INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

THAT PHYSIOLOGICAL STIMULATION of the human respiratory system could produce evoked responses measured on the scalp was first presented in 1986 by Davenport's group (7), who showed that inspiratory occlusions were effective in evoking demonstrable midlatency (<100 ms poststimulus) activity in normal human subjects. Other work has shown that afferents from the diaphragm and intercostal muscles project to the somatosensory cortex in cats and humans (8, 10), and it is reasonable to expect that physiological stimulation should produce respiratory-related evoked potentials (RREPs) with midlatency components reflective of the nature of afferent activation of the entire set of respiratory mechanoreceptors distributed throughout the respiratory system. Our laboratory has explored the character of the RREPs evoked by small, brief, negative pressure pulses applied at the onset of inspiration (4, 6, 15, 18) and has been able to separate reliably much of the contamination caused by concurrent activation of muscles of the face and upper airway. These muscles respond reflexively to the negative-pressure airway stimulus, and, by computing the global field power (GFP), we obtain an indication of the variation of evoked activity over the monitored electrode set (4). The GFP is a reference-independent measure of evoked activity that reflects activation of the underlying cortex (14). By focusing on the nature of the evoked responses within 100 ms poststimulus, we have a reliable index of the time course and magnitude of the afferent information from respiratory mechanoreceptors produced by the applied pressure-pulse stimulus.

Perception of respiratory mechanical stimuli depends on cognitive processing of afferent information from peripheral mechanoreceptors and can involve determination of the threshold value of a stimulus required for detection or estimation of magnitudes over a range of stimuli above the detection threshold. Signal detection may only require activation of the most sensitive of a set of parallel sensory pathways, but it is likely that magnitude estimation involves evaluation of information from multiple receptor pathways. Stevens' power law is commonly used to describe magnitude estimates as a function of stimulus intensity. That is

(1)

where  is the perceived magnitude, K is a constant,  is the stimulus intensity, and  is Stevens' exponent. This relationship is conveniently evaluated by taking the log of both sides to obtain a relationship linear in log-log coordinates for which the slope is  [i.e., log() = log(K) +  log()].

Integration of afferent information to form a magnitude estimate should depend on the character of the afferent signals available for cognitive processing in the cortex. It has been shown that, in a given subject, a direct relationship exists between the amplitude of midlatency evoked potentials and the  of the applied stimulus for tactile (9) and respiratory (11, 12) modes of sensation. This suggests that the features of the measured scalp-evoked activity are closely correlated with the afferent neural signal serving as the input to cognitive processing. In our experience with RREPs occurring in response to oral pressure pulses, we have seen a wide range in the magnitudes of the evoked responses across our sample of normal human subjects (4, 6), and we wondered if this would result in a similarly wide range of magnitude estimation characteristics. To our knowledge, there has been no previous comparison for any modality of perception and cortical evoked activity across subjects, although one report has described the relationship between afferent activity in peripheral nerves and perception of the evoking mechanical stimulation of receptors in the hand of human subjects (13).

We hypothesized that subjects who showed a greater amount of afferent activation as indicated by a large GFP signal in the time from the earliest appearance of afferent activity (~20 ms after oral pressure pulse onset through 100 ms poststimulus) would show greater perception of the applied pressure. Because we have found more intersubject variation in subjects' responses to applied stimuli than to the control activity observed when no stimulus is applied, we expected subjects with greater evoked activity to demonstrate greater sensitivity to the stimulus as evidenced by a larger estimated magnitude for a standard applied pressure pulse. The rationale for this hypothesis is simply that subjects with a greater amount of afferent activity to an applied stimulus would likely translate that increased afferent traffic into a larger perceptual estimate. The focus of this study concerns the way in which variation among subjects in the GFP (the afferent signal) explains variation in intersubject perception. Preliminary results have been reported previously based on a smaller data set (5).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Experimental Protocol

Subjects breathed on an apparatus designed to permit brief (200-ms) negative pressure pulses to be applied at inspiratory onset by a computer-controlled apparatus located outside the room in which the subject sat in a comfortable dental chair within a large Faraday cage. Pressure-pulse amplitudes of 5 to 30 cmH₂O relative to atmospheric pressure were generated by a variable vacuum source. For the evoked responses, a standard 10-cmH₂O pulse was used; the magnitude estimation tests applied pressures over the full range of 5 through 30 cmH₂O.

