014). The Executive RT Test has also been used to uncover alterations in executive function in various patient groups, such as impaired executive function in patients suffering from mild head injury with persistent symptoms (Hartikainen et al., 2010), improved cognitive flexibility after successful cardiac surgery (Liimatainen et al., 2016), improved working memory due to vagus nerve stimulation (Sun et al., 2017), and alternations in emotion–attention interaction in patients with mild head injury (Mäki-Marttunen et al., 2014) and orbitofrontal lesion (Mäki-Marttunen et al., 2016). We hypothesized that DBS at MD impairs executive functions, specifically working memory, in contrast to DBS at control location (ANT) or DBS off.

METHODS

Participants

Eight individuals (four men, four women; mean age = 32 years, SD = 11 years) treated for refractory epilepsy with ANT-DBS in Tampere University Hospital, Finland, participated in the study. The study was approved by the Tampere University Hospital regional review board, and participants gave their consent according to the Declaration of Helsinki governing the use of human subjects. This group is a good performer subset of the participants used in our previous study (Hartikainen et al., 2014), complemented with one additional individual tested recently. Table 1 summarizes participant medical history, etiology, medication, and implant location.

Table 1. 

Participant Information

Participant No.Type of EpilepsyEtiologyImaging FindingsMedicationMD Contact LocationANT Contact Location
LeftRightLeftRight
Occipital Cortical dysplasia MRI+ Carbamazepine, Clobazam mtt/iml MD ANT/eml ANT 
Multifocal Cortical dysplasia MRI+ Oxcarbamazepine, Clobazam, Zonisamide mtt/iml iml VA/eml VA/eml 
Frontal Cortical dysplasia MRI+ Phenytoin, Clonazepam MD MD iml MD 
Multifocal Encephalitis MRI− Clobazam, Zonisamide, Lacosamide VA/eml MD eml ANT/iml 
Multifocal Encephalitis MRI+ Oxcarbamazepine, Topiramate, Clobazam MD MD MD ANT/iml 
Multifocal Meningoencephalitis MRI− Sodium valproate MD MD MD ANT/iml 
12 Occipital Cortical dysplasia MRI+ Carbamazepine, Clobazam MD MD ANT VA 
13 Multifocal Unknown MRI− Oxcarbamazepine iml MD ANT ANT 
Participant No.Type of EpilepsyEtiologyImaging FindingsMedicationMD Contact LocationANT Contact Location
LeftRightLeftRight
Occipital Cortical dysplasia MRI+ Carbamazepine, Clobazam mtt/iml MD ANT/eml ANT 
Multifocal Cortical dysplasia MRI+ Oxcarbamazepine, Clobazam, Zonisamide mtt/iml iml VA/eml VA/eml 
Frontal Cortical dysplasia MRI+ Phenytoin, Clonazepam MD MD iml MD 
Multifocal Encephalitis MRI− Clobazam, Zonisamide, Lacosamide VA/eml MD eml ANT/iml 
Multifocal Encephalitis MRI+ Oxcarbamazepine, Topiramate, Clobazam MD MD MD ANT/iml 
Multifocal Meningoencephalitis MRI− Sodium valproate MD MD MD ANT/iml 
12 Occipital Cortical dysplasia MRI+ Carbamazepine, Clobazam MD MD ANT VA 
13 Multifocal Unknown MRI− Oxcarbamazepine iml MD ANT ANT 

MRI+ = pathological MRI findings; MRI− = no pathological MRI findings; ANT = anterior thalamic nuclei; MD = mediodorsal nucleus; VA = ventral anterior nucleus; mtt = mammillothalamic tract; iml = internal medullary lamina; eml = external medullary lamina.

Executive RT Test

The Executive RT Test is a modified go/no-go test with emotional distractors and task switching. The test requires participants to store the orientation of a triangle in working memory and press, or withhold from pressing, a corresponding response button according to go/no-go signal. The go/no-go signal is a traffic light where go/no-go is either green or red light and changes after every block throughout the experiment. The test is designed to engage multiple aspects of executive functions concurrently to mimic real-life requirements for human executive functions. Traditionally in neuroscience research experiments are designed to isolate one specific brain function at a time. However, because we wanted to study the effect of electric stimulation on different executive functions (working memory, inhibition and shifting) with one test, the use of this kind of “integrated test” was deemed applicable.

The experiment consisted of 32 blocks composed of 64 trials per block, totaling 2048 trials per participant. Each trial consists of four parts: a triangle at the center of the screen pointing up or down, a delay, a traffic light indicating a go or no-go situation and a response period (Figure 1). Each trial lasted approximately 2 sec, jittered 150 msec before the onset of the triangle.

Figure 1. 

Schematic presentation of the Executive RT Test (Hartikainen et al., 2010). An upright or downward triangle was presented for 150 msec, followed by a fixation cross for 150 msec and finally by a go/no-go signal (traffic light) for 150 msec indicating whether participants should respond by pressing one of the two buttons according to orientation of the triangle. After the traffic light, a fixation cross was shown until the 1700 msec response window ended. The middle circle of the traffic light contained either a neutral or an emotional distractor.

Figure 1. 

Schematic presentation of the Executive RT Test (Hartikainen et al., 2010). An upright or downward triangle was presented for 150 msec, followed by a fixation cross for 150 msec and finally by a go/no-go signal (traffic light) for 150 msec indicating whether participants should respond by pressing one of the two buttons according to orientation of the triangle. After the traffic light, a fixation cross was shown until the 1700 msec response window ended. The middle circle of the traffic light contained either a neutral or an emotional distractor.

In the case of a go signal, when the triangle was pointing down participants pressed the “Down” button with their index finder, and when the triangle was pointing up participants pressed the “Up” button with their middle finger. The “Up” button was located slightly higher than the “Down” button on the response pad, making responding easy and intuitive. The triangle orientation—button mapping remained the same throughout the study. Within a block, the orientation of the triangles was randomized, and half of the trials were go trials whereas the other half were no-go trials. The go/no-go signal changed at the beginning of each block, that is, in half of the blocks the go signal was green light and the no-go signal was red light and in the remaining half the go signal was red light and the no-go signal was green light. The block's go signal (response rule) was communicated to the participant at the beginning of each block via instruction text presented on the computer screen. On half of the trials, a neutral distractor was presented in the middle circle of the traffic light, whereas in the remaining half of the trials, an emotional distractor was presented in the middle circle of the traffic light. The distractor stimuli were composed of identical black line and circle elements to control for low-level visual differences and reconfigured to depict a spider (emotional distractor) or a flower (neutral control distractor; Vuilleumier & Schwartz, 2001; Figure 1).

The outcome measures of the Executive RT Test are RTs and the number of basic errors (incorrect response, missed response, or commission error) and composite errors (shift errors and total errors) under different conditions. Each basic error type is thought to reflect impairment in different cognitive process. The number of incorrect responses in go trials is thought to indicate working memory performance, missing responses in go trials lapses in sustained attention where the participant “forgets” to respond (Smallwood & Schooler, 2006; Smallwood et al., 2004; Manly, Robertson, Galloway, & Hawkins, 1999), and commission errors in no-go trials failures in top–down inhibitory control. For assessing rule switching, the go signal is changed in every block, alternating between a green traffic light and a red traffic light. If a participant has a rule-switching problem, they would continue responding with the previous response rule (go signal) instead of the new rule. This would be reflected as a simultaneous increase in the number of missing responses and commission errors at the beginning of a new block, so the sum of missing responses and commission errors would indicate a rule-switching problem (shift errors). The sum of all the basic error types reflects an overall efficiency of executive functions (total errors). The performance difference caused by the distractor, that is, neutral or emotional figure, is thought to reflect emotional reactivity.

