The DOI matches its Crossref record; 50 of the 51 references it cites were found in a registry or an index, and 1 are not checked yet; 13 of the 68 sources its claims cite were checked for support.
Unfamiliar face matching with photographs of infants and children
Open accessAbstract
Background: Infants and children travel using passports that are typically valid for five years (e.g. Canada, United Kingdom, United States and Australia). These individuals may also need to be identified using images taken from videos and other sources in forensic situations including child exploitation cases. However, few researchers have examined how useful these images are as a means of identification. Methods: We investigated the effectiveness of photo identification for infants and children using a face matching task, where participants were presented with two images simultaneously and asked whether the images depicted the same child or two different children. In Experiment 1, both images showed an infant (<1 year old), whereas in Experiment 2, one image again showed an infant but the second image of the child was taken at 4–5 years of age. In Experiments 3a and 3b, we asked participants to complete shortened versions of both these tasks (selecting the most difficult trials) as well as the short version Glasgow face matching test. Finally, in Experiment 4, we investigated whether information regarding the sex of the infants and children could be accurately perceived from the images. Results: In Experiment 1, we found low levels of performance (72% accuracy) for matching two infant photos. For Experiment 2, performance was lower still (64% accuracy) when infant and child images were presented, given the significant changes in appearance that occur over the first five years of life. In Experiments 3a and 3b, when participants completed both these tasks, as well as a measure of adult face matching ability, we found lowest performance for the two infant tasks, along with mixed evidence of within-person correlations in sensitivities across all three tasks. The use of only same-sex pairings on mismatch trials, in comparison with random pairings, had little effect on performance measures. In Experiment 4, accuracy when judging the sex of infants was at chance levels for one image set and above chance (although still low) for the other set. As expected, participants were able to judge the sex of children (aged 4–5) from their faces. Discussion: Identity matching with infant and child images resulted in low levels of performance, which were significantly worse than for an adult face matching task. Taken together, the results of the experiments presented here provide evidence that child facial photographs are ineffective for use in real-world identification.
Verified 2026-11-09 · 1 reference not checked yet
Overview
AI snapshot
Written from the paper’s full text.
TL;DR
Across four experiments, unfamiliar face matching with infant and child photographs was poor and significantly worse than adult face matching. The authors conclude that infant and child facial photographs are ineffective for real-world identification.
Key findings
- Matching two infant photos in Experiment 1 produced 72.0% accuracy (d′ = 1.26), while matching an infant photo to a 4–5-year-old child photo in Experiment 2 produced 64.4% accuracy (d′ = 0.78).
- In Experiments 3a and 3b, difficult infant tasks were near chance: infant–infant accuracy was 56.4% and 55.7%, and infant–child accuracy was 52.9% and 51.5%, compared with short-version Glasgow Face Matching Test adult accuracy of 73.3% and 70.4%.
- Experiment 4 found sex judgments were highest for children aged 4–5 (82.3%), above chance for Experiment 1 infants (56.7%), and at chance for Experiment 2 infants (50.7%).
- Same-sex versus different-sex mismatch pairings had little effect in Experiments 1 and 2, with mixed evidence in Experiment 3a.
Methods
- Participants completed same/different face matching tasks: Experiment 1 used 41 match and 41 mismatch trials with two infant images; Experiment 2 used 30 match and 30 mismatch trials pairing an infant image with a child aged 4–5 years.
- Experiments 3a and 3b used shortened, difficult versions of the infant and infant–child tasks plus the short Glasgow Face Matching Test; Experiment 4 asked participants to judge the sex of the same infant and child images.
- Analyses used percentage correct, hit rate, false alarm rate, d′ sensitivity, criterion, ANOVAs, and within-person correlations; samples included university students and MTurk community participants.
Limitations
- The paper states that Experiments 1 and 2 used random mismatch pairings, which may inflate accuracy relative to real-world fraud; Experiment 2 used celebrity children’s images with less control over pose and expression.
- The paper states that difficult mismatch trials in Experiments 3a and 3b were selected from few observations, and low infant-task performance created floor effects and restricted range for correlation analyses.
- The paper notes samples were mostly women and had little regular infant contact, so experience effects could not be tested; it also states that it cannot determine whether infant and adult matching rely on the same cognitive processes.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
13 of 68 cited sources checked · 12 supported · 0 flagged · accuracy not yet measured
- Premisecitation
In White et al. (2015), adult trials resulted in statistically better performance than both child and adolescent trials.
“the latter resulting in statistically better performance in comparison with both child and adolescent trials”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites White et al. (2015) source
- Premisecitation
Hit rates for adult faces decrease with only 17 months passing between two photographic sittings.
“hit rates (correct identification of targets) for adult faces decrease with only 17 months passing between two photographic sittings (Megreya, Sandford & Burton, 2013)”
Quote in the paperSubstantial claim1 of 1 sources checked- cites Megreya, Sandford & Burton, 2013 source
- Premisecitation
Passport officers are no better than the general population on unfamiliar face matching tasks.
“It is worth noting, for example, that passport officers are no better than the general population on such tasks (White et al., 2014).”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites White et al., 2014 source
- Premisecitation
Adults show worse recognition of previously seen infants based on their faces.
“adults showing worse recognition of previously seen infants based on their faces (Chance, Goldstein & Andersen, 1986)”
Quote in the paperSubstantial claim1 of 1 sources checked- cites Chance, Goldstein & Andersen, 1986 source
- Interpretationown data
This result appears to contradict previous work showing that infant and neonate faces could be categorised accurately by sex.
“This result appears to contradict previous work showing that infant and neonate faces could be categorised accurately by sex (Kaminski et al., 2011; Tskhay & Rule, 2016).”
Quote in the paperSubstantial claim1 of 2 sources checked
- Premisecitation
It is important to consider match and mismatch trials separately because previous research has found a dissociation between accuracies on these two trial types.
“However, it was important to consider the match and mismatch trials separately because previous research has found a dissociation between accuracies on these two trial types (Megreya & Burton, 2007).”
Quote in the paperSubstantial claim1 of 1 sources checked- cites Megreya & Burton, 2007 source
- Premisecitation
In previous studies, accuracy levels for sex categorisation of infant and neonate faces were always low, despite being statistically above chance.
“However, in these studies, accuracy levels were always low, despite being statistically above chance.”
Quote in the paperSubstantial claim1 of 2 sources checked - Interpretationreasoning
Infant and child facial photographs are ineffective for use in real-world identification.
“Taken together, and in combination with previous work (White et al., 2015), our results suggest that such low levels of accuracy mean infant and child facial photographs are ineffective for use in real-world identification”
Quote in the paperSubstantial claim1 of 1 sources checked- cites White et al., 2015 source
Experiment 1 Results report 39.0% accuracy for child target images, and its Discussion describes errors for children as especially high. But the study’s child applicants were aged 6–13 (Experiment 1 Method); it did not test infants. The statement that the photographs are ineffective for real-world identification is the citing authors’ synthesis, not a conclusion this source establishes for both infants and children. (LLM, conf=0.86)
Experimental — accuracy not yet measured
- Findingown data
Experiment 2 performance was lower than Experiment 1 performance.
“For Experiment 2, performance was lower still (64% accuracy) when infant and child images were presented, given the significant changes in appearance that occur over the first five years of life.”
Quote in the paperSubstantial claim - Findingown data
d' sensitivity on the GFMT was higher than the results presented here.
“The same pattern was also found for d’ sensitivity, t(55) = 18.25, p < 0.001, Cohen’s d = 2.06.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Burton, White & McNeill, 2010
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Findingown data
Participants were able to judge the sex of children aged 4-5 from their faces.
“As expected, participants were able to judge the sex of children (aged 4–5) from their faces.”
Quote in the paperNot a substantial claim - Findingown data
Accuracy on the GFMT was higher than the results presented here.
“As Table 1 illustrates, accuracy on the GFMT was higher than our results presented here, t(38) = 15.19, p < 0.001, Cohen’s d = 2.49.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Burton, White & McNeill, 2010
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Findingown data
In Experiment 1, response criterion was not significantly different from zero.
“Finally, criterion was not significantly different from zero, t(29) = 0.48, p = 0.632, Cohen’s d = 0.09, suggesting no bias in responses.”
Quote in the paperNot a substantial claim - Findingown data
Criterion was not significantly different between the present study and the GFMT.
“However, criterion was not significantly different in comparison with the GFMT, t(35) = 1.79, p = 0.082, Cohen’s d = 0.34.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Burton, White & McNeill, 2010
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Findingown data
In Experiment 1, there was no difference in percentage correct on mismatch trials where the two identities were the same versus different with respect to sex.
“Interestingly, we found no difference in performance, t(29) = 0.81, p = 0.425, Cohen’s d = 0.12.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 2, percentage correct was significantly higher than chance levels.
“We found that both percentage correct, t(29) = 10.22, p < 0.001, Cohen’s d = 1.87, and d’ sensitivity, t(29) = 9.46, p < 0.001, Cohen’s d = 1.73, were significantly higher than chance levels.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 2, hit rate was significantly better than chance levels.
“In addition, both hit rate, t(29) = 6.67, p < 0.001, Cohen’s d = 1.22, and false alarm rate, t(29) = 7.73, p < 0.001, Cohen’s d = 1.41, were significantly better than chance levels.”
Quote in the paperNot a substantial claim - Findingown data
In Experiments 3a and 3b, there was mixed evidence of within-person correlations in sensitivities across the three tasks.
“along with mixed evidence of within-person correlations in sensitivities across all three tasks”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 4, accuracy when judging the sex of infants was at chance levels for one image set.
“In Experiment 4, accuracy when judging the sex of infants was at chance levels for one image set and above chance (although still low) for the other set.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 2, d' sensitivity was significantly higher than chance levels.
“We found that both percentage correct, t(29) = 10.22, p < 0.001, Cohen’s d = 1.87, and d’ sensitivity, t(29) = 9.46, p < 0.001, Cohen’s d = 1.73, were significantly higher than chance levels.”
Quote in the paperSubstantial claim - Findingown data
The use of only same-sex pairings on mismatch trials, compared with random pairings, had little effect on performance measures.
“The use of only same-sex pairings on mismatch trials, in comparison with random pairings, had little effect on performance measures.”
Quote in the paperNot a substantial claim - Findingown data
In Experiments 3a and 3b, performance was lowest for the two infant tasks compared with adult face matching.
“In Experiments 3a and 3b, when participants completed both these tasks, as well as a measure of adult face matching ability, we found lowest performance for the two infant tasks, along with mixed evidence of within-person correlations in sensitivities across all three tasks.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3b, there was no association between the two infant tasks (r(117) = 0.14, p = 0.143).
“However, there was no association between the two infant tasks, r(117) = 0.14, p = 0.143, or the ‘infant-infant’ task and the GFMT, r(117) = 0.17, p = 0.067.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3b, for mismatch trials, accuracies were lower for the infant-child task than for the other two tasks.
