Child Abu~ & Neglect, Vol. 18. No. 12. pp. 1051-IIX)2. 19'94 Ct)pyngh) C 1994 El~vier Science Lid Pnnted in the USA. All rights reserved 0145-2134/t~4 $6 O0 + .130

Pergamon

0145-2134(94)00083-2

SPOTLIGHT ON PRACTICE

SHOULD THIS CHILD BE REMOVED FROM HOME? HYPOTHESIS GENERATION AND INFORMATION SEEKING AS PREDICTORS OF CASE DECISIONS DAVID

R. MANDEL,

DARRIN

R. LEHMAN,

AND JOHN C. YUILLE

University of British Columbia. Vancouver. BC, Canada

Abstract--Two vital aspects of tile investigative process in child abuse and neglect (CAN) cases are (a) generating as many plausible hyp, theses as possible and (b) seeking out as much uncontaminated information as ~)ssible. Ahernativcly. unwarranted assumptions about tile nature of CAN cases can impair investigative decision making. We examined whether tile numbers of (a) unwarranted assumptions. (b) hyl~)theses generated, and (c) requests flw additional infl)rmation concerning a hyp()thetical reported case of CAN predicted level of agreement with a premature decision to remove a child from home among a group of CAN professionals. As expected, lower levels of agreement with tile intervention were associated with (a) less unwarranted assumptions. (b) a greater number of hyp,thescs generated, and (c) more requests for inf(wmation concerning tile case. C()mparcd with a group of undergraduates, a signilicantly greater percentage of (.'AN professionals requested information, and a signilicantly smaller percentage of professionals made unwarranted assumptions. Interestingly. however, no significant diffi:rcnce in mean level of agreement with the intervention was ()bservctl between professionals and undergraduates. Directions fi)r future research are discussed.

Key Word~'--l)ecision making, tlyI~thcscs. Inft)nn:ttitm seeking. INTRODUCTION A N A L A R M I N G N U M B E R o f c h i l d r e n are a b u s e d o r n e g l e c t e d e a c h year. A h i g h l y c o n s e r v a tive e s t i m a t e f r o m a s t u d y o f t w o - p a r e n t f a m i l i e s is t h a t 1.7 m i l l i o n c h i l d r e n b e t w e e n the a g e s 3 - 7 y e a r s e x p e r i e n c e s e r i o u s p h y s i c a l a b u s e b y t h e i r p a r e n t s a n n u a l l y in the U.S. ( S t r a u s & K a n t o r . 1987). If t h e d e l i n i t i o n o f p h y s i c a l a b u s e is b r o a d e n e d to i n c l u d e c h i l d r e n w h o are hit w i t h a n o b j e c t b y t h e i r p a r e n t s , t h e figure r i s e s to a p p r o x i m a t e l y 6.5 m i l l i o n ( S t r a u s & K a n t o r , 1987). A l t h o u g h c o m p a r a b l e e s t i m a t e s o f t h e i n c i d e n c e o f c h i l d n e g l e c t are u n k n o w n , d a t a o n the c l a s s i l i c a t i o n a n d s u b s t a n t i a t i o n r a t e s o f reported c a s e s o f c h i l d m a l t r e a t m e n t are a v a i l a b l e . T h e N a t i o n a l C e n t e r o n C h i l d A b u s e P r e v e n t i o n R e s e a r c h ( N C C A P R , 1992) t b u n d t h a t a n e s t i m a t e d 2 . 6 9 4 J X ) 0 ( 4 2 p e r IJX)O) c h i l d r e n w e r e r e p o r t e d to U.S. C h i l d P r o t e c t i v e S e r v i c e s This work was supp.ncd in part by a Natural Sciences and Engineering Research Council of Canada (NSERC) Postgraduate Scholarship to David R. M.'mdel, grants from NSERC and the Social Sciences and Ilumanitics Research Council of Canada (SSIIRC) to Darrin R. Lehman. and a grant fr(un SStlRC to John C. Yuille. Submitted fl)r publication July 26, 1993; final revision received October 13. 1993: accepted October 24. 1993. Requests fi)r reprints should be addressed to David R. Mandcl. Department of Psychology. University of British Columbia. 2136 West Mall. Vanc.uvcr. BC, Canada V6T IZ.J,. 1051

