Search

A meta-analysis of 143 studies was conducted to explore how the social desirability response bias may influence sex effects on ratings on measures of ethical decision-making. Women rated themselves as more ethical than did men; however, this sex effect on ethical decision-making was no longer significant when social desirability response bias was controlled. The indirect questioning approach was compared with the direct measurement approach for effectiveness in controlling social desirability response bias. The indirect questioning approach was found to be more effective.

Whether or not socially desirable responding is a cause for concern in personality assessment has long been debated. For many researchers, McCrae and Costa laid the issue to rest when they showed that correcting for socially desirable responding in self-reports did not improve the agreement with spouse ratings on the Neuroticism, Extraversion, and Openness to Experience Personality Inventory. However, their findings rest on the assumption that observer ratings in general, and spouse ratings in particular, are an unbiased external criterion. If spouse ratings are also susceptible to socially desirable responding, correcting for the bias in self-rated measures cannot be assumed to increase agreement between self-reports and spouse ratings, and thus failure to do so should not be taken as evidence for the ineffectiveness of measuring and correcting for socially desirable responding. In the present study, McCrae and Costa’s influential study was replicated with the exception of measuring socially desirable responding with the Marlowe–Crowne Social Desirability Scale, in both self-reports and spouse ratings. Analyses were based on responses from 70 couples who had lived together for at least one year. The results showed that both self-reports and spouse ratings are susceptible to socially desirable responding and thus McCrae and Costa’s conclusion is drawn into question.

Negativity bias means that negative information is usually given more emphasis than comparable positive information. Under signal detection theory, recent research found that people more frequently and incorrectly identify negative task-related words as having been presented originally than positive words, even when they were not presented. That is, people have lax decision criteria for negative words. However, the response biases for task-unrelated negative words and for emotionally important words are still unclear. This study investigated response bias for these two kinds of words. Study 1 examined the response bias for task-unrelated negative words using an emotional Stroop task. Proportions of correct recognition to negative and positive words were assessed by non-parametric signal detection analysis. Participants have lower (i.e., more lax) decision criteria for task-unrelated negative words than for positive words. Study 2 supported and expanded this result by investigating participants' response bias for highly emotional words. Participants have lower decision criteria for highly emotional words than for less emotional words. Finally, possible evolutionary sources of the response bias were discussed.

For more than 30 years, the Interpersonal Reactivity Index (IRI) has been used to measure the multidimensional aspects of empathy. But the 28-item, 4-factor model of Davis (1980 ) is currently contested because of methodological issues and for theoretical reasons. Confirmatory (CFA) and exploratory factor analyses (EFA) were applied in two French-speaking Belgian student samples (1,244 participants in the first and 729 in the second study) to test this model and to propose a shortened version. A non-optimal fit was found with respect to the CFI value (Study 1). By splitting the student group into two random subsamples, EFA and then CFA were used to propose a 15-item, 4-factor model with good fit indices. A CFA on the second student group (Study 2) replicated this model. Results are discussed considering the influence of social desirability response bias, an absence of strong invariance across sex and the usefulness of self-report scales to measure empathy.

Critics of self-report methods suggest that participants are likely to underreport their own negative behaviors given concerns about social desirability. The current study examined the problem of self-report bias by comparing individuals' estimates of their own and others' aggressive behavior. Undergraduate students (95 women, 50 men) completed the Richardson Conflict Response Questionnaire with regard to either their own behavior or that of another person of the same sex. As expected, participants reported significantly less aggression for themselves than for others. However, self and other ratings showed the same pattern of sex differences. These results suggest that, although individuals may underreport their own aggressive behaviors, this tendency may not affect relationships among variables.