“For mismatch trials, accuracies were lower for the infant-child task in comparison with the other two tasks (both ps < 0.001).”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3a, split-half reliability for mismatch trials was significant for the infant-child task (r(91) = 0.41, p < 0.001).
“Carrying out the same process for mismatch trials, we again found significant associations for both the ‘infant-infant’ task, r(91) = 0.56, p < 0.001, and ‘infant-child’ task, r(91) = 0.41, p < 0.001.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 2, percentage correct was significantly lower than in Experiment 1.
“For this experiment, we found performance levels that were significantly lower than for Experiment 1: percentage correct, t(54) = 4.25, p < 0.001, Cohen’s d = 1.10, and d’ sensitivity, t(57) = 4.27, p < 0.001, Cohen’s d = 1.10.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3a, d' sensitivities for the short GFMT showed a medium-sized correlation with the infant-child task (r(91) = 0.40, p < 0.001).
“Here, we found that d’ sensitivities for the short version of the GFMT showed medium-sized correlations with both the ‘infant-infant’ task, r(91) = 0.35, p = 0.001, and ‘infant-child’ task, r(91) = 0.40, p < 0.001.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3b, there was a significant Task × Trial Type interaction on percentage correct.
“However, these effects were qualified by a significant Task × Trial Type interaction, F(2, 236) = 27.57, p < 0.001, η2p = 0.19.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3b, there was no association between the infant-infant task and the GFMT (r(117) = 0.17, p = 0.067).
“However, there was no association between the two infant tasks, r(117) = 0.14, p = 0.143, or the ‘infant-infant’ task and the GFMT, r(117) = 0.17, p = 0.067.”
Quote in the paperSubstantial claim - Findingown data
Performance was significantly higher than chance for Experiment 2 children, t(39) = 30.72, p < 0.001, Cohen's d = 4.86.
“and Experiment 2 children, t(39) = 30.72, p < 0.001, Cohen’s d = 4.86”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3a, there was a significant effect of task on d' sensitivity, with higher sensitivity on the short GFMT than on the other two tasks.
“We found a significant effect of task, F(2, 184) = 165.87, p < 0.001, η2p = 0.64, with pairwise comparisons revealing that participants showed higher sensitivity on the short version of the GFMT in comparison with the other two tasks (both ps < 0.001).”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3b, percentage correct was significantly lower in the infant-child task than in the infant-infant task.
“In addition, percentage correct scores were significantly lower in the ‘infant-child’ task in comparison with the ‘infant-infant’ task (p < 0.001).”
Quote in the paperSubstantial claim - Findingown data
In Experiment 2, criterion was significantly lower for the GFMT than in this experiment.
“In addition, criterion was significantly lower for the GFMT, t(44) = 2.36, p = 0.023, Cohen’s d = 0.32.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3b, there was a significant effect of task on criterion.
“An analysis of criterion found a significant effect of task, F(2, 236) = 24.78, p < 0.001, η2p = 0.17.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3a, there was a significant Task × Trial Type interaction on percentage correct.
“However, these effects were qualified by a significant Task × Trial Type interaction, F(2, 184) = 34.94, p < 0.001, η2p = 0.28.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3a, there was no main effect of Trial Type on percentage correct.
“A 3 (Task: infant only, infant-child, adult only) × 2 (Trial Type: match, mismatch) within-subjects ANOVA found a significant main effect of Task, F(2, 184) = 165.10, p < 0.001, η2p = 0.64, but no main effect of Trial Type, F(1, 92) = 1.90, p = 0.171, η2p = 0.02.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3b, criterion was significantly higher for the infant-infant task than for the other two tasks.
“Pairwise comparisons revealed that criterion was significantly higher for the ‘infant-infant’ task in comparison with the other two tasks (both ps < 0.001; see Table 1), with these two tasks not differing from each other (p = 0.996).”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3A, performance was only higher for different-sex in comparison with same-sex mismatch trials for Experiment 1 images.
“These results align with the findings of Experiment 3A, where performance was only higher for different-sex in comparison with same-sex mismatch trials for Experiment 1 images.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3b, there was a significant effect of task on d' sensitivity, with higher sensitivity on the short GFMT than on the other two tasks.
“We found a significant effect of task, F(2, 236) = 169.54, p < 0.001, η2p = 0.59, with pairwise comparisons revealing that participants showed higher sensitivity on the short version of the GFMT in comparison with the other two tasks (both ps < 0.001).”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3a, criterion was significantly higher for the infant-infant task than for the other two tasks.
“Pairwise comparisons revealed that criterion was significantly higher for the ‘infant-infant’ task in comparison with the other two tasks (both ps < 0.001; see Table 1), with these two tasks not differing from each other (p = 0.560).”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 2, d' sensitivity on the GFMT was higher than in this experiment.
“The same pattern was also found for d’ sensitivity, t(52) = 22.39, p < 0.001, Cohen’s d = 2.65.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3a, there was a significant effect of task on criterion.
“An analysis of criterion found a significant effect of task, F(2, 184) = 31.86, p < 0.001, η2p = 0.26.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3a, there was a significant difference in accuracy on same-sex versus different-sex mismatch trials for the infant-infant task.
“Paired samples t-tests showed no difference in performance for the ‘infant-child’ task, t(92) = 1.10, p = 0.275, Cohen’s d = 0.11, but a significant difference for the ‘infant-infant’ task, t(92) = 4.47, p < 0.001, Cohen’s d = 0.46.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 2, d' sensitivity was significantly lower than in Experiment 1.
“For this experiment, we found performance levels that were significantly lower than for Experiment 1: percentage correct, t(54) = 4.25, p < 0.001, Cohen’s d = 1.10, and d’ sensitivity, t(57) = 4.27, p < 0.001, Cohen’s d = 1.10.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 2, performance on the GFMT was higher than the percentage correct in this experiment.
“As Table 1 illustrates, performance on the GFMT was higher than our percentage correct presented here, t(34) = 17.36, p < 0.001, Cohen’s d = 3.48.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 2, there was no difference in accuracy on mismatch trials where the two identities were the same versus different with respect to sex.
“For each participant, we calculated their percentage correct on these two trial types, with a paired samples t-test showing no difference in performance, t(29) = 0.69, p = 0.496, Cohen’s d = 0.17.”
Quote in the paperNot a substantial claim - Findingown data
Match trials resulted in poor performance when image characteristics (e.g., lighting direction, image quality, facial expression) significantly differed across the two photographs.
“First, match trials resulted in poor performance when the image characteristics (e.g. lighting direction, image quality, facial expression) significantly differed across the two photographs.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3b, d' sensitivities for the short GFMT and the infant-child task showed a medium-sized correlation (r(117) = 0.26, p = 0.004).
“Here, we found that d’ sensitivities for the short version of the GFMT and the ‘infant-child’ task showed a medium-sized correlation, r(117) = 0.26, p = 0.004.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3b, there was a significant main effect of Trial Type on percentage correct.
“A 3 (Task: infant only, infant-child, adult only) × 2 (Trial Type: match, mismatch) within-subjects ANOVA found a significant main effect of Task, F(2, 236) = 157.17, p < 0.001, η2p = 0.57, and of Trial Type, F(1, 118) = 6.20, p = 0.014, η2p = 0.05.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3a, percentage correct was significantly lower in the infant-child task than in the infant-infant task.
“In addition, percentage correct scores were significantly lower in the ‘infant-child’ task in comparison with the ‘infant-infant’ task (p = 0.003).”
Quote in the paperSubstantial claim - Findingreasoning
Difficult mismatch trials presented two images where superficial characteristics were similar.
“Second, and as a consequence, difficult mismatch trials presented two images where these superficial characteristics were similar.”
Quote in the paperNot a substantial claim - Findingown data
Performance was significantly higher than chance for Experiment 1 infants, t(39) = 7.74, p < 0.001, Cohen's d = 1.22.
“We found that performance was significantly higher than chance levels for both Experiment 1 infants, t(39) = 7.74, p < 0.001, Cohen’s d = 1.22”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3a, d' values were very low and close to chance levels for both infant tasks.
“However, it is important to note that d’ values were very low and close to chance levels for both infant tasks.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3b, for mismatch trials, the adult only and infant only tasks did not differ.
“However, the adult only and infant only tasks did not differ (p = 0.520).”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 2, false alarm rate was significantly better than chance levels.
“In addition, both hit rate, t(29) = 6.67, p < 0.001, Cohen’s d = 1.22, and false alarm rate, t(29) = 7.73, p < 0.001, Cohen’s d = 1.41, were significantly better than chance levels.”
Quote in the paperNot a substantial claim - Findingown data
There was a significant effect of image set on percentage correct scores, F(2, 78) = 271.70, p < 0.001, η2p = 0.87.
“We found a significant effect of image set, F(2, 78) = 271.70, p < 0.001, η2p = 0.87”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3a, d' sensitivities for the short GFMT showed a medium-sized correlation with the infant-infant task (r(91) = 0.35, p = 0.001).
“Here, we found that d’ sensitivities for the short version of the GFMT showed medium-sized correlations with both the ‘infant-infant’ task, r(91) = 0.35, p = 0.001, and ‘infant-child’ task, r(91) = 0.40, p < 0.001.”
Quote in the paperSubstantial claim - Findingown data
Participants' accuracies were no different from chance for the Experiment 2 infants, t(39) = 0.60, p = 0.555, Cohen's d = 0.09.
“However, participants’ accuracies were no different from chance for the Experiment 2 infants, t(39) = 0.60, p = 0.555, Cohen’s d = 0.09.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3a, split-half reliability for match trials was significant for the infant-child task (r(91) = 0.31, p = 0.002).
“As a measure of split-half reliability, we correlated these two accuracies across participants, finding significant associations for both the ‘infant-infant’ task, r(91) = 0.36, p < 0.001, and ‘infant-child’ task, r(91) = 0.31, p = 0.002.”
Quote in the paperSubstantial claim - Findingown data
Participants performed differently across the three image sets (all pairwise ps < 0.001).
“with pairwise comparisons revealing that participants performed differently across the three sets (all ps < 0.001).”
Quote in the paperSubstantial claim - Findingown data
Task performance was significantly lower when a five-year age gap was introduced between the two images (64% accuracy in Experiment 2).
“As we might expect, task performance was significantly lower still when we introduced a five-year age gap between the two images (64%; Experiment 2).”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3a, there was a similar-sized association between the two infant tasks (r(91) = 0.28, p = 0.007).
“A similar-sized association was also found between the two infant tasks, r(91) = 0.28, p = 0.007.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3b, d' sensitivity was significantly lower in the infant-child task than in the infant-infant task.
“In addition, d’ sensitivities were significantly lower in the ‘infant-child’ task in comparison with the ‘infant-infant’ task (p < 0.001).”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3a, split-half reliability for match trials was significant for the infant-infant task (r(91) = 0.36, p < 0.001).