Participants practiced the experiment before the actual test until they felt confident performing the task. During the practice session, participants were given verbal feedback about their performance, but no feedback was provided during the experiment. Typically, participants needed one practice block to feel confident performing the task. The experiment was presented using Presentation software (Neurobehavioral Systems, Inc., Berkeley, CA).

DBS Lead Implantation and Imaging

Hospital neurosurgeons implanted patients with deep brain stimulator (Medtronic 3389, Medtronic, Inc., Minneapolis, MN) for the treatment of refractory epilepsy. The DBS lead was implanted transventricularly into the anterior thalamus (ANT) using Leksell stereotactic frame (Elekta AB, Stockholm, Sweden). The ANT-DBS stereotactic target was 5–6 mm lateral, 0–2 mm anterior, and 12 mm superior, respective to midcommisural point and adjusted individually using 3T MRI short tau inversion images (Siemens Healthcare GmbH, Erlangen, Germany) visualizing mamillothalamic tract and ANT. Postoperatively, the locations of the stimulator contacts were assessed relative to midcommisural point using preoperative MRI-postoperative CT fusion images. The DBS lead has four 1.5-mm contacts spaced 0.5 mm apart, and due to deep insertion, tip contacts in the lead penetrated through ANT reaching MD. These contacts were used to disrupt MD function. Contacts at ANT were used as a control location (Figures 2A and 3).

Figure 2. 

(A) CT + MRI fusion image with thalamic structures illustrating DBS implantation. Tip contacts at MD were used to disrupt MD function. Contacts at ANT were used as a control location for stimulation. (B and C) DTI images illustrating white matter tracts from MD (stimulation location) to pFC in one individual. Yellow tracts = from right MD; Red tracts = from left MD.

Figure 2. 

(A) CT + MRI fusion image with thalamic structures illustrating DBS implantation. Tip contacts at MD were used to disrupt MD function. Contacts at ANT were used as a control location for stimulation. (B and C) DTI images illustrating white matter tracts from MD (stimulation location) to pFC in one individual. Yellow tracts = from right MD; Red tracts = from left MD.

Figure 3. 

(A) Schematic presentation of the DBS lead implantation and stimulation sites. Inferior border of ANT was used as a borderline, and contacts above the borderline were considered ANT contacts and contacts below the borderline were MD contacts. (B) Stimulation protocol. Stimulator was ON for two blocks of testing, then it was turned off for another two blocks, etc. When stimulator was adjusted, there was a 2- to 3-min break between the blocks. This was repeated 8 times (16 blocks) for Location 1, for example, MD, after which the active electrodes were set to Location 2 (ANT) and another 16 blocks of testing was done. Starting location was counterbalanced between participants so that half of the participants started with MD location and another half with ANT location.

Figure 3. 

(A) Schematic presentation of the DBS lead implantation and stimulation sites. Inferior border of ANT was used as a borderline, and contacts above the borderline were considered ANT contacts and contacts below the borderline were MD contacts. (B) Stimulation protocol. Stimulator was ON for two blocks of testing, then it was turned off for another two blocks, etc. When stimulator was adjusted, there was a 2- to 3-min break between the blocks. This was repeated 8 times (16 blocks) for Location 1, for example, MD, after which the active electrodes were set to Location 2 (ANT) and another 16 blocks of testing was done. Starting location was counterbalanced between participants so that half of the participants started with MD location and another half with ANT location.

Neural connectivity from MD to pFC was demonstrated using diffusion tensor imaging (DTI). DTI data were obtained using Siemens Trio 3T scanner (Siemens Healthcare GmbH) using the following parameters: repetition time = 7600 msec, echo time = 91 msec, averaging 1, diffusion directions 30, b factors 0 and 1000 sec/mm2, field of view = 256 mm, matrix 128 × 128, slice thickness 2 mm, voxel size 2 × 2 × 2 mm. Fiber tracking was performed using the Stealthwiz software (Medtronic, Inc.) with the following parameters: fractional anisotrophy (FA) start value 0.2, apparent diffusion coefficient (ADC) stop value 0.1, seed density 5, maximal directional change 45–52°. Seed point for MD was delineated using short tau inversion recovery images used for fiber calculation (Figure 2B and C).

DBS Protocol

Therapeutic DBS for refractory epilepsy typically applies monopolar, constant voltage stimulation where one of the four contacts in a lead is used as a cathode and DBS stimulator implanted subcutaneously in the chest as an anode. However, in this study, we used a bipolar,constant current stimulation. In bipolar stimulation mode, one of the four contacts is used as an anode and another, typically a neighboring contact, as a cathode. The advantage of bipolar stimulation is that it provides a small and well-controlled volume of tissue affected, translating into a very focal neuromodulatory effect covering the cylindrical volume of a few millimeters away from the lead between active contacts, which are 0.5 mm apart. During the experiment, the DBS stimulator was programmed to 5-mA constant current and 140-Hz stimulation frequency.

Two stimulation sites, MD and ANT, were defined based on contact location at the superior-inferior axis where the inferior border of ANT served as a borderline (Figure 3). The MD site included contacts below the borderline in MD or in its immediate horizontal neighboring structures. ANT site, that is, control location, included contacts above the borderline in ANT or in its immediate horizontal neighboring structures. Therapeutic stimulation is done bilaterally, that is, one lead is in the left-side thalamus and one in the right thalamus (Figure 2). Of the eight participants, four had both left- and right-side MD contacts bilaterally inside MD. Three participants had one contact in one side inside MD and in another side in the neighboring structure. One participant had MD contacts at both sides in the neighboring structure (Table 1).

The stimulation protocol was divided into two parts. First, one stimulation site, for example, ANT, was stimulated, and then another site, MD, was stimulated. The starting order of the sites was counter balanced between participants. At each site, the stimulator was turned ON for two blocks of testing (approximately 5 min) and then turned OFF for another two blocks, and so forth. There was a 2- to 3-min break between ON and OFF periods when the stimulator was adjusted. This was repeated 8 times at both sites, totaling 16 blocks and 1024 trials per stimulation site. Participants were blind to the stimulator status and to the active contact location (Figure 3).

Statistical Analysis

Errors were analyzed using generalized binary logistic regression with Stimulation condition (ON at MD, ON at ANT [control location], and OFF) and Distractor valence (emotional, neutral) as fixed effect predictors and Participants as random effect predictor. Trials in the OFF condition (OFF at MD and OFF at control location) were pooled together. In the logistic regression model, ON at MD was used as a reference value as it enabled comparison of MD stimulation condition to both control conditions (ON at ANT and OFF) in a single analysis. Errors were dichotomized for binary logistic regression so that for total errors the outcome variable was “error” or “correct,” for incorrect response “incorrect” or “other” (correct or missing response), for missing responses “miss” or “other” (correct or incorrect response) and for commission errors “commission error” (correct or incorrect response when no response was expected in no-go trial) or “correct” (no response in no-go case). Separate models were created for each error type where a model predicted a participant's probability to make an error of that type. To analyze shift errors, a generalized logistic regression model similar to other error types was used, but with an additional predictor, tertile (first, other), to account for the beginning of a block. Before the analysis, the data were checked for potential blocks where the participant apparently responded applying an incorrect rule but there were no indications of a switching problem. The criterion to exclude these blocks was that the sum of missing responses and commission errors in the entire block was larger than 75% of all trials. One block from two participants was excluded from the analysis because of the participant responding with an incorrect rule without other indications of shifting problem. Normal distribution of the random predictors was ensured with Q–Q plots.

RTs were analyzed with repeated-measures ANOVA using Stimulation condition and Distractor valence as within-subject factors. In case of significant interaction, data were stratified per group, and groups were analyzed separately. For RT analysis, only trials with correct response were used.

