Are judgments of the positional frequencies of letters systematically biased due to availability?

Abstract

R a l p h H e r t w i g a n d G e r d G i g e r e n z e r Max Planck Institute for Psychological Research How do people estimate whether a particular letter is more frequent in the 1st versus in a later position? The authors tested 2 precise versions of the availability hypothesis, a hypothesis that assumes that frequency processing occurs on the level of the phonological classes of vowels and consonants, and the regressed-frequencies hypothesis, which assumes monitoring of individual letters. Across 3 studies, it was found that (a) judgments of whether a letter is more frequent in the 1st or the 2rid position generally followed the actual proportions and (b) the estimated relative frequencies in the 1st versus the 2rid position closely agreed with the actual rank ordering, except for an overestimation of low and underestimation of high values. These results favor the regressed-frequencies hypothesis and challenge the conclusions about frequency judgments in the heuristics and biases literature.

Judgments of Risk Frequencies: Tests of Possible Cognitive Mechanisms.

“…13 We used contrast analysis as the measure for the covariation of predictions and estimates (Rosenthal & Rosnow, 1985;Sedlmeier et al, 1998). Table 5 shows the results of the contrast analysis (MS contrast , MS error , df error , and F value).…”

“…In contrast to Lichtenstein et al's (1978) view, which assumes that people have biased knowledge of risk frequencies because of "disproportionate exposure, memorability, or imaginability of various events" (p. 551), the regressed-frequency mechanism assumes that people's frequency knowledge is roughly accurate except for the estimates' tendency to regress toward the mean. It should be noted that not only is this tendency akin to the primary bias observed by Lichtenstein et al but also it is ubiquitous in studies of other types of frequency judgments (e.g., Begg, Maxwell, Mitterer, & Harris, 1986;Greene, 1984;Hintzman, 1969Hintzman, , 1988Sedlmeier et al, 1998;Shanks, 1995;Williams & Durso, 1986;Zacks & Hasher, 2002).…”

“…This subjective judgment of ease of retrieval serves as a cue on whose basis the frequency of each risk can be inferred. Although ease of retrieval has been effectively manipulated in recent studies (e.g., Schwarz & Vaughn, 2002), it has rarely been measured (but see Sedlmeier, Hertwig, & Gigerenzer, 1998).…”

“…Our analysis was based on a total of 1,211,000,000 words. 6 To determine the amount of regression, we followed the procedure used by Sedlmeier et al (1998). First, both the actual frequencies of the diseases and the geometric mean judgments were transformed to percentages (of the 42 diseases in Christensen-Szalanski et al, 1983, we excluded 7 as no definite actual frequencies were reported).…”

How to Improve the Diagnostic Inferences of Medical Experts

“…Minds appear to be tuned to make inferences from natural frequencies rather than from probabilities and percentages. This argument is consistent with developmental studies indicating the primacy of reasoning with discrete numbers over fractions, and studies of adult humans and animals indicating the ability to monitor frequency information in natural environments in fairly accurate and automatic ways (e.g., Gallistel & Gelman, 1992;Jonides & Jones, 1992;Real, 1991;Sedlmeier, Hertwig, & Gigerenzer, 1998). For most of their existence, humans and animals have made inferences from information encoded sequentially through direct experience, and natural frequencies are the final tally of such a process.…”