The Temperament and Character Inventory was translated into Japanese, and, to confirm the psychometric properties of the inventory, three samples were recruited from a nonpatient population. In nonpatient population A ( N = 555), the full version (240 items) of the inventory with dichotomous measuring, along with the General Health Questionnaire and the Social Desirability Scale, were distributed to the subjects. Factor analyses of the subscales showed that the factor structure of the inventory was consistent with Cloninger's theory. Correlations of the scale scores with the General Health Questionnaire and the Social Desirability Scale scores were almost negligible, indicating that the scale is resistant to the current psychopathology and response bias. In this and the other two university student samples ( ns = 395 and 377), Cronbach coefficients α of the scale scores were substantially high except for the short version (125 items) of the inventory with dichotomous measures. The Japanese version of the inventory appears to have internal reliability and content and construct validity in a Japanese population.

A ratio is often used to compare the magnitudes of two related variables and can form an important basis for judgments and decisions as in financial ratio analysis in accounting. However, in psychophysics research shows that ratio judgments of sensory stimuli can produce a logarithmic response bias wherein subjects do not respond to differences in the ratio components but to differences in the relative magnitudes of the resulting ratios. Subjects were required to predict the results of basketball matches when supplied summary information regarding each team's performance. Those subjects who received this information as ratios of wins to losses produced predictions with margins consistent with the logarithmic response bias. Each subject's awareness of or sensitivity to the logarithmic response bias was evaluated from his preferences for various methods of presenting such summary information, including the ratios used in the present study. The ratios which produced the response bias were deemed acceptable although less preferable methods, and these ratings were independent of previous experience with comparisons of such ratios. The results of this study highlight the problems associated with comparing ratios, and the lack of awareness individuals have of such problems.

To limit research costs, former parent-participants from a completed pediatric neurology study were asked to assist with new research on behavioral contributors to childhood injury. An evaluation of the response bias associated with this recontact effort showed that 224 parents who completed the survey did not differ from the 51 who failed to complete it in race, age, education, or income. Nonresponders were significantly less likely to be married than responders.

The current study examined the effect of jury deliberation on the tendency for mock jurors to find attractive defendants guilty less often. It was expected that there would be an interaction between group deliberation (yes or no) and defendant's appearance (plain-looking or attractive). It was hypothesized that mock jurors who did not deliberate would be more likely to find a plain-looking defendant guilty and that deliberation would mitigate this effect. The study was a 2 × 2 between-subjects factorial design. Participants were assigned randomly to one of four conditions: attractive defendant/deliberation, attractive defendant/no deliberation, plain-looking defendant/deliberation, and plain-looking defendant/no deliberation. A total of 172 undergraduates from a small, rural college in Vermont contributed to this study: mock jurors were 70 men and 52 women, ages ranged from 18 to 52 years ( M=20.5, SD = 4.9). The hypothesis was supported. Mock jurors who did not deliberate were more likely to find the plain-looking defendant guilty, whereas mock jurors who deliberated were more likely to find the attractive defendant guilty.

The Kogan Attitudes Toward Old People scale has been used in gerontological research for more than 35 years. Criticisms of the scale, including dated wording and response bias, have increased in the 1990s. Researchers must examine the scale and decide whether a revision is now necessary.

Of the trauma theories available, Herman's has captured the essential features common to most traumatic etiologies. Continued development of this theory has included dimensions within stages of recovery. The Trauma States of Recovery is presented here as a way of assessing stages of recovery and inherent components. Three samples of about 500 college students each and a clinical sample ( n = 34) were recruited. Independent principal components analyses, employing a scree plot, gave a five-component solution accounting for 40.2% of the total variance. The component solution employed Promax rotation for the correlated components. The five components were named Emotional Awareness and Control, Symptom Mastery, Coping, Memory, and Relationships. Internal consistencies, test-retest reliability, and split-half reliability were acceptable. Response bias was low. Convergent validity (with various subscales from the Beck Depression Inventory, the Self-esteem Scale, the Spielberger Anger-Expression Scale, and the Dissociative Experiences Scale) and discriminant validity (with various subscales of the AIDS Knowledge and Attitudes Survey and Jackson's Personality Research Form Dominance Scale) were promising. The Trauma Stages of Recovery also significantly discriminated between clinical and nonclinical samples on three of the five components, Symptom Mastery, Coping, and Memory.

