- Îðàíãóòàíû íàó÷èëèñü èñïîëüçîâàòü îðóäèÿ ñ âûãîäîé äëÿ ñåáÿ [2019-03-22]
- Ìîçã ñîáàê îòëè÷èë íàñòîÿùèå ñëîâà îò òàðàáàðùèíû [2019-03-22]
- Êàêàäó Ãîôôèíà âûêëåâàëè ïàëî÷êè èç êàðòîíêè è èñïîëüçîâàëè èõ â áûòó [2019-03-22]
- Îðàíãóòàíîâ óëè÷èëè â ñîçäàíèè êðþ÷êîâ [2019-03-22]
- Ëþáîâü ñàìîê ãóïïè ê ÿðêèì ñàìöàì îáúÿñíèëè ãåíàìè è îñâåùåíèåì [2019-03-22]
- Îðàíãóòàíû ðàññêàçûâàþò äðóã äðóãó î ïðîøëîì [2019-03-22]
Ïîíðàâèëàñü ëè îíà Âàì? Íóæíî ëè äåëàòü äðóãèå âèäåî-ëåêöèè ïî òåìå ýòîëîãèè?
Íàì âàæíî âàøå ìíåíèå.
Mealey, Daood, and Krage (1996) reported that old–new discrimination of faces 1 week after they were first encountered in an attractiveness rating task varied as a function of whether a face was described as that of (a) a person who cheated on a particular occasion (e.g., “a bishop who was caught embezzling money from his own church,” p. 122), (b) a person who did something that was irrelevant to the cheating–trustworthiness dimension, and (c) a person who did something that was explicitly trustworthy (e.g., “a vendor at baseball games who, after finding a wallet containing $250, located the owner using the driver's license,” p. 122). At a descriptive level, the cheater episodes were associated with better old–new face discrimination than the irrelevant episodes, which, in turn, were associated with better old–new face discrimination than the trustworthy episodes. This pattern corresponded to Mealey et al.'s prediction derived from their extension of the social contract theory (Cosmides, 1989, Cosmides & Tooby, 1989). Very briefly, the theory postulates brain mechanisms that are functionally specialized in the detection of cheaters to have been selected during human evolution. Integrated into a “cheater detection module,” these mechanisms should allow the individual to quickly draw inferences on whether someone has cheated in prior exchanges. Mealey et al. assumed that such a module might also help avoid future interactions with cheaters if it served to remember such individuals particularly well.
However, closer inspection of the Mealey et al. (1996) study reveals a number of problems. First, memory for faces associated with different types of episodes (cheating, irrelevant, trustworthy) was in the predicted order only for faces associated with low-status occupations (e.g., the vendor in the introductory example) but was in the reverse order for high-status occupations (e.g., the bishop). This inconsistency is disconcerting and does not fit any of the predictions Mealey at al. derived from the social contract theory.
Second, Mealey et al.'s (1996) assumption that a cheater detection module implies a memory component led them to predict that cheater faces should be remembered better than all other faces. However, the difference in the number of faces recognized between faces associated with an episode of trustworthy behavior (5.12/12 on average) and faces associated with irrelevant information (5.54/12) was more than twice as large as the difference between the latter types of faces and the cheater faces (5.7/12). Thus, by far, the largest observed difference was one that was not predicted by Mealey et al., which is puzzling.
Third, Mealey et al. (1996) only reported that the overall difference in recognition memory for all three types of faces was significant. They did not report whether the critical difference in memory between cheater faces and faces associated with irrelevant information was at all significant. Given its small size (see the previous paragraph), this theoretically most important difference most likely was not significant.
Fourth, Mealey et al. (1996) did not report details or possibly important characteristics of the behavior descriptions presented with the photographs. It is thus impossible to know whether there were confounds in the materials that may account for the observed episode-related differences and their puzzling inconsistencies across status differences. For instance, from the examples presented in the first paragraph of this section, it is obvious that the behavior descriptions differed in length and amount of detail provided (and, hence in how easy they were to be encoded and remembered) and possibly in other critical features as well (e.g., the valence of the described behavior and the social status of the professions used in the descriptions). Thus, separate norming studies in the relevant participant population would have been necessary to identify possible confounds empirically.
Fifth, and most importantly, it seems inappropriate to derive from the social contract theory the prediction that old–new recognition should be better for faces of cheaters than for those of trustworthy persons. Improved old–new recognition of a face per se cannot provide an evolutionary benefit as long as the cheater context in which the face was encountered is not remembered concurrently. If anything, the social contract theory may allow derivation of the prediction that source memory is improved for individuals with a history of cheating relative to individuals encountered in other situations.