“As a measure of split-half reliability, we correlated these two accuracies across participants, finding significant associations for both the ‘infant-infant’ task, r(91) = 0.36, p < 0.001, and ‘infant-child’ task, r(91) = 0.31, p = 0.002.”
Quote in the paperSubstantial claim - Findingown data
For the infant-child task, changes in hair colour or style were often present.
“For the ‘infant-child’ task, changes in hair colour or style were often present”
Quote in the paperSubstantial claim - Findingown data
In Experiments 1 and 2, participants were no better on mismatch trials when the two identities were of different sexes.
“In Experiments 1 and 2, we found that participants were no better on mismatch trials when the two identities were of different sexes.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3b, there was a significant main effect of Task on percentage correct.
“A 3 (Task: infant only, infant-child, adult only) × 2 (Trial Type: match, mismatch) within-subjects ANOVA found a significant main effect of Task, F(2, 236) = 157.17, p < 0.001, η2p = 0.57, and of Trial Type, F(1, 118) = 6.20, p = 0.014, η2p = 0.05.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3a, participants' accuracies were higher on mismatch trials where a male and a female infant were presented together (M = 68.9%) than on two same-sex infants (M = 59.6%) for the infant-infant image set.
“This result suggests that, in contrast with Experiment 1, participants’ accuracies were higher on mismatch trials where a male and a female infant were presented together (M = 68.9%) in comparison with two same-sex infants (M = 59.6%) for this image set.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 2, criterion did not differ from Experiment 1.
“However, we found no difference between the two experiments with regard to criterion, t(50) = 0.13, p = 0.897, Cohen’s d = 0.03.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3a, d' sensitivity was significantly lower in the infant-child task than in the infant-infant task.
“In addition, d’ sensitivities were significantly lower in the ‘infant-child’ task in comparison with the ‘infant-infant’ task (p = 0.001).”
Quote in the paperNot a substantial claim - Findingown data
Performance levels in Experiments 3A and 3B were closer to chance (52-56%).
“Unsurprisingly, performance levels in these experiments were closer to chance (52–56%).”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3a, there was a significant main effect of Task on percentage correct in the 3x2 ANOVA.
“A 3 (Task: infant only, infant-child, adult only) × 2 (Trial Type: match, mismatch) within-subjects ANOVA found a significant main effect of Task, F(2, 184) = 165.10, p < 0.001, η2p = 0.64, but no main effect of Trial Type, F(1, 92) = 1.90, p = 0.171, η2p = 0.02.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3a, split-half reliability for mismatch trials was significant for the infant-infant task (r(91) = 0.56, p < 0.001).
“Carrying out the same process for mismatch trials, we again found significant associations for both the ‘infant-infant’ task, r(91) = 0.56, p < 0.001, and ‘infant-child’ task, r(91) = 0.41, p < 0.001.”
Quote in the paperSubstantial claim - Findingown data
In Experiment 3b, simple main effects of Task were significant for both match and mismatch trials.
“These simple main effects were significant for both match, F(2, 236) = 82.42, p < 0.001, η2p = 0.41, and mismatch trials, F(2, 236) = 42.98, p < 0.001, η2p = 0.27.”
Quote in the paperNot a substantial claim - Findingown data
In Experiment 3a, there was no difference in accuracy on same-sex versus different-sex mismatch trials for the infant-child task.
“Paired samples t-tests showed no difference in performance for the ‘infant-child’ task, t(92) = 1.10, p = 0.275, Cohen’s d = 0.11, but a significant difference for the ‘infant-infant’ task, t(92) = 4.47, p < 0.001, Cohen’s d = 0.46.”
Quote in the paperSubstantial claim - Findingown data
There was some evidence of within-person correlations in performance across the two infant tasks and the short version of the GFMT.
“We found some evidence of within-person correlations in performance across our two infant tasks and the short version of the GFMT.”
Quote in the paperNot a substantial claim - Premisecitation
The sex of infant faces may be perceived automatically.
“Indeed, the sex of infant faces may be perceived automatically (Tskhay & Rule, 2016).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Tskhay & Rule, 2016 source
- Interpretationreasoning
The poor performance levels with infant matching reported here may represent an upper estimate in applied situations because pairing identities based upon visual similarity would increase the difficulty of mismatch trials.
“As such, the poor performance levels with infant matching reported here may represent an upper estimate in applied situations (e.g. border control), in that pairing identities based upon visual similarity would increase the difficulty of mismatch trials (see Experiment 3).”
Quote in the paperSubstantial claim- Rests on an assumption: Identities are not paired at random in real-world contexts; fraudulent passports would be selected in order to most resemble individuals.
- Premisecitation
Adult participants were able to categorise the sex of 1-24 month old infants at levels above chance.
“adult participants were able to categorise the sex of neonates (Kaminski et al., 2011; Porter, Cernoch & Balogh, 1984; Round & Deheragoda, 2002) and 1–24 month old infants (Tskhay & Rule, 2016) at levels above chance”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Tskhay & Rule, 2016 source
- Interpretationown data
In Experiment 1, participants showed no response bias.
“suggesting no bias in responses”
Quote in the paperNot a substantial claim - Premisecitation
Infant faces appear more alike than adult faces for adult viewers.
“infant faces appear more alike than adult faces for adult viewers (based on subjective ratings; Chance, Goldstein & Andersen, 1986)”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Chance, Goldstein & Andersen, 1986 source
Checked before — its earlier verdict is held back until a new check runs.
- Premisenone given
Only two studies have reported on face matching with images of children.
“we were surprised to find only two studies reporting on face matching with images of children”
Quote in the paperSubstantial claim - Interpretationreasoning
Age gaps of up to five years likely result in significant difficulties in face matching.
“Such age gaps likely result in significant difficulties because infants’ faces change substantially at a young age (Chakravarty et al., 2011; Farkas et al., 1992; Ferrario et al., 1999), and unfamiliar face matching is closely bound to the visual properties of the particular images (Hancock, Bruce & Burton, 2000).”
Quote in the paperSubstantial claim - Premisecitation
Familiar face matching results in rapid and accurate assessments.
“Research has repeatedly shown that deciding whether two different face photographs are of the same person, or whether a person standing in front of you is the same person depicted in a photograph, results in rapid and accurate assessments for familiar faces (Bruce et al., 2001) and inaccurate assessments for unfamiliar faces (Bruce et al., 1999, 2001; Kemp, Towell & Pike, 1997; Megreya & Burton, 2”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Bruce et al., 2001 source
- Premisecitation
Infants' faces change substantially at a young age.
“infants’ faces change substantially at a young age (Chakravarty et al., 2011; Farkas et al., 1992; Ferrario et al., 1999)”
Quote in the paperSubstantial claim0 of 3 sources checked - Interpretationown data
Identity matching with infant and child images resulted in low levels of performance.
“Identity matching with infant and child images resulted in low levels of performance, which were significantly worse than for an adult face matching task.”
Quote in the paperSubstantial claim - Premisecitation
Adult participants were able to categorise the sex of neonates at levels above chance.
“adult participants were able to categorise the sex of neonates (Kaminski et al., 2011; Porter, Cernoch & Balogh, 1984; Round & Deheragoda, 2002) and 1–24 month old infants (Tskhay & Rule, 2016) at levels above chance”
Quote in the paperNot a substantial claim0 of 3 sources checked - Premisecitation
Yadav et al. (2014) found poor face matching performance for full-face images of children (60%, where chance level was 50%) when the two photographs depicted an individual once in the age range 0-5 years and again in the range 6-10 years.
“In a study on machine face recognition, Yadav et al. (2014) found poor face matching performance for full-face images of children (60%, where chance level was 50%) when the two photographs depicted an individual once in the age range 0–5 years and again in the range 6–10 years.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Yadav et al. (2014) source
Checked before — its earlier verdict is held back until a new check runs.
- Premisecitation
Unfamiliar face matching is closely bound to the visual properties of the particular images.
“unfamiliar face matching is closely bound to the visual properties of the particular images (Hancock, Bruce & Burton, 2000)”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Hancock, Bruce & Burton, 2000 source
- Interpretationown data
Child facial photographs are ineffective for use in real-world identification.
“Taken together, the results of the experiments presented here provide evidence that child facial photographs are ineffective for use in real-world identification.”
Quote in the paperSubstantial claim - Premisecitation
Unfamiliar face matching results in inaccurate assessments.
“Research has repeatedly shown that deciding whether two different face photographs are of the same person, or whether a person standing in front of you is the same person depicted in a photograph, results in rapid and accurate assessments for familiar faces (Bruce et al., 2001) and inaccurate assessments for unfamiliar faces (Bruce et al., 1999, 2001; Kemp, Towell & Pike, 1997; Megreya & Burton, 2”
Quote in the paperSubstantial claim0 of 3 sources checked- cites Bruce et al., 1999, 2001
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Kemp, Towell & Pike, 1997
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Megreya & Burton, 2006, 2008
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationreasoning
The lower performance in Experiment 2 is attributable to significant changes in appearance over the first five years of life.
“given the significant changes in appearance that occur over the first five years of life”
Quote in the paperSubstantial claim - Interpretationown data
Border control officers would incorrectly identify 27% (95% CI [23%, 32%]) of the 400,000 infant cases if they were all presented in mismatch contexts.
“At best, this suggests that border control officers would incorrectly identify 27% (95% CI [23%, 32%]) of the 400,000 infant cases mentioned earlier if they were all presented in ‘mismatch’ contexts.”
Quote in the paperSubstantial claim- Rests on an assumption: In many countries, infants are required to travel using their own passports as photographic ID in order to combat trafficking.
- Premisecitation
Larger age gaps result in worse matching performance.
“larger age gaps result in worse matching performance (using three images each of four female students; Seamon, 1982)”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Seamon, 1982 source
- Study design and conductreasoning
In Experiment 1, random pairing of identities on mismatch trials meant that faces could differ in hair colour, eye colour, and even sex.
“Here, the random pairing of identities for mismatch trials meant that it was possible for faces to differ in terms of hair colour, eye colour, and even sex.”
Quote in the paperSubstantial claim - Interpretationown data
Identity matching with infant and child images was significantly worse than for an adult face matching task.
“Identity matching with infant and child images resulted in low levels of performance, which were significantly worse than for an adult face matching task.”
Quote in the paperSubstantial claim - Premisecitation
A benchmark test of unfamiliar face matching found performance levels of around 90%.
“Indeed, a benchmark test of unfamiliar face matching found performance levels of around 90% (Burton, White & McNeill, 2010), representing a ‘best case’ scenario since images were high quality and taken only minutes apart.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Burton, White & McNeill, 2010
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationreasoning
Performance in the current work suggests that any effective method of improving matching will continue to fall short of practical requirements for what is acceptable in terms of accuracy.
“Indeed, performance in the current work certainly suggests that any effective method of improving matching will continue to fall short of practical requirements for what is acceptable in terms of accuracy.”
Quote in the paperNot a substantial claim- Rests on an assumption: Children’s faces are more homogeneous than adult faces, with lower between-face variability.
