"Recognition ROCs are curvilinear—Or are they? On premature arguments against the two-high-threshold model of recognition": Correction to Bröder and Schütz (2009).

Abstract

authors reconstructed 59 data sets from published studies in which response bias had been manipulated experimentally. These data were fit with signal detection theory (SDT) and a twohigh-threshold model (2HTM). However, the G 2 statistic for SDT fits was misspecified, and there were some errors in the reconstructed frequencies. (The authors are greatly indebted to David Kellen, who noticed these issues and cross-checked the corrections.) Below are corrected results; the authors also (a) separately analyzed read and anagram word data from Buchner et al.'s (1995) Experiments 1 and 2, and (b) included reconstructed binary response results from Curran et al. (2007), Experiment 3. All data sets were reanalyzed by minimizing G 2 for SDT and 2HTM using various routines with converging results. Rather than presenting a corrected version of the complete 4-page appendix table in Bro ¨der and Schu ¨tz, the new G 2 values are reported following the same order as the originally tabled studies from top to bottom, but with four instead of two data sets from Buchner et al. and four instead of three from Curran et al. Square brackets indicate which values stem from the same literature source to further facilitate their correct assignment. The G 2 values for the 2HTM are (

Perceptual stimuli with novel bindings interfere with visual working memory

“…One explanation for this apparent discrepancy may relate to the small set of easily identifiable stimuli used in those experiments, and the use of only two memory stimuli on each trial. If participants retain representations of the identities and locations of stimuli presented in the memory array, even when the bindings between these features are lost-as suggested by Cali et al (2015)-this may allow them to respond to a memory test with an informed guess about the nature of the relevant binding (see, e.g., Rhodes, Cowan, Hardman, & Logie, 2018 for evidence of informed guessing in tests of visual WM). Further, if the distractor stimulus is sometimes mistaken for a memory stimulus, this would reduce the success of such a guessing process.…”

A critical comparison of discrete-state and continuous models of recognition memory: Implications for recognition and beyond.

“…As a result, the following discussion of empirical ROCs focuses on distinguishing between threshold MPT and continuous SDT models. Schütz (2009, 2011) conducted a meta-analysis^ of 62 studies that reported binary ROCs from memory tasks, and they collected new data in three experiments to more clearly address the issue. Empirical ROCs.…”