Subjects generally performed the magnitude estimation task first and returned on a later day to perform the evoked potential experiment, although occasionally these experiments were performed on the same day. Perception was evaluated by using a cross-modality technique patterned after that described by Muza and Zechman (16) using a handgrip dynamometer with electrical output. Magnitude estimation of pressure pulses was accompanied by a line-length estimation task to evaluate how individual subjects used the handgrip force transducer with a nonrespiratory perception task. Horizontal bars of six different lengths were projected on the wall of the darkened experimental room. Six sequences of the six lengths (randomized order) were presented. Subjects were given standardized instructions to match their grip force to the length of the projected line and, when they had achieved the appropriate force, to press a button to indicate that their choice was valid. The grip-force and button data were recorded on an on-line computer running an electronic strip-chart program (CODAS, DataQ, Cleveland, OH) for later analysis.

After the line-length test, subjects were given a series of thirty-six 200-ms pressure pulses at six levels, from 5 through 30 cmH₂O, randomized within six replicate sequences. The pressure-pulse magnitude was modulated by using a variable transformer to vary the vacuum source; the replicated stimuli were close to, but not exactly identical to, the target for each pressure level. A light-emitting diode lit during expiration alerted subjects that a pressure pulse would accompany the next inspiration, and they were instructed to match their grip force to the perceived size of the pressure pulse and to press the button when their evaluation was satisfactory. The same electronic strip chart used for line-length tests was used to record the pressure amplitude, grip force, and button signal. A pressure pulse resulting in airway collapse was immediately apparent in the oscillation of mouth pressure. Such a trial was marked as defective and repeated.

After perception tests were complete, evoked potential measurements were obtained in different experiments for sets of ~100 trials with 10-cmH₂O pressure pulses. When inspiratory pressure exceeded 0.2 cmH₂O, the pressure pulses were applied by an on-line computer data-acquisition and control program (LabVIEW, National Instruments, Austin, TX). This system monitored mouth pressure, activated a set of computer-controlled balloon valves to apply the pressure pulse, and digitized 200-ms sequences of 30 or 60 electroencephalogram (EEG) channels and mouth pressure at 2 kHz. All 30 or 60 EEG channel samples plus mouth pressure were displayed on a dual-monitor system running on a laboratory computer (Macintosh G3, Apple, Cupertino, CA) for quality evaluation before being saved to disk. Only trials in which the EEG was uncontaminated by eye movements or other artifact, and for which the pressure pulse was of the correct amplitude and uncontaminated by airway collapse, were saved to disk for later analyses. Each experiment included a control set of 100 trials performed under conditions identical to the test conditions but with the vacuum source turned off.

Data Analysis

Perception experiments. The line-length experiments were analyzed by measuring the grip force exerted in response to each projected line length and by fitting a linear regression line to the log-log plot of grip force as a function of line length. The first sequence of six line lengths, the training set, was discarded for each subject. The slope of this log-log regression was interpreted as the  and the antilog of the intercept as the K. We noted that, in virtually every subject, the grip force for the longest of the six lines lay consistently above the regression line that was a reasonable fit to the other five line-length results, and it was apparent that subjects were induced to squeeze harder when they noted that the line filled the screen onto which it was projected. We, therefore, omitted the longest line length from all analyses, leaving 25 line-length trials to comprise the response data set. If the subject responses resulted in a fitted line that was visually adequate and had a correlation coefficient of 0.5, the result was retained for further analysis. Acceptable line-length data were obtained from 13 of the 18 subjects, and the median correlation coefficient for the regressions included was 0.82. Replicate tests were performed in a few subjects, and their results were averaged over the number of their adequate trials to provide a single estimate.

Evoked response analysis. Off-line analyses included trial-by-trial band-pass filtering (10-160 Hz) and ensemble averaging of the 30 channels of EEG. Filtering was performed by using routines programmed in a general-purpose analysis package (MATLAB, Mathworks, Natick, MA) and was designed to filter each trial twice, once forward and then backward, to eliminate any phase shift due to filtering. Ensemble averaging was also performed by routines programmed in MATLAB, and these routines returned the mean respiratory-related evoked responses with 95% confidence intervals to evaluate the reliability of the response from each channel.