- Interpretationown data
In Experiment 3A, for infant-infant trials, sex information was beneficial when making same/different judgements.
“However, the results of Experiment 3A suggested that, at least for ‘infant-infant’ trials, sex information was indeed beneficial when making same/different judgements.”
Quote in the paperSubstantial claim - Interpretationreasoning
In Experiment 2, the failure to use sex information was most likely due to the infant faces, since child photographs included hair and clothing information that made sex categorisation fairly easy.
“Most likely, this was a failing with the infant faces (see Experiment 1) since the child photographs (aged 4–5 years) included both hair and some clothing information that made sex categorisation fairly easy.”
Quote in the paperNot a substantial claim - Interpretationown data
Limited information regarding infant sex is present in some cases but not others.
“Taken together, we can conclude only that limited information (accuracy was 57% for Experiment 1 images) regarding infant sex is present in some cases but not others”
Quote in the paperSubstantial claim - Premisecitation
External features (hair, facial outline, etc.) contain less identity information.
“given that the external features (hair, facial outline, etc.) contain less identity information (Kramer et al., 2018).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Kramer et al., 2018 source
- Interpretationreasoning
The accuracy estimates in Experiments 1 and 2 may be higher than those found in similar real-world contexts because characteristics such as sex and hair colour differed on some mismatch trials.
“Importantly, our estimates of accuracy in these two experiments may even be higher than those found in similar real-world contexts since characteristics such as sex, hair colour, etc. differed on some mismatch trials.”
Quote in the paperNot a substantial claim - Interpretationown data
With such poor performance, any apparent associations between tasks in Experiment 3a may simply be due to noise.
“With such poor performance, any apparent associations between tasks may simply be due to noise.”
Quote in the paperSubstantial claim - Interpretationown data
Experiments 1 and 2 provided no evidence that participants were able to perceive or utilise sex information to increase performance on mismatch trials where a male and a female identity were presented together.
“Experiments 1 and 2 provided no evidence that participants were able to perceive or utilise sex information in order to increase performance on mismatch trials where a male and a female identity were presented together.”
Quote in the paperSubstantial claim - Premisecitation
The short version of the GFMT provides a more difficult test of adult face matching.
“Second, we wanted to compare performance directly with the short version GFMT (Burton, White & McNeill, 2010), which provides a more difficult test of adult face matching.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Burton, White & McNeill, 2010
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationreasoning
Border control officers are faced with a significantly error-prone situation when comparing infant photographs to real-world images.
“Taken together, it seems clear that comparing an infant’s photograph to a real-world image (a little more variation than a typical passport photo) is a highly difficult task, suggesting that border control officers are faced with a significantly error-prone situation.”
Quote in the paperSubstantial claim- Rests on an assumption: Performance of MTurk participants on these tasks is indicative of border control officers' performance.
- Interpretationown data
It is more difficult to match identities using infant faces than adult faces.
“here we have demonstrated that, on the whole, it is more difficult to match identities using infant faces than adult faces.”
Quote in the paperSubstantial claim - Premisecitation
Unfamiliar matching relies heavily on the visual properties of the particular images.
“Previous research has demonstrated that unfamiliar matching relies heavily on the visual properties of the particular images (Hancock, Bruce & Burton, 2000)”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Hancock, Bruce & Burton, 2000 source
- Interpretationreasoning
Such images could simply be insufficient for useful identification and matching in real-world situations.
“With less information to distinguish between identities, such images could simply be insufficient for useful identification and matching in real-world situations.”
Quote in the paperSubstantial claim- Rests on an assumption: Children’s faces are more homogeneous than adult faces, with lower between-face variability.
- Interpretationown data
Although not as large as the correlation between adult matching tasks previously reported, these results may suggest that an underlying ability with faces supports both adult and infant matching performance.
“Although not as large as the correlation between adult matching tasks previously reported, these results may suggest that an underlying ability with faces supports both adult and infant matching performance.”
Quote in the paperSubstantial claim - Study design and conductreasoning
The difficulty manipulation in Experiment 3a may not have been as powerful as the one used by Burton, White & McNeill (2010).
“As a result, our difficulty manipulation may not have been as powerful as the one used by Burton, White & McNeill (2010).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Burton, White & McNeill (2010)
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationcitation
The performance level in Experiment 2 (64%) was similar to low levels of accuracy found in previous research investigating matching across age gaps where individuals were depicted once in the age range 0-5 years and again in the range 6-10 years.
“Our performance level here (64%) was similar to the low levels of accuracy found in previous research investigating matching across age gaps where individuals were depicted once in the age range 0–5 years and again in the range 6–10 years (Yadav et al., 2014).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Yadav et al., 2014 source
Checked before — its earlier verdict is held back until a new check runs.
- Interpretationreasoning
Participants found it hard to ignore changes in hair colour or style even though such differences could be expected for an infant over a five-year period.
“suggesting that participants found it hard to ignore these details even though such differences could be expected for an infant over a five-year period.”
Quote in the paperSubstantial claim - Interpretationown data
In Experiment 3A, participants were able to use information regarding the sex of infants to perform more accurately on mismatch trials where a male and a female infant were presented together.
“The results of Experiment 3A suggested that, in contrast with Experiment 1, participants were able to use information regarding the sex of infants in order to perform more accurately on mismatch trials where a male and a female infant were presented together.”
Quote in the paperSubstantial claim - Interpretationown data
Matching two images of infants was significantly more difficult than matching two images of adult faces taken from a university population (around 87-90%).
“and significantly more so than with two images of adult faces taken from a university population (around 87–90%; Bobak, Dowsett & Bate, 2016; Burton, White & McNeill, 2010).”
Quote in the paperSubstantial claim0 of 2 sources checked- cites Bobak, Dowsett & Bate, 2016
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Burton, White & McNeill, 2010
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationreasoning
Image characteristics and external features strongly influenced matching decisions in the current experiments.
“Taken together, it is no surprise that image characteristics and external features strongly influenced matching decisions here.”
Quote in the paperSubstantial claim - Interpretationown data
These findings provide mixed support for previous research where infant sex was shown to be accurately judged from facial images.
“These findings provide mixed support for previous research where infant sex was shown to be accurately judged from facial images (Tskhay & Rule, 2016).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Tskhay & Rule, 2016 source
- Interpretationown data
Comparing an infant's photograph to a real-world image (a little more variation than a typical passport photo) is a highly difficult task.
“Taken together, it seems clear that comparing an infant’s photograph to a real-world image (a little more variation than a typical passport photo) is a highly difficult task, suggesting that border control officers are faced with a significantly error-prone situation.”
Quote in the paperSubstantial claim - Interpretationown data
Adult face matching is easier than infant matching tasks.
“This would provide us with a full replication of the main result (adult face matching is easier than infant matching tasks) while simulating more real-world contexts, where fraudulent passports would be selected in order to most resemble individuals.”
Quote in the paperSubstantial claim - Interpretationown data
Performance in both infant-infant and infant-child matching tasks was noticeably lower than with typical adult face matching tests.
“Performance in both infant-infant and infant-child matching tasks was noticeably lower than with typical adult face matching tests.”
Quote in the paperSubstantial claim - Interpretationown data
The low levels of accuracy observed were only minimally aided (if at all) by the availability of sex category information.
“Despite the randomised pairing of identities, the low levels of accuracy we observed were only minimally aided (if at all) by the availability of sex category information.”
Quote in the paperNot a substantial claim
- Probably not a substantial claim. 27
- It may state two separate claims as one. 48
- The quote only partly states this claim. 3
Probably not a substantial claim.
PremisecitationThe own-age bias seems to be at least partially based on experience.
“This bias seems to be at least partially based on experience (Harrison & Hole, 2009)”
Quote in the paperSubstantial claim1 of 1 sources checked- cites Harrison & Hole, 2009 source
It may state two separate claims as one.
PremisecitationIn White et al. (2015), overall performance on child matching trials was 39%, with accuracy on adolescent (41%) and adult trials (45%) also very poor.
“Overall performance on these trials was 39%, with accuracy on adolescent (41%) and adult trials (45%) also very poor”
Quote in the paperSubstantial claim1 of 1 sources checked- cites White et al. (2015) source
Probably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductcitationSample size in Experiment 2 was based on past research using a face matching paradigm.
“Sample size was again based on past research using a face matching paradigm (Dowsett & Burton, 2015; Estudillo & Bindemann, 2014).”
Quote in the paperNot a substantial claim1 of 2 sources checked- cites Dowsett & Burton, 2015
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Estudillo & Bindemann, 2014 source
Probably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductcitationSample size in Experiment 1 was based on past research using a face matching paradigm.
“Sample size was based on past research using a face matching paradigm (Dowsett & Burton, 2015; Estudillo & Bindemann, 2014).”
Quote in the paperNot a substantial claim1 of 2 sources checked- cites Dowsett & Burton, 2015
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Estudillo & Bindemann, 2014 source
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
PremisecitationIn adult face matching, images taken only minutes apart produced significant levels of error, and performance decreased even further as months passed between photographic sittings.
“Research with adult face matching has shown that images taken only minutes apart produced significant levels of error (Burton, White & McNeill, 2010), with performance decreasing even further as months passed by between photographic sittings (Fysh & Bindemann, 2018; Megreya, Sandford & Burton, 2013).”
Quote in the paperNot a substantial claim1 of 3 sources checked- cites Burton, White & McNeill, 2010
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Fysh & Bindemann, 2018
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Megreya, Sandford & Burton, 2013 source
It may state two separate claims as one.
Findingown dataIn Experiment 1, hit rate and false alarm rate were significantly better than chance levels.
“In addition, both hit rate, t(29) = 9.83, p < 0.001, Cohen’s d = 1.79, and false alarm rate, t(29) = 9.93, p < 0.001, Cohen’s d = 1.81, were significantly better than chance levels.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 1, percentage correct and d' sensitivity were significantly higher than chance levels.
“For this experiment, we found that both percentage correct, t(29) = 19.98, p < 0.001, Cohen’s d = 3.65, and d’ sensitivity, t(29) = 16.44, p < 0.001, Cohen’s d = 3.00, were significantly higher than chance levels.”
Quote in the paperSubstantial claimThe quote only partly states this claim.
Findingown dataIn Experiment 1, mean overall accuracy for matching two infant photos was 72.0% (SD 6.03).
“In Experiment 1, we found low levels of performance (72% accuracy) for matching two infant photos.”
Quote in the paperSubstantial claimThe quote only partly states this claim.
Findingown dataIn Experiment 2, mean overall accuracy for matching an infant photo with a child photo was 64.4% (SD 7.71).
“For Experiment 2, performance was lower still (64% accuracy) when infant and child images were presented, given the significant changes in appearance that occur over the first five years of life.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 4, accuracy when judging the sex of infants was above chance, although still low, for the other image set.
“In Experiment 4, accuracy when judging the sex of infants was at chance levels for one image set and above chance (although still low) for the other set.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 1, matching two images of infants was difficult, with accuracy of 72%.
