Transcranial Direct Current Stimulation (tDCS) of Wernicke's and Broca's Areas in Studies of Language Learning and Word Acquisition

Summary

Here, we describe a protocol for using transcranial direct current stimulation for psycho- and neurolinguistic experiments aimed at studying, in a naturalistic yet fully controlled way, the role of cortical areas of the human brain in word learning, and a comprehensive set of behavioral procedures for assessing the outcomes.

Abstract

Language is a highly important yet poorly understood function of the human brain. While studies of brain activation patterns during language comprehension are abundant, what is often critically missing is causal evidence of brain areas' involvement in a particular linguistic function, not least due to the unique human nature of this ability and a shortage of neurophysiological tools to study causal relationships in the human brain noninvasively. Recent years have seen a rapid rise in the use of transcranial direct current stimulation (tDCS) of the human brain, an easy, inexpensive and safe noninvasive technique that can modulate the state of the stimulated brain area (putatively by shifting excitation/inhibition thresholds), enabling a study of its particular contribution to specific functions. While mostly focusing on motor control, the use of tDCS is becoming more widespread in both basic and clinical research on higher cognitive functions, language included, but the procedures for its application remain variable. Here, we describe the use of tDCS in a psycholinguistic word-learning experiment. We present the techniques and procedures for application of cathodal and anodal stimulation of core language areas of Broca and Wernicke in the left hemisphere of the human brain, describe the procedures of creating balanced sets of psycholinguistic stimuli, a controlled yet naturalistic learning regime, and a comprehensive set of techniques to assess the learning outcomes and tDCS effects. As an example of tDCS application, we show that cathodal stimulation of Wernicke's area prior to a learning session can impact word learning efficiency. This impact is both present immediately after learning and, importantly, preserved over longer time after the physical effects of stimulation wear off, suggesting that tDCS can have long-term influence on linguistic storage and representations in the human brain.

Introduction

The neurobiological mechanisms of human language function are still poorly understood. As the bedrock of our communication ability, this unique human neurocognitive trait plays a particularly important role in our personal and socio-economic lives. Any deficits affecting speech and language are devastating for the sufferers and expensive for the society. At the same time, in the clinic, procedures for treatment of speech deficits (such as aphasia) remain suboptimal, not least due to poor understanding of the neurobiological mechanisms involved¹. In research, the recent advent and rapid development of neuroimaging methods have led to multiple discoveries describing activation patterns; yet, causal evidence is often still lacking. Furthermore, language areas of the brain are located somewhat suboptimally for application of mainstream neurostimulation approaches which can provide causal evidence, most importantly the transcranial magnetic stimulation technique (TMS). Whereas offline TMS protocol, such as theta burst stimulation, can cause pain due to the close proximity of the muscles to the point of stimulation, "online" TMS protocols can introduce sound artifacts from stimulation, which is undesirable due to interference with linguistic stimulus presentation². Even though TMS is widely used in language studies despite such inconveniences, a welcome alternative may be provided by other stimulation methods, most notably transcranial direct-current stimulation (tDCS). In recent years, tDCS has seen a remarkable growth in its use due to its accessibility, ease of use, relative safety and often rather striking outcomes³. Even though the exact mechanisms underpinning tDCS influence on neural activity are not understood completely, the mainstream view is that, at least at low intensity levels (typically 1-2 mA for 15-60 min), it does not cause any neural excitation or inhibition per se, but instead modulates the resting transmembrane potential in a graded way towards de- or hyperpolarization, shifting the excitation thresholds up or down and thereby making the neural system more or less susceptible to modulations by other events, stimuli, states or behaviors^4,5. Whereas most of the applications reported to date have focused on the motor function⁶ and/or motor system deficits, it has been increasingly applied to higher-level cognitive functions and their respective disabilities. There has been a rise in its application to speech and language, mostly in research aimed at the recovery of post-stroke aphasia^7,8,9, even though it has so far led to mixed results with respect to the therapeutic potential, stimulation sites and hemispheres, and optimal current polarity. As this research, and particularly the application of tDCS in cognitive neurobiology of normal language function, is still in its infancy, it is crucial to delineate procedures for stimulating at least the core language cortices (most importantly Wernicke's and Broca's areas) using tDCS, which is one of the main aims of the current report.