(2)

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Perception data and fitted lines on log-log plots are shown in Fig. 1 for line-length and pressure pulse responses from a subject with a high  for pressure pulse stimuli (_PP). The pressure pulse responses in Fig. 1A were best fitted on the log-log plot by a line with a slope of 1.08 and intercept of 1.87 with a correlation coefficient of 0.88. The line-length responses for the same subject are shown in Fig. 1B and were best fitted by a log-log slope of 1.16 with an intercept of 1.32 and a correlation coefficient of 0.89. Figure 2 shows analogous pressure pulse (Fig. 2A) and line-length (Fig. 2B) magnitude estimation responses for a subject with a low _PP. Pressure pulse estimates were best fitted with a log-log slope of 0.34 with an intercept of 0.62 and a correlation of 0.74, whereas the line-length estimates were best characterized by a slope of 0.69, an intercept of 0.55, and a correlation of 0.81. The pressure pulse responses for the 18 subjects yielded an average slope of 0.70 ± 0.07 (SE). Line-length responses from the 13 subjects with acceptable line-length data yielded a mean slope of 0.91 ± 0.08. 

View larger version (14K): [in this window] [in a new window]   Fig. 1.   Individual subject line-length and pressure pulse perception examples for a subject with a high Stevens exponent () for pressure. A: pressure pulse results. B: line-length responses. Pressure values are in cmH2O, and grip force values are in V, both from the respective transducers (the grip force was uncalibrated). Solid line shows the best fitted regression line through the points, and the regression statistics are listed. INTCPT, intercept; CORREL, correlation.

View larger version (13K): [in this window] [in a new window]   Fig. 2.   Individual subject line-length and pressure pulse perception examples for a subject with a low  for pressure. A: pressure pulse results. B: line-length responses and fitted lines. Pressure values are in cmH2O, and grip force values are in V, both from the respective transducers (the grip force was uncalibrated). Solid line shows the best fitted regression line through the points, and the regression statistics are listed.

The predicted ME10, based on each subject's individual estimated  and intercept are shown in Fig. 3 and are plotted as a function of the individual values for the GFP20-100. The plot shows the predicted magnitude vs. the log of the GFP20-100 as that transformation produced a reasonably linear relationship and a significant regression; the plot vs. the untransformed data seemed to be best fit by a hyperbolic-like function and did not produce a significant linear relationship. We do not now have a theoretical basis for why the logarithmic transformation adequately captures a linear data trend.

View larger version (10K): [in this window] [in a new window]   Fig. 3.   Predicted magnitude estimates (ME) of a standard oral pressure pulse stimulus of 10 cmH2O (ME10) vs. the log of the summed global field power (GFP) over a period of 20-100 ms poststimulus (GFP20-100). ME10 estimates for individual subjects are based on each subject's  and intercept for perception of pressure pulses. ME10 is in V corresponding to the uncalibrated but constant handgrip dynamometer; summed GFP is in µV. Solid line shows the best fitted linear regression.

The relationship of the predicted magnitude estimates with the estimates of afferent activity shown in Fig. 3 is not consistent with our original hypothesis, and this will be discussed later. We examined the extent to which the relationship between the predicted perception and the afferent signal was affected by the separate contributions of the slopes and the intercepts of the fitted perception responses, and Figs. 4 and 5 show the relationships between each aspect of the perception response (slope, Fig. 4, and intercept, Fig. 5) vs. the GFP20-100. A highly significant, inverse relationship holds between the  (Fig. 4) and the afferent signal, but there is no apparent relationship between the intercept and each subject's afferent activity (Fig. 5).

View larger version (11K): [in this window] [in a new window]   Fig. 4.   Individual subject  values for pressure pulse stimulation (PP) plotted vs. GFP20-100. The PP is dimensionless; the summed GFP is in µV. The solid line is the best fitted linear regression.

View larger version (8K): [in this window] [in a new window]   Fig. 5.   Individual subject Stevens power law intercepts for pressure pulse stimulation (INTPP) plotted vs. GFP20-100. The pressure pulse intercept is dimensionless; the summed GFP is in µV. The solid line is the best fitted linear regression, which was not significant (NS).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The most important finding of this study is that subjects who appear to have greater afferent information from respiratory mechanoreceptors do not necessarily perceive a standard pressure pulse stimulus to have a greater magnitude, compared with subjects seeming to have less afferent traffic. This contradicts our original hypothesis that there would be a direct relationship between afferent traffic and perception. The conclusion that, across subjects, an inverse relationship exists between afferent traffic and perception is dependent on a number of assumptions, which we discuss below.