Response bias refers to systematic error in the data produced by patients or research subjects, e.g., due to erroneous recall or reporting. In forensic settings, response bias appears to be associated with the nature of the proceedings. The present study tested whether patients in litigation exhibit different response patterns than nonlitigating patients when asked about their preinjury problems. 34 litigants and 80 nonlitigants rated various areas of cognitive and emotional functioning as problematic in the past and currently. The primary finding was that plaintiffs reported preinjury functioning superior to that of controls. These findings suggest the need for caution in inferences that litigants are either reliable or deceitful because response biases may affect self-reports in a misleading fashion.

In a split-ballot experiment ( N = 458) we tested, first, to what extent the logical presentation of the Photo Anxiety Questionnaire forces respondents into the usual response pattern, that is, a higher reported anxiety on items mentioning a time closer to dental treatment. Secondly, we tested whether subjects use the nonverbal response scale or merely use the position of the photographs on the scale when answering the questions. The original questionnaire had the highest reliability and was least influenced by response bias. Changing the order of the photographs on the response scale did not alter the response pattern, implying that subjects do react to the intensity displayed in the photographs and not just to the position of the photograph on the scale. Because changing the order of the items altered the response pattern only marginally, the response pattern usually found does not seem to be the direct consequence of the logical presentation of the Photo Anxiety Questionnaire.

The shortened versions of the 16 PF (Forms C or D) were given to 227 college students (who took Form D twice), 33 prisoners (who took Forms C and D), and 95 substance abusers (who took Form C twice). All were asked to answer honestly, then retake these tests to appear in a better light. Significant differences in Motivational Distortion scores were found between test and retest in each group.

Summary.-Through a task analysis of the position of Resident Hall Assistant, 97 incumbents and 19 supervisors provided ratings on task liking and three task characteristics-frequency of occurrence, level of difficulty, and consequence of error-for 88 separate tasks. A significant relation was demonstrated between task liking (a potential response bias) and the ratings on these task characteristics. Within job analysis, incumbents and supervisors provide ratings of task characteristics and other measures which form an empirical base for the development of personnel systems, e.g., employee selection, performance evaluation, etc. (Levine, 1983). Response bias may contaminate these ratings, providing an erroneous foundation for the construction of the systems that

The psychic or paranormal phenomenon termed “remote viewing” was investigated using a signal-detection experimental procedure. “Remote viewing” involves the reported ability of a person (the receiver) physically separated from another person (the sender) to describe the surroundings (the target) of the sender. 90 volunteer college students were used in an experimental condition designed to measure the accuracy of “remote viewing” and 25 students served in a control condition designed to provide a baseline for guessing or response bias. The accuracy of “remote viewing” was objectively measured by having receivers select color photographs of the sender's location and rate the confidence of their selections. The results offered no statistical support for “remote viewing” either in terms of the proportions of correct and incorrect selections or in terms of the confidence ratings attached to correct and incorrect selections. The finding that successful receivers offered reliably more selection responses or guesses than did nonsuccessful receivers provided a basis for possible interpretation of success in demonstrations of “remote viewing.”

Mental Measarements Yearbook (Buros, 1972), Wiggins deplored the fact that the test manual did not provide any information on the comparability of the need scales from the Personality Research Form with those of its close competitors, the Adjective Check List (Gough & Heilbmn, 1965) and Edwards Personal Preference Schedule (Edwards, 1954). However, such comparative data were not provided in the 1974 edition of the Personality Research Form manual, and lack of such information makes it difficult to integrate various research findings based on different inventories. Although the three inventories were developed using the same theoretical framework provided by Murray's ( 1938) theory of personality needs, they are substantially different methodologically. The Personality Research Form has true-false statements, and its scales are relatively independent, internally consistent, and free from response biases due to acquiescence and social desirability. The Adjective Check List, on the other hand, contains trait-descriptive adjectives in a checklist format, and its need scales show considerable item overlap which inflates the scale intercorrelations, while the Edwards Personal Preference Schedule's statements are presented in forced-choice format in an attempt to control social desirability response bias.

