The utility of multilevel analysis to understanding psychological processes and psychosomatic relations is illustrated by our recent work on loneliness. Social isolation and loneliness are potent but little understood risk factors for broad-based morbidity and mortality (Seeman, 2000). Although loneliness has a heritable component, differences in social cognition provide a better explanation for the physiological characteristics of lonely versus nonlonely individuals than does a model based on invariant traits or genetic determinism. Lonely individuals tend to construe their world, including the behavior of others, as punitive or potentially punitive. Consequently, lonely individuals are more likely to be socially anxious and to adopt a prevention focus rather than a promotion focus in their social interactions (Ernst & Cacioppo, 1999). Lonely individuals are more likely to appraise stressors as threats rather than challenges and to cope in a passive, isolative fashion rather than an active fashion that includes seeking the help and support of others. These differences in social cognition predictably result in an increased likelihood of lonely individuals acting in self-protective and, paradoxically, self-defeating ways (Cacioppo, Berntson, et al., 2000).
From the dual clues that isolation and loneliness are associated with broad-based mortality and with a higher death rate across the adult life span, one can surmise that the underlying mechanism operates on a wide range of bodily systems, or that there is more than one mechanism through which loneliness influences health. In the absence of lower level analyses, however, the relations between loneliness and health remain mere empirical associations. Recent evidence suggests that different transduction pathways account for acute effects of loneliness on morbidity and mortality (e.g., suicide) and chronic effects (e.g., heart diseases, cancers). Two of the neurobehavioral mechanisms that contribute to the association between loneliness and chronic disease are (a) catabolic processes—lonely individuals perceive more hassles and stressors in daily life and are characterized by higher tonic levels of peripheral resistance in the cardiovascular system, which over time may have damaging effects on body organs and systems; and (b) anabolic processes—lonely individuals show physiological repair and maintenance processes (e.g., wound healing, sleep) that are less efficient than nonlonely individuals (Cacioppo, Hawkley, & Berntson, 2003).
A variety of autonomic differences have been found to distinguish lonely and nonlonely individuals. The bivariate model of autonomic control outlined previously represents a significant advance in our understanding of higher neural influences on autonomic substrates, considerably clarifies psychophysiological relations, and increases the fidelity of mappings from psychological processes to autonomic cardiac control.
Characterization of response modes. As discussed earlier, pharmacological blockade analyses represent a gold standard for the quantification of patterns of sympathetic and parasympathetic control of end organs such as the heart. Blockade analyses are not always possible, however, so noninvasive measures are desirable. Noninvasive measures of parasympathetic and sympathetic control have now been validated, at least for the heart.
Respiratory sinus arrhythmia (RSA) is a fluctuation in heart rate in phase with respiration, inspiration being associated with an increase in heart rate and expiration with a decrease. RSA arises from pulmonary and thoracic stretch receptor afferents to brain stem reflex substrates that trigger inhibition of vagal outflow and excitation of sympathetic outflow (Berntson, Cacioppo, & Quigley, 1993). Both of these changes synergistically act to increase heart rate, but there are differences in the time constants of the sympathetic and parasympathetic synapses at the cardiac sinoatrial node pacemaker (see Berntson et al., 1997). The consequence of these temporal dynamics is that respiratory rhythms in the parasympathetic cardiac innervation is translated into rhythmical fluctuations in heart rate, whereas the sympathetic synapses are sufficiently slow that respiratory fluctuations are filtered out. Respiratory rhythms in heart rate thus reflect vagal cardiac control, with larger fluctuations associated with higher vagal tone (Berntson et al., 1993). RSA has been repeatedly validated as an index of vagal control of the heart, with some caveats (see Berntson et al., 1997; Berntson, Cacioppo, Binkley, et al., 1994; Cacioppo et al., 1994).
An additional noninvasive measure, pre-ejection period (PEP) is available to index sympathetic control of the heart. Pre-ejection period is the time between the electrical invasion of the ventricular myocardium (Q wave of the EKG) to the opening of the aortic valve and the onset of ventricular ejection. The pre-ejection period is a standard marker of ventricular myocardial contractility, as more forceful myocardial contractions result in the more rapid rise of intraventricular pressure and hence earlier ventricular ejection. A decrease in PEP thus indicates an increase in contractility. The sympathetic innervation enhances myocardial contractil ity, whereas the parasympathetic system plays only a minor role. Consequently, variations in sympathetic control yield corresponding changes in contractility and inverse changes in PEE With appropriate controls and caveats, PEP has been validated as an index of sympathetic cardiac control (Berntson, Cacioppo, Binkley, et al., 1994b; Cacioppo et al., 1994).
