Overview

This study established an anatomical correlate in the white matter of the brain for SPS individuals: those with the highest neuroticism-adjusted HSP-proxy scores showed the greatest RD, FA, and MD effects in several neocortical areas compared to those with the lowest HSP scores (not highly sensitive). The interpretation of the direction of the effects, negative or positive correlations, must be limited only to identifying a locus of change (see Microstructural differences: the physiological impact of positive MD/RA and negative FA correlations). Whole-brain, exploratory effects were greatest in brain regions involved in (a) higher order emotion and reward processing: the ventromedial prefrontal cortex; and (b) in regions involved in empathy, self-other processing, attention and flexible coding of the environment: the premotor cortex and supramarginal gyrus. Compared to whole brain results, smaller ROI effects were also seen in self-other processing areas (IFG, precuneus, fusiform, angular gyrus), higher order visual processing regions of the ventral and dorsal pathways (precuneus, inferior parietal, temporoparietal junction, STSb), and primary sensory processing areas, such as the lateral occipital and transverse temporal (primary) auditory cortex.

All of these results are consistent with behavioral observations for SPS, which include sensory sensitivity, a tendency to be overwhelmed by sensory stimuli, and more attention to emotional and visual details of stimuli than others who do not show these traits (Aron and Aron 1997; Aron et al. 2012). The localization of the higher order visual processing effects is also consistent with the previous studies, as reviewed in the Introduction, that used functional MRI to assess details of visual scenes and emotional reactions in highly sensitive people (Jagiellowicz et al. 2011; Acevedo et al. 2014, 2017). Thus, regional functional effects previously described were confirmed by microstructural effects. In addition, this is the first study to suggest involvement of primary sensory processing areas in the cortex, such as the visual and auditory cortex. The study also highlights the premotor cortex, with its connections to the supramarginal gyrus and attention functions. Somatosensory as well as environmental stimuli like vision are new possible highlights. Furthermore, the study suggests a novel focus on the functions of the ventromedial prefrontal cortex. Finally, this is one of few studies to show small DTI anatomical effects within axonal tracts for normal-range behavioral traits, in this case sensory processing. Other studies have investigated neurologically normal subjects for DTI effects, mostly using the “big five” personality traits (Xu and Potenza 2012), but these traits overlap very little with the HSP Scale, other than with neuroticism (which is controlled for in this study). In this study of a normal trait, the results indicate novel behavior-related brain regions to explore in future studies.

Ventromedial prefrontal cortex

The largest effects were in white matter of the ventromedial and ventrolateral prefrontal cortex (De La Vega et al. 2016), on both right and left sides, but with more voxels on the right. These tracts connect major limbic system components: the hippocampus/peri-hippocampal cortex and amygdala to the medial prefrontal cortex. The subcallosal cingulum area, functionally connected with the ventromedial and ventrolateral prefrontal cortex (Dunlop et al. 2017), is known for its influence on mood, especially depression (Mayberg et al. 2013; Dunlop et al. 2017). The cingulum is also structurally connected to several of the ROIs that were correlated with HSP-proxy scores in this study: the precuneus, bank of the superior temporal sulcus, and supramarginal gyrus (Bathelt et al. 2019).

The ventromedial and lateral prefrontal cortex areas are described by Hiser and Koenigs (2018) as having three broad domains of psychological function: decision-making based on reward and value (Sescousse et al. 2013); generation and regulation of negative emotions; and, social cognition, such as facial recognition, theory of mind, and processing self-relevant information. These of course interact strongly, especially given the social nature of humans, as in the function of the vmPFC in making moral decisions (Cameron et al. 2018) or in the “intuitive feeling of rightness” that guides decision-making, often social in nature, as a function of memory retrieval (Hebscher and Gilboa 2016).