“Our results provide compelling evidence that matching two images of infants was difficult (72%; Experiment 1)”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3a, for mismatch trials, accuracies on all three tasks significantly differed from each other, in descending order of adult only, infant only, then infant-child.
“For mismatch trials, accuracies on all three tasks also significantly differed from each other (all ps < 0.051), in descending order of adult only, infant only, then infant-child.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3b, for match trials, accuracies on all three tasks significantly differed from each other, in descending order of adult only, infant-child, then infant only.
“Pairwise comparisons showed that, for match trials, accuracies on all three tasks significantly differed from each other (all ps < 0.001), in descending order of adult only, infant-child, then infant only.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3a, criterion was significantly above zero in the infant-infant task, suggesting a bias towards responding 'different'.
“However, this measure was significantly above zero in the ‘infant-infant’ task, t(92) = 5.97, p < 0.001, Cohen’s d = 0.62, suggesting a bias towards responding ‘different’.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataAccuracy was highest for Experiment 2 children (M = 82.3%, SD = 6.6%), followed by Experiment 1 infants (M = 56.7%, SD = 5.5%), and then Experiment 2 infants (M = 50.7%, SD = 7.1%).
“Accuracy was highest for the Experiment 2 children (M = 82.3%, SD = 6.6%), followed by the Experiment 1 infants (M = 56.7%, SD = 5.5%), and then finally the Experiment 2 infants (M = 50.7%, SD = 7.1%).”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3a, percentage correct and d' sensitivity remained significantly higher than chance levels on each task.
“Although performance was very low in the two infant tasks, we did find that both percentage correct (both ps < 0.002) and d’ sensitivity (both ps < 0.003) remained significantly higher than chance levels on each task.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3b, criterion was significantly above zero in the infant-infant task, suggesting a bias towards responding 'different'.
“However, this measure was significantly above zero in the ‘infant-infant’ task, t(118) = 6.33, p < 0.001, Cohen’s d = 0.58, suggesting a bias towards responding ‘different’.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3a, there was a significant effect of task on percentage correct, with participants performing better on the short version of the GFMT than on the other two tasks.
“We found a significant effect of task, F(2, 184) = 165.10, p < 0.001, η2p = 0.64, with pairwise comparisons (Dunn–Šidák corrected here and below) revealing that participants performed better on the short version of the GFMT in comparison with the other two tasks (both ps < 0.001).”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 2, criterion was not significantly different from zero, suggesting no bias in responses.
“Finally, criterion was not significantly different from zero, t(29) = 0.58, p = 0.567, Cohen’s d = 0.11, suggesting no bias in responses.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3a, criterion did not differ from zero in the infant-child task, suggesting no bias.
“Criterion did not differ from zero in the ‘infant-child’ task, t(92) = 1.10, p = 0.273, Cohen’s d = 0.11, suggesting no bias in responses.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3a, for match trials, accuracies on all three tasks significantly differed from each other, in descending order of adult only, infant-child, then infant only.
“Pairwise comparisons showed that, for match trials, accuracies on all three tasks significantly differed from each other (all ps < 0.001), in descending order of adult only, infant-child, then infant only.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3b, there was a significant effect of task on percentage correct, with participants performing better on the short GFMT than on the other two tasks.
“We found a significant effect of task, F(2, 236) = 157.17, p < 0.001, η2p = 0.57, with pairwise comparisons revealing that participants performed better on the short version of the GFMT in comparison with the other two tasks (both ps < 0.001).”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3b, percentage correct and d' sensitivity remained significantly higher than chance on each task.
“Although performance was very low in the two infant tasks, we did find that both percentage correct (both ps < 0.030) and d’ sensitivity (both ps < 0.037) remained significantly higher than chance on each task.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3b, criterion did not differ from zero in the infant-child task, suggesting no bias.
“Criterion did not differ from zero in the ‘infant-child’ task, t(118) = 0.44, p = 0.658, Cohen’s d = 0.04, suggesting no bias in responses.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataIn Experiment 3a, simple main effects of Task were significant for both match and mismatch trials.
“These simple main effects were significant for both match, F(2, 184) = 120.34, p < 0.001, η2p = 0.57, and mismatch trials, F(2, 184) = 47.06, p < 0.001, η2p = 0.34.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataThe associations either decreased or were absent when the experiment was repeated in Experiment 3B.
“Indeed, these associations either decreased or were absent when the experiment was repeated (Experiment 3B).”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Findingown dataThe majority of participants in Experiments 1 and 2 were women.
“We note that the majority of participants in Experiments 1 and 2 were women.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationUp to 400,000 children are trafficked across international borders annually.
“It is estimated that up to 400,000 children are trafficked across international borders annually (U.S. Department of State, 2007).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites U.S. Department of State, 2007
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
The quote only partly states this claim.
InterpretationreasoningThe present study will likely result in inflated estimates of accuracy because random pairing of identities on mismatch trials allowed faces to differ in hair colour, eye colour, and sex.
“As such, the present study will likely result in inflated estimates of accuracy.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductreasoningExperiment 2 used photos of celebrities' children because no database containing photos of infants across the age range 0-5 years was known.
“As we were unaware of any database containing photos of infants across the age range 0–5 years, we elected to use photos of celebrities’ children due to their widespread availability online.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationThe quality of exposure to infant faces may be more important than the quantity.
“although the quality of exposure to infant faces may be more important than the quantity (Yovel et al., 2012)”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Yovel et al., 2012 source
Probably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationBaby-selling and illegal adoption have been reported in Europe, Africa, Central and South America, Central Asia and East Asia.
“Furthermore, baby-selling and illegal adoption have been reported in Europe, Africa, Central and South America, Central Asia and East Asia (UNODC, 2016).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites UNODC, 2016
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
Study design and conductreasoningThe final image set for Experiment 1 comprised 22 male and 19 female infants (age range 3-11 months), each with two different images that met international standards for passport portraits.
“Our final image set comprised 22 male and 19 female infants (age range: 3–11 months), each with two different images that met international standards for passport portraits.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Premisenone givenNeither of the previous two studies specifically examined face matching for infants or focused on the age range considered valid for passports.
“Importantly, neither of these previous studies specifically examined face matching for infants or focussed on the age range considered valid for passports.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductreasoningThe approach in Experiment 2 resulted in less control over the photographs with regard to pose and facial expression.
“This approach resulted in less control over the photographs with regard to pose and facial expression.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may state two separate claims as one.
PremisecitationAn own-race bias is present in both face recognition and matching tasks.
“evidence has shown that an own-race bias is present in both face recognition (Meissner & Brigham, 2001) and matching tasks (Megreya, White & Burton, 2011; Meissner, Susa & Ross, 2013)”
Quote in the paperSubstantial claim0 of 3 sources checkedProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductnone givenThere was no overlap between the Experiment 2 sample and the Experiment 1 sample.
“There was no overlap between this sample and those who participated in Experiment 1.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
Study design and conductnone givenAll participants in Experiments 1 and 2 provided written informed consent and were verbally debriefed at the end of the experiment.
“All participants in Experiments 1 and 2 provided written informed consent and were verbally debriefed at the end of the experiment.”
Quote in the paperNot a substantial claimProbably not a substantial claim.
Study design and conductreasoningIn Experiments 1 and 2, there was no attempt to make it difficult to detect matches and mismatches in the facial identities presented.
“In these experiments, there was no attempt to make it difficult to detect matches and mismatches in the facial identities presented (e.g. identity pairs were not systematically matched for sex, hair colour, etc.).”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Interpretationown dataEither participants were unable to categorise our infant faces by sex or they neglected to use this information when making same/different judgements.
“Therefore, either participants were unable to categorise our infant faces by sex or they neglected to use this information when making same/different judgements.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
PremisecitationSex characteristics are more evident after puberty and are likely subtle where present in infants.
“However, as mentioned earlier, sex characteristics are more evident after puberty and are likely subtle where present in infants.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Enlow, 1982
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
Probably not a substantial claim.
PremisecitationViewers show an own-age bias when recognising faces, being better at recognising previously seen faces of their own age group.
“viewers show an own-age bias when recognising faces, where people are better at recognising previously seen faces of one’s own age group (Rhodes & Anastasi, 2012)”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Rhodes & Anastasi, 2012 source
It may state two separate claims as one.
PremisecitationCraniofacial shape cues to an identity's sex are more pronounced after puberty, resulting in prepubescent children displaying less between-face variability.
“For example, craniofacial shape cues to an identity’s sex are more pronounced after puberty (Enlow, 1982), resulting in prepubescent children displaying less between-face variability.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Enlow, 1982
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
Study design and conductnone givenTrent University's ethics committee approved all experiments presented here (ref: 22305), and they were carried out in accordance with the World Medical Association Declaration of Helsinki.
“Trent University’s ethics committee approved all experiments presented here (ref: 22305), which were carried out in accordance with the provisions of the World Medical Association Declaration of Helsinki.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Study design and conductreasoningIn Experiment 1, there was no attempt to pair identities in mismatch trials based upon visual similarity, permitting an upper-bound for accuracy.
“There was no attempt to pair the identities in mismatch trials based upon visual similarity, permitting us to establish an upper-bound for accuracy.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Interpretationown dataThese results cast doubt on the correlational findings of Experiment 3A, suggesting that any underlying face matching ability may be weaker than was found earlier.
“These results cast doubt on the correlational findings of Experiment 3A, suggesting that any underlying face matching ability may be weaker than was found earlier.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Study design and conductown dataNo data were excluded from the analysis because the minimum percentage correct score for the children's images was 67%, and there was no reason to believe any participants were not paying attention.
“No data were excluded because the minimum percentage correct score for the children’s images (aged 4–5 years old and hence expected to be easier to judge accurately) was 67%. We therefore had no reason to believe that any participants were not paying attention during the experiment.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationReliance on internal facial features (eyes, nose, and mouth) increases as we become more familiar with a face.
“evidence suggests an increasing reliance on the internal facial features (eyes, nose, and mouth) as we become more familiar with a face (Clutterbuck & Johnston, 2002, 2004, 2005)”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Clutterbuck & Johnston, 2002, 2004, 2005
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationBobak, Dowsett & Bate (2016) reported a correlation of 0.72 between participants' d' values on the GFMT and a second face matching task.
“For instance, Bobak, Dowsett & Bate (2016) reported a correlation of 0.72 between participants’ d’ values on the GFMT and a second face matching task.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Bobak, Dowsett & Bate (2016)
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
It may state two separate claims as one.
PremisecitationRecent adult matching tasks specifically constructed to be challenging showed higher levels of accuracy: 66% for the short version of the Kent face matching test, 72% for matching with male models, and 83% for other-race faces.
“Importantly, even recent adult matching tasks, specifically constructed to be challenging, showed higher levels of accuracy—66% for the short version of the Kent face matching test (Fysh & Bindemann, 2018), 72% for matching with male models (Dowsett & Burton, 2015), and 83% for other-race faces (Kokje, Bindemann & Megreya, 2018).”