All statistical analysis was done using R v.3.2.1 (R Core Team, 2016) and lme4 package v.1.1-9 for regression analysis (Bates, Mächler, Bolker, & Walker, 2015) and ez package 4.4-0 for repeated-measures ANOVA (Lawrence, 2016).

RESULTS

Error rates are shown in Table 2 and Figure 4, and RTs are shown in Table 2.

Table 2. 

Performance in Different Stimulation Conditions

StimulationDistractorMean RT (SD)Median Total Errors (iqr)Median Incorrect Responses (iqr)Median Missing Responses (iqr)Median Commission Errors (iqr)Median Shift Errors (iqr)
ON at MD Emotional 528 msec (143) 6.2% (6.4) 5.5% (5.3) 1.2% (1.6) 1.2% (3.1) 2.3% (2.7) 
Neutral 523 msec (139) 6.6% (5.9) 4.7% (4.1) 0.8% (0.6) 1.2% (2.5) 2.0% (1.8) 
ON at ANT Emotional 507 msec (148) 4.0% (0.9) 1.6% (1.6) 0.4% (0.8) 0.9% (1.6) 2.3% (0.9) 
Neutral 487 msec (149) 4.5% (1.6) 2.6% (1.8) 0.0% (1.0) 1.3% (2.1) 1.6% (1.3) 
OFF Emotional 485 msec (128) 3.9% (2.5) 2.5% (1.2) 0.4% (0.6) 1.1% (1.6) 1.8% (1.8) 
Neutral 492 msec (129) 4.0% (3.5) 2.5% (0.8) 0.2% (0.2) 1.1% (1.5) 1.4% (1.4) 
StimulationDistractorMean RT (SD)Median Total Errors (iqr)Median Incorrect Responses (iqr)Median Missing Responses (iqr)Median Commission Errors (iqr)Median Shift Errors (iqr)
ON at MD Emotional 528 msec (143) 6.2% (6.4) 5.5% (5.3) 1.2% (1.6) 1.2% (3.1) 2.3% (2.7) 
Neutral 523 msec (139) 6.6% (5.9) 4.7% (4.1) 0.8% (0.6) 1.2% (2.5) 2.0% (1.8) 
ON at ANT Emotional 507 msec (148) 4.0% (0.9) 1.6% (1.6) 0.4% (0.8) 0.9% (1.6) 2.3% (0.9) 
Neutral 487 msec (149) 4.5% (1.6) 2.6% (1.8) 0.0% (1.0) 1.3% (2.1) 1.6% (1.3) 
OFF Emotional 485 msec (128) 3.9% (2.5) 2.5% (1.2) 0.4% (0.6) 1.1% (1.6) 1.8% (1.8) 
Neutral 492 msec (129) 4.0% (3.5) 2.5% (0.8) 0.2% (0.2) 1.1% (1.5) 1.4% (1.4) 

Stimulation at MD was associated with significantly more incorrect button presses and errors in general (total errors) compared with control conditions (stimulation at ANT or stimulation OFF). Percentage values are calculated from all trials. RT is for correct responses starting from the onset of the triangle and ending at the button press.

Figure 4. 

High-frequency stimulation at MD increased total errors and incorrect responses compared with stimulation at the control location (ANT) and stimulation OFF. Increase in total errors was mostly driven by increase in incorrect responses. Additionally, there was a statistically significant increase in missing responses, but only when compared with stimulation OFF. There was no statistically significant difference in probability to make a commission error or shift errors.

Figure 4. 

High-frequency stimulation at MD increased total errors and incorrect responses compared with stimulation at the control location (ANT) and stimulation OFF. Increase in total errors was mostly driven by increase in incorrect responses. Additionally, there was a statistically significant increase in missing responses, but only when compared with stimulation OFF. There was no statistically significant difference in probability to make a commission error or shift errors.

Total Errors

Generalized binary logistic regression for total errors revealed a main effect of stimulator ON at MD. MD stimulation increased participants' overall probability to make an error 2.26 times compared with stimulation of the control location (OR = 2.26, 95% CI [1.69, 3.01]) and 1.80 times compared with stimulator OFF (OR = 1.80, CI [1.45, 2.24]), reflecting disrupted executive functions in general when MD was stimulated. “Total errors” is a sum of basic error types, and it was mainly driven by incorrect responses.

Incorrect Responses

MD stimulation increased participants' probability to respond incorrectly 2.88 times compared with stimulation of the control location (OR = 2.88, CI [1.95, 4.24]) and 2.08 times compared with stimulator OFF (OR = 2.08, CI [1.57, 2.75]), reflecting compromised working memory performance when MD was stimulated.

To control for alternative factors that could explain increased probability to respond incorrectly, we added incorrect response model following covariates: orientation of the previous triangle (opposite, same) to control for the potential competition between motor responses when the previous triangle orientation was noncongruent with the current orientation, go signal color (green-go, red-go) to control for increased cognitive effort to respond with red traffic light instead of overlearned green traffic light, and a tertile information (first, other) to control for the response rule change as in shift error analysis. Even after controlling for these covariates, MD stimulation remained significant when compared with stimulation of the control location (OR = 9.52, CI [1.36, 66.67]) and stimulation OFF (OR = 5.51, CI [1.73, 17.51]) situations. Previous triangle orientation, go signal color, and tertile (beginning of the block) were not statistically significant predictors nor were there any significant interactions.

Working memory and attention are related constructs, where target must first be attended before it can be stored in working memory (Awh, Vogel, & Oh, 2006). If a participant has an attention problem, there should be simultaneous missing responses. To test for possible interaction between working memory and attention, we split each block where MD was stimulated into eight subblocks (eight trials each, approximately 16 sec) and calculated a local correlation between incorrect responses and missing responses. There was no statistically significant correlation between missing responses and incorrect responses (r = .01, p = .79), that is, working memory impairment cannot be explained by impairment in attention.

Other Error Types: Missing Responses, Commission Errors, and Shifting Errors

MD stimulation increased participants' probability to miss responding 1.90 times compared with stimulation OFF (OR = 1.90, CI [1.08, 3.33]), but difference in missing responses between MD stimulation and control location stimulation did not reach significance. Thus, increase in missing responses is not regionally specific.

There were no statistically significant predictors for commission errors or shifting errors.

Reaction Time

Repeated-measures ANOVA found no main effects for RT, but there was an interaction effect between Stimulation and the Valence of emotional distractor, F(2, 14) = 11.30, p = .001. Post hoc ANOVA done separately for each stimulation condition resulted in a main effect of emotional distractor for OFF condition, F(1, 7) = 28.50, p = .001, and ON at ANT condition, F(1, 7) = 12.34, p = .01, but not for ON at MD condition, F(1, 7) = 1.17, p = .31. When ANT was stimulated, RTs were longer in context of emotional distractor (507 msec, SD = 137 msec) in comparison with neutral distractor (488 msec, SD = 138 msec). In contrast, when there was no stimulation (OFF), RTs were slightly faster for emotional distractor, 486 msec (SD = 118 msec), compared with neutral distractor, 492 msec (SD = 119 msec).

Subgroup Comparison

Four participants had MD contacts bilaterally inside MD, whereas another four participants had either one or both contacts in a neighboring structure. To assess whether there is a difference between the two groups, we created separate error models for total errors and incorrect button presses, which included a Group term (bilateral MD group, unilateral MD group). Group and Participant terms had a hierarchical relationship, that is, Participant (random effect predictor) belonged to one of the two groups (fixed effect predictor). As distractor valence was not significant in the main analysis, it was not included in the subgroup comparison. Interaction effect between Group and Stimulation was statistically significant for incorrect responses (OR = 2.0, CI [1.10, 3.4]) and was approaching significance for total errors (OR = 1.5, CI [1.0, 2.2]) when comparing stimulator ON at MD and stimulator ON at ANT. Stimulator ON at MD comparison to stimulator OFF did not yield significances for total errors nor for incorrect responses.