This research assessed the likelihood for recipients of mail questionnaires who have higher interest in the topic to reply more frequently to the initial questionnaire than to a follow-up. For a sample of 440 commercial respondents, and in a situation where those who were likely to be more interested in the survey had to answer more questions, the normal trend was reversed: 86% of respondents on follow-up were in the interested group, compared to 71% of the initial respondents. The results highlight the importance of a follow-up when a researcher requires responses from the ‘interested’ group.

In a simulated jury experiment, 158 university students were asked to render certain decisions as jurors in a criminal case. Subjects were exposed to damaging (to the defendant) pretrial publicity or to neutral pretrial publicity. Subjects were asked if they could remain impartial and also completed the F-scale questionnaire. Results indicated that even when subjects claim they are unbiased by pretrial publicity, their decisions as simulated jurors reveal bias. Their F-scores, however, appeared unrelated to judicial decisions.

Split-half and retest reliabilities were computed on Comrey Personality Scales administered to 153 Navy Hospital Corpsmen on two occasions separated by about 5 mo. Corrected split-half reliabilities for the 8 personality trait scales were between .66 and .94, with an average of .87, on the first administration and between .68 and .94, with an average of .85, on the second administration. Retest reliabilities for these eight scales were between .59 and .74, with an average of .66. t tests comparing scale means for each administration yielded significant differences on seven of the eight personality trait scales, the Validity Check scale, and the Response Bias scale. On all of these scales, except the Validity Check, the first administration means were higher than the second administration means.

The study sought to evaluate the influence of extreme response style on mood state factors isolated. The factors obtained after partialling out response bias from intercorrelations of adjectives were compared with those obtained from the original set. A sample of the 349 high school students were administered a 63-adjective feeling-and-mood inventory. A principal axis analysis of the adjective intercorrelations disclosed seven mood states, all monopolar. After extreme response-bias score was partialled out of the intercorrelations the matrix of partial correlations was analyzed and seven factors were identified. Three of the mood states identified were bipolar and four were monopolar.

Depressed, formerly depressed, and never depressed female college students were randomly assigned to success, failure, or neutral experiences on a sham measure of empathy ( ns = 12/cell). They then read descriptions of their supposed positive and negative intellectual and personality characteristics, and took a recognition memory test. Based on Beck's (1967) theory, it was predicted that depressed subjects would display negative biasing of memory after neutral or failure experiences and that the latent negative biases of the formerly depressed would be activated by failure. The predictions for depressed subjects were not confirmed. However, the memories of the formerly depressed for information about their intellectual qualities became negatively biased after failure. Beck's view that those predisposed to depression harbor latent negative cognitive biases received some support, but the results also indicated that his conception of global negative biasing of cognition during depression is over-simplified.

Structural convergences among two sets of measures of response styles were examined in two separate factor analytic studies. Three factors accounting for 61.7% of the total variance were extracted in the first study involving 11 psychometric measures and two factors accounting for 56.7% of the total variance in the second study of five measures. Emerging as a separate factor in both studies (accounting for 10.1 and 22.0% of the total variance and 8.4 and 36.5% of the common variance, respectively) was a factor labelled Satiation. This variance was attributed to differences in boredom/enthusiasm among individuals completing the psychometric questionnaires. This source of variance warrants attention in investigations involving questionnaires with numerous items, especially when the items are on a limited number of topics.

The reported studies investigated two issues of importance to researchers and practitioners interested in the Marlowe-Crowne Social Desirability Scale, the sensitivity or social desirability of the individual items comprising the scale and the continued appropriateness of the original scoring scheme presented in 1960. Analysis indicates that over half of the items may no longer be sensitive enough to be useful. Also, it appears that the original keying may no longer be appropriate for a number of items. Finally, based on these findings, implications for the continued use of the Marlowe-Crowne Scale are discussed.

