APPENDIX
A
Interepretations of Findings of Controlled-Exposure and Social Survey Studies
DIFFERENCES BETWEEN STUDY TYPES
As summarized in Table 5, social surveys differ from controlled-exposure field studies (whether conducted in laboratory or field settings) in a number of ways that impede direct comparisons of the data collected in the two types of studies. It is useful to keep these differences in mind when developing a dosage-response relationship for predicting community response on the basis of a combination of information from the two types of studies.
DIFFICULTIES IN LINKING OF LABORATORY AND FIELD DATA
Estimates of the annoyance of individual sounds as judged in controlled-exposure studies and of the prevalence of noise-induced annoyance in a community are both derived from information about the relative frequencies of occurrence of self-reports of annoyance, based on a pooling of the opinions of either (a) test subjects about the immediate annoyance of individual test signals or (b) survey respondents about the long-term annoyance of their neighborhood noise environment. The similarity in the pooling of annoyance judgments does not constitute a logical link between the probability that a single noise intrusion will lead to an individual's report of annoyance in some degree (as in a controlled-exposure study) and the proportion of residents highly annoyed by cumulative noise exposure in a community (as in a social survey).
Probabilities of individual reports of high annoyance to controlled signal presentations are inferred from the relative frequencies of judgments of the an
TABLE 5 Differences Between Controlled-Exposure and Social Survey Studies of Noise-Induced Annoyance
Factor |
Controlled-Exposure Studies |
Community Surveys |
Customary goal |
Determine perceptual function |
Estimate opinions of population |
Study design |
Experimental (causal relationship between independent and dependent variables inferable in principle) |
Observational (correlational evidence only, no strict inference of causality possible) |
Sampling method |
Self-selection |
Variable; often random representation |
Test participants |
Paid test subjects providing informed consent |
Neighborhood residents |
Dependent variable (quantity judged) |
Immediate, solicited judgments of short-term annoyance of temporally compressed presentations of novel signals during a test session of relatively short duration |
Delayed, retrospective judgment of long-term annoyance of habitual exposure to familiar sounds occurring at times of day and intervals appropriate to their generating mechanisms |
Basic datum |
Relative frequency of occurrence of specific, solicited judgments |
Prevalence among respondents of self-reports of annoyance in similar degrees |
Data collection setting |
Contrived (laboratory or modified residential) listening environment |
Unmodified residential |
Characterization of independent variable |
Precise knowledge of signal characteristics, reasonable information about individual exposures |
No control over noise exposure other than site selection; imprecise knowledge of place exposure, little or no knowledge of personal exposure |
Low-frequency content of sounds judged |
Often lacking in ability to accurately reproduce low-frequency content of high-energy impulsive sounds and/or secondary emissions |
Appropriate to source and residential setting; no decorrelation of low-frequency content and absolute level of individual events |
noyance of single presentations of test signals. Suppose, for example, that 1,000 annoyance judgments are solicited by asking 50 test subjects 20 times each during the course of a laboratory experiment to judge the annoyance of a particular signal presented at a constant level. Suppose further that the annoyance judgments are solicited on a five-category absolute judgment scale (e.g., “not at all annoying,” “slightly annoying,” “moderately annoying,” “very annoying,” and “extremely annoying”). If 200 of these 1,000 judgments fall in the categories “very” or “extremely” annoying, few would quibble with characterizing the probability as 0.2 that the signal when presented at the level in question would be judged “highly” (that is, either “very” or “extremely”) annoying.
The resulting dichotomy of annoyance judgments (“highly annoyed” versus “not highly annoyed”) seems consistent with Schultz's (1978) prevalence of annoyance scale. However, the prevalence ofhigh annoyance in a community is estimated from a count of the number of respondents expressing a shared opinion about the long-term annoyance of neighborhood noise exposure, not about the annoyance of individual noise intrusions. For example, if 40 of 200 survey respondents at an interview site with homogeneous noise exposure describe themselves as either “very” or “extremely” annoyed by neighborhood noise exposure, the proportion of highly annoyed respondents associated with the neighborhood noise exposure is said to be 0.2.
Both dichotomized annoyance judgments from controlled-exposure studies and annoyance prevalence estimates from social surveys may be treated as binomial proportions. Each person in a community is assumed to be either highly annoyed (p) or not highly annoyed (q = 1 − p) by each flyover. The expectation of the binomial distribution is simply Np, the product of the number of people exposed and the probability of high annoyance per flyover. Expectations of the prevalence of annoyance can then be based simply on total population.
Even though these binomial proportions can be manipulated in a similar manner, the two types of information on which they are based are not necessarily comparable. The binomial estimate of the prevalence of annoyance induced by a given noise intrusion cannot be directly interpreted as the long-term annoyance of repeated exposures.