1052

D.R. Mandel. D. R. Lehman. and J. C. Yuille

agencies in 199 ! - - a 40% increase over the estimated number obtained in 1985. Of the primary allegations made in child maltreatment cases reported in II of the 21 U.S. states whose social service agencies use the categories physical abuse, sexual abuse, neglect, and emotional maltreatment, 43% were categorized as neglect, 28% as physical abuse, 16% as sexual abuse. 5% as emotional abuse, and I1% as other (NCCAPR, 1992). Of the substantiated primary allegations made in the remaining 10 states, 53% were categorized as neglect, 21% as physical abuse, 13% as sexual abuse, 8% as child maltreatment, and 7% as other. Although nationwide incidence estimates of various types of child maltreatment in Canada do not exist (Wachtei, 1989). the general trend indicated by data from provincial child abuse registries parallels that found in the U.S. (Wachtel, 1989). Given the importance of thoroughly investigating reported cases of child abuse and neglect (CAN), a significant concern centers on the decision-making skills of CAN investigators, such as police officers and social workers, who are responsible for such investigations. Unfortunately, police officers generally receive only superficial training in investigative skills that are appropriate for use with children (Yuille, 1984), although the increased awareness of police training facilities has led to some improvements (Yuille, 1986). Given social workers' role as child advocates ( C o n t e & Berliner, 1981), their professional training is likely to emphasize clinically-oriented interview techniques that, while fostering support for the child, may undermine its investigative purpose (Hunter, Yuille. & Harvey, 1990). Indeed. social workers have commonly assumed the primafitcia truth of children's statements, viewing the interview as a means of confirming rather than testing the veracity of CAN allegations (see, e.g., Failer, 1984). Two vital aspects of thorough CAN investigations are (a) generating as many plausible hypotheses as possible (Ceci, Leichtman. & White. in press) and (b) seeking as much uncontaminated information as possible (Yuille, tlunter. Joffc, & Zaparniuk, 1993) concerning the nature of a case. The investigative process can be impaired signiticantly when CAN professionals make selective infi)rmation searches or fail to consider multiple hypotheses. Increases in either false-positive error rates (i.e., situations in which an action was taken when it should not have been taken) or false-negative error rates (i.e., situations in which no action was taken when one should have bccn taken), or the misidentification of perpetrators when abuse has occurred (Ceci et al., in press), can result. Rigid adherence to a particular a priori assumption about a case increases the likelihood that leading interview questions will be asked. And, leading questions c:m elicit lalse information. Some investigators' attributions for case events strongly conform to their a priori beliet:s about the general nature of the types of cases they investigate. As early as 1929, Rice provided striking case examples of how the beliefs of experienced interviewers from various social service agencies inlluenced their attributions for the causes of the destitution of the 2,000 homeless men they interviewed. One interviewer was known by his associates as an ardent prohibitionist, whereas another interviewer was regarded as a socialist. In accounting for the causes of the homeless men's destitution, the prohibitionist was more than twice as likely as the socialist to attribute the cause to alcohol consumption, whereas the socialist was more than twice as likely as the prohibitionist to attribute the cause to industrial filctors beyond their control (e.g., layoffs, seasonal work, plant shutdowns). Interestingly, it seems that these interviewers biased the homeless men's own attributions for why they were destitute. Men who were interviewed by the prohibitionist were over three times more likely than men interviewed by the socialist to attribute the cause of their destitution to alcohol consumption, whereas men interviewed by the socialist were about I 1/2 times more likely than men interviewed by the prohibitionist to attribute the cause of their destitution to industrial factors. Researchers also have found that some professionals rely on the "representativeness" (Kahncman & Tversky. 1972) of extra-legal factors (Howe, Herzberger, & Tennen, 1988), such as