Here, we will consider application of tDCS to language areas in a word-learning experiment. In general, the case of word learning is taken here as one example of a neurolinguistic experiment, and the tDCS part of the procedure should not change substantially for other types of language experiments targeting the same areas. Yet, we use this opportunity to also highlight major methodological considerations in a word acquisition experiment per se, which is the second main aim of the current protocol description. Brain mechanisms underpinning word acquisition – a ubiquitous human capacity at the core of our linguistic communication skill – remain largely unknown¹⁰. Complicating the picture, existing literature differs widely in how experimental protocols promote word acquisition, in control over stimulation parameters, and in tasks used to assess learning outcomes (see, e.g., Davis et al.¹¹). Below, we describe a protocol that uses highly controlled stimuli and presentation mode, while ensuring a naturalistic context-driven acquisition of novel vocabulary. Furthermore, we use a comprehensive battery of tasks to assess the outcomes behaviorally at different levels, both immediately after learning and following an overnight consolidation stage. This is combined with sham and cathodal tDCS of language areas (we make a particular example using Wernicke's area stimulation) which can provide causal evidence on underlying neural processes and mechanisms.

Protocol

All procedures were approved by the local research ethics committee of St. Petersburg State University, St. Petersburg, with consent obtained from all participants.

NOTE: All participants must sign the informed consent and fill in a questionnaire to attest the absence of any contraindications for tDCS stimulation (see Technique and Considerations in the Use of 4 x 1 Ring High-definition Transcranial Direct Current Stimulation (HD-tDCS) by Willamar and colleagues¹²) and to collect other data relevant to the study such as vision acuity, demographics, language experience and handedness. For the latter, the seminal work by Oldfield¹³ is recommended.

1. Subjects and experimental environment

1. In a typical language experiment, ensure that all subjects are right-handed and have no record of language deficits, neurological or psychiatric disorders. Their native language and bilingual/multilingual status must be controlled.
2. Conduct all measurements in a sound-proof or at least sound-attenuated chamber. Sound insulation is very important, since any extraneous sound, noise, human speech, etc. can significantly affect the performance and thus influence the data (Figure 1).
3. To avoid interference by unnecessary subject-experimenter contact, place only the screen, headphones/speakers and any input devices (keyboard, button boxes) inside the chamber. Have all interaction with the experimenter over intercom unless personal contact is required.
4. Use the following optimal parameters, based on extensive piloting, for background color and font size: grey background color (RGB: 125, 125, 125), black text color (RGB: 0; 0; 0), Arial font face, size 27.
5. To reduce delays and jitter in visual presentation, use a video card and a monitor with a refresh rate of 100 Hz and higher.
6. To measure reaction times, use research-grade response pads, which have better ergonomics and more precise timing in comparison with conventional keyboards.

2. Stimulus preparation

1. Choose words of the language in question, which are controlled for their duration, lexical frequency and overall structure (to avoid any basic effects of surface stimulus properties on higher-level processing). Here, all base words were eight phonemes/letters long and consisted of three syllables with the CVCCVCVC structure (where C is consonant, and V is vowel).
2. To create multiple lists, divide the words into sets, which should not differ statistically (as measured with, e.g., t-tests) on their lemma, bigram and/or trigram as well as syllabic frequency. These can be obtained from language-specific psycholinguistic databases; here, Russian National Corpus was used (http://www.ruscorpora.ru/en/). Here, one set was used for creating (through modification) orthographically similar novel words and pseudowords, another set for creating unrelated control pseudowords, and a further set used as unrelated control words (Figure 2A). This led to five sets of 10 items each (50 stimuli in total). Modify these procedures in accordance with your exact experimental requirements.
3. To minimize any effects of surface forms on newly acquired semantics, counterbalance the sets across the subject sample, such that they play different experimental roles for different subjects.
4. Create novel word forms such that they follow the rules of phonology and phonotactics and resemble existing words in terms of orthographic and phonological structure.
   NOTE: To make sure that the new words can enter into competition with existing words, the current procedures were based on those developed in a series of experiments by Gaskell and colleagues^11,14 and aimed at keeping the word onsets (CVCCV-) stable, while rotating their offsets (-CVC) across different items in the set. That is, we preserved the first two syllables of an existing word and varied the ultimate syllable such that a new, previously unfamiliar novel word form was created (e.g., mandarin -> mandanal*, where the last CVC was taken from another word in the list, cardinal, to create a new item).
5. Repeat the procedure described above for creating as many novel word forms as needed. For the current demonstration, we created lists of novel word forms to be learnt and of similar unlearnt pseudowords (e.g. mandarin -> mandanal*, mandaket*, all three potentially entering into a lexical competition post-learning, as neighbors) as well as further control lists of real words and novel pseudowords that did not share this similarity and thus would not produce a lexical competition with the main stimuli (e.g., circular, muskenal*; Russian examples are used throughout, transliterated from Cyrillic to Latin script for ease of understanding).

