Comparison of the triangle and a self-defined two alternative forced choice test

Comparison of the triangle and a self-defined two alternative forced choice test McClure Lawless

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

Comparison of the triangle and a self-defined two alternative forced choice test 2010 Food Quality and Preference 10.1016/j.foodqual.2010.02.005 McClure2010-comparison-of-the-triangle-and-a-self-define McClure2010-comparison-of-the-triangle-and-a-self-define.txt, the project's staged copy 51eefcb6a4992321c4cf5ab64f2f40e021bacf13 not-redistributable sensory-science

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

0 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them method Quotation withheld because this record is not cleared for redistribution. sensory discrimination testing Quotation withheld because this record is not cleared for redistribution. behavioural-experiment behavioural-experiment Quotation withheld because this record is not cleared for redistribution. mechanisms choosing an irrelevant attribute in the self-defined test adds irrelevant variation, which degrades the panellist's later discrimination Quotation withheld because this record is not cleared for redistribution. mechanisms wider discriminal dispersion in some subjects makes rank-order reversals (inversions) of the two products more likely, so their responses oppose the majority Quotation withheld because this record is not cleared for redistribution. mechanisms the triangle test requires three implied paired comparisons, whose combined variability lowers its power relative to a directional AFC test Quotation withheld because this record is not cleared for redistribution. Re-tasting during the preview was permitted but not recorded, so the procedure may have drifted toward a full warm-up protocol for some subjects. what this limits: the comparison between the two test procedures: RE-TASTING WAS ALLOWED DURING THE PREVIEW PHASE, BUT WAS NOT RECORDED, so drift toward a fuller tasting cannot be excluded Quotation withheld because this record is not cleared for redistribution. Variation between the food systems may reflect differences between the judges who happened to participate rather than the methods. what this limits: the between-food variation: SOME OF THE VARIATION AMONG THE DIFFERENT FOODS WAS UNDOUBTEDLY CONTRIBUTED BY DIFFERENCES AMONG JUDGES Quotation withheld because this record is not cleared for redistribution. Participants were not excluded from later product tests, so the same people could recur across product systems. what this limits: independence across product systems: NO ATTEMPT WAS MADE TO EXCLUDE EARLIER PARTICIPANTS FROM A SUBSEQUENT FOOD PRODUCT TEST, ALTHOUGH THEY WERE OF COURSE RESTRICTED TO PARTICIPATION IN ONLY ONE CONDITION PER PRODUCT Quotation withheld because this record is not cleared for redistribution. The two measures of the same thing disagree in direction: the method with the statistical advantage returns the LOWER estimate of absolute sensory difference, and the paper calls this paradoxical. what this limits: choosing a discrimination test on statistical power: the paper reports the SD-AFC as statistically superior to the triangle while returning lower Thurstonian d-prime values for the same products Quotation withheld because this record is not cleared for redistribution. Below-chance performance was not recorded as below chance: those groups were assigned a d-prime of zero so that the comparison could be run. what this limits: any d-prime of zero in this paper's subgroup table: it may be an assigned floor rather than a measured value, since below-chance groups were set to zero for the comparison Quotation withheld because this record is not cleared for redistribution. What counts as the CORRECT attribute is set by the majority of subjects, not by the manipulation, and a subject can be scored as atypical for disagreeing with the plurality. what this limits: the typical/atypical split as a measure of attending to the right attribute: the groups were defined by what a plurality of subjects said, and all other subjects were categorised as atypical Quotation withheld because this record is not cleared for redistribution. Who turns up decides part of the result, and the authors say so, crediting a reviewer. what this limits: comparing discrimination results across product systems: the authors state the variation was partly due to differences among judges and to who shows up on a given day Quotation withheld because this record is not cleared for redistribution. The central question the paper set out to inform is left open. what this limits: the value of an inspection or preview phase in discrimination testing: the paper states this remains an open question and may depend on the product Quotation withheld because this record is not cleared for redistribution. A result relied on in this literature is reported as untested: an overall difference was claimed by taking three non-significant comparisons together, with no test of that claim reported. what this limits: citing Rousseau et al. (1999) for a familiarization effect: this paper reports that none of the individual comparisons were significant and that no test of the combined claim was reported Quotation withheld because this record is not cleared for redistribution. d-prime as the measure of sensory difference, so that food-discrimination tests are scored on a psychophysical sensitivity scale rather than by proportion correct Quotation withheld because this record is not cleared for redistribution. subjects were randomly assigned across the five discrimination procedures for simple stimuli eliminates: that a difference between procedures is a difference between the subjects who happened to use each · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. d-prime d-prime Quotation withheld because this record is not cleared for redistribution. sensory difference sensory difference Quotation withheld because this record is not cleared for redistribution. preview samples preview samples Quotation withheld because this record is not cleared for redistribution. familiarization samples familiarization samples Quotation withheld because this record is not cleared for redistribution. triangle test triangle test Quotation withheld because this record is not cleared for redistribution. inversion inversion Quotation withheld because this record is not cleared for redistribution. discrimination - as STATISTICAL SIGNIFICANCE: how reliably a method returns a significant result, on which the SD-AFC beats the triangle Quotation withheld because this record is not cleared for redistribution. discrimination - as ABSOLUTE SENSORY DISTANCE: the Thurstonian d-prime, on which