We have interpreted the GFP20-100 as an estimate of afferent activity from respiratory mechanoreceptors over that period. This assumes that somatosensory activity over that period is relatively uncontaminated by endogenous processing of the afferent activity. Such an assumption is clearly not tenable in detail, based on the long-standing knowledge that sleep affects components of similar latency in median nerve somatosensory evoked responses (1, 17, 21). More pertinent to RREPs, sleep has been shown to affect components in the evoked responses with latencies >100 ms postocclusion, but sleep effects were much less apparent with regard to shorter latency components (19). Similarly, a recent report by Webster and Colrain (20) on the effects of attention on mid- and longer latency components of the evoked response to occlusion pressure stimuli showed that the only effect of attention on components with latencies within the time period of interest to us was to delay the P₁ component from 54 to 67 ms; there was no effect of attention on the amplitudes of any components with latencies <100 ms. In a previous report (18) looking at evoked responses to pressure pulses in attending and nonattending conditions, our laboratory did not see any effect on the evoked responses that occurred <100 ms poststimulus. It is likely that subtle modulation of some aspects of RREPs occurs within the 100-ms poststimulus span of interest, but that does not preclude the use of that information to examine exogenous evoked activity, provided the state of arousal is reasonably consistent within and between experiments (as was the case here). Therefore, we feel that our interpretation of the GFP20-100 as an index of afferent activity is reasonable, especially as it is less contaminated by electromyographic activity from concurrently activated facial and upper airway muscles than are direct measures of evoked responses (4).

Our data (Fig. 3) indicate that the relationship between perception and afferent activity, as captured in the relationship between the predicted estimate to a standard 10-cmH₂O pressure pulse stimulus and the GFP20-100, is opposite to what we hypothesized. Subjects with greater GFP activity to 10-cmH₂O pressure pulses had smaller predicted magnitude estimates to that pressure stimulus. Rather than conversion of afferent activity into  in a direct relationship as we had expected, it appears that subjects receiving greater afferent information attenuate their perception of the causative stimulus. There was also an inverse relationship between the _PP and the summed GFP (Fig. 4), but there was no apparent relationship between the power law intercept and the amount of afferent traffic as estimated by the GFP.

There are at least two complicating issues that need to be considered with regard to the measurements. The first is whether the magnitude estimation parameters for pressure pulse stimuli are valid. This is not an easy question to answer because, to our knowledge, there are no published data regarding the perception of this stimulus. Muza and Zechman (16) reported a  for peak mouth pressure of 1.22, measured during loaded breathing, and we can compare our  for the peak applied pressure pulse to that value if we correct our value for the grip force exponent of 1.7 (16). Thus, if we multiply our average _PP of 0.70 by 1.7, we obtain a value of 1.20, quite close to the result of Muza and Zechman. We, therefore, feel that our magnitude estimates are reasonable.

The second issue is whether our GFP results could be systematically affected by volume conductivity differences among subjects that might create an inverse relationship between scalp estimates of afferent activity and magnitude estimates that otherwise might actually follow our hypothesis. For example, what if the subjects with high measured GFP values had a lower value for skull resistance, the major resistive component between the cortical sites of sensory activity and the scalp where activity is measured? If this were true, for any given level of local cortical activation, subjects with lower skull resistance would have greater local radial currents through the skull, and this would increase the local variation in scalp potentials and augment the GFP. To compensate for this, we would have to correct the high-GFP measurements downward and the low-GFP measurements upward. Such a correction would only serve to steepen the inverse slope for the relationship of ME10 or _PP vs. GFP and could only produce the hypothesized direct relationship between perception and afferent activity if the corrections were large enough to actually reverse the sign of the slope. We think that this is very unlikely because it would require subjects with high GFP measured on the scalp to have less cortical activity than subjects with low scalp GFP values.