Quote in the paperSubstantial claim0 of 3 sources checked- cites Fysh & Bindemann, 2018
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Dowsett & Burton, 2015
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Kokje, Bindemann & Megreya, 2018 source
It may state two separate claims as one.
Interpretationown dataIn Experiment 2, there was no evidence that participants were able to categorise faces by sex and/or use this information when matching.
“Again, we found no evidence that participants were able to categorise our faces by sex and/or use this information when matching.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
Study design and conductown dataIn Experiment 2, very few participants had regular contact with infants, so this factor could not be explored in analyses.
“Unfortunately, as in Experiment 1, very few of our (university student) sample had such experience, and so we were unable to explore this further in our analyses.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Interpretationown dataAt its most difficult, using face photographs of infants provided almost no useful information, with accuracies significantly lower than those found with adult faces comparatively selected for difficulty.
“This clearly demonstrated that, at its most difficult, using face photographs of infants provided almost no useful information, with accuracies significantly lower than those found with adult faces comparatively selected for difficulty.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningIn Experiment 3b, the low level of performance for the infant tasks and the possibility of a floor effect provide a restricted range with which to investigate any associations, with the risk that any apparent relationships may simply be the result of noise.
“As discussed above, the low level of performance for the infant tasks, and the possibility of a floor effect, provide a restricted range with which to investigate any associations, with the risk that any apparent relationships may simply be the result of noise.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductown dataIn Experiment 3b, data from 11 participants were excluded because they responded incorrectly to one or both checking trials.
“Data from 11 participants were excluded because they responded incorrectly to one or both of the ‘checking’ trials.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationPerformance with matching unfamiliar adult faces can be improved by displaying only the internal features.
“Indeed, there is some evidence to suggest that performance with matching unfamiliar adult faces can be improved by displaying only the internal features (Kemp et al., 2016).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Kemp et al., 2016 source
It may state two separate claims as one.
PremisecitationThe advantage of displaying only internal features was limited to the most difficult trials only and failed to generalise to a card sorting task.
“However, this advantage was limited to the most difficult trials only and failed to generalise to a card sorting task (Kramer et al., 2018).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Kramer et al., 2018 source
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
Study design and conductown dataIn Experiment 2, one participant recognised only one identity, so no data were excluded from analyses.
“One participant responded that they had only recognised one identity. We therefore decided not to exclude any data from the subsequent analyses.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Interpretationown dataSex information was present in the Experiment 1 infant and Experiment 2 children image sets.
“suggesting that sex information was present in these image sets.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Study design and conductreasoningIn Experiment 3a, selecting difficult mismatch trials was problematic because identities were paired randomly, so specific pairings occurred infrequently and accuracies were based on small numbers of observations.
“However, this was more problematic for mismatch pairs because identities were paired randomly in these experiments. As such, specific pairings occurred infrequently and accuracies were therefore based on small numbers of observations.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductown dataIn Experiment 3a, data from 21 participants were excluded because they responded incorrectly to one or both checking trials.
“Data from 21 participants were excluded because they responded incorrectly to one or both of the ‘checking’ trials.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Interpretationown dataParticipants may have benefitted to some degree on different-sex trials, but this advantage failed to produce any noticeable gain in accuracy.
“Therefore, participants in the current work may have benefitted to some degree on different-sex trials, but this advantage failed to produce any noticeable gain in accuracy (see Experiments 3A versus 3B in Table 1)”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
PremisecitationSuper-recognisers are remarkably good at both recognition and matching with adult faces.
“While these individuals are remarkably good at both recognition and matching with adult faces (Bobak, Hancock & Bate, 2016)”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Bobak, Hancock & Bate, 2016
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
It may state two separate claims as one.
Interpretationown dataThe correlations were small to medium in Experiment 3A, unlike the large correlations found in previous work for two adult face matching tasks.
“However, these small to medium effects (Experiment 3A) were unlike the large correlations found in previous work when researchers considered two tasks of adult face matching (Bobak, Dowsett & Bate, 2016).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Bobak, Dowsett & Bate, 2016
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
It may state two separate claims as one.
PremisecitationThe quality of exposure may be crucial, with improvement found only when people are required to individuate faces of a particular category.
“Importantly though, evidence suggests that the quality of exposure may be crucial (Yovel et al., 2012), with improvement found only when people are required to individuate faces of a particular category.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Yovel et al., 2012 source
Probably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationExperience with infants may result in improved infant face recognition.
“While previous research suggests that experience with infants may result in improved infant face recognition (Cassia et al., 2009)”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Cassia et al., 2009 source
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
PremisecitationWomen, but not men, demonstrate an own-sex advantage on match trials, and women perform better than men on mismatch trials depicting either sex of face.
“Previous research has shown that women, but not men, demonstrate an own-sex advantage on match trials, and it also seems that women perform better than men on mismatch trials depicting either sex of face (Megreya, Bindemann & Havard, 2011).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Megreya, Bindemann & Havard, 2011 source
It may state two separate claims as one.
Interpretationown dataSex information is present in some image sets, although even in those cases performance remained close to chance levels.
“The results of Experiments 3A and 4 provided additional evidence more in line with past research, suggesting that sex information is present in some image sets, although even in those cases, performance remained close to chance levels.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningThe low correlations can be explained by the generally low performance on the infant tasks, which introduces restricted range issues and raises the possibility that much of the variability in performance is due to noise.
“However, the low correlations can also be explained by the generally low performance on the infant tasks. The low accuracy introduces restricted range issues and also raises the possibility that much of the variability in performance is due to noise.”
Quote in the paperSubstantial claim
The paper relies on this without support.
Passports for infants and children are typically valid for five years in Canada, the United Kingdom, the United States and Australia, and many countries allow them to be used for five years before expiration.
“Infants and children travel using passports that are typically valid for five years (e.g. Canada, United Kingdom, United States and Australia).”
“Many countries that require infants to have their own passports allow them to be used for five years before expiration (e.g. Canada, United Kingdom, United States and Australia).”
The paper relies on this without support.
Infants and children may need to be identified using images taken from videos and other sources in forensic situations including child exploitation cases.
“These individuals may also need to be identified using images taken from videos and other sources in forensic situations including child exploitation cases.”
The paper relies on this without support.
In many countries, infants are required to travel using their own passports as photographic ID in order to combat trafficking.
“In many countries, infants are required to travel using their own passports as photographic ID in order to combat trafficking, and so it is important to determine the efficacy with which infants can be identified using these images.”
The paper relies on this without support.
Identities are not paired at random in real-world contexts; fraudulent passports would be selected in order to most resemble individuals.
“Importantly, and as discussed below, identities are not paired at random in real-world contexts, where fraudulent passports would be selected in order to most resemble individuals.”
The paper relies on this without support.
Professionals are required to carry out face matching comparisons with images that were taken up to five years beforehand.
“In reality, professionals are required to carry out face matching comparisons with images that were taken up to five years beforehand.”
The paper relies on this without support.
Evidence collected in child abuse cases often features more unconstrained images than passports allow.
“Similarly, evidence collected in child abuse cases often features more unconstrained images than passports allow.”
The paper relies on this without support.
Comparing two passport-style images mirrors passport replacement and renewal procedures, which are typically carried out online or by post and do not involve live matching to a person.
“Comparing two passport-style images mirrors passport replacement and renewal procedures, which are typically carried out online or by post and do not involve ‘live matching’ to a person.”
The paper relies on this without support.
The child photographs, including minor deviations from passport guidelines and uncropped, mirror real-world appearances that might be presented in border control contexts.
“We felt that some minor deviations from passport photograph guidelines were acceptable for this task and mirrored real-world appearances that might be presented in border control contexts.”
The paper relies on this without support.
Performance of MTurk participants on these tasks is indicative of border control officers' performance.
“Taken together, it seems clear that comparing an infant’s photograph to a real-world image (a little more variation than a typical passport photo) is a highly difficult task, suggesting that border control officers are faced with a significantly error-prone situation.”
The paper relies on this without support.
Selecting the most difficult trials from previous participants mimics realistic cases of fraud where identities are typically selected to be the most likely to fool authorities.
“By selecting the most difficult trials in our infant and child matching tasks, we mimic realistic cases of fraud where identities are typically selected to be the most likely to fool the authorities.”
The paper relies on this without support.
Fraudsters would select identity pairings on sex if sex information is available in infant facial photographs.
“If sex information is available in infant facial photographs then identity pairings would certainly be matched by fraudsters on this dimension.”
The paper relies on this without support.
Fraudulent passports would be selected in order to most resemble individuals.
“where fraudulent passports would be selected in order to most resemble individuals.”
The paper relies on this without support.
People who use fraudulent passports choose the person to pair with a document/photograph based on facial similarity, making mismatches harder to spot.
“For those who make use of fraudulent passports (altered in some way, or simply not their own), the choice of who should be paired with which document/photograph will be driven by facial similarity, which makes the job of spotting mismatches that much harder.”
The paper relies on this without support.
Children’s faces are more homogeneous than adult faces, with lower between-face variability.
“This is because children’s faces appear to be more homogeneous than adult faces, displaying lower levels of between-face variability.”
The paper relies on this without support.
The majority of undergraduate university students have minimal experience with infants.
“Unfortunately, the majority of undergraduate university students have minimal experience with infants.”
The paper relies on this without support.
This is the first focused investigation of the utility of infant and child facial photographs for use in identification.
“Our experiments represent the first focussed investigation of the utility of infant and child facial photographs for use in identification.”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Only two studies to date have investigated the performance of typical adults when matching children's faces, and difficulties were apparent in both.
“although only two studies to date have investigated the performance of typical adults when matching children’s faces, difficulties were apparent in both.”
From Enhanced Matching of Children’s Faces in “Super-Recognisers” But Not High-Contact Controls (2020) · cites this paper as Kramer et al., 2018
Kramer et al. (2018) found poorer performance when pairs comprised two photographs of infants (Experiment 1) or one infant and one child face (Experiments 2 and 3), than when pairs comprised two adult faces.
“In a simultaneous matching task, Kramer et al. (2018) also found poorer performance when pairs comprised two photographs of infants (Experiment 1) or one infant and one child face (Experiments 2 and 3), than when pairs comprised two adult faces.”
From Enhanced Matching of Children’s Faces in “Super-Recognisers” But Not High-Contact Controls (2020) · cites this paper as Kramer et al., 2018
Statistics recomputed
Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.