The major results in the vmPFC in this study, as well as in the preceding fMRI studies (Acevedo et al. 2014, 2017), point to perhaps the most important aspect of SPS, which is “depth of processing” (Aron et al. 2012). The term is based on cognitive conceptualizations of levels or processing, with the idea that processing to deeper levels with more detailed cognitive contexts leads to better memory and better learning overall (Lockhart et al. 1976; Leow 2018). The hypothesized evolutionary development of depth of processing in SPS is based on a computer simulation demonstrating that unusual responsivity to the environment will evolve when there are enough payoffs for an individual difference in noticing details, as long as most individuals do not notice these details (Wolf et al. 2008). (If all individuals did, there would be no special benefit.) That is, SPS is considered fundamentally an individual difference in “depth of processing” through careful observation of situation/time A to compare those details in memory to situation/time B and gain any potential benefits others miss. Individuals without the trait are thought not to process A as carefully, a strategy which is often equally or more effective, since B may bear no resemblance to A or there may be little reward in noticing any resemblances.

This type of careful processing relies on emotional motivation, the desire for rewards, such as winning, and the desire to avoid fear-related stimuli, such as losing (Baumeister et al. 2007). Hence, understanding SPS as fundamentally about depth of processing for decision-making based on reward value and social value is consistent with the significant differences that were found in the vmPFC, with its close connections to emotion-related memory processing areas such as the hippocampus and amygdala. Memory enhanced by motivation is, again, key to SPS, and specific activations occur for those high in SPS when processing emotional stimuli. For example, Acevedo et al. (2017) found in their comparison of responses to positive and negative stimuli that there were considerable differences in regional brain activation for high and low SPS in the vmPFC, in the same areas where microstructural differences were found in this study (Fig. 2), and in regions that mediate memory, attention, awareness, and reflective thinking.

Theory of Mind

In a meta-analytic review of the Theory of Mind, Mar (2011) highlighted a core mentalizing network: the mPFC, precuneus, bilateral pSTS, bilateral angular gyri, and the right IFG. These structures showed microstructural differences associated with SPS in this study, notably the mPFC, right IFG, precuneus, the bank of the STS, and angular gyrus region. Second, this mentalizing network overlaps with the narrative comprehension network in a number of areas, including the mPFC, bilateral pSTS/TPJ, precuneus, and possibly the right IFG, again areas implicated by this study. Theory of Mind is central to the core concept of SPS, in that this survival strategy would require more reflection on another’s behavior to accurately see from their perspective and correctly attribute to them their motivations and intentions.

Premotor cortex, attention, and somatosensory processing

Finally, regarding the extensive cluster from the left premotor cortex to the post-central somatosensory and supramarginal gyrus, it could be expected that the arc of axonal effects we found in sensory processing areas from posterior to anterior (Figs. 2B, 4A, 5A) would include the premotor cortex, where the hypothesized deep processing associated with SPS would result in the preparation for action. This potential somatosensory/posterior parietal/premotor cortex involvement in SPS is a novel contribution to its understanding and may be helpful in future studies.

Fig. 5

Primary sensory, higher order sensory processing, and cognitive processing regions were identified as correlated with the HSP-proxy scores. Freesurfer-based ROIs are shown for white matter where correlation coefficient effect size was − 0.165 to 0.148, rendered in 3D on a template brain. Areas include the general path of the dorsal and ventral visual pathways. A Sagittal view of the right side. B Posterior view. Red: MD positive correlation with HSP-proxy questionnaire scores. Green: RD positive correlation. Blue: FA negative correlation. Gold: MD and RD positive correlation. Aquamarine in IFG: RD and FA correlations. *p < 0.05. Otherwise, p values were 0.06–0.09. See Table 3. A primary auditory cortex, F fusiform, IFG inferior frontal gyrus, IP inferior parietal including angular gyrus/temporoparietal junction/supramarginal gyrus, LO lateral occipital cortex, P precuneus, STS B-bank of the superior temporal sulcus, MD mean diffusivity, RD radial diffusivity, FA fractional anisotropy