When data were stratified per group and incorrect responses were analyzed separately, the bilateral MD group was 4.0 times more probable to make an incorrect response when stimulator was ON at MD compared with stimulator ON at control location. Respectively, the unilateral MD group was 2.0 times more probable to make an incorrect response when stimulator was ON at MD compared with stimulator ON at control location (Table 3). Thus, subjects having MD contact inside MD bilaterally were twice as probable to respond incorrectly compared with participants having MD contact inside MD only in one side (Figure 5).

Table 3. 

Logistic Regression Results

Total ErrorsIncorrect ResponsesMissed ResponsesCommission ErrorsShift Errors
A. Main Analysis 
(Intercept × ON at MD/emotional distractor) 0.07 (0.04–0.13) 0.09 (0.04–0.19) 0.02 (0.01–0.04) 0.02 (0.01–0.05) 0.02 (0.01–0.04) 
Neutral distractor 0.87 (0.69–1.09) 0.84 (0.63–1.13) 0.66 (0.35–1.24) 1.10 (0.68–1.77) 1.46 (0.79–2.71) 
ON at ANT 0.44 (0.33–0.59)*** 0.35 (0.24–0.51)*** 0.57 (0.29–1.12) 0.66 (0.38–1.16) 0.68 (0.32–1.46) 
OFF 0.55 (0.45–0.69)*** 0.48 (0.36–0.64)*** 0.53 (0.30–0.92)* 0.81 (0.52–1.25) 0.82 (0.45–1.49) 
Neutral distractor × ON at ANT 1.17 (0.78–1.76) 1.15 (0.66–2.01) 1.52 (0.56–4.12) 1.01 (0.46–2.20) 1.02 (0.38–2.77) 
Neutral distractor × off 1.04 (0.76–1.42) 1.12 (0.74–1.68) 0.54 (0.20–1.45) 0.98 (0.53–1.81) 0.65 (0.29–1.46) 
 
B. Subgroup Analysis, Incorrect Responses 
(Intercept × ON at MD/Unilateral MD group) 0.08 (0.04–0.19) 0.11 (0.04–0.29) 0.02 (0.01–0.05) 0.02 (0.00–0.09) 0.02 (0.01–0.05) 
Bilateral MD group 0.63 (0.19–2.05) 0.56 (0.13–2.37) 0.55 (0.11–2.75) 0.77 (0.10–5.91) 1.41 (0.33–6.09) 
ON at ANT 0.56 (0.43–0.74)*** 0.49 (0.34–0.70)*** 0.56 (0.27–1.15) 0.76 (0.46–1.24) 1.07 (0.39–2.92) 
OFF 0.61 (0.49–0.75)*** 0.54 (0.41–0.71)*** 0.38 (0.20–0.71)** 0.91 (0.62–1.34) 1.09 (0.47–2.54) 
Bilateral MD group × ON at ANT 0.68 (0.45–1.03) 0.51 (0.29–0.91)* 1.50 (0.55–4.07) 0.70 (0.31–1.57) 0.37 (0.08–1.71) 
Bilateral MD group × OFF 0.84 (0.61–1.16) 0.88 (0.58–1.32) 1.30 (0.52–3.24) 0.70 (0.37–1.31) 0.56 (0.17–1.83) 
 
 Bilateral MD Group Unilateral MD Group    
(Intercept: ON at MD) 0.06 (0.01–0.25) 0.10 (0.06–0.18)    
ON at ANT 0.25 (0.16–0.40)*** 0.49 (0.35–0.70)***    
Total ErrorsIncorrect ResponsesMissed ResponsesCommission ErrorsShift Errors
A. Main Analysis 
(Intercept × ON at MD/emotional distractor) 0.07 (0.04–0.13) 0.09 (0.04–0.19) 0.02 (0.01–0.04) 0.02 (0.01–0.05) 0.02 (0.01–0.04) 
Neutral distractor 0.87 (0.69–1.09) 0.84 (0.63–1.13) 0.66 (0.35–1.24) 1.10 (0.68–1.77) 1.46 (0.79–2.71) 
ON at ANT 0.44 (0.33–0.59)*** 0.35 (0.24–0.51)*** 0.57 (0.29–1.12) 0.66 (0.38–1.16) 0.68 (0.32–1.46) 
OFF 0.55 (0.45–0.69)*** 0.48 (0.36–0.64)*** 0.53 (0.30–0.92)* 0.81 (0.52–1.25) 0.82 (0.45–1.49) 
Neutral distractor × ON at ANT 1.17 (0.78–1.76) 1.15 (0.66–2.01) 1.52 (0.56–4.12) 1.01 (0.46–2.20) 1.02 (0.38–2.77) 
Neutral distractor × off 1.04 (0.76–1.42) 1.12 (0.74–1.68) 0.54 (0.20–1.45) 0.98 (0.53–1.81) 0.65 (0.29–1.46) 
 
B. Subgroup Analysis, Incorrect Responses 
(Intercept × ON at MD/Unilateral MD group) 0.08 (0.04–0.19) 0.11 (0.04–0.29) 0.02 (0.01–0.05) 0.02 (0.00–0.09) 0.02 (0.01–0.05) 
Bilateral MD group 0.63 (0.19–2.05) 0.56 (0.13–2.37) 0.55 (0.11–2.75) 0.77 (0.10–5.91) 1.41 (0.33–6.09) 
ON at ANT 0.56 (0.43–0.74)*** 0.49 (0.34–0.70)*** 0.56 (0.27–1.15) 0.76 (0.46–1.24) 1.07 (0.39–2.92) 
OFF 0.61 (0.49–0.75)*** 0.54 (0.41–0.71)*** 0.38 (0.20–0.71)** 0.91 (0.62–1.34) 1.09 (0.47–2.54) 
Bilateral MD group × ON at ANT 0.68 (0.45–1.03) 0.51 (0.29–0.91)* 1.50 (0.55–4.07) 0.70 (0.31–1.57) 0.37 (0.08–1.71) 
Bilateral MD group × OFF 0.84 (0.61–1.16) 0.88 (0.58–1.32) 1.30 (0.52–3.24) 0.70 (0.37–1.31) 0.56 (0.17–1.83) 
 
 Bilateral MD Group Unilateral MD Group    
(Intercept: ON at MD) 0.06 (0.01–0.25) 0.10 (0.06–0.18)    
ON at ANT 0.25 (0.16–0.40)*** 0.49 (0.35–0.70)***    

(A) Main analysis. Participants' probability to respond incorrectly or make an error in general is much higher when MD is stimulated compared with stimulation of the control location (ANT) or when stimulation is OFF. (B) Subgroup analysis for incorrect responses. When MD stimulation was compared with ANT stimulation, participants having electrodes bilaterally in MD were approximately twice as probable to make a working memory-related error, that is, respond incorrectly, compared with participants having only one electrode in MD.

*p < .05.

**p < .01.

***p < .001.

Figure 5. 