- Consistentt(29) = 19.98reported p < .001 · recomputed p < .001
“t(29) = 19.98, p < .001”
- Consistentt(29) = 16.44reported p < .001 · recomputed p < .001
“t(29) = 16.44, p < .001”
- Consistentt(29) = 9.83reported p < .001 · recomputed p < .001
“t(29) = 9.83, p < .001”
- Consistentt(29) = 9.93reported p < .001 · recomputed p < .001
“t(29) = 9.93, p < .001”
- Consistentt(29) = 0.48reported p = .632 · recomputed p = .635
“t(29) = 0.48, p = .632”
- Consistentt(38) = 15.19reported p < .001 · recomputed p < .001
“t(38) = 15.19, p < .001”
- Consistentt(55) = 18.25reported p < .001 · recomputed p < .001
“t(55) = 18.25, p < .001”
- Consistentt(35) = 1.79reported p = .082 · recomputed p = .082
“t(35) = 1.79, p = .082”
- Consistentt(29) = 0.81reported p = .425 · recomputed p = .425
“t(29) = 0.81, p = .425”
- Consistentt(29) = 10.22reported p < .001 · recomputed p < .001
“t(29) = 10.22, p < .001”
- Consistentt(29) = 9.46reported p < .001 · recomputed p < .001
“t(29) = 9.46, p < .001”
- Consistentt(29) = 6.67reported p < .001 · recomputed p < .001
“t(29) = 6.67, p < .001”
- Consistentt(29) = 7.73reported p < .001 · recomputed p < .001
“t(29) = 7.73, p < .001”
- Consistentt(29) = 0.58reported p = .567 · recomputed p = .566
“t(29) = 0.58, p = .567”
- Consistentt(54) = 4.25reported p < .001 · recomputed p < .001
“t(54) = 4.25, p < .001”
- Consistentt(57) = 4.27reported p < .001 · recomputed p < .001
“t(57) = 4.27, p < .001”
- Consistentt(50) = 0.13reported p = .897 · recomputed p = .897
“t(50) = 0.13, p = .897”
- Consistentt(34) = 17.36reported p < .001 · recomputed p < .001
“t(34) = 17.36, p < .001”
- Consistentt(52) = 22.39reported p < .001 · recomputed p < .001
“t(52) = 22.39, p < .001”
- Consistentt(44) = 2.36reported p = .023 · recomputed p = .023
“t(44) = 2.36, p = .023”
- Consistentt(29) = 0.69reported p = .496 · recomputed p = .496
“t(29) = 0.69, p = .496”
- ConsistentF(2, 184) = 165.1reported p < .001 · recomputed p < .001
“F(2, 184) = 165.10, p < .001”
- ConsistentF(1, 92) = 1.9reported p = .171 · recomputed p = .171
“F(1, 92) = 1.90, p = .171”
- ConsistentF(2, 184) = 34.94reported p < .001 · recomputed p < .001
“F(2, 184) = 34.94, p < .001”
- ConsistentF(2, 184) = 120.34reported p < .001 · recomputed p < .001
“F(2, 184) = 120.34, p < .001”
- ConsistentF(2, 184) = 47.06reported p < .001 · recomputed p < .001
“F(2, 184) = 47.06, p < .001”
- ConsistentF(2, 184) = 165.87reported p < .001 · recomputed p < .001
“F(2, 184) = 165.87, p < .001”
- ConsistentF(2, 184) = 31.86reported p < .001 · recomputed p < .001
“F(2, 184) = 31.86, p < .001”
- Consistentt(92) = 1.1reported p = .273 · recomputed p = .274
“t(92) = 1.10, p = .273”
- Consistentt(92) = 5.97reported p < .001 · recomputed p < .001
“t(92) = 5.97, p < .001”
- Consistentt(92) = 1.1reported p = .275 · recomputed p = .274
“t(92) = 1.10, p = .275”
- Consistentt(92) = 4.47reported p < .001 · recomputed p < .001
“t(92) = 4.47, p < .001”
- Consistentr(91) = 0.36reported p < .001 · recomputed p < .001
“r(91) = 0.36, p < .001”
- Consistentr(91) = 0.31reported p = .002 · recomputed p = .002
“r(91) = 0.31, p = .002”
- Consistentr(91) = 0.56reported p < .001 · recomputed p < .001
“r(91) = 0.56, p < .001”
- Consistentr(91) = 0.41reported p < .001 · recomputed p < .001
“r(91) = 0.41, p < .001”
- Consistentr(91) = 0.35reported p = .001 · recomputed p < .001
“r(91) = 0.35, p = .001”
- Consistentr(91) = 0.4reported p < .001 · recomputed p < .001
“r(91) = 0.40, p < .001”
- Consistentr(91) = 0.28reported p = .007 · recomputed p = .007
“r(91) = 0.28, p = .007”
- ConsistentF(2, 236) = 157.17reported p < .001 · recomputed p < .001
“F(2, 236) = 157.17, p < .001”
- ConsistentF(1, 118) = 6.2reported p = .014 · recomputed p = .014
“F(1, 118) = 6.20, p = .014”
- ConsistentF(2, 236) = 27.57reported p < .001 · recomputed p < .001
“F(2, 236) = 27.57, p < .001”
- ConsistentF(2, 236) = 82.42reported p < .001 · recomputed p < .001
“F(2, 236) = 82.42, p < .001”
- ConsistentF(2, 236) = 42.98reported p < .001 · recomputed p < .001
“F(2, 236) = 42.98, p < .001”
- ConsistentF(2, 236) = 169.54reported p < .001 · recomputed p < .001
“F(2, 236) = 169.54, p < .001”
- ConsistentF(2, 236) = 24.78reported p < .001 · recomputed p < .001
“F(2, 236) = 24.78, p < .001”
- Consistentt(118) = 0.44reported p = .658 · recomputed p = .661
“t(118) = 0.44, p = .658”
- Consistentt(118) = 6.33reported p < .001 · recomputed p < .001
“t(118) = 6.33, p < .001”
- Consistentr(117) = 0.26reported p = .004 · recomputed p = .004
“r(117) = 0.26, p = .004”
- Consistentr(117) = 0.14reported p = .143 · recomputed p = .129
“r(117) = 0.14, p = .143”
- Consistentr(117) = 0.17reported p = .067 · recomputed p = .065
“r(117) = 0.17, p = .067”
- ConsistentF(2, 78) = 271.7reported p < .001 · recomputed p < .001
“F(2, 78) = 271.70, p < .001”
- Consistentt(39) = 7.74reported p < .001 · recomputed p < .001
“t(39) = 7.74, p < .001”
- Consistentt(39) = 30.72reported p < .001 · recomputed p < .001
“t(39) = 30.72, p < .001”
- Consistentt(39) = 0.6reported p = .555 · recomputed p = .552
“t(39) = 0.60, p = .555”
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Child-trafficking estimate is described as infant cases
MinorThe earlier estimate concerns children trafficked across borders, not infants specifically. The later passage incorrectly describes those 400,000 cases as infant cases.
“At best, this suggests that border control officers would incorrectly identify 27% (95% CI [23%, 32%]) of the 400,000 infant cases mentioned earlier if they were all presented in ‘mismatch’ contexts.”
“It is estimated that up to 400,000 children are trafficked across international borders annually (U.S. Department of State, 2007).”
Citations
Cites
Thirty years of investigating the own-race bias in memory for faces: a meta-analytic review
Meissner, C. A., & Brigham, J. C. (2001) · Psychology, Public Policy, and Law · doi:10.1037/1076-8971.7.1.3 · Cited by 1,204
Office to Monitor and Combat Trafficking in Persons (2006) · Trends in Organized Crime · doi:10.1007/s12117-006-1022-7 · Cited by 876
Enlow, D. H., & BOSTWICK, J. (1977) · Plastic & Reconstructive Surgery · doi:10.1097/00006534-197701000-00025 · Cited by 499
The Glasgow Face Matching Test
Burton, A. M., White, D., & McNeill, A. (2010) · Behavior Research Methods · doi:10.3758/brm.42.1.286 · Cited by 472
Recognition of unfamiliar faces
Hancock, P. J. B., Bruce, V., & Burton, A. M. (2000) · Trends in Cognitive Sciences · doi:10.1016/s1364-6613(00)01519-9 · Cited by 464
Super-recognizers: People with extraordinary face recognition ability
Russell, R., Duchaine, B., & Nakayama, K. (2009) · Psychonomic Bulletin & Review · doi:10.3758/pbr.16.2.252 · Cited by 429
Unfamiliar faces are not faces: Evidence from a matching task
Megreya, A. M., & Burton, A. M. (2006) · Memory & Cognition · doi:10.3758/bf03193433 · Cited by 361
The own-age bias in face recognition: a meta-analytic and theoretical review
Rhodes, M. G., & Anastasi, J. S. (2012) · Psychological Bulletin · doi:10.1037/a0025750 · Cited by 346
Verification of face identities from images captured on video.
Bruce, V., Henderson, Z., Greenwood, K., Hancock, P. J. B., Burton, A. M., & Miller, P. (1999) · Journal of Experimental Psychology: Applied · doi:10.1037/1076-898x.5.4.339 · Cited by 333
Passport officers’ errors in face matching
White, D., Kemp, R. I., Jenkins, R., Matheson, M., & Burton, A. M. (2014) · PLoS ONE · doi:10.1371/journal.pone.0103510 · Cited by 241
Matching identities of familiar and unfamiliar faces caught on CCTV images
Bruce, V., Henderson, Z., Newman, C., & Burton, A. M. (2001) · Journal of Experimental Psychology: Applied · doi:10.1037/1076-898x.7.3.207 · Cited by 233
Hayward, A. F. (1983) · Journal of Dentistry · doi:10.1016/0300-5712(83)90224-5 · Cited by 221
Matching faces to photographs: Poor performance in eyewitness memory (without the memory).