The focus of the current research was on development of a short scale to measure work-related fear of AIDS for use in field research and on assessment of the extent to which its ability to predict managers' intentions to discriminate were free of social-desirability response-bias effects. Based upon a sample of 198 managers from southern US manufacturing, service, and government organizations, the study yielded a 5-item scale with particularly strong test-retest reliability of 88 and internal consistency indicated by Cronbach α of .92. While initial evidence was promising regarding predictive validity, social desirability was correlated with both fear of AIDS and one of the criterion variables but accounted for very little common variance. Techniques suggested by Ganster, Hennessey, and Luthans (1983) to detect social desirability response bias were employed and did not yield moderator, spurious, or suppression effects. Supplementary analysis showed that respondents in manufacturing exhibit significantly stronger fear of AIDS than either service or government managers. Significant but weak correlations of −.16 and −.20 were found between fear of AIDS and subject's sex and education. Males and the less well educated tend to express greater fear of AIDS. Implications are discussed regarding research and practice in management.

Tests of the job characteristics model using the Job Diagnostic Survey have been criticized in the literature for having single-source response bias. To test this criticism, undergraduate and graduate students used the Job Diagnostic Survey to describe their job as “student” (the pretest). The same students then worked at and described a contrived job using the survey. Results from the current study suggested that personality and instrument characteristics had relatively minimal effects on interscale correlations of the scores in the survey within and across situations. However, response biases attributable to priming, consistency, and implicit theories artificially inflated interscale correlations.

College students interacted via the Taylor reaction-time method of studying aggression with an opponent who attacked the subject with maximum shocks for six trials and switched to a matching strategy for 1, 6, or 12 additional trials. At the end of the task, subjects responded to a questionnaire that measured awareness of their own shock settings, the opponent's shock settings, and the relationship between shock settings made by the subject and those made by the opponent. While a statistically significant proportion of subjects correctly recalled the shock settings and relationships between shock settings, a substantial proportion of subjects did not. Results were discussed in relation to a “trade-off” between the use of “masking” tasks, such as the reaction-time procedure, to control subjects' response bias, and the risk that stimulus contingencies may be obscured by such “masking” tasks. One solution to the dilemma may be to replicate the findings both with and without a “masking” task.

To test the notion that extreme responding on Semantic Differential scales is related to maladjustment 3 groups of 12-yr.-old boys rated socially meaningful Semantic Differential Concepts. A psychopathic delinquent and non-psychopathic delinquent group checked the neutral space more frequently than a control “normal” group.

In order to evaluate the effects of bias among psychologists, two virtually identical abstracts of presumably empirical research in the area of astrology were prepared, one with positive findings and the other with negative findings. The data collected by mail from a random selection of professional psychologists indicated that those Ss receiving the negative abstract rated the study as better designed, more valid, and as having more adequate conclusions than those receiving the positive abstract. The many emotionally charged spontaneous comments included with the responses suggests that strong affective reactions are involved, as well as prejudgments.

Increasing the 16-item Rydell-Rosen Ambiguity Tolerance Scale to 20 items raised the reliability from .64 to .86 (split-half, corrected by Spearman-Brown). The 20-item scale (AT-20) was cross-validated on a sample of 789 undergraduate students. Retest reliability was estimated at .63 ( p < .01) for a 6-mo. interval. As predicted, AT-20 related to (a) performance in a complex task, (b) the F Scale, (c) dogmatism, (d) rigidity, and (e) church attendance. The AT-20 was not affected by social desirability response bias as measured by the Marlowe-Crowne.