Hyp~0thesisgeneration and intbrmation seeking

H')53

the race of a victim (Hampton & Newberger, 1985; O'Toole. Turbett. & Nalepka, 1983; Williams & Farreli, 1990) or a child victim's age (Winefield & Bradley, 1992; Zellman. 1992). in making decisions about CAN cases. O'Toole et al. (1983) found that some CAN professionals selectively attend to intbrmation, such as a family's race or socioeconomic status (SES), that confirms their beliefs about the nature of CAN, despite the fact that there are no empirically validated reasons to base decisions on such information. In one study comparing nurses' and physicians' beliefs about the indicators of abuse, physicians emphasized the importance of considering suspicious injuries (e.g., cigarette bums), race, and SES, whereas nurses emphasized only suspicious injuries as signs of abuse. In a second study. O'Toole et al,, found that, consistent with physicians" belietk, suspiciousness of injury, race, and SES predicted their recognition and reporting of child abuse, whereas only suspiciousness of injury predicted nurses" judgments. The open-ended responses made by some physicians in these studies were striking. One physician replied, "'1 don't see child abuse in my private practice. It doesn't happen in middle-class neighborhoods." Another physician stated. "It was a large, poor family; the child was dirty and c o m p l a i n i n g ~ a typical case of abuse (italics added)" (p. 353). These examples illustrate respectively how biased investigations can lead to increases in lalse-negative errors and false-positive errors. The Present Study Despite the apparent importance of thorough information seeking and hypothesis generation for the CAN domain, coupled with the growing evidence that CAN investigations are fraught with decisional biases, no study, to our knowledge, has examined systematically the relation between information seeking, hyp,,thcsis generation and investigative decision making in the context of a reported case of CAN. In this study, we cxanfincd whether the numbers of (a) tmwarrantcd assumptions, (b) hypotheses generated, and (c) requests for additional information predicted decisions concerning the rcmowd of a child from home in a hypothetical case of alleged CAN. The premature removal of a child from home can present serious problems. Premature decisions to remove a child from home can result in unnecessary trauma for the child and family (MacFarlane & Bulklcy, 1982). Prcnmture remowds also place unnecessary additional strain on linancially and otherwise overburdened soeial service agencies (NCCAPR, 1992; U.S. Advisory Board on Child Abuse and Neglect, 1990), limiting resources available to deal with cases where assistance is more clearly needed (Meddin, 1984). Thus, it seems useful to examine how CAN investigators' inl'ormation-sceking and hypothesis-generating behaviors predict their decisions concerning the removal of a child from home. The vignette employed in this study was designed such that, although stereotypic pieces of infommtion that might prime respondents to believe that abuse and neglect were occurring emerge, no evidence that would substantiate such allegations is detected. Moreover, all of the information that does emerge was designed such that multiple interpretations of its signilicance could be offered. For example, the social worker in the scenario "sees that the child lapproximately 8 or 9 years oldl is dirty. There are bruises on both the child's knees and there is a scrape on the child's right elbow." Is the child so dirty that he or she laces a health threat'? Or is the dirt that the social worker sees simply grass stains on the child's clothes? Although the scrape and bruises might be the result of abuse, their locations are also consistent with normal play activity. Without further questioning and further evidence, however, the delinitive source of their occurrence remains unknown, in short, given the inh)rmation provided, the hyl~thetic:d case is in&'terminate. Nevertheless, the social worker decides to remove the child from home. Respondents wcrc asked to indicate their level of agreement with the social worker's deci-