In addition to all these problems, Barclay and Lalumière (2006) recently reported failure to replicate the results of Mealey et al. (1996). In sum, then, the status of the findings reported by Mealey et al. is less than clear. It thus seemed important to test whether the original results could be replicated when possibly problematic properties of the materials were controlled and equated across conditions.
2. Experiment 1
As in the study by Mealey et al. (1996), participants rated the attractiveness of facial photographs presented together with short descriptions of each depicted person's behavioral history and social status. One week later, previously seen and new faces were again rated for attractiveness and were subsequently judged as old or new.
Participants were 64 females and 32 males, most of whom were students at the Heinrich Heine University of Düsseldorf. They were paid for participating. Their ages ranged from 19 to 53 years (mean=26, S.D.=5.9).
2.1.2. Apparatus and materials
Seventy-two facial photographs of males (256 bits, 116×164-pixel grayscales) were randomly assigned to two sets of 36 photographs each (henceforth, Sets 1 and 2). Brief descriptions typed below the photographs conveyed the behavioral history (cheating, irrelevant, trustworthy) and the social status (low, high) of each person shown. For instance, the description “R.O. is a cashier. Again and again, he would shortchange and keep the rest of the money for himself” would convey a history of cheating of a low-status person, that of “B.G. is a surgeon. He works at a hospital on the outskirts and performs many different surgeries every day” would convey irrelevant information of a high-status person, and that of “E.K. is an architect. Any structural damages he might have caused he would get repaired at his own expenses” would convey a history of trustworthiness of a high-status person. Social status was conveyed through the profession of the person shown. Based on an independent norming study (n=24), professions with the lowest (20/82) and those with the highest (20/82) ratings were selected and linked with cheating, irrelevant, or trustworthy behaviors. An independent group (n=21) rated the valence of each behavior description to make sure that instances of cheating, irrelevant, and trustworthy behavior were perceived as negative, neutral, and positive, respectively. Finally, 6 sentences were selected for each of the categories of the 3 (behavior: cheating, irrelevant, trustworthy) by 2 (social status: low, high) design. The properties of these sets of sentences are documented in Table 1. Photographs and descriptions were combined randomly for each participant.
|History of cheating||Irrelevant information||History of trustworthiness||Average|
|Experiments 1 and 2|
|Low social status||−2.41 (0.68)||0.27 (0.31)⁎||1.67 (0.80)||−0.16 (0.23)|
|High social status||−2.29 (0.84)||0.17 (0.35)⁎||1.84 (0.75)||−0.10 (0.26)|
|Average||−2.35 (0.72)||0.22 (0.28)⁎||1.75 (0.73)|
|Low social status||2.04 (0.41)||1.62 (0.45)⁎||1.82 (0.44)||1.83 (0.40)|
|High social status||4.06 (0.40)||4.00 (0.45)⁎||3.96 (0.48)||4.00 (0.33)|
|Average||3.05 (0.24)||2.81 (0.25)⁎||2.89 (0.28)|
|Low social status||20.0 (2.19)||20.0 (2.19)⁎||20.0 (2.19)||20.0 (2.19)|
|High social status||20.0 (2.19)||20.0 (2.19)⁎||20.0 (2.19)||20.0 (2.19)|
|Average||20.0 (2.19)||20.0 (2.19)⁎||20.0 (2.19)|
|Low social status||−2.36 (0.50)||–||1.61 (0.66)||−0.38 (0.29)|
|High social status||−2.08 (0.51)||–||2.05 (0.62)||−0.01 (0.34)|
|Average||−2.22 (0.44)||–||1.83 (0.54)|
|Low social status||2.12 (0.39)||–||2.06 (0.44)||2.09 (0.39)|
|High social status||3.96 (0.48)||–||3.82 (0.42)||3.88 (0.37)|
|Average||3.04 (0.31)||–||2.93 (0.31)|
|Low social status||20.6 (2.32)||–||20.6 (2.32)||20.6 (2.32)|
|High social status||20.6 (2.32)||–||20.6 (2.32)||20.6 (2.32)|
|Average||20.6 (2.32)||–||20.6 (2.32)|
Values are expressed as mean (S.D.). Valence ratings ranged from −3 (negative) to +3 (positive), whereas status ratings ranged from 1 (low status) to 5 (high status).
Used in Experiment 1 only.