Incorrect responses for each participant. Participants 2, 3, 5, 6, 12, and 13 made significantly more errors when MD was stimulated compared with ANT stimulation or no stimulation conditions. Participants 1 and 4 have a deviant incorrect responses profile, reflecting their slightly different contact locations compared with other Participants (Table 1). We speculate that the reason for Participant 1 having essentially no difference between MD stimulation and ANT stimulation is due to the left MD location being in the mammillothalamic tract, which is connected to ANT, and thus, MD stimulation affects ANT rather than MD. Participant 4 has a very different profile compared with all other participants, but we do not have a plausible explanation for that. Participant 2 has similar response profile to others even if her contact locations deviate from others. We speculate that she has her left MD contact in the internal medullary lamina and stimulation reaches MD, causing similar incorrect response profile like the other participants who have the contacts in MD. In the figure, it seems like some participants would make less errors when ANT is stimulated compared with no stimulation condition. The difference is not, however, significant. We speculate that the phenomena are an indication of the carryover effect where stimulation given during the previous MD stimulation block still effects the succeeding OFF block, thus increasing incorrect responses in that block. The presence of carryover effect suggests that the break (cool-off period) after stimulator adjustment was not long enough for the effect of stimulation to disappear entirely.

Figure 5. 

Incorrect responses for each participant. Participants 2, 3, 5, 6, 12, and 13 made significantly more errors when MD was stimulated compared with ANT stimulation or no stimulation conditions. Participants 1 and 4 have a deviant incorrect responses profile, reflecting their slightly different contact locations compared with other Participants (Table 1). We speculate that the reason for Participant 1 having essentially no difference between MD stimulation and ANT stimulation is due to the left MD location being in the mammillothalamic tract, which is connected to ANT, and thus, MD stimulation affects ANT rather than MD. Participant 4 has a very different profile compared with all other participants, but we do not have a plausible explanation for that. Participant 2 has similar response profile to others even if her contact locations deviate from others. We speculate that she has her left MD contact in the internal medullary lamina and stimulation reaches MD, causing similar incorrect response profile like the other participants who have the contacts in MD. In the figure, it seems like some participants would make less errors when ANT is stimulated compared with no stimulation condition. The difference is not, however, significant. We speculate that the phenomena are an indication of the carryover effect where stimulation given during the previous MD stimulation block still effects the succeeding OFF block, thus increasing incorrect responses in that block. The presence of carryover effect suggests that the break (cool-off period) after stimulator adjustment was not long enough for the effect of stimulation to disappear entirely.

DISCUSSION

This study provides direct, causal evidence from humans for the role of MD in working memory. By disrupting and recovering MD's function in human participants with focal, high-frequency electric stimulation we could modulate an individual's working memory performance. The effect was strongest in participants who had electrodes precisely inside MD bilaterally compared with participants having only one electrode precisely in MD and another electrode in the close vicinity of MD. The stimulation effect was region- and task-specific. The increased number of incorrect errors was only observed when MD was stimulated and not when the control location (ANT), located a few millimeters above MD, was stimulated. The observed effect of MD stimulation was specific to working memory as there was no evidence for disruption of response inhibition, attention, or cognitive flexibility.

Previous studies investigating MD's role in working memory and executive function have used indirect methods, that is, human vascular lesion studies, human imaging studies, and animal studies. In contrast to these previous studies, we compared the performance of the same human individuals when high-frequency electric stimulation, that is, “lesion,” was alternated ON and OFF. Using a within-subject design, we could control for individual variability, such as participants' baseline cognitive performance and different medications. Furthermore, compared with the previous human vascular lesion studies, the bipolar stimulation mode used in this study provided very precise control for the location of a “lesion”.

To make anatomically and functionally specific conclusions on the effect of focal neuromodulation on brain functions, it is necessary to have an anatomical control location and a control task. In the current study, ANT served as an anatomical control location, and we observed the effects on working memory only when MD was stimulated but not when ANT was stimulated. This suggests a location-specific effect of the stimulation rather than regionally unspecific general effect of DBS. In the Executive RT Test, different subtasks can be considered as control tasks to one another. Different subtasks test different executive processes, and they are associated with different error types. We discovered that an increase in the error type reflecting working memory performance, incorrect responses, had a statistical significance when compared with both control conditions (no stimulation and stimulation of the control location), and it also had the largest effect size. Stimulation had no effect on response inhibition (commission errors) or shifting, and there was only a borderline effect on attention (missed responses). There was only a barely significant difference in probability to miss responding (95% CI lower limit 1.08) when stimulation of MD was compared with no stimulation condition and no significant difference when compared with stimulation of the control location. Observing stimulation having a major effect on working memory, only a borderline effect on attention and no effect on response inhibition or shifting suggest a specific effect of electric MD stimulation on working memory rather than an unspecific effect on executive functions in general. Taken together, both a control task and a control location in the experiment allowed for conclusions on anatomically and functionally specific effect of high-frequency DBS at MD, that is, the impairment of working memory. Finally, conclusions of causality can be made when cognitive performance is altered online as a result of DBS delivery.

Each error type in the Executive RT Test is thought to reflect the efficiency of a specific cognitive process, but there are potential alternative explanations. The number of incorrect responses in go trials is thought to indicate working memory performance, but it could be impacted also by lapses in early selective attention where triangle's orientation is not stored properly in working memory (Awh et al., 2006). Incorrect responses could also be impacted by problems related to cognitive flexibility, more specifically to shifting or inhibiting competing motor responses, that is, when the current triangle and previous triangle are incongruent or the response rule has just changed. On the other hand, a nonintuitive “red-go” response rule requires a greater amount of cognitive control than overlearned “green-go” response rule and could result in an increased number of incorrect responses in “red-go” trials. To control for cognitive flexibility, we used triangle congruency, go signal color, and beginning of the block as a covariate in the incorrect responses regression model and found they had no impact to incorrect responses. If incorrect responses would be explained by lapses in early attention, there should be simultaneous lapses in responding, that is, increase in missed responses and prolonged RTs. This was not the case, because there was no correlation between missing responses and incorrect responses at the subblock level and there was no significant difference in RT when MD was stimulated compared with ANT stimulation or no stimulation. Moreover, MD stimulation did not differ significantly from ANT stimulation with regard to number of missed responses. Thus, MD stimulation resulting in increased number of incorrect responses cannot be explained by impaired attention, shifting, or competing motor responses, and results suggest primary disruption in working memory performance.

At the core of this study is the specificity of the stimulated location. Knowledge of the DBS volume of the tissue affected, that is, its shape and size, is largely based on computational models and simulations (Buhlmann, Hofmann, Tass, & Hauptmann, 2011; Montgomery, 2010). In bipolar stimulation mode, the electrical field is located between the positive and the negative contacts in a lead. Because two neighboring contacts were used resulting in an electric field that is cylindrically shaped and covering a volume of a few millimeters away from the lead in horizontal direction and 3.5 mm in vertical direction (one contact is 1.5 mm, and contacts are spaced 0.5 mm apart), we could obtain very precise location information. The vertical electric field is precise because it is limited by the contacts, whereas horizontally the electric field attenuates quickly in tissue until it reaches zero. Thus, the “lesion” caused by the bipolar high-frequency stimulation is very focal around the active contacts.

It is challenging to implant DBS leads with high accuracy in 3-D space even with the modern stereotactic devices. As a result, contacts may not always be precisely inside the target nucleus and instead only be in the proximity of the target. As described earlier, four participants had MD contacts bilaterally exactly in MD whereas others had one (three participants) or both (one participant) contacts in the structure adjacent to MD. The adjacent structure was typically internal medullary lamina separating different nuclei but may have also involved the mamillothalamic tract in two cases or ventral anterior nucleus in one case. In these cases, it is possible that the electric field in horizontal direction might reach the neighboring structure adjacent to MD affecting results. However, this is unlikely because only a minority of the contacts, 5 of 16 “MD” contacts, were not in MD, and those locations were distributed between different nuclei so that impact to any single neighboring structure would be minimal. In the vertical direction, a contact defines the boundary of the electric field so only the case where the electric field can reach ANT from MD or MD from ANT is when one of the bipolar contacts is in the “wrong” side of internal medullary lamina separating ANT and MD. This was not the case with any of the MD locations. In summary, the effect of the stimulation is focal, and in those few cases where the contact was not precisely in MD, the effect to adjacent structures was not significant. The specificity of the stimulated location was further demonstrated when subgroups were compared and participants having electrodes bilaterally inside MD showed twofold probability to respond incorrectly compared with participants having only one contact in MD. This finding is in line with the previous literature, where bilateral stimulation has been shown to be more efficient than unilateral stimulation (Bastian, Kelly, Revilla, Perlmutter, & Mink, 2003; Ondo, Almaguer, Jankovic, & Simpson, 2001; Kumar, Lozano, Sime, Halket, & Lang, 1999).