1054

D.R. Mandel. D. R. Lehman. and J. C. Yuille

sion. We anticipated that respondents who realized that considerably more information pertinent to the case could potentially be obtained would be more skeptical of the social worker's "'quick" decision to remove the child. Similarly, we expected that respondents who generated hypotheses about the case would be less likely to agree with the social worker's decision, because the testing of those hypotheses are precluded by the circumscribed information that was presently available. In this study, statements that reflected an unreasonably high degree of decisional closure were coded as unwarranted assumptions. In contrast with hypothesis generation, which opens up multiple interpretations for case events, unwarranted assumptions limit such possibilities. Because the "biasing factors" in this case (i.e., the predominantly Black and poor neighborhood, the social worker's negative descriptions of the child's appearance and condition of the residence) would most likely lead one to conclude that abuse and/or neglect were occurring, we hypothesized that unwarranted assumptions would be associated with agreement with the premature intervention taken by the social worker. Thus, we predicted that lower numbers of unwarranted assumptions, higher numbers of hypotheses generated, and higher numbers of requests for additional information concerning the case would be associated with lower levels of agreement with the action taken by the social worker in this vignette to remove the child from home. In addition to CAN professionals (the sample of primary interest in this study), we also asked a comparison group of undergraduates to respond to the same vignette. One might expect that CAN professionals, owing to their greater experience, would be more skeptical of the premature intervention made in the vignette, would make less unwarranted assumptions, would generate more hypotheses, and would request more additional information than would students. However, given the findings illustrating decisional biases among CAN professionals, one might expect to lind no significant differences in investigative decisions between professionals and undergraduates.

MFTHOD

Overview A group of CAN professionals and a group of undergraduates read a vignette of a reported case of CAN. Respondents were asked to state their levels of agreement with the hypothetical social worker's decision to remove the child from home, and to state the features of the case that led them to their decisions.

Re.wondents Professional sample. One hundred and fifty professionals enrolled in one of fi)ur, 3-day workshops focusing on issues related to the investigation of CAN cases voluntarily participated in the study. These workshops focused on characteristics of particular types of male sex offenders of children, child development (e.g., language and cognitive development), the Step-Wise Interview procedure (Yuille et al., 1993), and Statement Validity Analysis (Raskin & Yuille, 1989). The criterion fi~r inclusion in the study was that respondents had to have professional experience that involved dealing with child abuse cases. Nine respondents did not meet this criterion. The remaining 141 (99 female and 42 male) respondents attended a workshop in either Salamanca, New York (n = 34), Hamilton, Ontario (n = 42), Goderich, Ontario (n = 41 ), or Vancouver, British Columbia (n = 24). Almost three quarters of the sample were social workers (n = 103); other professional groups included police officers (n = 19), administrators (n = 10), and others (n = 9). Respondents' mean age was 36 years (SD = 9 years), and their professional experience in dealing with CAN cases was highly variable, ranging from less

Hypothesis generation and information seeking

1055

than 1 year to 23 years (M = 5 years, SD = 5 years). When asked what their primary duties were in dealing with reported CAN cases, 91 reported investigation, 28 reported either treatment or counseling. 21 reported child protection, 18 reported intake duties, 13 reported supervisory duties, and 3 reported referral duties. (Frequencies sum to more than 141 because some respondents reported more than one primary duty. Only the first two duties reported were coded.) Respondents were surveyed within a 2-month period during May and June, 1992.

Student sample. One hundred and forty undergraduates enrolled in a social psychology course at the University of British Columbia voluntarily participated in this study for course credit. Data from nine respondents were dropped because they either indicated having previous professional experience with CAN cases or did not provide information on their previous experience. The remaining 131 respondents (81 females and 50 males) had a mean age of 21 years (SD = 2 years). Respondents represented the Faculties of Arts (n = 79), Science (n = 20), Commerce (n = 23). and other faculties (n = 9).

Materials Child abuse and neglect vignette. Respondents were asked to carefully read and consider the following vignette: A s~:ial work agency receives an anonymous call on May 28th from a caller who states that he is aware of a case of child neglect and abuse. The caller gives the name of the family and the address. Chris. a social v,orker with many years experience, resl'~nds to the call. The address that was given is kx:ated in a very i~'a~r area of the city that is predominantly populated by blacks. At 4:15 p.m.. the social worker knocks on the d ~ r . It is answered by a young child approximately 8 or 9 years old. At tirst glance Chris sees that the child is dirty. There are bruises on both the child's knees and there is a scrape on the child's right elbow, l,tmking past the child, Chris views the residence. It seems quite messy. Upton closer inspection Chris finds that there is little fo
After reading the vignette, respondents were asked to indicate their level of agreement with the social worker's intervention by responding along a strongly disagree (1) to strongly agree (7) scale. Next, respondents were asked to list in an open-ended manner the features of the case that led them to their decisions.