Participants were tested individually. Under the disguise of a study on the retest reliability of attractiveness measures, participants were asked to rate the attractiveness of 36 (Set 1 or 2, counterbalanced across participants) facial photographs that were presented in random order. Each trial started with a headline (“How attractive do you find this person?”) and a photograph. The behavior description was shown 2 s later, followed 4.5 s later by the attractiveness rating scale [ranging from 1 (not attractive at all) to 6 (extremely attractive)]. Participants rated attractiveness using the computer mouse and then initiated the next trial.
One week later, participants saw a random sequence of 72 photographs, half of which had been presented in the first phase (Set 1 or 2, depending on the Phase 1 assignment), and the other half were new (Set 2 or 1). Each trial started with a headline (“How attractive do you find this person?”) and a photograph. The attractiveness rating scale appeared 1.5 s later. After the rating, a new headline appeared (“Is this face old or new?”), followed by “old” and “new” checkboxes, one of which participants selected depending on whether they thought that they had seen a face 1 week before or not. They initiated the next trial at their own discretion.
The within-subject independent variables were behavioral history (cheating, irrelevant, trustworthy) and social status (low, high). The dependent measures were attractiveness ratings and old–new recognition in terms of d′. Hit and false alarm rates were adjusted as suggested by Snodgrass and Corwin (1988) to avoid undefined values.
Given a sample size (N) of 96, an α value of .05, and the assumption that the average population correlation between the levels of the behavioral history repeated-measures variable (ρ) is .60 (estimated from pilot data), effects of size f=.15 [i.e., between small (f=.10) and medium (f=.25) effects as defined by Cohen (1977)] could be detected for this variable with a probability of 1−β=.95. All power calculations reported in this article were conducted using the G•Power program (Erdfelder, Faul, & Buchner, 1996). A multivariate approach was used for all within-subject comparisons. In the present application, all multivariate test criteria correspond to the same (exact) F statistic, which is reported. Partial η2 is reported as a measure of the size of an effect. The level of α was set to .05, except for post hoc tests, for which the significance level was Bonferroni–Holm corrected (Holm, 1979).
Following Mealey et al. (1996), all analyses were first run with participant sex as an additional between-subjects variable. There was no significant interaction of participant sex with any other variable. There is also no theoretical reason why participant sex should make a difference in the ability to remember cheaters. For these reasons and to keep the results sections free of unnecessary clutter, we dropped the participant sex variable from all analyses reported here.
The upper panel of Fig. 1 illustrates the mean Phase 1 attractiveness ratings. A 3×2 multivariate analysis of variance (MANOVA) showed that these ratings differed as a function of the behavioral history variable [F(2, 94)=42.08, p<.001, η2=.47]. Post hoc contrasts showed that faces associated with cheating were rated as significantly less attractive than the other two types of faces [F(1, 95)=85.01, p<.001, η2=.47] and that faces associated with irrelevant information were rated as less attractive than faces associated with trustworthiness [F(1, 95)=7.40, p<.01, η2=.07]. There was also a significant main effect of social status [F(1, 95)=9.78, p<.01, η2=.10]. Faces associated with a high-status profession were rated as more attractive than faces associated with a low-status profession. The interaction between both variables was not significant [F(2, 94)=0.38, p=.69, η2<.01]. These results show that the behavior descriptions were attended to and processed—necessary preconditions for analyzing Phase 2 effects of the descriptions.
The lower panel of Fig. 1 shows that recognition memory was clearly above chance (d′=0 would indicate no sensitivity to the old–new distinction). A 3×2 MANOVA showed that there was no main effect of behavioral history [F(2, 94)=0.48, p=.62, η2=.01] and no main effect of social status [F(1, 95)=2.38, p=.13, η2=.02]. The interaction between these variables was also not significant [F(2, 94)=2.80, p=.07, η2=.06].
Whereas recognition judgments represent explicit memory assessment, Phase 2 attractiveness ratings may be considered an implicit memory measure, possibly reflecting after-effects of the experience with a face independent of whether the face was recognized. Finding a person unattractive might indicate a generally negative reaction toward that person, discouraging possibly costly social exchanges with him or her. However, as the data illustrated in Table 2 show, Phase 2 attractiveness ratings reflected neither an effect of behavioral history [F(2, 94)=0.07, p=.93, η2<.01] nor that of social status [F(1, 95)=2.10, p=.15, η2=.03]. There was also no interaction between these variables [F(2, 94)=0.11, p=.90, η2<.01].
|History of cheating||Irrelevant information||History of trustworthiness||History of cheating||Irrelevant information||History of trustworthiness|
|Low-status profession||High-status profession|
|Experiment 1||2.55 (0.08)||2.54 (0.08)||2.52 (0.08)||2.49 (0.07)||2.49 (0.08)||2.49 (0.08)|
|Experiment 2||2.62 (0.07)||–||2.57 (0.06)||2.65 (0.06)||–||2.63 (0.06)|
Values are expressed as mean (S.E.M.).