What could explain the inverse relationship between sensation and perception observed among subjects? The conversion of an afferent stimulus into a perception involves a sensory and a cognitive aspect, and the observed response could have its explanation in either or both of these. One of us has proposed a model for the sensory afferent process (2, 3), the sensory aggregate model (SAM), which describes the expected response of a population of sensory afferents with different activation thresholds. Figure 6 demonstrates the expected afferent responses from two subjects having afferent mechanoreceptor populations that consist of the same number of receptors, each with identical stimulus-response shapes (logistic) and mean values for their activation thresholds but that differ only in the variance of the distribution of activation thresholds. Figure 6 shows the population responses to identical applied (fixed) stimuli at any level below some maximum. We assume that the afferent activity arriving at the somatosensory cortex is proportional to the summation of receptor activity from all receptors whose activation threshold lies at or below the level of the applied pressure stimulus. Clearly, the summed afferent activity (GFP) will be greater for the more broadly distributed population (left) than for the more narrow distribution (right). This corresponds to the response for the 10-cmH₂O applied pressure pulse. In Fig. 7, we illustrate the effect of this difference in the distribution of activation thresholds on perception as measured by magnitude estimation. We hypothetically apply a series of stimuli over a range of magnitudes and assume that the slope of the magnitude estimate will reflect how much the afferent activity changes with an increase (or decrease) in applied stimulus magnitude. Because the population on the right is more narrowly distributed, the full range of afferent activity will be encompassed by a smaller range of applied pressures than will be true for the more broadly distributed population on the left. Thus the more narrowly distributed activation threshold population would lead to a steeper slope between applied pressure and magnitude estimates. This analysis shows that the SAM model is consistent with our observation that subjects with higher GFP tend to produce a lower estimate of a standard pressure pulse stimulus (Fig. 3) and a lower exponent (Fig. 4).

View larger version (29K): [in this window] [in a new window]   Fig. 6.   Scheme based on the sensory aggregate model (SAM) for relating sensory receptor population characteristics to the afferent signal as measured by the GFP. Populations in the examples on the left and right have identical numbers of mechanosensors with identical stimulus-response characteristics and the same mean activation threshold. The right and left examples differ only in the dispersion of activation thresholds about the mean value. The same stimulus is applied to both populations, and the resulting GFP is proportional to the summed activity from all activated receptors. More receptors are activated at a higher firing rate in the population on the left with the wider dispersion of thresholds, and this leads to a higher GFP than is the case for the more narrowly distributed population on the right.

View larger version (25K): [in this window] [in a new window]   Fig. 7.   Scheme based on the SAM to relate perceptual performance to sensory population characteristics. The same populations are used here as in Fig. 6 to examine the effect of a difference in dispersion of activation thresholds alone. Here the pressure pulse stimuli are applied over a range of values, and perceptual judgments are made based on some sort of internal comparison of how much the afferent activity varies as the stimulus changes. , Perceived magnitude; , stimulus intensity.

However, this consistency does not mandate acceptance of the SAM model as the explanation for our observations. On the basis of our results, it is likely that subjects do differ with respect to the afferent activity that they receive from their respiratory mechanoafferents, because we cannot think of a reasonable alternative to explain why they differ so in their GFP responses to a 10-cmH₂O stimulus. If we omit the two smallest and two largest GFP responses to minimize the effect of outliers, we still find nearly an order of magnitude difference between the remaining highest and lowest GFP responses to the standard 10-cmH₂O stimulus. With respect to the cognitive processing of that afferent information, however, subjects may yet modulate their magnitude estimation based on learned experiences and other cognitive factors. We measured subjects' magnitude estimation characteristics with respect to both a respiratory task and a nonrespiratory task (evaluation of line lengths). Figure 8 (which compares the individual  values for these two modalities, which process completely different sensory inputs) and Fig. 9 (which compares the fitted intercepts) indicate surprising and significant correlations between the power law parameters for these two quite distinct sensory modalities. Subjects with a higher  for line length (_LL) also have a higher _PP (and likewise for intercepts), suggesting that some aspect of the judgment process consequent to sensory afferent processing determines a portion of the magnitude estimation, regardless of the sensory modality. Over the ranges of pressure pulse and line-length stimuli tested, the ratio of the highest to lowest _LL was ~3.3, and the ratio for _PP was ~3.0, indicating a similar range of judgment effects for both modalities. Considering the power law intercept variation, the ratio of maximum to minimum for line length was ~5.4, whereas the ratio for pressure pulse stimuli was nearly 7.0. There was no significant relationship between the exponent and intercept within either stimulus modality.