3. Sentence stimuli for contextual semantic learning

1. Create novel meanings to be associated with the new words in the process of learning. This could be made-up, obsolete or rare objects or concepts not present in the subjects' native language or culture.
2. For contextual learning of novel semantics, the procedures used by Mestrez-Misse and colleagues¹⁵ are recommended. Create several unique sentences that describe situations through which one can understand the meaning of each of the novel words (e.g., "To control insects in medieval times, people used mandaket"). Use a sequence of such sentences for each of the novel words (here, a total of 5 sentences per word), and gradually reveal the meaning of each new concept from a more general to more specific sentential context.
3. Present novel words ideally in their dictionary form (i.e., uninflected, e.g., singular nominative or accusative case in Russian), such that the surface form is not inflected differently in different sentences (Table 1), unless inflection rule learning is also required.
4. Control and balance the length of the sentences and the number of words between conditions. Here, each sentence consisted of 8 words. Always place novel words at the end of the sentences. Such placement allows build-up of necessary contextual information (further, this allows implementing this design, if needed, in an EEG or MEG setting to record evoked brain responses unmasked by further word stimuli).
5. Present word-specific sentences in word-specific sub-blocks, gradually revealing the meaning of each new word, without interleaving or randomizing sentences related to different novel words.
6. Randomize the order of the sub-blocks across the subject group. Word-by-word sentence presentation is recommended if the visual modality is used.
7. Determine the interstimulus interval based on specific stimulus properties to allow their convenient presentation (Figure 2B); make sure to separate different sub-blocks with additional intervals and give regular breaks.

4. Tasks to assess acquisition of new word forms and novel meanings

NOTE: Use several tasks to assess different levels of acquisition and comprehension of both surface word forms and lexical semantics. Five tasks are used in the present protocol: free recall, cued recognition, lexical decision, semantic definition and semantic matching. The tasks are applied in the order they are listed below, which was optimized to reduce any carryover between successive tasks.

1. In the free recall task, have each participant reproduce as many new word forms as they could remember by typing them into the prepared spreadsheet. The instruction is as follows: "Please write down in the column all the new words that you can remember."
2. Include the same stimuli in the recognition and lexical decision (second and third tasks, respectively) and use the same presentation rate.
   1. These tasks include all items (novel words, real competitor words the novel ones are derived from, untrained pseudoword competitors derived from the same real words, unrelated control pseudowords and unrelated control existing words).
   2. For the recognition task, use the following instruction: "You will be presented with words sequentially. Press button "1" with the middle finger of the left hand if you have encountered the word during the experiment, or press "2" with the index finger of the left hand if you have not." Modify the response coding, hand and fingers in accordance with your specific requirements.
   3. The instruction for the lexical decision task is: "You will be presented with real and meaningless words sequentially. Press "1" with the middle finger of the left hand if the word makes sense, or press "2" with the index finger of the left hand if it does not." Modify these as necessary.
3. Use the semantic definition task to estimate the acquisition of novel meaning and the correspondence between the meaning and the surface form.
   1. Give participants a list of the learnt items (i.e., those presented previously in the learning phase) with the instruction above: "Here is a list of new words presented to you previously. Try to define each of them and type their definitions into the spreadsheet".
   2. To assess the completeness and accuracy of the given definitions, engage independent experts to rate the responses; agreement between experts could be tested using, e.g., Kendall's coefficient of concordance (W).
4. Use semantic matching task to assess the acquisition of semantics through making explicit links between the newly learnt word forms and their meanings in a simplified manner.
   1. Use the following instruction: "You will be presented a word and three definitions. You should choose one correct definition for each word by pressing the corresponding button". Only one of the definitions is correct, with the other two corresponding to the other novel items. In addition to the three optional definitions, including "none of this" or/and "not sure" options is also recommended.