Working memory is defined as a construct for storing information for short periods of time with the online manipulation of this information (Cowan, 2008; Baddeley, 1992). Funahashi, Watanabe, and colleagues have studied the nature and the manipulation of this information in pFC with monkeys using cell population analysis while monkeys performed an oculomotor delayed-response task (Takeda & Funahashi, 2002; Niki & Watanabe, 1976). They showed that there are temporal changes in the direction of the cell population vectors during the oculomotor delayed-response task and concluded that during the task vector directions coded first retrospective sensory information, that is, the location of the visual cue, and later the prospective motor activity information, that is, forthcoming saccade direction. The same experiment was later repeated while MD activity was observed (Watanabe & Funahashi, 2004a, 2004b). When comparing changes in the population vectors between pFC and MD, the authors noticed that population vectors in pFC and MD followed a similar pattern of change, but the change from coding past sensory information to coding forthcoming motor activity started earlier in MD compared with pFC (Funahashi, 2013; Watanabe & Funahashi, 2012; Funahashi, Takeda, & Watanabe, 2004). Authors concluded that MD participates more on the prospective motor aspect of working memory than retrospective sensory aspect. Similarly, Mitchell, Browning, and colleagues have shown with nonhuman primates the critical role of MD and its connections to pFC in executive functions and adaptive decision-making using object-in-place scene discrimination task (Browning et al., 2015) and Parnaudeau et al. have demonstrated that pharmacogenetic diminishing of MD neuronal activity interfered with MD–pFC beta range synchronicity in mice, which further correlated with the impairment of working memory (Parnaudeau et al., 2013). Building on these findings, we argue that high-frequency stimulation at MD impairs subject's working memory by interfering retrospective sensory and prospective motor action information processing in MD, which is reflected in pFC via extensive MD–pFC connections (Jang & Yeo, 2014; Mitchell & Chakraborty, 2013).

Before visual information enters working memory, it is stored temporarily in iconic memory. Iconic memory, or visual information persistence, is a very brief sensory buffer storing visual information after the actual physical stimulus has disappeared (Coltheart, 1980; Sperling, 1960). Iconic memory is thought to decay completely in 200–300 msec (Loftus, Duncan, & Gehrig, 1992; Irwin & Yeomans, 1986). The experiment used a 150-msec delay period between the triangle offset and a traffic light onset, so one could speculate that the triangle's orientation is stored in iconic memory and no working memory needs to be recruited. However, there are arguments why storing information only in iconic memory and not in working memory is not probable. First, information in iconic memory is overwritten immediately when a new stimulus arrives (Enns & Di Lollo, 2000; Kovács, Vogels, & Orban, 1995). The go or no-go signal (traffic light) onset was 150 msec after the triangle offset, so the initial visual stimuli (triangle) in iconic memory is replaced at this time point by the visual stimuli conveying go/no-go information. Second, iconic memory has been associated with the anterior STS ( Keysers, Xiao, Földiák, & Perrett, 2005), and there is only little evidence for sparse connections from MD to STS in rhesus monkeys (Giguere & Goldman-Rakic, 1988), so the plausible neurophysiological mechanism for how MD stimulation could interfere with iconic memory is missing. In contrast, there are vast direct neural connections from MD to pFC (Jang & Yeo, 2014; Mitchell & Chakraborty, 2013), and MD has been linked with working memory (Parnaudeau et al., 2013; Watanabe & Funahashi, 2012). Connections from MD to pFC are demonstrated also in Figure 2. In conclusion, the previous arguments point out that successfully indicating the direction of the previously presented triangle mandates the recruitment of working memory, and furthermore, there is a plausible neurophysiological mechanism of how high-frequency electric stimulation at MD can result in the increased probability to respond incorrectly by interfering with working memory performance.

Since MD has wide connections to different areas in pFC, it has been linked with executive functions in general (Browning et al., 2015; Parnaudeau et al., 2015; Mitchell & Chakraborty, 2013). As the Executive RT Test requires cognitive control, it could be argued that MD stimulation interfered with cognitive control instead of working memory. Although working memory and cognitive control are closely linked and interacting constructs, there was no evidence for difficulties in cognitive control. RTs in the Executive RT Test did not change due to stimulation nor did the errors measuring cognitive control and inhibition (commission errors). To that end, we can conclude that the detected effect indeed was due to high-frequency electric stimulation disrupting MD functioning and resulting in interference in working memory performance.

In conclusion, we showed that online high-frequency stimulation at the MD impaired working memory performance in humans. The finding is in accordance with the previous research and extends it by providing direct, causal evidence from humans for MD's role in human cognition. The results have scientific relevance by confirming the role of MD in human working memory causally. Although high-frequency electric stimulation of MD disrupts working memory performance, it is possible that other kinds of neuromodulation of MD could improve working memory performance with vast potential implications for restoring or enhancing cognition (Bick & Eskandar, 2016; Sankar, Lipsman, & Lozano, 2014). Furthermore, the results have clinical relevance in treating refractory epilepsy with DBS by highlighting the importance of the precise contact locations. If active electrodes are located too deep, the treatment effect (Lehtimäki et al., 2016), as well as working memory performance as reported in the current study, may be compromised. The current results suggest that the cognitive performance profile could be used as an additional indicator of the active electrode location supporting anatomically accurate targeting of the electrodes. The results further demonstrate the feasibility of using clinically implanted deep brain stimulators as a neuroscientific research tool that allows for causal evidence for brain structure–function relationship in humans.

Acknowledgments

This research was supported by the Academy of Finland and the Competitive Research Fund of Pirkanmaa Hospital District.

Reprint requests should be sent to Kaisa M. Hartikainen, Behavioral Neurology Research Unit, Finn-Medi 6-7, Tampere University Hospital, P.O. Box 2000, FI-33520, Tampere, Finland, or via e-mail: kaisa.hartikainen@uta.fi.