Coding of open-ended responses. Respondents' open-ended responses were coded using three categories: requests for additional information (INFO), hypotheses generated (HYPOTHESES), and unwarranted assumptions (ASSUMPTION). The number of statements that represented each category was recorded. INFO included general requests for more information (e.g., "I would need more information" or " a more thorough investigation is needed") and specific requests for information (e.g., "the social worker should have contacted neighbors for more information," "'was it a weekday?," "what was the child's developmental level?"). Because each general request amounts to the same assertion, only the first general request was counted as a statement in this category. In contrast, each different specific request for additional information was counted as a statement in this category. HYPOTHESES included hypotheses generated about the case (e.g., "'conditions in the home may reflect community standards rather than signs of neglect," "'the parents may have just stepped out for a brief period," "neglect may be occurring") or hypotheses about specific facts presented in the case (e.g., "the bruises and scrape may have been from normal play activities," " t o the child, 'all day' may mean since he or she got home from school," "It's May 28th; if the family is poor, they

1056

D. R. Mandel, D. R. Lehman, and J. C. Yuille

might only be able to buy food after pay checks are issued"). Requests for information that clearly suggested a hypothesis were coded under HYPOTHESES (e.g., "it would be good to know if the child got the bruises from playing outside") and not under INFO. Finally, ASSUMPTION included statements that reflected unwarranted assumptions (e.g., "'the social worker was biased against blacks," "further investigations would confirm that the child was being neglected," "the child is probably not receiving regular nutrition"j. As can be seen in the previous examples, unwarranted assumptions often stem from the degree of inferential certainty or closure conveyed in a statement. For example, the statement "the social worker is making assumptions due to the address being in a poor, Black neighborhood" would be coded under ASSUMPTION, whereas the statement "the social worker m a y be making a s s u m p t i o n s . . . " would be coded under HYPOTHESES. Open-ended responses were coded by the first author. Employing a second coder, interrater reliabilities for the three categories were calculated on a random sample of 25% (n = 35) of the professional group. The percentages of agreement were 71% for INFO. 86% for HYPOTHESES. and 77% for ASSUMPTION.

RESULTS C A N Professionals

Approximately half of the professional sample was surveyed on the first day of the seminar, whereas the others were surveyed on the third day. Because no significant differences in mean level of agreement with the intervention, mean number of unwarranted assumptions, mean number of hypotheses generated, or mean number of requests for information between these two groups were obaincd lall ts < I I I I, we have collapscd across this variable for all subsequent analyscs. CAN professionals' level of agreement with the decision to remove the child from home was distributed fairly evenly over the possible range of ratings (ns = 24, 27, 21, 15, 20, 19, and 15 for ratings I - 7 , respectively), X"(6, N = 141) = 5.80, p > .40, with a mean of 3.69 (SD = 2.(X~), indicative of slight disagreement with the intervention. Of the 139 respondents who provided open-ended responses, 71% requested more information, 39% generated hypotheses, and 12% made unwarranted assumptions. There was a weak, but signilicant, negative correlation between the number of requests for information and the number of unwarranted assumptions, r(137J = - . 2 0 , p = .(12; neither of these wlriables were correlated signiticantly with the number of hypotheses generated. Respondents provided a mean of 2.57 (SD = 1.96) codable open-ended responses. Although mean level of agreement with the intervention did not differ significantly between males and females, significantly more requests for additional information were made by females ( M = 2.00, SD = 1.75) than by males (M = 1.12, S D = !.33), t'(101) = 3.25, p < .005. (In eases where heterogeneity of variances was detected, t' was calculated based on the separate variance estimates. The Welch-Satterthwaite solution [Howell, 19871 was used to calculated df, which was rounded to the nearest integer.) Consistent with the previous tinding, a marginally significant difference in the number of hypotheses generated by females (M = 0.80, SD = 1.19) and males (M = 0.48, S D = 0,86), t'(106) = 1.83, p = .07, also was observed. Level of agreement with the intervention, as well as each of the three coded variables, was not correlated significantly with respondents' age or with their professional experience in dealing with CAN cases. Multiple regression analysis was used to predict respondents' level of agreement with the social worker's intervention. A statistical model was built by entering the number of unwar-