Through a combination of measurements and calculations, additional parameters of cardiody-namic and hemodynamic processes can be derived noninvasively. Heart rate (HR) and stroke volume (SV) each has sympathetic and parasympathetic contributions, and together these parameters determine cardiac output (CO) or the amount of blood expelled by the heart into the vascular system each minute (i.e., CO = HR*SV). Blood pressure, which must be maintained within relatively narrow ranges to maintain adequate circulation, is a function of the cardiac output and total peripheral resistance (TPR; the resistance to blood flow through the circulatory system). Systolic and diastolic blood pressure (SBP and DBP, respectively) can be measured noninvasively, and mean arterial pressure (MAP) can be calculated from these measures (e.g., MAP = ,33*SBP + ,67*DBP). TPR, therefore, can be calculated from MAP and CO (i.e., TPR = BP/CO).
Psychophysiological patterns in loneliness. In the pursuit of psychophysiological mappings across levels of organization and analysis, the corollary of proximity specifies that the complexity of mapping generally increases across more disparate levels. Consequently, simple isomorphic relations between events or processes are less likely to hold across more disparate levels of organization. Rather, relations across levels may need to consider patterns of multivector mappings to achieve more isomorphic mappings. Work on loneliness highlights this issue.
The cardiovascular autonomic features of lonely individuals do not organize simply on a single sym-pathetic-parasympathetic dimension. Lonely and nonlonely young adults have comparable blood pressure, but the underlying physiology differs between these groups: lonely individuals have been found to be characterized by higher total peripheral resistance and lower cardiac output than nonlonely individuals.
This difference is equally apparent at rest (baseline) as when performing orthostatic or psychological stressors (Cacioppo, Hawkley, Crawford, et al., 2002), and ambulatory recordings further revealed that this difference is evident not only in the laboratory but also during a typical day in their lives (Hawkley, Burleson, Berntson, & Cacioppo, 2003). These physiological differences and their links to health would go unrecognized with measures of blood pressure alone, which highlights the importance of theoretical systems that aid in the selection of appropriate measures to effectively bridge across levels.
Chronic elevations in total peripheral resistance not only mean that the heart muscle must work harder to distribute the same amount of blood through the circulatory system, but the reduced diameter of the blood vessels may also increase turbulence in and potential damage to the vasculature. Both central (e.g., baroreceptor reflex) and peripheral (e.g., vascular elasticity) mechanisms may degrade over time, further diminishing the ability to maintain normotensive pressure even during rest. Consistently elevated levels of vascular resistance, coupled with age-related decreases in vascular compliance, may set the stage for the development of hypertension. A study of older adults in a south Chicago apartment development confirmed this hypothesis. Because the sample size was relatively small, participants were categorized into low or high lonely groups by a median split on their scores on the UCLA loneliness scale. Results indicated that age was positively and significantly correlated with systolic blood pressure among lonely individuals, whereas there were no age-related increases in systolic blood pressure among nonlonely individuals.
Corollary of asymmetry. The patterns of neuroendocrine and autonomic control in lonely individuals are not intuitively obvious, but may have substantial basic and health significance. Further studies will be necessary to elucidate the neurophysiologi-cal and neurobehavioral origins of these patterns, and their health implications. Both of these efforts will require multilevel analyses, through which information at multiple levels can provide converging perspectives and insights. There is no single level of analysis that would permit meaningful pur suit of these questions. Moreover, there is no single level that can be universally assumed to be preeminent in multilevel analyses. On the other hand, the levels of organization and analysis cannot be viewed as "equivalent," and conceptualizations of the relationships among levels are not simply a matter of preference. Rather, there is a fundamental asymmetry across levels in multilevel research.
The most optimal approach may be to conceptually guide research by the data and constructs deriving from the level that confers the greatest organization on the problem. For the question of how social stress impacts glucocorticoid resistance, for example, the most useful organizing level of organization and analysis may revolve around neu-roimmune systems. In contrast, for the question as to physiological features of loneliness, the more salient level of analysis may be the psychosocial. Indeed, the organization in the literature is apparent only by parsing populations on the dimension of loneliness. In the absence of that, the lawful variance associated with loneliness would be relegated to the error term, and there would be no way of identifying this source of variance based on physiological studies alone.
This discussion is not intended to foster largely meaningless debates as to the "ultimate" level of analysis, nor is it intended to deny reductionism or support substitutionism. Rather, the organizing level of analysis is that which serves most effectively to structure knowledge and guide research and theory. In the case of loneliness, there is a natural asymmetry, with preeminence of the social psychological level that defines the primary conceptual dimension. This does not imply that the research and findings of this level of analysis are any more important that those of other levels, as all are required. Moreover, the optimal level of analysis may change over time, as constructs at the social-psychological level come to be implemented or integrated at lower levels, whereby the focus of research may shift to a lower level.
The corollary of asymmetry asserts that the optimal level of analysis for guiding research and theory may be that at which the major conceptual dimensions are implemented or realized. Studies of the relations between physical stress and disease may not necessarily need to appeal to the social-psychological level of organization, at least not initially. Social-psychological processes are likely important modulators of such relations, however, as social relations have been shown to be important moderator variables in stress-immune relations. This is an illustration of where a shift in focus toward higher levels of organization and analysis may be as informative as a reductionistic shift toward lower levels.
Was this article helpful?