REFERENCES

REFERENCES
Aggleton
,
J. P.
, &
Mishkin
,
M.
(
1983
).
Memory impairments following restricted medial thalamic lesions in monkeys
.
Experimental Brain Research
,
52
,
199
209
.
Awh
,
E.
,
Vogel
,
E. K.
, &
Oh
,
S.-H.
(
2006
).
Interactions between attention and working memory
.
Neuroscience
,
139
,
201
208
.
Baddeley
,
A.
(
1992
).
Working memory
.
Science
,
255
,
556
559
.
Bastian
,
A. J.
,
Kelly
,
V. E.
,
Revilla
,
F. J.
,
Perlmutter
,
J. S.
, &
Mink
,
J. W.
(
2003
).
Different effects of unilateral versus bilateral subthalamic nucleus stimulation on walking and reaching in Parkinson's disease
.
Movement Disorders
,
18
,
1000
1007
.
Bates
,
D.
,
Mächler
,
M.
,
Bolker
,
B.
, &
Walker
,
S.
(
2015
).
Fitting Linear mixed-effects models using lme4
.
Journal of Statistical Software
,
67
,
1
48
.
Benazzouz
,
A.
,
Gao
,
D. M.
,
Ni
,
Z. G.
,
Piallat
,
B.
,
Bouali-Benazzouz
,
R.
, &
Benabid
,
A. L.
(
2000
).
Effect of high-frequency stimulation of the subthalamic nucleus on the neuronal activities of the substantia nigra pars reticulata and ventrolateral nucleus of the thalamus in the rat
.
Neuroscience
,
99
,
289
295
.
Bick
,
S. K. B.
, &
Eskandar
,
E. N.
(
2016
).
Neuromodulation for restoring memory
.
Neurosurgical Focus
,
40
,
E5
.
Browning
,
P. G. F.
,
Chakraborty
,
S.
, &
Mitchell
,
A. S.
(
2015
).
Evidence for mediodorsal thalamus and prefrontal cortex interactions during cognition in macaques
.
Cerebral Cortex
,
25
,
4519
4534
.
Buhlmann
,
J.
,
Hofmann
,
L.
,
Tass
,
P. A.
, &
Hauptmann
,
C.
(
2011
).
Modeling of a segmented electrode for desynchronizing deep brain stimulation
.
Frontiers in Neuroengineering
,
4
,
15
.
Coltheart
,
M.
(
1980
).
Iconic memory and visible persistence
.
Perception & Psychophysics
,
27
,
183
228
.
Cowan
,
N.
(
2008
).
What are the differences between long-term, short-term, and working memory?
Progress in Brain Research
,
169
,
323
338
.
D'Esposito
,
M.
, &
Postle
,
B. R.
(
2014
).
The cognitive neuroscience of working memory
.
Annual Review of Psychology
,
66
,
115
142
.
Enns
,
J.
, &
Di Lollo
,
V.
(
2000
).
What's new in visual masking?
Trends in Cognitive Sciences
,
4
,
345
352
.
Fisher
,
R.
,
Salanova
,
V.
,
Witt
,
T.
,
Worth
,
R.
,
Henry
,
T.
,
Gross
,
R.
, et al
(
2010
).
Electrical stimulation of the anterior nucleus of thalamus for treatment of refractory epilepsy
.
Epilepsia
,
51
,
899
908
.
Funahashi
,
S.
(
2013
).
Thalamic mediodorsal nucleus and its participation in spatial working memory processes: Comparison with the prefrontal cortex
.
Frontiers in Systems Neuroscience
,
7
,
36
.
Funahashi
,
S.
,
Takeda
,
K.
, &
Watanabe
,
Y.
(
2004
).
Neural mechanisms of spatial working memory: Contributions of the dorsolateral prefrontal cortex and the thalamic mediodorsal nucleus
.
Cognitive, Affective & Behavioral Neuroscience
,
4
,
409
420
.
Gazzaley
,
A.
,
Rissman
,
J.
, &
D'Esposito
,
M.
(
2004
).
Functional connectivity during working memory maintenance
.
Cognitive, Affective & Behavioral Neuroscience
,
4
,
580
599
.
Giguere
,
M.
, &
Goldman-Rakic
,
P. S.
(
1988
).
Mediodorsal nucleus: Areal, laminar, and tangential distribution of afferents and efferents in the frontal lobe of rhesus monkeys
.
The Journal of Comparative Neurology
,
277
,
195
213
.
Goldman-Rakic
,
P.
(
1995
).
Cellular basis of working memory
.
Neuron
,
14
,
477
485
.
Hartikainen
,
K. M.
,
Sun
,
L.
,
Polvivaara
,
M.
,
Brause
,
M.
,
Lehtimäki
,
K.
,
Haapasalo
,
J.
, et al
(
2014
).
Immediate effects of deep brain stimulation of anterior thalamic nuclei on executive functions and emotion–attention interaction in humans
.
Journal of Clinical and Experimental Neuropsychology
,
36
,
540
550
.
Hartikainen
,
K. M.
,
Waljas
,
M.
,
Isoviita
,
T.
,
Dastidar
,
P.
,
Liimatainen
,
S.
,
Solbakk
,
A. K.
, et al
(
2010
).
Persistent symptoms in mild to moderate traumatic brain injury associated with executive dysfunction
.
Journal of Clinical and Experimental Neuropsychology
,
32
,
767
774
.
Irwin
,
D. E.
, &
Yeomans
,
J. M.
(
1986
).
Sensory registration and informational persistence
.
Journal of Experimental Psychology. Human Perception and Performance
,
12
,
343
360
.
Isseroff
,
A.
,
Rosvold
,
H. E.
,
Galkin
,
T. W.
, &
Goldman-Rakic
,
P. S.
(
1982
).
Spatial memory impairments following damage to the mediodorsal nucleus of the thalamus in rhesus monkeys
.
Brain Research
,
232
,
97
113
.
Jang
,
S. H.
, &
Yeo
,
S. S.
(
2014
).
Thalamocortical connections between the mediodorsal nucleus of the thalamus and prefrontal cortex in the human brain: A diffusion tensor tractographic study
.
Yonsei Medical Journal
,
55
,
709
714
.
Keysers
,
C.
,
Xiao
,
D.-K.
,
Földiák
,
P.
, &
Perrett
,
D. I.
(
2005
).
Out of sight but not out of mind: The neurophysiology of iconic memory in the superior temporal sulcus
.
Cognitive Neuropsychology
,
22
,
316
332
.
Kocabicak
,
E.
,
Temel
,
Y.
,
Höllig
,
A.
,
Falkenburger
,
B.
, &
Tan
,
S. K.
(
2015
).
Current perspectives on deep brain stimulation for severe neurological and psychiatric disorders
.
Neuropsychiatric Disease and Treatment
,
11
,
1051
1066
.
Kovács
,
G.
,
Vogels
,
R.
, &
Orban
,
G. A.
(
1995
).
Cortical correlate of pattern backward masking
.
Proceedings of the National Academy of Sciences, U.S.A.
,
92
,
5587
5591
.
Kumar
,
R.
,
Lozano
,
A. M.
,
Sime
,
E.
,
Halket
,
E.
, &
Lang
,
A. E.
(
1999
).
Comparative effects of unilateral and bilateral subthalamic nucleus deep brain stimulation
.
Neurology
,
53
,
561
566
.
Lawrence
,
M. A.
(
2016
).
ez: Easy analysis and visualization of factorial experiments
.
R package version 4.4-0
. https://CRAN.R-project.org/package=ez.
Lehtimäki
,
K.
,
Möttönen
,
T.
,
Järventausta
,
K.
,
Katisko
,
J.
,
Tähtinen
,
T.
,
Haapasalo
,
J.
, et al
(
2016
).
Outcome based definition of the anterior thalamic deep brain stimulation target in refractory epilepsy
.
Brain Stimulation
,
9
,
268
275
.
Liimatainen
,
J.
,
Peräkylä
,
J.
,
Järvelä
,
K.
,
Sisto
,
T.
,
Yli-Hankala
,
A.
, &
Hartikainen
,
K. M.
(
2016
).
Improved cognitive flexibility after aortic valve replacement
.
Interactive Journal of Cardiovascular and Thoracic Surgery
,
23
,
630
636
.
Loftus
,
G. R.
,
Duncan
,
J.