The importance of this broad premotor area continues to evolve (Rizzolatti et al. 1987). Rizzolatti et al. suggested a theory of attention focused on this area, presenting evidence that the premotor areas were the source of attention rather than a separate attention-directing mechanism. That is, attention is turned to a stimulus within the premotor area, so that attention consists of nothing more than preparation for a motor activity (e.g., an eye movement toward a stimulus deemed important or auditory areas preparing for a sound). Although refinements and extensions (Wollenberg et al. 2018) have occurred, this view of attention is still tenable, according to experiments by (Schubotz and Von Cramon 2003).

Schubotz and Cramon distinguished the left premotor cortex, correlated in our study with SPS, as associated with nonspatial tasks and rapid acquisition of new motor sequences. Overall, evidence regarding the premotor area suggests “environmental features do not have to remind us of specific actions or movements to induce premotor activation on a more or less conscious level. Rather, features are represented in a highly fragmented format that allows for instant recombination and very flexible coding of any currently attended environment” (p. 126). Schubotz (2007) presented considerable evidence that the premotor area (along with most of the areas in the brain associated with SPS) helps in the prediction of events.

Furthermore, a mean diffusivity study by Takeuchi et al. (2019) found that an area including the premotor cortex plays a major role in emotional salience and empathy. This area was also activated in the (Acevedo et al. 2014) fMRI study finding empathy for happy and distressed partners and strangers.

Attention, flexible coding, prediction, somatosensory processing, and empathy all fit with the theory that SPS involves attending to subtle stimuli that may be relevant for survival and predicting those environmental details that what will be relevant in future environments. Overall, these whole brain, exploratory analyses provide a picture that is consistent with behavior associated with SPS.

ROI results

The particular ROIs with clear statistical significance were the left precuneus (RD, r = 0.148, p = 0.030), part of the dorsal visual stream, and the right parsopercularis/IFG (FA, r = − 0.165, p = 0.012), associated with empathy. For the precuneus, the other side was significant too (RD, r = 0.144, p = 0.038), while for the IFG, the same side was marginally significant for RD, as well (r = 0.137, p = 0.053). Also, the transverse temporal gyrus/primary auditory cortex (A1), part of the dorsal auditory stream was marginally statistically significant (MD, r = 0.132, p = 0.072).

As for the precuneus, fMRI studies (Cavanna and Trimble 2006) of healthy subjects suggest that it plays a major role in visuo-spatial imagery, episodic memory retrieval, and self-processing tasks, such as the experience of agency and taking the first-person perspective. All of these activities are more prominent in those high in SPS, and the precuneus was another area often correlated with SPS in fMRI studies. There is also a hypothesized role for the precuneus in consciousness itself (Cavanna 2007), along with areas nearby in the posteromedial parietal cortex. It is especially active during the conscious resting state, but is deactivated when consciousness decreases (e.g., sleep, anesthesia, Alzheimer’s disease). Indeed, it has been proposed that it is part of a larger network that correlates with self-consciousness, as it engages in self-related mental representations, self-reflection, and autobiographical memory retrieval. Meditation, which creates states of restful alertness, is associated with microstructural differences in the precuneus in practitioners compared to controls (Shao et al. 2016; Avvenuti et al. 2020). Without suggesting that SPS somehow results in more consciousness, it may well demonstrate an internal tendency for more awareness and integration of diverse aspects of inner and outer experience. Although the insula was not a factor in this study, it has a similar role in the brain and was found to be more active in the fMRI studies of SPS already cited, and has also sometimes been described as the “seat of consciousness” (Craig 2009).