Hypothesis generation and information seeking

1057

Table !. Regression of Coded Variables on Level of Agreement with the Intervention

Predictor

B

R

R: Change

F

ASSL'MPTION HYPOTHESES [NFO

.22 -.35 -31

.22 .41 .51

.05 .13 .09

4.73* 23.86** 16.70"*

Note. Adjusted R-" = .25. *p < .05; **p < .0~)1.

ranted assumptions, the number of hypotheses generated, and the number of requests for information, respectively. This order o f entry reflects the theoretical premise that any rigid assumptions held by investigators would thwart the generation of alternative hypotheses, and that the assumptions held and the hypotheses generated by investigators, in turn, would guide subsequent information searches. As can be seen in Table 1, all three variables significantly predicted respondents" level of agreement with the intervention. The full model was highly significant, F(3. 135) = 16.23, p < .0001, accounting for 25% of the variance in the criterion. All statistical assumptions were met. Comparison of CAN Pr~,ssionaL~ with Undergra~htates Unlike C A N professionals, undergraduates" level o f agreement with the intervention was bimodally distributed across the possible range of ratings (ns = 13, 27, 26, 9, 20, 24, and 12 for ratings I - 7 , respectively), X"(6, N = 131) = 17.28, p < .01, with a mean of 3.89 (SD = 1.91), This mean is not significantly different from C A N professionals' mean level of agreement, t(270) < J I J. Nor was the mean number of undergraduates" codable open-ended responses (M = 2.34, SD = 1.95) signilicantly different from that of C A N professionals, t(268)

< Ill. ]'he numbers of respondents who made one or more (a) requests fl~r information and (b) unwarranted asst, tnptio,ls were nonindcpendcnt of the subject sample: The percentage of undergraduz,tes who requested additional information (47%) was signilicantly lower than that of C A N professionals, jr"( I, N = 269) = 16.46, p < .001; and the percentage of undergraduates who made unwarranted assumptions (34%) was significantly higher than that of C A N professionals, X"( I, N = 269) = 17.90, p < .001. The percentage of undergraduates who generated hypotheses (47%) was not significantly different from that of C A N professionals, XZ(I, N = 2697 = 1.79, p > . 15. A m o n g the respondents who made at least one request for information, undergraduates made significantly fewer requests (M = 1.51, SD = 0.74) than did C A N professionals (M = 2.43, SD = 1.50), t'(152) = - 5 . 1 9 , p < .001. A m o n g the respondents who made at least one unwarranted assumption, undergraduates made significantly more unwarranted assumptions (M = 1.39, SD = 0.58) than did C A N professionals M = 1.06, SD = 0.24), t'(59) = 3.11, p < .005. Finally, among respondents who generated at least one hypothesis, undergraduates generated signilicantly more hypotheses (M = 2.48, SD = 1.27) than C A N professionals (M = 1.81, SD = 1.07), t(113) = 3.(X), p < .(X)5. In contrast with C A N professionals, female undergraduates had a significantly higher mean level of agreement (M = 4.23, SD = 1.90) with the intervention than male undergraduates (M = 3.32, SD = 1.80), t(129) = 2.73, p < .01; females generated significantly fewer hypotheses (M = 0.96, SD = 1.38) than did males (M = 1.49, SD = 1.67), t(128) = - I . 9 4 , p = .05; and females made marginally fewer requests fi~r information (M = 0.60, SD = 0.89) than did males (M = 0.88, SD = 0.93). t(128) = - 1 . 6 7 . p = .10. The same multiple regression strategy employed with C A N professionals was used to predict level of agreement with the intervention among undergraduates. This analysis yielded two