Experiment 1 was designed as a replication of the study reported by Mealey et al. (1996) with the exception that the properties of the descriptions used to characterize the displayed faces were more closely controlled. These descriptions clearly affected participants' evaluations of the photographs, as the Phase 1 attractiveness ratings show. In particular, the rated attractiveness of faces associated with cheating was much lower than that for the other types of faces. In sharp contrast, and contrary to the findings of Mealey et al., recognition memory was not enhanced for faces associated with cheating. This was so despite the fact that participants' old–new face recognition performance was quite good. The results with respect to the Phase 2 attractiveness ratings were parallel to those for the recognition judgments.
Given these results, we decided to replicate Experiment 1 under conditions that should maximize the contrast along the cheating–trustworthiness dimension. We omitted the faces associated with irrelevant information based on the consideration that this would eliminate a source of distraction away from the cheating–trustworthiness dimension. We also increased the sample size so as to amplify the statistical power of the relevant statistical tests.
3. Experiment 2
Participants were 84 females and 39 males who were paid for participating. Most of them were students at the Heinrich Heine University of Düsseldorf. Their ages ranged from 18 to 57 years (mean=25, S.D.=7.1). They had not participated in Experiment 1.
The used apparatus and materials, procedure, and design for Experiment 2 were identical to those for Experiment 1, with the following exceptions: Of the photographs used in Experiment 1, 48 were randomly selected and assigned to two sets of 24 photographs each. Of the behavior descriptions used in Experiment 1, those describing a history of cheating (6 with low-status professions and 6 with high-status professions) and those describing a history of trustworthiness (6 with low-status professions and 6 with high-status professions) were used. Given N=123, α=.05, and ρ=.60, effects of size f=.15 of the behavioral history variable could be detected with a probability of 1−β=.96.
The upper panel of Fig. 1 illustrates the mean Phase 1 attractiveness ratings. A 2×2 MANOVA showed that faces associated with cheating were rated as significantly less attractive than faces associated with trustworthiness [F(1, 122)=148.70, p<.001, η2=.55]. Faces associated with a high-status profession were rated as more attractive than faces associated with a low-status profession [F(1, 122)=9.51, p<.01, η2=.07]. The interaction between both variables was not significant [F(1, 122)=2.57, p=.11, η2=.02].
The lower panel of Fig. 1 illustrates the recognition memory data. A 2×2 MANOVA showed neither a significant main effect of behavioral history [F(1, 122)=0.38, p=.54, η2<.01] nor that of social status [F(1, 122)=0.30, p=.59, η2<.01]. The interaction between both variables was also not significant [F(1, 122)=2.29, p=.13, η2=.02]. The Phase 2 attractiveness ratings shown in Table 2 reflected neither an effect of behavioral history [F(1, 122)=1.18, p=.28, η2=.01] nor that of social status [F(1, 122)=1.54, p=.22, η2=.01]. There was also no interaction between these variables [F(1, 122)=0.37, p=.55, η2<.01].
Experiment 2 closely replicated the results of Experiment 1. Faces associated with a history of cheating were rated as much less attractive than faces associated with a history of trustworthiness. In terms of the standardized effect size measure, this effect was even larger in Experiment 2 (η2=.55) than in Experiment 1 (η2=.47), indicating that our attempt to amplify the contrast along the cheating–trustworthiness dimension was successful. Still, recognition memory was not enhanced for faces associated with a history of cheating.
An additional aspect to consider is the length of the retention interval. Experiments 1 and 2 were designed to be analogous in this respect to the study of Mealey et al. (1996). However, it may be the case that the source information conveyed in the behavioral descriptions does not survive, to a sufficient degree, the retention interval of 1 week. For instance, Jacoby, Kelley, Brown, and Jasechko (1989) showed that nonfamous names presented once in an experiment were mistakenly judged as famous 24 h later. On an immediate test, no such false fame occurred, presumably because the source of the names' familiarity could be recollected on the immediate test but was unavailable 24 h later, as a consequence of which familiarity was misattributed to the current judgment dimension (fame).