Megreya, A. M., & Burton, A. M. (2008) · Journal of Experimental Psychology Applied · doi:10.1037/a0013464 · Cited by 211
Face Recognition by Metropolitan Police Super-Recognisers
Robertson, D. J., Noyes, E., Dowsett, A. J., Jenkins, R., & Burton, A. M. (2016) · PLoS ONE · doi:10.1371/journal.pone.0150036 · Cited by 197
Super‐recognisers in Action: Evidence from Face‐matching and Face Memory Tasks
Bobak, A. K., Hancock, P., & Bate, S. (2015) · Applied Cognitive Psychology · doi:10.1002/acp.3170 · Cited by 157
Exploring Levels of Face Familiarity by Using an Indirect Face-Matching Measure
Clutterbuck, R., & Johnston, R. (2002) · Perception · doi:10.1068/p3335 · Cited by 150
Error Rates in Users of Automatic Face Recognition Software
White, D., Dunn, J. D., Schmid, A. C., & Kemp, R. I. (2015) · PLOS ONE · doi:10.1371/journal.pone.0139827 · Cited by 141
Growth Patterns of the Nasolabial Region: A Morphometric Study
Farkas, L. G., Posnick, J. C., Hreczko, T. M., & Pron, G. E. (1992) · The Cleft Palate-Craniofacial Journal · doi:10.1597/1545-1569_1992_029_0318_gpotnr_2.3.co_2 · Cited by 128
Hits and false positives in face matching: a familiarity-based dissociation
Megreya, A. M., & Burton, A. M. (2007) · Perception & Psychophysics · doi:10.3758/bf03193954 · Cited by 124
Fysh, M. C., & Bindemann, M. (2018) · British Journal of Psychology · doi:10.1111/bjop.12260 · Cited by 118
Demonstrating how unfamiliar faces become familiar using a face matching task
Clutterbuck, R., & Johnston, R. (2004) · The European Journal of Cognitive Psychology · doi:10.1080/09541440340000439 · Cited by 115
Evidence for a contact-based explanation of the own-age bias in face recognition
Harrison, V., & Hole, G. J. (2009) · Psychonomic Bulletin & Review · doi:10.3758/pbr.16.2.264 · Cited by 112
Matching as an index of face familiarity
Clutterbuck, R., & Johnston, R. (2004) · Visual Cognition · doi:10.1080/13506280444000021 · Cited by 101
Matching face images taken on the same day or months apart: the limitations of photo ID
Megreya, A. M., Sandford, A., & Burton, A. M. (2013) · Applied Cognitive Psychology · doi:10.1002/acp.2965 · Cited by 92
Bobak, A. K., Dowsett, A. J., & Bate, S. (2016) · PLOS ONE · doi:10.1371/journal.pone.0148148 · Cited by 85
The Other-Race Effect does not Rely on Memory: Evidence from a Matching Task
Megreya, A. M., White, D., & Burton, A. M. (2011) · Quarterly Journal of Experimental Psychology · doi:10.1080/17470218.2011.575228 · Cited by 84
Unfamiliar face matching: Pairs out-perform individuals and provide a route to training
Dowsett, A. J., & Burton, A. M. (2015) · British Journal of Psychology · doi:10.1111/bjop.12103 · Cited by 77
Disguising Superman: How Glasses Affect Unfamiliar Face Matching
Kramer, R. S. S., & Ritchie, K. L. (2016) · Applied Cognitive Psychology · doi:10.1002/acp.3261 · Cited by 66
Self-citationGlobal Report on Trafficking in Persons 2016
United Nations Office on Drugs and Crime (2017) · Global report on trafficking in persons · doi:10.18356/6b57c9bb-en · Cited by 63
Sex differences in unfamiliar face identification: Evidence from matching tasks
Megreya, A. M., Bindemann, M., & Havard, C. (2011) · Acta Psychologica · doi:10.1016/j.actpsy.2011.03.003 · Cited by 62
Generalization across view in face memory and face matching
Estudillo, A. J., & Bindemann, M. (2014) · i-Perception · doi:10.1068/i0669 · Cited by 56
Cassia, V. M., Picozzi, M., Kuefner, D., & Casati, M. (2009) · Quarterly Journal of Experimental Psychology · doi:10.1080/17470210802617654 · Cited by 53
Kramer, R. S. S., Manesi, Z., Towler, A., Reynolds, M. G., & Burton, A. M. (2017) · Perception · doi:10.1177/0301006617725242 · Cited by 50
Self-citationThe City Infant Faces Database: A validated set of infant facial expressions
Webb, R., Ayers, S., & Endress, A. D. (2017) · Behavior Research Methods · doi:10.3758/s13428-017-0859-9 · Cited by 37
Can I see your passport please? Perceptual discrimination of own-and other-race faces
Meissner, C. A., Susa, K. J., & Ross, A. B. (2013) · Visual Cognition · doi:10.1080/13506285.2013.832451 · Cited by 36
Cross-race correlations in the abilities to match unfamiliar faces
Kokje, E., Bindemann, M., & Megreya, A. M. (2018) · Acta Psychologica · doi:10.1016/j.actpsy.2018.01.006 · Cited by 29
Recognition of neonates by facial-visual characteristics
Porter, R. H., Cernoch, J. M., & Balogh, R. D. (1984) · Pediatrics · doi:10.1542/peds.74.4.501 · Cited by 28
Ferrario, V. F., Sforza, C., Serrao, G., Colombo, A., & Ciusa, V. (1999) · American Journal of Orthodontics and Dentofacial Orthopedics · doi:10.1016/s0889-5406(99)70220-9 · Cited by 25
Recognizing age-separated face images: humans and machines
Yadav, D., Singh, R., Vatsa, M., & Noore, A. (2014) · PLoS ONE · doi:10.1371/journal.pone.0112234 · Cited by 22
Automated analysis of craniofacial morphology using magnetic resonance images
Chakravarty, M. M., Aleong, R., Leonard, G., Perron, M., Pike, G. B., Richer, L., Veillette, S., Pausova, Z., & Paus, T. (2011) · PLoS ONE · doi:10.1371/journal.pone.0020241 · Cited by 22
Improving unfamiliar face matching by masking the external facial features
Kemp, R. I., Caon, A., Howard, M., & Brooks, K. R. (2016) · Applied Cognitive Psychology · doi:10.1002/acp.3239 · Cited by 20
Unfamiliar Face Matching With Frontal and Profile Views
Kramer, R. S. S., & Reynolds, M. G. (2018) · Perception · doi:10.1177/0301006618756809 · Cited by 19
Self-citationCan massive but passive exposure to faces contribute to face recognition abilities?
Yovel, G., Halsband, K., Pelleg, M., Farkash, N., Gal, B., & Goshen-Gottstein, Y. (2012) · Journal of Experimental Psychology: Human Perception and Performance · doi:10.1037/a0027077 · Cited by 18
Dynamic facial recognition: examination of a natural phenomenon
Seamon, J. G. (1982) · The American Journal of Psychology · doi:10.2307/1422130 · Cited by 18
Is it a he or a she? Behavioral and computational approaches to sex categorization
Kaminski, G., Méary, D., Mermillod, M., & Gentaz, E. (2011) · Attention, Perception, & Psychophysics · doi:10.3758/s13414-011-0139-1 · Cited by 16
Recognition memory for infant faces: An analog of the other-race effect
Chance, J. E., Goldstein, A. G., & Andersen, B. (1986) · Bulletin of the Psychonomic Society · doi:10.3758/bf03330133 · Cited by 12
Davis, J. P., Jansari, A. S., & Lander, K. (2013) · Research Explorer (The University of Manchester) · Cited by 11
People automatically extract infants’ sex from faces
Tskhay, K. O., & Rule, N. O. (2016) · Journal of Nonverbal Behavior · doi:10.1007/s10919-016-0235-y · Cited by 7
Round, J. E. C., & Deheragoda, M. (2002) · BMJ · doi:10.1136/bmj.325.7378.1446 · Cited by 5
Kemp, R. (1997) · doi:10.1002/(sici)1099-0720(199706)11:3
Farkas (1992) · doi:10.1597/1545-1569(1992)0292.3.co;2
Flagged first, then by citations
Cited by
Sorted by citations
Rezlescu, C., Danaila, I., Miron, A., & Amariei, C. (2020) · Progress in brain research · doi:10.1016/bs.pbr.2020.06.005 · Cited by 124
Unfamiliar face matching, within-person variability, and multiple-image arrays
Sandford, A., & Ritchie, K. L. (2021) · Visual Cognition · doi:10.1080/13506285.2021.1883170 · Cited by 28
Searching for faces in crowd chokepoint videos
Kramer, R. S. S., Hardy, S. C., & Ritchie, K. L. (2019) · Applied Cognitive Psychology · doi:10.1002/acp.3620 · Cited by 24
Self-citationEnhanced Matching of Children’s Faces in “Super-Recognisers” But Not High-Contact Controls
Bate, S., Bennetts, R., Murray, E., & Portch, E. (2020) · i-Perception · doi:10.1177/2041669520944420 · Cited by 12
Unfamiliar Face Matching With Driving Licence and Passport Photographs
Kramer, R. S. S., Mohamed, S., & Hardy, S. C. (2019) · Perception · doi:10.1177/0301006619826495 · Cited by 11
Self-citationKramer, R. S. S., Jones, A. L., & Gous, G. (2020) · Applied Cognitive Psychology · doi:10.1002/acp.3754 · Cited by 10
Self-citationIndividual Differences Between Observers in Face Matching
Bate, S., Mestry, N., & Portch, E. (2021) · Oxford University Press eBooks · doi:10.1093/oso/9780198837749.003.0006 · Cited by 8
The performance of practitioners conducting facial comparisons on images of children across age
Michalski, D., Heyer, R., & Semmler, C. (2019) · PLoS ONE · doi:10.1371/journal.pone.0225298 · Cited by 6
Diagnostic Features for Human Categorisation of Adult and Child Faces
Faghel-Soubeyrand, S., Kloess, J. A., Gosselin, F., Charest, I., & Woodhams, J. (2021) · Frontiers in Psychology · doi:10.3389/fpsyg.2021.775338 · Cited by 5
Recognizing newly learned faces across changes in age
Sexton, L., Mileva, M., HOLE, G. J., Strathie, A., & Laurence, S. (2023) · Visual Cognition · doi:10.1080/13506285.2024.2315813 · Cited by 4
Kramer, R. S. S., & McIntosh, R. D. (2024) · Quarterly Journal of Experimental Psychology · doi:10.1177/17470218241254209 · Cited by 3
Self-citationDiagnostic Features for Human Categorisation of Adult vs. Child FacesPreprint
Faghel-Soubeyrand, S., Kloess, J. A., Gosselin, F., Charest, I., & Woodhams, J. (2021) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/578bd
Correction to Tsantani et al. (2022).
(2023) · Journal of Experimental Psychology General · doi:10.1037/xge0001373
FEATURES OF IDENTIFICATION OF THE CHILD BY APPEARANCE
Zernova, O. (2024) · KELM (Knowledge Education Law Management) · doi:10.51647/kelm.2024.4.17
Attard‐Johnson, J., Dark, O., Murray, E., & Bate, S. (2026) · Applied Cognitive Psychology · doi:10.1002/acp.70294
Metadata
- Authors
- Kramer, R. S. S., Mulgrew, J., & Reynolds, M. G.
- Affiliations
- University of Lincoln (GB)
- Trent University (CA)
- Corresponding author
- Kramer, Robin S.S.
- Venue
- PeerJ
- Publisher
- PeerJ
- Published
- 2018-06-11
- Volume
- 6
- Pages
- e5010
- ISSN
- 2167-8359 (ISSN-L)
- PMC
- PMC6001712
- OpenAlex
- W2808640403
- Cited by
- 15
- References
- 50
- Citations by year
- 2026: 1 · 2024: 2 · 2023: 2 · 2021: 4 · 2020: 3 · 2019: 3
- FWCI
- 1.32
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- PeerJ (publishedVersion, cc-by)
- Europe PMC (PubMed Central) (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- Lincoln Repository (University of Lincoln) (acceptedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
- Figshare (submittedVersion, cc-by)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › Face Recognition and Perception
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face recognition and analysis
- Social Sciences › Psychology › Social Psychology › Deception detection and forensic psychology
- Keywords
- face matching, photo-identification, facial appearance changes, forensic identification, child exploitation
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-11-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