Summary.-The Henderson Environmental Learning Process Scale (HELPS) was administered to the primary caregivers of 185 elementary school children enrolled in self-contained classes for the learning handicapped. Scores on four annual administrations of the instrument were aggregated, and the data were subjected to a principal-factor analysis. The resulting six factorsReinforcement, Community Involvement, Intellectual Guidance, Role Playing, Expectations, and Literacy--did not correspond closely to those found by Henderson, Bergan, and Hunt in a previous study. The pioneering research of Dave (1963) and Wolf (1964) led directly to a number of efforts to identify variables in the home environment that are related to children's intelligence and academic achievement (Hanson, 1975; Mosychuk, 1966; Radin, 1969). The Henderson Environmental Learning Process Scale (HELPS) (Henderson, Bergan, & Hunt, 1972) is one product of these efforts. In its original form, it was a 55-item questionnairedesigned to elicit responses bearing on environmental stimulation, parental guidance, models, reinforcement, and level of aspiration. Henderson, et al. (1972) administered the HELPS to the mothers of 126 first graders and selected the 25 items that were most highly correlated with the total score for further study. Raw scores on these items were adjusted to control for response bias (Bergan, 1968), and the data were subjected to a principal-component analysis. The resulting five factors-Extended Interests and Community Involvement, Valuing Language and School Related Behavior, Intellectual Guidance, Providing a Supportive Environment for School Learning, and AttentionAid not generally correspond to the factors hypothesized. The purpose of the present study was to reexamine the factor structure of the HELPS. For this purpose, we elected to return to the original item pool but eliminated four items that depended upon the presence of a father or father surrogate in the home. We also chose to use raw scores rather than adjusting them to control for response bias. Our sample consisted of 185 primary caregivers, almost all of them the mothers, of elementary school children enrolled in self-contained classes for the learning handicapped (educable mentally retarded and educationally handicapped) in four Southern California counties.

As part of a comprehensive evaluation of an in-school parenting education curriculum, 281 junior high school students in a major urban school district completed both the Nowicki-Strickland Locus of Control Scale for Children and the Crandall, Crandall, and Katkovsky Children's Social Desirability Questionnaire. Potential differences in influence of social-desirability response bias on the locus scores of students of differing race and sex were explored via zero-order correlations and bivariate regression. Although mean scores of black respondents were significantly higher than those of their white counterparts on measures of social-desirability response bias and external locus of control, Children's Social Desirability scores never accounted for more than 1% of the variance in the locus of control scores of any subgroup.

68 This article is the third in a series of surveys designed to provide a concise statement of objective findings on alcoholism as a stimulus for research and an aid in clinical practice. The survey of the literature in 1974 is less detailed and has been divided into different topical sections than used previously (58, 59). This change was dictated by the nature of objective information about alcoholic Americans located in U.S. journals. Many of the 133 articles surveyed could be considered demographic in nature, and fewer articles contained tests of statistical - significance. Objective information has been divided into two major categories: treatment orientation and behavior and attitudes of alcoholics. Treatment orientation includes: identification, admission to treatment, treatment, treatment outcome, morbidity, and attitudes toward alcoholics. Behavior and attitudes of alcoholics includes: response bias, women, minority groups, behavioral studies, verbal learning and interrelationships among variables. References and cross-references follow each sub-title. Differences reported in the text were significant at the .05 level of confidence or better. When statistical differences were not calculated by the authors, differences reported were of sufficient magnitude to warrant consideration by those interested in the given topic. Two appendices give references to reviews of the literature in 1974 and to objective information about student

