Participants

Ethics permission came from the Ethics Commission of the German Society of Psychology (SK 012013_5) according to the principles expressed in the Declaration of Helsinki. MRI time was to be provided by the Max Planck Institute for Human Development, so the study took place in Berlin. Interviews were to be led by RTH, the (English speaking) creator of DES, so we sought native English speakers. We recruited them by telephone from the databank of potential volunteers maintained by the Max Planck Institute. Informed consent was obtained both in writing and (repeatedly) orally. Given the substantial time and scanner commitment of this exploratory study, we determined a priori that we could apply the procedure to no more than seven individuals in the time available, so we recruited seven healthy volunteers. Two withdrew after the first sampling day, one for scheduling difficulties and one for unspecified reasons. The remaining five completed all phases of the study. All participants had normal or corrected-to-normal vision. No participant had a history of neurological, major medical, or psychiatric disorder. The participants (3 women and 2 men) had a mean age of 22.4 (ranging from 18 to 30). Four were right-handed; one male was left-handed (with laterality index -87.5 [30]).

Design and overview

Each of the five participants was scheduled individually for 19 sessions, generally across a two-week period, which were divided into four phases as illustrated in Fig 1. In Phase 1 (in-scanner elicited phase, the first day of participation), we fully explained the study, administered a short battery of questionnaires not relevant to the present report, and familiarized the participant with the MRI scanner and procedures. The participant entered the scanner, where we conducted a 10 min structural scan and a 5 min resting state scan according to standard MRI research procedures (instruction: “please close your eyes and relax, without falling asleep”). Then we presented the elicitation task (adapted to include inner speech items from a recent fMRI imagery paradigm used by Belardinelli et al. [31]), where we visually presented short written prompts to imagine seeing (e.g., “to see a pencil”), saying (e.g., “to say ‘elephant’), hearing (e.g., “to hear a tinkling”), feeling (e.g., “the feeling of anxiety”), or sensing something (e.g., “the sensation of shiver”). Each category of prompt (seeing, saying, etc.) included 8 potential stimuli; the complete list of stimuli is in S1 Table. As described in Belardinelli et al. [31], the stimuli were presented in mini-blocks consisting of four prompts of one of the five categories, each shown for 7 s with 1 s inter-stimulus interval (= 32 s in total). Thus one mini-block consisted of four seeing prompts; another mini-block consisted of four saying prompts; and so on. Participants were instructed to vividly imagine what was requested by the prompt for the duration of the presentation of the prompt (7 s). After four prompts, a fixation cross was shown for 19 s before the next mini-block of prompts was presented. Each participant was presented with two mini-blocks for each of the five categories.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Schematic of the experimental design.

https://doi.org/10.1371/journal.pone.0147932.g001

In Phase 2 (natural-environment spontaneous phase, typically beginning the first day and occupying the remainder of the first week), we instructed the participant in the use of the DES beeper and the sampling task [5, 18]: participants wore the beepers in their natural environments for approximately three hours, during which they would hear (through an earphone) six randomly occurring beeps. Immediately after each beep, the participant jotted down notes about whatever inner experience happened to be ongoing (was “in flight”) at the moment of the onset of the beep. Later that day or the next day, the participant returned for a DES expositional interview about those six beeped experiences; this interview was conducted by RTH and at least one other member of the research team including SK, BA-D, and/or CF as well as others. The expositional interview was “iterative” [5, 32], designed to increase skill in apprehending and describing inner experience across sessions. The participant then completed three more natural-environment sampling periods, each involving six random beeps, and each followed by an expositional interview. If communication about phenomena did not seem adequate after four sampling/interview days, additional sampling and interview sessions could be conducted; we invoked this option with one participant for one additional sampling period and expositional interview.

In Phase 3 (in-scanner spontaneous phase, typically occurring during the second week), the participant entered the scanner for a 25-min session with resting-state instructions (“keep your eyes open, stay relaxed and calm”). Otherwise, the participant’s experience was free to roam. At four quasi-random times, the participant received a DES beep through an MR compatible headphone (Visuastim), and then immediately jotted notes about the ongoing-at-the-moment-of-the-beep-onset experience on a clipboard positioned on the participant’s lap (viewable through a mirror). Immediately after exiting the scanner, we (RTH and at least one other interviewer, including SK, BA-D, and CF) conducted a DES expositional interview with the participant about the experiences that had been ongoing at the four randomly beeped moments. Thus the in-scanner sampling procedure was designed to be similar to the natural-environment sampling procedure. This sequence (25 min scan / four random beeps / expositional interview) was repeated a total of nine times (typically twice a day during the second week), resulting for each participant in 4 × 9 = 36 quasi-random samples of experience occurring in 25 × 9 = 225 minutes of fMRI scanning.