View larger version (8K): [in this window] [in a new window]   Fig. 8.   Comparison of values obtained for the Stevens exponents for pressure (PP) and line length (LL) in each subject for whom adequate line-length and pressure pulse data were obtained. The solid line describes the best fitted linear regression;  values are dimensionless.

View larger version (9K): [in this window] [in a new window]   Fig. 9.   Comparison of values obtained for INTPP and the Stevens power law intercepts for line length (INTLL) in each subject for whom adequate line-length data were obtained. The solid line describes the best fitted linear regression. Intercepts are dimensionless.

We used the regression results for _PP regressed on _LL from Fig. 8 and for the intercept results from Fig. 9 to correct each subject's _PP and intercept for perception of pressure pulses (INT_PP) for the individual tendency of each subject to over- or underexpress his or her perception using the force-grip cross-modality approach. This correction reduced each subject's pressure pulse slope and INT_PP by an amount computed using the regression results of the pressure pulse slope and INT_PP from all 13 subjects for whom both pressure pulse and line-length data were available, together with each of the 13 subjects' line-length slope and intercept. The corrected  and corrected intercept for each subject were used to compute a corrected estimate of the ME10 stimulus that accounts for each subject's individual characteristics using the cross-modality expression of line length. The corrected ME10 is plotted vs. the GFP20-100 in Fig. 10, together with the best fit regression line. The resulting regression is significant (P < 0.01), and the inverse relationship is still apparent. The cognitive component related to using grip force to express perception of respiratory pressure pulses adds variability to the relationship between the  and the afferent signal, and accounting for the nonrespiratory variation in magnitude estimation in this manner reduces the range of variation in the ME10 responses from a maximum-to-minimum ratio of 3.29 to 1.82. A significant effect of afferent magnitude remains, and this may be attributable to the SAM explanation offered earlier or have other cognitive sources.

View larger version (11K): [in this window] [in a new window]   Fig. 10.   Corrected ME10 (dimensionless) vs. the summed GFP20-100 (µV) for subjects with adequate line-length data. The corrected ME10 was obtained from the power law after subtracting the effect of line length from each subject's estimated PP using the regression equation of Fig. 8 with each subject's line-length exponent and after subtracting the line-length effect from each subject's power law intercept using the regression equation of Fig. 9 with each subject's line-length intercept. The solid line describes the best fitted linear regression.

Our laboratory has previously shown that information from upper airway mechanoreceptor afferents arrives in normal subjects roughly between 50-80 ms poststimulus (6), and we wondered how well information from that time span correlated with the corrected estimates of perception of the 10-cmH₂O pressure pulse stimulus. Figure 11 shows this relationship and indicates that a virtually identical relationship holds for both the 20- to 100-ms and the 50- to 80-ms summations of afferent activity. It is reasonable to speculate, therefore, that supralaryngeal mechanoreceptor activity may largely contribute to the perception of pressure pulse stimuli in our subjects.

View larger version (11K): [in this window] [in a new window]   Fig. 11.   Corrected ME10 (dimensionless) vs. the summed GFP over a period of 50-80 ms poststimulus (GFP50-80) (µV) for subjects with adequate line-length data. Correction for nonrespiratory perception was done as described in Fig. 10. The solid line describes the best fitted linear regression.

The net effect of the relatively shallow, inverse relation between afferent activity and perceptual sensitivity is that subjects will tend to return magnitude estimates that are much more similar across the subject population than would be the case if our hypothesized direct relationship pertained with steep slope. The shallow relationship that we found between perception and somatosensory activation by respiratory stimuli is similar to the lack of correlation reported by Knibestöl and Vallbo (13) between perception and afferent nerve responses to mechanical stimulation of receptors in the hand, although our finding of the small but significant negative relationship is unexplained. It is possible that humans adapt to the amount of afferent information provided by their individual mechanoreceptor characteristics. It might be quite distracting for those with highly sensitive mechanosensation to be as responsive as those with less-sensitive sensory mechanisms to ongoing mechanical information consequent to breathing. Therefore, the shallow, inverse relation between perception and afferent activity may result from hereditary or adaptive variation in cognition that serves to modulate the impact of natural stimuli on brain function.