, &
Gehrig
,
P.
(
1992
).
On the time course of perceptual information that results from a brief visual presentation
.
Journal of Experimental Psychology. Human Perception and Performance
,
18
,
530
549
.
Mäki-Marttunen
,
V.
,
Kuusinen
,
V.
,
Brause
,
M.
,
Perakyla
,
J.
,
Polvivaara
,
M.
,
Ribeiro
,
R.
, et al
(
2014
).
Enhanced attention capture by emotional stimuli in mild traumatic brain injury
.
Journal of Neurotrauma
,
279
,
1
30
.
Mäki-Marttunen
,
V.
,
Kuusinen
,
V.
,
Peräkylä
,
J.
,
Ogawa
,
K. H.
,
Brause
,
M.
,
Brander
,
A.
, et al
(
2016
).
Greater attention to task-relevant threat due to orbitofrontal lesion
.
Journal of Neurotrauma
,
34
,
400
413
.
Manly
,
T.
,
Robertson
,
I. H.
,
Galloway
,
M.
, &
Hawkins
,
K.
(
1999
).
The absent mind: Further investigations of sustained attention to response
.
Neuropsychologia
,
37
,
661
670
.
McIntyre
,
C. C.
,
Grill
,
W. M.
,
Sherman
,
D. L.
, &
Thakor
,
N. V.
(
2004
).
Cellular effects of deep brain stimulation: Model-based analysis of activation and inhibition
.
Journal of Neurophysiology
,
91
,
1457
1469
.
Mitchell
,
A. S.
,
Browning
,
P. G. F.
, &
Baxter
,
M. G.
(
2007
).
Neurotoxic lesions of the medial mediodorsal nucleus of the thalamus disrupt reinforcer devaluation effects in rhesus monkeys
.
Journal of Neuroscience
,
27
,
11289
11295
.
Mitchell
,
A. S.
, &
Chakraborty
,
S.
(
2013
).
What does the mediodorsal thalamus do?
Frontiers in Systems Neuroscience
,
7
,
37
.
Monchi
,
O.
,
Petrides
,
M.
,
Petre
,
V.
,
Worsley
,
K.
, &
Dagher
,
A.
(
2001
).
Wisconsin Card Sorting revisited: Distinct neural circuits participating in different stages of the task identified by event-related functional magnetic resonance imaging
.
Journal of Neuroscience
,
21
,
7733
7741
.
Montgomery
,
E. B. J.
(
2010
).
Deep brain stimulation programming: Principles and practice
(1st ed.).
New York
:
Oxford University Press
.
Niki
,
H.
, &
Watanabe
,
M.
(
1976
).
Prefrontal unit activity and delayed response: Relation to cue location versus direction of response
.
Brain Research
,
105
,
79
88
.
Ondo
,
W.
,
Almaguer
,
M.
,
Jankovic
,
J.
, &
Simpson
,
R. K.
(
2001
).
Thalamic deep brain stimulation
.
Archives of Neurology
,
58
,
218
.
Parnaudeau
,
S.
,
oneill
,
P. K.
,
Bolkan
,
S. S.
,
Ward
,
R. D.
,
Abbas
,
A. I.
,
Roth
,
B. L.
, et al
(
2013
).
Inhibition of mediodorsal thalamus disrupts thalamofrontal connectivity and cognition
.
Neuron
,
77
,
1151
1162
.
Parnaudeau
,
S.
,
Taylor
,
K.
,
Bolkan
,
S. S.
,
Ward
,
R. D.
,
Balsam
,
P. D.
, &
Kellendonk
,
C.
(
2015
).
Mediodorsal thalamus hypofunction impairs flexible goal-directed behavior
.
Biological Psychiatry
,
77
,
445
453
.
Piras
,
F.
,
Caltagirone
,
C.
, &
Spalletta
,
G.
(
2010
).
Working memory performance and thalamus microstructure in healthy subjects
.
Neuroscience
,
171
,
496
505
.
Preuss
,
T.
(
1995
).
The argument from animals to humans in cognitive neuroscience
.
The Cognitive Neurosciences
,
81
,
1227
1241
.
R Core Team
. (
2016
).
R: A language and environment for statistical computing
.
Vienna, Austria
:
R Foundation for Statistical Computing
.
Rorden
,
C.
, &
Karnath
,
H.-O.
(
2004
).
Using human brain lesions to infer function: A relic from a past era in the fMRI age?
Nature Reviews Neuroscience
,
5
,
813
819
.
Salanova
,
V.
,
Witt
,
T.
,
Worth
,
R.
,
Henry
,
T. R.
,
Gross
,
R. E.
,
Nazzaro
,
J. M.
, et al
(
2015
).
Long-term efficacy and safety of thalamic stimulation for drug-resistant partial epilepsy
.
Neurology
,
84
,
1017
1025
.
Sankar
,
T.
,
Lipsman
,
N.
, &
Lozano
,
A. M.
(
2014
).
Deep brain stimulation for disorders of memory and cognition
.
Neurotherapeutics
,
11
,
527
534
.
Smallwood
,
J.
,
Davies
,
J. B.
,
Heim
,
D.
,
Finnigan
,
F.
,
Sudberry
,
M.
,
O'Connor
,
R.
, et al
(
2004
).
Subjective experience and the attentional lapse: Task engagement and disengagement during sustained attention
.
Consciousness and Cognition
,
13
,
657
690
.
Smallwood
,
J.
, &
Schooler
,
J. W.
(
2006
).
The restless mind
.
Psychological Bulletin
,
132
,
946
958
.
Sperling
,
G.
(
1960
).
The information available in brief visual presentations
.
Psychological Monographs: General and Applied
,
74
,
1
29
.
Sun
,
L.
,
Peräkylä
,
J.
,
Holm
,
K.
,
Haapasalo
,
J.
,
Lehtimäki
,
K.
,
Ogawa
,
K. H.
, et al
(
2017
).
Vagus nerve stimulation improves working memory performance
.
Journal of Clinical and Experimental Neuropsychology
. doi:10.1080/13803395.2017.1285869.
Sun
,
L.
,
Peräkylä
,
J.
,
Polvivaara
,
M.
,
Öhman
,
J.
,
Peltola
,
J.
,
Lehtimäki
,
K.
et al
, (
2015
).
Human anterior thalamic nuclei are involved in emotion–attention interaction
.
Neuropsychologia
,
78
,
88
94
.
Takeda
,
K.
, &
Funahashi
,
S.
(
2002
).
Prefrontal task-related activity representing visual cue location or saccade direction in spatial working memory tasks
.
Journal of Neurophysiology
,
87
,
567
588
.
Van der Werf
,
Y. D.
,
Scheltens
,
P.
,
Lindeboom
,
J.
,
Witter
,
M. P.
,
Uylings
,
H. B. M.
, &
Jolles
,
J.
(
2003
).
Deficits of memory, executive functioning and attention following infarction in the thalamus; A study of 22 cases with localised lesions
.
Neuropsychologia
,
41
,
1330
1344
.
Vuilleumier
,
P.
, &
Schwartz
,
S.
(
2001
).
Beware and be aware: Capture of spatial attention by fear-related stimuli in neglect
.
NeuroReport
,
12
,
1119
1122
.
Watanabe
,
Y.
, &
Funahashi
,
S.
(
2004a
).
Neuronal activity throughout the primate mediodorsal nucleus of the thalamus during oculomotor delayed-responses. I. Cue-, delay-, and response-period activity
.
Journal of Neurophysiology
,
92
,
1738
1755
.
Watanabe
,
Y.
, &
Funahashi
,
S.
(
2004b
).
Neuronal activity throughout the primate mediodorsal nucleus of the thalamus during oculomotor delayed-responses. II. Activity encoding visual versus motor signal
.
Journal of Neurophysiology
,
92
,
1756
1769
.
Watanabe
,
Y.
, &
Funahashi
,
S.
(
2012
).
Thalamic mediodorsal nucleus and working memory
.
Neuroscience and Biobehavioral Reviews
,
36
,
134
142
.
Zoppelt
,
D.
,
Koch
,
B.
,
Schwarz
,
M.
, &
Daum
,
I.
(
2003
).
Involvement of the mediodorsal thalamic nucleus in mediating recollection and familiarity
.
Neuropsychologia
,
41
,
1160
1170
.