the same SD-AFC comes out LOWER than the triangle for the same products Quotation withheld because this record is not cleared for redistribution. preview - a single presentation of both test stimuli before testing, explicitly contrasted with the more extensive warm-up Quotation withheld because this record is not cleared for redistribution. d-prime - stated here as the Thurstonian measure of sensory difference, and used precisely because it is claimed to allow comparison ACROSS methods Quotation withheld because this record is not cleared for redistribution. inversion - a subject using a relevant attribute but ranking the two products the wrong way round, which the paper distinguishes from choosing an irrelevant attribute Quotation withheld because this record is not cleared for redistribution. builds-on → Pfaffmann (1953) Quotation withheld because this record is not cleared for redistribution. extends → O'Mahony and Goldstein (1986); O'Mahony et al. (1988) warmed-up paired comparison Quotation withheld because this record is not cleared for redistribution. contradicts → Angulo et al. (2007) Quotation withheld because this record is not cleared for redistribution. sensory difference —operationalised-as→ d-prime sensory difference d-prime strength of the relation: 1 Quotation withheld because this record is not cleared for redistribution. preview samples —renamed-across-fields→ familiarization samples preview samples familiarization samples strength of the relation: 1 Quotation withheld because this record is not cleared for redistribution. chi-squared chi-squared Quotation withheld because this record is not cleared for redistribution. z-test z-test Quotation withheld because this record is not cleared for redistribution. beta-binomial model beta-binomial model Quotation withheld because this record is not cleared for redistribution. effect size (Thurstonian d-prime) effect size (Thurstonian d-prime) Quotation withheld because this record is not cleared for redistribution. variance variance Quotation withheld because this record is not cleared for redistribution. t-test t-test Quotation withheld because this record is not cleared for redistribution. ANOVA ANOVA Quotation withheld because this record is not cleared for redistribution. Tukey post hoc test Tukey post hoc test Quotation withheld because this record is not cleared for redistribution. binomial test binomial test Quotation withheld because this record is not cleared for redistribution. The three AFC formats, including the new self-defined 2-AFC, detected the sensory difference more reliably than the triangle test did. Quotation withheld because this record is not cleared for redistribution. Poorer discrimination in the self-defined 2-AFC went with panellists naming the wrong critical attribute or judging the difference in the reverse direction during the inspection phase. Quotation withheld because this record is not cleared for redistribution. Letting panellists preview both samples before a triangle or 2-AFC test brought no measurable benefit. Quotation withheld because this record is not cleared for redistribution. Although statistically more powerful, the self-defined 2-AFC produced smaller Thurstonian d-prime values than the triangle test, i.e. less absolute sensory differentiation. Quotation withheld because this record is not cleared for redistribution. Panellists who chose the same attribute as the majority reached higher d-prime values than those who chose an atypical attribute, significantly so for the fruit drink. Quotation withheld because this record is not cleared for redistribution. Panellists who reversed the majority direction of the attribute (calling the sweeter sample less sweet) discriminated worse than typical non-inverted panellists in the fruit drink and tea. Quotation withheld because this record is not cleared for redistribution. The attribute a panellist stated during the preview predicted whether that panellist would answer correctly on later trials. Quotation withheld because this record is not cleared for redistribution. Self-reported confidence after a preview barely predicted subsequent discrimination performance. Quotation withheld because this record is not cleared for redistribution. For all 2-AFC tests, presenting the weaker sample before the stronger gave more correct answers than the reverse order. Quotation withheld because this record is not cleared for redistribution. self-defined 2-AFC (SD-2-AFC) self-defined 2-AFC (SD-2-AFC) Quotation withheld because this record is not cleared for redistribution. preview sample preview sample Quotation withheld because this record is not cleared for redistribution. paradox of the discriminatory nondiscriminators paradox of the discriminatory nondiscriminators Quotation withheld because this record is not cleared for redistribution. total number correct total number correct Quotation withheld because this record is not cleared for redistribution. d-prime d-prime Quotation withheld because this record is not cleared for redistribution. attribute selected by subjects attribute selected by subjects Quotation withheld because this record is not cleared for redistribution. randomized order of presentation randomized order of presentation Quotation withheld because this record is not cleared for redistribution. Over 600 — number of individual discrimination tests run across the four food systems Quotation withheld because this record is not cleared for redistribution. 30 to 40 — number of subjects used per testing condition, chosen to match industrial discrimination panels Quotation withheld because this record is not cleared for redistribution. z = 2.04, p < .05 — size of the deficit in d-prime for the self-defined 2-AFC relative to the triangle in the soup system Quotation withheld because this record is not cleared for redistribution. v2 = 13.81, p < .05 — chi-squared advantage of the self-defined 2-AFC over the triangle in correct responses for the tea Quotation withheld because this record is not cleared for redistribution. 9% or 10% (w/v) — the sucrose levels distinguishing the two fruit-drink samples Quotation withheld because this record is not cleared for redistribution. 3.5 g/L — monosodium glutamate added to the modified vegetable broth Quotation withheld because this record is not cleared for redistribution. over 600 — discrimination tests conducted across the four food and beverage systems Quotation withheld because this record is not cleared for redistribution.

The full text of this article is not redistributable, so it is not included here. The evidence sentences are withheld with it; the tag layer and its grounding status remain. The article itself is at https://doi.org/10.1016/j.foodqual.2010.02.005.