1058

D, R. Mandel. D. R. Lehman. and J. C. Yuille

significant predictors: the number of hypotheses generated (/~ = - . 4 6 , F[ I, 127] = 35.28, p < .(XX)I) and the number of requests for additional information (,8 = - . 4 2 , F[ I, 1271 = 33.06, p < .0~X)I ). This model, accounting for 37% of criterion variance, was highly significant, F(2, 127) = 38.59. p < .(X)OI. All statistical assumptions were met.

DISCUSSION It is striking that, given the spotty evidence presented in the vignette, almost 40% of CAN professionals agreed with the decision to remove the child from home, and less than 20% strongly disagreed with the intervention, if one systematically examines the information presented in the vignette in light of the allegations made, it should be apparent that, without further information, the case is indeterminate. That is, the evidence presented neither confirms nor refutes the allegations of child abuse and neglect. First, the anonymity of the caller should raise suspicions regarding his motivations for making a report. Second, the fact that the home is located in a very poor neighborhood should alert us to the fact that the standard of living in a large proportion of the homes in this area is likely to be below what many professionals might bc accustomed to and consider normative. Third, the description of the child as " d i r t y " is highly anlbiguous. For example, is the child filthy? Or, are there merely grass stains on the child's shirt? Fourth, the origin and extent of the scrape and bruises are similarly unknown. If anything nmy be inferred concerning these injuries, however, it is that their locations arc more consistent with normal play injuries than with locations where one would strike a child (Failer, Bowdcn, Jones, & tlildebrandt, 1981). Filth, the social worker's initial impressions of the home as being "quite messy'" and having "'little food in the refrigerator" also arc highly subjective and ambiguous. Because the social worker's visit took place at the end of the month, it is possible that the family, especially if poor. only stocks up on fl)od after pay checks arc i~;sucd. Alternatively. there may have been plenty of nonperishable food in a pantry. Finally, without further qucstioning, the child's cursory responses to the social worker's two questions can have any one of a number of meanings. For example, it is not uncommon for children to say. "'1 don't know," especially when talking to a stranger whom they may not trust. In a similar vein, to the child, "All d a y " might mean since hc or she got home from school. In short, there is no convincing evidence of child abuse or neglect at this point in the case investigation. One might ~.rgue that even if there is no clear evidence of abuse or neglect, because the child is alone, there is a protection issue at hand that warrants the child's temporary removal from home. We suggest that this is not so. The lack of inlormation in this case should preclude the assumptions that the child is not being cared for or is unable to care for himself or herself. For all the hypothetical social worker knows, the parent(s) may have left the child for 10 minutes, provided the child with emergency instructions, or arranged fl)r a caretaker to check in on the child periodically. Therefore, the first step taken by the social worker should have bccn to collect information that would clarify the nature of the child's situation, including the dcgrcc of risk the child laces. Another possible interpretation is that, although respondents realized the weakness of the evidence presented in the vignette, they assumed that " a social worker with many years of experience'" would know when it was appropriate to remove a child, in fact, only two respondents mentioned that the experience of the social worker was an important consideration underlying their level of agreement with the intervention. And, in one of these two cases the respondent actually argued that "'sometimes being in a business lsicl 'for a number of years' v,c can become lax in finding all of the facts.'" As hypothesized, the fewer unwarranted assumptions CAN professionals made, and the