In Phase 4 (final phase), we administered a short battery of questionnaires not relevant to the present report. The participant was then candidly debriefed.

Phenomenon selection

When an investigation aims at naturally occurring phenomena, the investigators cannot decide in advance what phenomena to study but must let the data dictate what will be (actually, already has been) investigated. In the present study we were prepared to study any of the five phenomena supposedly elicited by our modification of the Belardinelli et al. [31] tasks: seeing, saying, hearing, feeling, or sensing something. Sensory awareness was the most frequent in-scanner spontaneous phenomenon, occurring in 59 percent of the 180 samples; however, its varying modality (visual, bodily, auditory, etc.) made it an unlikely candidate for correlation with brain regions. Inner seeing was second most frequent in the scanner, occurring in 35 percent of the 180 in-scanner samples; however, inner seeing had zero frequency in two of the five participants’ natural environment DES sampling, and the correlation between inner seeing frequency in the natural environment and in the scanner for our participants was essentially zero (-.08); thus it seemed likely to be an artifact of the scanner situation. Inner speaking was next most frequent, occurring in 29 percent of the in-scanner samples; it occurred in all participants’ natural environment DES sampling and on at least 5 occasions for each participant in the scanner, and there was a high (.73) correlation across participants between the frequency of natural environment and in-scanner inner speaking, so inner speaking seemed a good candidate for further consideration. Inner hearing was rare (it occurred with appreciable frequency for only one participant, and occurred zero times or twice for three participants). Feelings were also rare (occurring in the scanner on only one occasion for three of the participants and never for the other two). Thus, as a result of the data that occurred, we focus here on inner speech-related neural processing, because there were enough such events to allow meaningful investigations.

Scanning procedure

Images were collected on a 3T Magnetom Trio MRI scanner system (Siemens Medical Systems, Erlangen, Germany) using a 32-channel radio frequency head coil. Structural images were obtained using a three-dimensional T1-weighted magnetization-prepared gradient-echo sequence (MPRAGE) based on the ADNI protocol (www.adni-info.org) (repetition time [TR] = 2500 ms; echo time [TE] = 4.77 ms; TI = 1100 ms, acquisition matrix = 256 × 256 × 176, flip angle = 7°; 1 × 1 × 1 mm voxel size). Functional images were collected using a T2*-weighted echo planar imaging (EPI) sequence sensitive to blood oxygen level dependent (BOLD) contrast (TR = 2000 ms, TE = 30 ms, image matrix = 64 × 64, FOV = 216 mm, flip angle = 80°, voxel size 3 × 3 × 3 mm³, 36 axial slices, interleaved data acquisition).

fMRI data pre-processing and main analysis

The fMRI data were analyzed using SPM8 software (Wellcome Department of Cognitive Neurology, London, UK). The first four volumes of all EPI series were excluded from the analysis to allow the magnetization to approach a dynamic equilibrium. Data processing started with slice time correction and realignment of the EPI datasets. A mean image for all EPI volumes was created, to which individual volumes were spatially realigned by means of rigid body transformations. The structural image was co-registered with the mean image of the EPI series. Then the structural image was normalized to the Montreal Neurological Institute (MNI) template using the segmentation-normalization procedure in SPM8, and the normalization parameters were applied to the EPI images to ensure an anatomically informed normalization. A commonly applied filter of 8 mm FWHM (full-width at half maximum) was used. Low-frequency drifts in the time domain were removed by modeling the time series for each voxel by a set of discrete cosine functions to which a cut-off of 128 s was applied. The statistical analyses were performed using the general linear model (GLM).

The elicitation task was modeled (following Belardinelli et al. [31]) as consisting of blocks with duration of 32 s length. In the spontaneous phase, the beeps of the DES procedure (i.e., instances of spontaneous inner speech) were modeled as events with a stick function on the onset of the beep and a duration of 0 s. Taking into account the standard 3 to 5 s delay in the hemodynamic response, this allowed us to model events in the brain in the seconds immediately prior to the moment of the beep. We also tried modeling with stick functions 1 s and 2 s before the beep onset and durations of 1 to 2 s for the “window” surveyed at each beep; these results were very similar to those reported below. These vectors were convolved with a canonical hemodynamic response function (HRF) and its temporal derivatives. Then, for the DES procedure, regressors were built coding the categories that raters (RTH with at least one other) assigned to the individual’s 36 events.