5. Procedures

1. Ensure that the tDCS stimulation precedes the behavioral task it is intended to modulate.
   1. Wernicke's area.
      NOTE: The stimulation electrode placement that best corresponds to Wernicke's area is CP5 according to the extended International 10-20 system for EEG^16,17.
      1. To locate this location in the absence of an electrode cap, follow the standard 10-20 system procedures.
      2. Measure the head with a tape from the inion to the nasion, and note the middle of this distance. Then, measure the distance from the left preauricular point to the right preauricular point, and mark the crosspoint of the two measurements.
      3. To find the CP5 location, measure 30% of the distance between the preauricular points from the crosspoint down the left hemisphere and mark it. Measure 10% of the distance between the inion and the nasion from the marked point to the back of the head. This point is the CP5 location for the active electrode (Figure 3).
   2. Broca's area
      NOTE: Closest to Broca's area is the F5 electrode site¹⁸ according to the 10-20 system.
      1. In the absence of an EEG cap, follow the standard 10-20 system procedures to find and mark the crosspoint between inion-nasion and preauricular points, as described above.
      2. To find the F5 location, measure 20% of the distance between the inion and the nasion from the crosspoint to the front of the head. Measure 30% of the distance between the preauricular points from the recently marked point down the left hemisphere. This point corresponds to the F5 location for the active electrode (Figure 3).
   3. Homologous locations in the right hemisphere: for right-hemispheric homologues of Wernicke's and Broca's areas, use the same procedures as above, with the exception of measuring the distance from the midline down the right side of the scalp. Electrode locations are: CP6 for the RH Wernicke homologue and F6 for the Broca homologue.
   4. Use spongy electrodes measuring 5 cm x 5 cm as this size is a good compromise between focal stimulation (which causes more irritation and discomfort) and larger electrodes that lack focality. Soak the electrodes in physiological saline solution for 5 min before application.
   5. In order to minimize the effect of stimulation on other areas of the brain, place the reference electrode at the base of the neck on the left (right for homologues) side (see Figure 3 and Figure 4). Use spongy electrodes measuring 5 cm x 5 cm as well.
      NOTE: Particular attention should be paid to preventing the spreading of the solution beyond the boundaries of the electrode application zone. Special care should be taken to keep the surrounding electrode area dry.
   6. For optimal cathodal stimulation, use 1.5 mA current for 15 min. At the onset, the current gradually rises from 0 to 1.5 mA over 30 s, and at the end of the stimulation it drops back to zero over 30 s.
   7. For anodal stimulation, use the same procedure as cathodal stimulation, except the polarity is reversed, and the anodal electrode is placed at the active site, while the cathode is used as the reference electrode located outside the scalp area.
2. Sham stimulation
   1. Perform the sham stimulation procedure generally as described above except that the current is only applied briefly in the beginning and the end of the sham session. To this end, during the first and the last 30 s of the session, apply an electric pulse of a triangular shape with a maximum of 1.5 mA, as used in the present protocol.
3. Main behavioral task: contextual semantic learning
   1. Present sets with contextual sentences for the novel words in a random order. Start each sentence with a word-by-word presentation.
   2. After this, display the entire sentence on the screen to ensure its full understanding. Have participants press the spacebar with the index finger of the left hand after reading the whole sentence. Duration of sentence presentation is 5000 ms.
      NOTE: The sets of the sentences are separated from each other by appearance of three crosshairs ("+++") for 2000 ms. Each new concept presentation starts with a single fixation cross ("+") present for 500 ms before the sentence words are flashed. Each word is presented for 500 ms, and the empty screen in the background color between words within one sentence is 300 ms long.
4. Acquisition assessment procedure
   1. To assess learning effects both immediately and following the overnight consolidation stage, break the stimulus set into two subsets, equally distributed across stimulus conditions and counterbalanced across the subject group, and run the assessment task immediately after the learning protocol on one subset, and after a 24 h delay on the other one.
      NOTE: This strategy is based on the literature that highlights the importance of overnight memory consolidation for the acquisition of new words^19,20.
   2. Use all developed tasks in the order described in section 3 above to assess different levels of word/concept acquisition. Choose the order of the tasks to minimize any carryover effects from one task to the following ones.
   3. For Tasks 1 and 4 use spreadsheets to be filled by subjects (by hand or using a text or spreadsheet processor); present the other tasks using temporally precise simulation software.
      NOTE: Each stimulus in Tasks 2 and 3 is presented for 600 ms, with a fixation cross ("+") present in the interstimulus interval (1400 ms); see Figure 3. For the other tasks the response time is not limited.

6. Data analysis

1. Perform data analysis using different tests comparing two sets of samples coming from continuous distributions (such as Wilcoxon signed rank test or Mann-Whitney U-test) or medians (two-sample t-test, if the distribution is normal).

Results

While the data were analyzed for the specific set of tasks, it should be emphasized that the developed set of tests and the paradigm could be adapted to a variety of psycholinguistic experiments. The results were analyzed in terms of accuracy scores (number of correct answers) and the reaction time (RT) using non-parametric Wilcoxon signed rank test and Mann-Whitney U test across groups (cathodal and sham stimulation conditions). Significant differences for tasks within each group are pre...

Discussion

The results highlight a few important points that need to be taken into account when conducting psycholinguistic research in general, and neurolinguistics tDCS studies in particular. Stimulation of language cortices (exemplified here by Wernicke's area) produces a complex pattern of behavioral outcomes. Unlike the TMS technique, where it is possible to fully disrupt speech processing (e.g., the so-called "speech arrest" protocol)²¹, this method enables a possibly more complex, graded a...