4. Experiment 3
Participants were 41 females and 23 males, most of whom were students at the Heinrich Heine University of Düsseldorf. Their ages ranged from 20 to 58 years (mean=27, S.D.=6.4). They had not participated in Experiments 1 and 2 and were paid for their participation.
The used apparatus and materials, procedure, and design were identical to those for Experiment 2, with the following exceptions: The number of faces in each cell of the design was increased from 6 to 10. For this purpose, the 72 photographs of faces used in Experiment 1 were complemented by 8 photographs from the same source. The photographs were then randomly assigned to two sets of 40 photographs each. Similarly, the 24 professions and behavioral histories used in Experiment 2 were complemented by 16 new professions and behavioral histories; these had been selected, in an independent norming study (n=33), with respect to their social status and valence ratings, respectively. The properties of the final sets of sentences are documented in Table 1. For each participant, the photograph–description combinations were determined randomly.
The first change to the procedure was that there was no delay between the acquisition phase (Phase 1) and the test phase (Phase 2). Second, the Phase 2 attractiveness ratings were dropped because they had proven to be insensitive to any of the experimental variables in Experiments 1 and 2. Given n=64, α=.05, and ρ=.60, effects of size f=.20 of the behavioral history variable could be detected with a probability of 1−β=.94.
The upper panel of Fig. 1 illustrates the mean Phase 1 attractiveness ratings. As in Experiments 1 and 2, faces associated with cheating were rated as significantly less attractive than faces associated with trustworthiness [F(1, 63)=92.40, p<.001, η2=.60]. The effect of social status failed to reach statistical significance [F(1, 63)=1.93, p=.17, η2=.03], which is different from what was found in Experiments 1 and 2. The interaction between both variables was not significant [F(1, 63)=1.60, p=.21, η2=.03].
The lower panel of Fig. 1 illustrates the recognition memory data. A 2×2 MANOVA showed no significant main effect of behavioral history [F(1, 63)=0.02, p=.90, η2<.01]. In contrast to the previous findings, faces associated with high-status professions were recognized somewhat better than faces associated with low-status professions [F(1, 63)=4.45, p=.04, η2=.07, mean=0.70 vs. 0.66, SE=0.02 vs. 0.02]. Considering the solitary status of this effect, together with its small size, we think it best to conceive of it as a chance effect. The interaction between both variables was not significant [F(1, 63)=3.12, p=.08, η2=.05].
Faces associated with a history of cheating were again rated as much less attractive than faces associated with a history of trustworthiness. This effect was even larger in Experiment 3 (η2=.60) than in Experiments 2 (η2=.55) and 1 (η2=.47). Nevertheless, and although the retention interval was minimal, recognition memory was again not affected by this variable.
5. General discussion
The present series of experiments yielded a consistent pattern of results. First, the behavioral histories associated with the photographs of faces had very large effects on the attractiveness ratings. Thus, participants attended to and processed the descriptions.
Second, and in contrast to the findings originally reported by Mealey et al. (1996), old–new recognition was unaffected by the behavioral histories. This was the case (a) in an experiment that was maximally similar to the study reported by Mealey et al. (Experiment 1), (b) in an experiment designed to maximize the contrast along the cheating–trustworthiness dimension by omitting the “irrelevant behavior” category (Experiment 2), and (c) in an experiment that minimized possible effects of forgetting by reducing the retention interval from 1 week to just a few minutes (Experiment 3). Thus, the failure to replicate Mealey et al.'s results occurred under a variety of circumstances in a highly consistent way. This is emphasized by the fact that even a combined analysis of the old–new recognition data from all three experiments (N=283) did not reveal a difference between faces associated with a history of cheating and faces associated with a history of trustworthiness [F(1, 280)=0.87, p=.35, η2<.01]. These results are parallel to those reported by Barclay and Lalumière (2006).
The most obvious difference between the present experiments and the study reported by Mealey et al. (1996) is that we determined empirically and then equated across conditions the properties of our behavioral descriptions, whereas Mealey et al. apparently did not exert the same amount of control over their materials, which may have led to unwanted influences confounding their experimental manipulations. This suspicion coincides with their puzzling finding that the overall difference among faces associated with different types of episodes (cheating, irrelevant, and trustworthy) was in the predicted order only for faces that were associated with low-status occupations but was in the reverse order for high-status occupations. This disconcerting inconsistency cannot be the consequence of the operating of the memory component of a hypothetical cheater detection module.
