Abstract
People often fail to detect the anomalous word in questions such as How many animals of each kind did Moses take on the Ark?, and incorrectly answer “two” despite knowing that Noah rather than Moses launched the Ark. The current study tests an account of this “Moses illusion” in which Moses mistakes reflect miscomprehension of the presented word (Moses) as the expected word (Noah) due to bottom-up (phonological) priming, top-down (semantic) priming, or both. Two experiments supported this miscomprehension account: Lexical- and proposition-level information contributed autonomously to miscomprehensions and Moses mistakes in Experiment 1, and prior presentation of nonanomalous information reduced subsequent anomaly detection in Experiment 2. Present results contradict accounts in which Moses mistakes involve semantic but not phonological processes, involve mechanisms different from everyday language comprehension, or involve special anomaly detection mechanisms for calculating the coherence between the Moses question and the anomalous word.
Acknowledgements
The authors gratefully acknowledge support from the Samuel A. MacKay Memorial Research Fund and NIA grant R01AG 09755. Dr. Meredith A. Shafto is currently at the Centre for Speech, Language, and the Brain, Department of Experimental Psychology, University of Cambridge, Downing Street, Cambridge CB2 3EB, UK. The authors thank Dr. Michael Shafto for advice on statistics and experimental design, Dr. Deborah Burke for helpful comments on an earlier draft, and Irwin Stein for recruiting participants.
Notes
1This effect is relevant to the original Moses question in Erickson and Mattson (Citation1981) because of the low-level phonological similarity between their critical words, Moses and Noah (which overlap in stress pattern, bisyllabicity, initial vowels, onset-nasality, and onset-voicing). Because the low-level phonological overlap between the phonologically similar and nonanomalous words in the present study did not affect anomaly detection, this suggests that Erickson and Mattson's results involve a simple Moses effect rather than a mega-Moses effect. However, caution is warranted on this issue. If low-level phonological overlap has weak but reliable effects for “online” anomaly detection responses during sentence presentation (as in Erickson and Mattson), then Erickson and Mattson's discovery is more appropriately characterised as a mega-Moses effect (see Shafto & MacKay, Citation2000).
2We agree with Kamas et al. (Citation1996) and Park and Reder (Citation2004) that sensitivity and bias measures may be relevant to understanding Moses mistakes and related phenomena. However, with the design of Experiment 2 we could not legitimately compute sensitivity and bias by condition (see Stanislaw & Todorov, Citation1999) and recent criticisms (see Pastore, Crawley, Berens, & Skelly, Citation2003) call into question the validity of the nonparametric measures used in Kamas et al. For example, Pastore et al. (Citation2003) question the common assumption that A' and B'' are independent and demonstrate that the types of designs that we and others use systematically underestimate A', an error that increases with higher sensitivities. Taken together, these criticisms indicate that nonparametric analyses may be particularly problematic for specific comparisons between sensitivity and bias measures across conditions.