ROI analysis

To investigate inner speech, we selected two anatomical ROIs taken from the Automatic Anatomical Labeling (AAL) atlas [33]; left inferior frontal gyrus (IFG), because of its association with speech production in right handers, and Heschl’s gyrus, i.e. primary auditory cortex. Although some studies have focused on specific subsections of the IFG such as BA 44 (e.g., [2]), we included the whole IFG based on the fact that previous investigations of inner speech have often included multiple Brodmann Areas [8, 27, 34]. We used right IFG for the left handed participant, whose laterality index was -87.5 and who clearly showed contralateral activation during speech production in a whole-brain analysis. Furthermore, the results are similar if we exclude this participant. Separately for each participant, each anatomical ROI, and each condition, the beta weights from the model detailed above were extracted [35] and used for further analysis. We chose these ROIs because any difference between spontaneous and elicited inner speech might show up as a different patterning of activation across either or both speech production and auditory perception areas. The results presented below are largely independent of the exact definition of the auditory ROI, because similar results were found when using a ROI spanning the functionally defined primary auditory cortex.

DES coding

Within 24 hours of each DES expositional interview, one of the interviewers wrote a description of each of that day’s samples. These descriptions were then circulated to the others for comment, with any disagreement resolved or left as an explicit disagreement, usually within 48 hours of the original interview. RTH and at least one additional person present at the interview independently judged whether any (or several) of the five phenomena designed to be elicited by the Belardinelli et al. [31] tasks (seeing, saying, hearing, feeling, or sensing something) was present at each sample; discrepancies were resolved by discussion. These ratings were thus constrained by the descriptions: the descriptions had been written and agreed to before the rating process began. However, the ratings were not intended to be ratings of the written descriptions; they were intended to be ratings of the experiences that had occurred at the moment of the beep. On a few (approximately six) occasions, the rating procedure highlighted an ambiguity, omission, or misleading aspect of the written description, in which case a return to the interview videotape might result in a revision of the written description. All such adjustments of the written descriptions were performed blind to the brain activations and analyses thereof. Idiographic characteristics (those that emerged distinctively for particular participants) were also described, but are not discussed here. Ratings for each participant are found in S2–S6 Tables.

Neurophysiological considerations

We observed that elicited inner speaking caused a decrease in activation of Heschl’s gyrus, in accord with prior elicited-inner-speech research and with some understandings of external speech, which hold that when we talk, there is a reduction in sensitivity to our own speech sounds [36]. External speaking is often associated with the dampening of neural responses to expected sounds in auditory cortex [37, 38], and similar effects are evident during silent articulation [39], suggesting that articulatory processes are sufficient to induce auditory suppression even in the absence of external speech. Similarly, in the silent-speech case, it may be that elicited inner speech involves a sufficient level of articulation—albeit silent—to inhibit responses in Heschl’s gyrus.

By contrast, we found that spontaneous inner speaking was associated with an increase in Heschl’s gyrus activation. That is the more surprising because inner speaking is phenomenologically speaking, not hearing [6]. One interpretation of these results is that perhaps Heschl’s gyrus is more involved in representing speech than is usually credited. Though more commonly linked with non-verbal auditory imagery [40], Heschl’s gyrus activation is also evident during silent lip-reading [41] and in some cases of auditory verbal hallucinations [42]. Surrounding areas of middle and superior temporal cortex often activate during voice imagery tasks [17, 43], and similar processes could have driven the activation patterns observed here.

Alternatively, the activation of left IFG by task-elicited inner speech may reflect more the elicitation task demands than the speech itself. Though classically linked to speech production, left IFG is also thought to be integral to planning and execution of hierarchical sequences and complex actions more generally [44, 45]. Task-elicited inner speech, unlike spontaneous inner speaking, also involves the reading of the command, its decoding, and the acquiescence thereto as well as the covert speaking of the commanded words. That we did not find evidence of IFG involvement during the spontaneous inner speech task may be the result of insufficient power to detect IFG involvement or the result of a cognitive difference between elicited and spontaneous inner speaking.

The small-n of this study makes it possible that our 5 participants were an unusual group, not representative of the wider population, and so our neurophysiological findings may have limited generalizability. However, this is made unlikely by two observations. First, our results did replicate the usual findings for task-elicited inner speaking, so our participants are likely to be not grossly dissimilar from the wider population. Second, the Heschl’s gyrus reversal (decrease when elicited, increase when spontaneous) was found for each of the five participants individually. Our participants were a heterogeneous group: male and female, right and left-handed, ranging in age from 18 to 30, all unacquainted with each other. They were not, seemingly, a very specialized group. On the other hand, they were all Caucasian native English speakers, so we can make no cross cultural claims.

Methodological considerations

The importance of this study arises as much from the exploration of method as from the exploration of neurophysiology. We turn to that discussion now.