One could argue that independent evidence for such a component came from a study of old–new recognition memory for imagined opponents in a single-shot prisoner's dilemma game (Oda, 1997). Oda reported that “the biased face recognition of potential cheaters is too robust to be affected by the sex and strategy of subjects, whereas a difference results from the sex of faces presented to subjects” (p. 309). Unfortunately, there are a number of problems with this “biased face recognition of potential cheaters” interpretation of those findings. For instance, of the male and female cooperators and defectors, all types of faces were recognized equally well with the single exception of male “cooperators,” who were recognized less well than all other characters—a data pattern for which a reasonable explanation does not exist. At the very least, it is problematic to portray this data pattern in terms of a “cheater bias.” Additionally, as the author admitted, but only in the discussion section of his article, a defector in a prisoner's dilemma game is by no means a cheater. This is so because “confessing to the police” is not equivalent to breaking a social rule (do not lie). Quite to the contrary, “confessing” may represent compliance with a social rule. The payoff matrix implemented by Oda was also a very rational choice that was preferred by the vast majority of his study participants.
In essence, then, the present results, together with those of Barclay and Lalumière (2006), suggest that the findings reported by Mealey et al. (1996) are singular and cannot be replicated, which is an important conclusion given that those findings are referred to in many evolutionary psychology handbooks and textbooks (e.g., Burnstein, 2005; Cartwright, 2000; Gaulin & McBurney, 2001; Palmer & Palmer, 2002), albeit sometimes incorrectly (Buss, 2004, Cummins, 2005). It also follows that, in contrast to the standard interpretation (e.g., Buss, 2004, Cartwright, 2000), the original Mealey et al. findings cannot be counted as evidence in favor of the social contract theory.
Barclay & Lalumière, 2006 1.Do people differentially remember cheaters?. . Human Nature. 2006;17:98–113.
Burnstein, 2005 2.Altruism and genetic relatedness. . In: Buss DM editors. The handbook of evolutionary psychology. Hoboken, NJ, US: John Wiley & Sons Inc; 2005;p. 528–551.
Buss, 2004 3.Evolutionary psychology: The new science of the mind. . Boston, MA: Allyn and Bacon; 2004;.
Cartwright, 2000 4.Evolution and human behavior: Darwinian perspectives on human nature. . Cambridge, MA: The MIT Press; 2000;.
Cohen, 1977 5.Statistical power analysis for the behavioral sciences (revised ed.). . Hillsdale, NJ: Lawrence Erlbaum Associates; 1977;.
Cosmides, 1989 6.The logic of social exchange: Has natural selection shaped how humans reason? Studies with the Wason selection task. . Cognition. 1989;31:187–276.
Cosmides & Tooby, 1989 7.Evolutionary psychology and the generation of culture: II. Case study: A computational theory of social exchange. . Ethology and Sociobiology. 1989;10:51–97.
Cummins, 2005 8.Dominance, status, and social hierarchies. . In: Buss DM editors. The handbook of evolutionary psychology. Hoboken, NJ, US: John Wiley & Sons Inc; 2005;p. 676–697.
Erdfelder et al., 1996 9.GPOWER: A general power analysis program. . Behavior Research Methods, Instruments, and Computers. 1996;28:1–11.
Gaulin & McBurney, 2001 10.In: Gaulin SJC, McBurney DH editor. Psychology: An evolutionary approach. Upper Saddle River, NJ: Prentice Hall; 2001;.
Holm, 1979 11.A simple sequentially rejective multiple test procedure. . Scandinavian Journal of Statistics. 1979;6:65–70.
Jacoby et al., 1989 12.Becoming famous overnight: Limits on the ability to avoid unconscious influences of the past. . Journal of Personality and Social Psychology. 1989;56:326–338.
Mealey et al., 1996 13.Enhanced memory for faces of cheaters. . Ethology and Sociobiology. 1996;17:119–128.
Oda, 1997 14.Biased face recognition in the prisoner's dilemma game. . Evolution and Human Behavior. 1997;18:309–315.
Palmer & Palmer, 2002 15.In: Palmer JA, Palmer LK editor. Evolutionary psychology: The ultimate origins of human behavior. Needham Heights, MA: Allyn and Bacon; 2002;.
Snodgrass & Corwin, 1988 16.Pragmatics of measuring recognition memory: Applications to dementia and amnesia. . Journal of Experimental Psychology: General. 1988;117:34–50.
Department of Experimental Psychology, Heinrich Heine University, D-40225 Düsseldorf, Germany
© 2008 Elsevier Inc. All rights reserved.