The right parsopercularis/IFG region was negatively associated with HSP-proxy scores and FA (p = 0.02), and positively associated with RD (p = 0.07). IFG functions may be particularly important to recognize in further studies. In an fMRI study, the IFG region was positively associated with HSP scores during positive emotion conditions while looking at a spouse or stranger (Acevedo et al. 2014). It has been identified with a mirror neuron system (Iacoboni et al. 1999; Jabbi and Keysers 2008; Van Overwalle and Baetens 2009) that responds to the movements of others, and may facilitate the understanding of others’ intentions and feeling of empathy. We previously suggested that this system’s activation is consistent with HSPs’ bias toward noticing positive expressions in others and high empathy (Acevedo et al. 2014).

Some of the ROI effects are along the dorsal and ventral visual/auditory pathways, which is especially noteworthy and we speculate that these pathways contribute to depth of processing in SPS. These pathways are described as identifying the “what” (ventral stream) and “where” (dorsal stream) of what is seen and heard, taking sensory experience beyond its initial input (Milner and Goodale 2008). The ventral stream in particular is associated with object recognition and form representation, the “what,” and is strongly connected to the medial temporal lobe, which stores long-term memories; the limbic system, which controls emotions; and joins with the dorsal stream, which identifies the “where.” The dorsal stream is said to guide actions and recognize where objects are in space. It stretches from the primary visual cortex in the occipital lobe into the parietal lobe. It contains a detailed map of the visual field and serves to detect and analyze movements. Thus, it commences with purely visual functions, ending with spatial awareness at its termination. As with the ventral stream, processing of sensory input along the dorsal stream becomes “deeper” or more elaborate. It ends up contributing to recognizing spatial relations, body image, and physical coordination. Again, as SPS has been described, the trait is not characterized by better initial sensory perception, better hearing or eyesight, but by more complete processing of what is perceived along these two visual/auditory pathways. However, the potential involvement of primary visual, auditory, and somatosensory areas suggested in this study leads to other questions for study of the most basic sensory detection and discrimination functions of these areas that may impact the higher order processing regions.

The A1 cortex that we included as an ROI for this study is thought to operate very early in the recognition of sounds. For example, a study by Warrier et al. (2009) found that non-Mandarin-speaking subjects who could successfully form an association between Mandarin Chinese “pitch patterns” and word meaning were found to have transverse temporal gyri (A1) with larger volume than subjects who had difficulty learning these associations. Successful completion of the task also was associated with a greater concentration of white matter in the left A1 of the subject. In general, larger transverse temporal gyri seemed to be associated with more efficient processing of speech-related cues, which could aid the learning and perceiving of new speech sounds. The A1 cortex is also associated with inner speech, what Hurlburt et al. (2016) might be considered a more advanced level of processing, but still preceding speech production. Although, to date, there are no studies of auditory functioning associated with SPS, it would seem to be a fruitful area for future research.

The superior temporal sulcus (STS bank, p = 0.06) is seen primarily as an area for higher visual processing. Hein and Knight (2008), in a review of carefully selected fMRI studies, concluded that the majority of findings implicate the STS in broader tasks involving theory of mind, audiovisual integration, motion processing, speech processing, and face processing. They conclude that rather than trying to pinpoint where in the STS these occur, it is best to view the function of the STS as varying according to the nature of network coactivations with different regions in the frontal cortex and medial temporal lobe during a particular task. This view is more in keeping with the notion that the same brain region can support different cognitive operations depending on task-dependent network connections, emphasizing the important role of network connectivity analysis in neuroimaging. It is consistent with current hypotheses about SPS that those high in SPS would show greater microstructural differences in an area associated with diverse types of processing (motion, speech, face, and audiovisual) as well as theory of mind.

Microstructural differences: the physiological impact of positive MD/RA and negative FA correlations

The physiological impact and thus psychological effects of positive MD/RA and negative FA correlations in normal brain are unclear (Soares et al. 2013). Therefore, the interpretation of negative or positive DTI effects must be limited to identifying only a locus of change in SPS rather than identifying higher or lower speed of processing in SPS. For example, a negative FA correlation in one-fiber system may indicate release of activity in a target area that increases depth of processing. It is different from disease states.