Si~nmary.-A rathrr general S-R paradigm is proposed in which stimulus biases, response biases, activation or general responsiveness, and association are defined as patterns of response strength or occurrence when a variety of responses are recorded to a variety of stimuli (an S-R matrix). Patterns of differences between similar matrices obtained on different occasions or with different groups of Ss are similarly defined as increments in stimulus biases, responses biases, activation, and association. Applications of the general paradigm to conditioning, verbal learning, and transfer srudies are suggested. In a recent experimenc (Babich, Jacobson, Bubash, e( Jacobson, 1965) riboniicleic acid (RNA) was excracted from the brains of rats previously crained co approach a food cup at the sound of a click. Untrained rats injected with this RNA made significantly more cup-approach responses when the click was sounded than a concrol group injected with RNA from uncrained rat brains. The authors state that various alternative interpretations of their results are possible, including memory coding in the RNA molecule. The ambiguity of the results of this experiment illustrates a methodological shortcoming common to many current S-R research efforcs: E is unablc to discriminate between general activation, stimulus and response biases, and S-R concingency or association. Although it is possible thac che parclculsr response of approaching the food cup when che click sounded was coded in the RNA from trained racs, there are ocher interpretations of the results which are more parsimonious. The RNA from trained animals might somehow increase the general activity of animals into which it is injected; perhaps the animals become more responsive in general to all stimu1i.l The obtained results might have been obtained by sensitizing the animals to the particular stimulus used; perhaps the RNA from trained animals made the experimental animals more sensitive to the click so that all responses to the click, including che cup-approach response, were made with greater frequency after injeccion. Finally, the RNA from crained animals might have produced a response bias; that is, approach responses might have been increased to all stimuli. In the - simple experimental design in which the observer records only a single response to a single stimulus, these differenc and potentially differentiable kinds of behavior are not discriminable. Without trying different stimuli, E cannoc show . that the response is stimulus specific; without recording a variety of possible responses, E cannot show that any response specificity exists, and without both 'An analogous activation effect might be obtained with white rats by injecting them with the blood of animals which hd undergone stressful training. The blood from the stressed animals might contain adrendin or some other activating substance, and the injected animals, becoming more active, might make the response of interest with greater frequency; such transfer of training a1ou;d hardly be regarded as memory coding.

Summary.-The purpose was to compare a priori and d' measures using data from Swanson and Watson's (1976) study with 10 mentally retarded children on a serial recognition task. Treatment main effects were simi!ar using both measures. Significant primacy and recency effects were found with d' measures, however, this was not found in a porr hoc analysis of proportion correct scores. The results suggested that future invest~garlons of serial recognition performance with emotionally mentally retarded ch~ldren use an unbiased index of memory strength. Recent scudies (e.g., Berch, 1975; Keely, 1971) have noted that the a priori (proportion correct) index commonly used in the probe-type serial memory task as developed by Atkinson, Hansen, and Bernback (1964) is sensitive to a variation in both memory strength and response bias. Thus, to the extent that subjects possess differential serial position response biases, the a prior; measure may lead to false conclusions regarding the relative levels of retention for different positions. The purpose of this study was to test this assumption by directly comparing a priori and unbiased measures (d') derived from the theory of signal decection on data reported in a recent study by Swanson and Watson ( 1976). In the preceding study the d' index was used to differentiace between strength of memory and response bias of 10 mentally retarded children on serial-recognition performance. Using a task outlined by Atkinson, et nl. (1364), significantly better recognition was found with three-dimensional than with two-dimensional representations, as reported by Swanson and Watson (1976). Response biases based on the Cj index for the middle position were found for both the two- and three-dimensional conditions. For the present analysis, raw scores from the above study were converted to the proportion correct at each serial position under both dimensionalities for each child's performance. A 2 X 6 (stimulus dimensionality X serial position) repeatedmeasures analysis of variance was performed on the a priori first-choice scores. As with Swanson and Watson's (1976) d' data, the analysis yielded a significant main effect of stimulus dimensionality ( F1,$ = 9.63, p < .05 ), indicating that