Indeed, the findings in this study are smaller than those seen in previous studies of disease progression or aging (Voineskos et al. 2012; Nir et al. 2013). There is no neurological or behavioral pathology in the group that we studied. The psychological traits measured are subtle and part of the normal range of human behavior. Thus, the effects are part of a normal variability in the population, but may be markers of slight anatomical differences in axon size and organization, perhaps impacting the speed of communication among regions (Horowitz et al. 2015).

Explaining complex human behavioral traits with imaging-derived biomarkers remains challenging (Young et al. 2020), because different microstructural features can contribute to a similar signal profile. Therefore, it is not possible to exclusively attribute a particular processing effect to the observed statistical negative or positive correlations. However, animal studies in traumatic brain injury (TBI) showed reduced FA and increased MD at the impact site due to demyelination and edema (Bigler and Maxwell 2012; Pasternak et al. 2016). Decreased FA can be the result of decreased diffusion hindrance as well as axonal loss (Harsan et al. 2006; Budde et al. 2011), while an increase in RD and unchanged AD was observed during myelin loss alone (Song et al. 2002; Roosendaal et al. 2009; Stricker et al. 2009), which increases MD as MD is the weighted average of AD and RD. Results from the ENIGMA group showed widespread decrease in FA and increases in MD and RD (Young et al. 2020) among people with schizophrenia compared to healthy controls. However, we cannot state that relatively higher HSP score volunteers are affected by edema or loss of myelin at the particular locations, because their MRIs were normal. Again, it is the locus of a change that is important; a decrease in FA does not indicate schizophrenia or any other mental disorder, either because the anatomical distribution of the changes is so different.

This study joins others that have looked at microstructural correlates of normal individual differences, such as personality traits (Xu and Potenza 2012) and cognitive abilities (Bathelt et al. 2019). Since such phenotypic differences can be caused by multiple genetic and environmental effects, looking for common microstructure may be another useful way to identify such differences, even though they may be small.

Limitations

A limitation of the study is its reliance on a proxy measure of SPS. A few questions in the YA-HCP questionnaire dataset addressed sensory processing directly. However, the proxy measure was found in independent samples to have a strong correlation with the standard measure (r = 0.79; r = 0.89 adjusting for reliabilities). Another limitation is the fact that we assessed microstructure properties with only two analysis methods (voxelwise and region-based) using one particular model (DTI). There are a number of additional analysis (David et al. 2021) and dMRI modeling techniques (Tournier et al. 2011; Mori and Tournier 2014) which can assess the current research questions from different perspectives and would be useful in future studies of SPS. Throughout this work, we did not consider a whole family of analysis modes, which utilizes the virtually reconstructed structural connections using the underlying diffusion orientation information, also known as fiber tractography (FT) (Jones 2010). The application of FT opens up a number of new ways to perform statistical comparisons: tract-based (Lebel et al. 2008); along-the-track (Reijmer et al. 2013); connectivity or connectome (Rubinov and Sporns 2010) as well as disconnectome (Thiebaut de Schotten et al. 2020); and tract geometry-based (Yeh 2020) among others, which would be useful in future studies. Also, from a modeling perspective, DTI has the theoretical limitation that it cannot resolve multiple fiber orientations within a voxel. As a result, tractography may provide inadequate results for certain pathways (Jeurissen et al. 2019), while relatively short association fibers are challenging or nearly impossible to map with DTI (David et al. 2019). One of the most notable solutions is called constrained spherical deconvolution (CSD) by Tournier et al. (2007). In CSD, a set of spherical harmonics are calculated to model the fiber orientation distributions (FODs). For tractography-based applications, modeling with CSD and like methods is necessary, since nearly all white matter voxels in the brain contain multiple fiber orientations (Jeurissen et al. 2013). Future aims of research more generally will be to replicate these findings and to examine the relation of SPS to brain structure with DTI and other dMRI-based techniques in younger and older age groups and in other populations.