Commonality, the tendency to respond to word-association stimuli with primary or popular responses, seems to be a reliable verbal trait. Peterson and Jenkins (1957) found more consistent, stable, and independent word clusters in the high-commonality S, more variable verbal chaining in the low-commonality S. Canestrari (1966) found that, regardless of S's age, hi&-commonality Ss learned high-strength paired associates faster, low-commonaliry Ss, low-strength pairs. In the present study, priming was compared for highand low-commonality Ss. Priming (recency) measures che increased occurrence on an association test of specified associates which have just been presented in a prerask (Storms, 1958); it may reflect latent response tendencies. Previously reported data (Segal, 1966, in press a, in press b) were rescored to compare priming in highand low-commonality Ss depending on (a) usual associative strength of the specified pair, and ( b ) nature of the pretask. Method.-Ss were 1267 psychology students at Brooklyn College and New York University. They were all glven a "pretask" to preactivate 14 key associares, then an association test containing ( a ) 14 stimuli known to elicit rhe key words and ( b ) 21 words with a high-strength primary response. Thirty different groups were tested with variants of this procedure using different pretasks or different sets of the 14 word-pairs. The prerasks included: ( a ) learning 14 words, ( b ) matching 14 words against a matrix of geometric shapes, (c) associating to 14 words, (d ) solving an analogies test which used 14 words and 35 other irrelevant words. The norm frequencies of the 14 preactivared words as associates to the 1-i cues ranged from .03 to 2.31. Resrrlrs.-Ss who gave 10 c~r more primary associares to the 21 stimuli were grouped as high-commonality Ss, those who gave 9 or less as low-commonality Ss. The priming measure was the number of preactivatecl words associated to the 14 cue stimuli, after S's response rate for 14 matched wo~ds (equal norm frequency) was subtracted. The matched words controlled both for the norm frequency of the word pairs and for any response bias of an individual S. Table 1 reports priming for highand low-commonality Ss as a function of norm associative strength. As anticipated, low-commonality Ss reveal more priming wich very low-strength associates, high-commonality Ss, with high-srrengrh associates. The table also shows the effects of the pretaslr, that high-commonaliry Ss are less variable and excel where the word is activated in an explicitly associative context and on the analogy test where semantic associations provide the task solution. Low-commonaliry Ss are more variable and are favored in the intentional and incidental learning tasks.

normal behavior, and providing scores on 8 personality dimensions and two validity scales. Braun and Iervolino (1972) found that mean scores were significantly higher under faking vs standard instructions for the two validity scales, and for five of the eight personality scales. Also, 71% of faking instruction Ss had unacceptable scores on at least one of the rwo validity scales, compared with only 20% of Ss having standard instructions. In the present study, 155 university students completed the scales under instructions identical with those used by Braun and Iervolino. Thus, 75 had instructions to be honest and forthright, and 80 had instructions to fake so as to make an excellent impression but to also conceal their faking. Results of t tests based on independent groups showed that mean scores on the two validity scales, Validity Check (V) and Response Bias (R), were significantly higher under faking vs standard instructions, as was also true for six of the eight personality scales: Trust, Social Conformity, Activity, Emotional Stability, Extraversion, and Empathy. These findings are highly similar to those of the original study. Detection of faking was studied as follows. As in the original investigation, the percent of Ss equaling or exceeding the following cut-off points was determined for each group separately: V,19; V,26; R,57; R,70. These represent values that would be either one or two SDs above the manual means for both men and women. This procedure had to be used in the original study since Ss were not asked to indicate their sex. Also determined in both studies were the percents scoring one or more SDs above the mean on eirher V or R. This latter method is the one which correctly detected 71 % o[ the faking group in the original study at the expense of 20% of the control group's false positives. The present smdy again found 20% of the control group's false positives, but only 54% of faking Ss detected, and with the difference between control and faking percents significant (p < .01). However, the percent of faking group detections in the present study was significantly less than in the original invesr~gauon (54 vs 71; p < .01). Since in the present study, S did record own sex, a modified faking detection analysis was also run using V and R norms specific to a given S's sex. While the percent of faking Ss detected increased from 54 to 64, the percent of control false positives also increased from 20 to 26, resulting in no substantial net gain in effectiveness of detection. In conclusion, the present results confirm those of the original investigation. The Comrey scales may be faked, but the detection keys show some degree of usefulness even when Ss are aware of their existence and try to beat them.