H~p~)thcsis generation and information ~ecking

1059

more they generated hypotheses and requested additional information, the more likely they were to disagree with the decision to remove the child from home. Although the correlational nature of these data preclude firm causal statements, taken together, they lend support to the notion that the tendency to adopt rigid, unwarranted assumptions, and the failure to consider multiple, plausible hypotheses and to gather as much uncontaminated information as is feasible can lead to premature decisional closure in CAN investigations. The approach taken in the present study is unique to this area of investigation in that the structure, rather than the content, of respondents" open-ended statements was coded. For instance, respondents who wrote, "there is little food," as a reason for their decisions essentially were reiterating a fact presented in the vignette. They were neither making an unwarranted assumption, generating hypotheses, nor requesting information. Alternatively, respondents who wrote. "there is no food," actually distorted the inlbrmation presented, despite the fact that the issue discussed (i.e., the availability of food) is the same in both statements. In the latter example, an unwarranted assumption is made, and the situation is represented in cognitively simple, categorical terms (see Schr~v,ter. Driver. & Streufert, 1967: Suedfeld & Rank. 1976, for a description of other structural-coding systems of cognitive complexity)--eithcr there is food or there is no food. By stating. "there is no food," all perceived ambiguity is removed. If there is no food, the child simply cannot eat, and thus is at serious risk. It is easy to see how investigators could arrive at an unwarranted conclusion simply by forcing the inlormation they receive into "black-or-white'" terms, Our results suggcst that professional experiencc in the CAN domain d(ves not mitigatc the tendency to prematurely accept an intrt.sivc fl)rm of intervention. The number of years of experience CAN professionals had in dealing with CAN cascs was not signiticantly corrclatcd with their level of ;,tgrecnlcnt with the intervention. And, consistent with this linding, level of agreement did not differ signilicantly between CAN profcssit)nals and undergraduates who had no experience in this domain. These findings suggest tlmt tllcrc is no empirical justilication for using CAN professionals" cxpcricncc in the CAN domain as a basis fi)r justifying their investigative dccisions or actions. Indeed, l)cnt (1982) found that cxpcricnccd policc investigators elicited more intt('cttral(" descriptive inl'ornmtion in inlerviewing children than their less experienced counterparts. Several interesting differences between CAN professionals and undergraduates emerged. Although these two groups of respondents did not differ in their mean level of agreement with the intervention, when compared with undergraduates, a signilicantly greater percentage of CAN professionals made at least one request for additional information, and a signilicantly smaller percentage of CAN profcssionals made at least one unwarranted assumption. Moreovcr, anlong respondents who requested nlorc information, CAN professionals made signiticantly more such requests than undergraduates; and among respondents who made unwarranted assumptions, CAN professionals made signilicantly fewer unwarr:mted assumptions than undergraduates. Taken together, these lindings suggest that, despite the lact that we found no evidence that CAN professionals make bcttcr investigative decisions than undcrgraduatcs, CAN professionals are somewhat morc critical than undergraduates in processing case infornmtion. This interpretation, however, is weakened somewhat by the linding that, among respondents who generated hypotheses, CAN professionals generated signilicantly fewer than undergraduates. Perhaps because of CAN professionals' greater experience with the uncertainties involved in interpreting case tindings, they are somewhat less inclined than undergraduates to jump to lirm conclusions or to express their hypotheses in certain terms (both of which would increase the number of unwarranted assumptions in this study). In any case, that the percentage of CAN professionals who made unwarranted assumptions was small in absolute tcrms (only 12%) is a positive linding. Similarly, because of their past investigative experience, CAN

1060

D. R. Mandel. D. R. Lehman, and J. C. Yuille

professionals are likely to think of a greater number of specific questions that might be pursued in a real case investigation than would undergraduates. It is interesting to note that while female CAN professionals requested significantly more information and generated marginally more hypotheses than male CAN professionals, female undergraduates had a significantly higher mean level of agreement with the intervention, generated significantly fewer hypotheses, and made marginally fewer requests for information than did male undergraduates. The reason why female CAN professionals appear more critical of the case material than male professionals, or why a reversed pattern of sex differences emerged in our undergraduate sample, is unclear at present. Moreover, 73% of the CANprofessional sample were social workers, and 78% of this subgroup were female. It will thus be instructive in future studies to examine, for instance, whether female police officers are more critical in processing case information than are male police officers. Future Research

Respondents in the present study stated their level of agreement with the action taken by the social worker in the vignette. It will be important to examine in future research how respondents" own decisions concerning CAN cases are predicted by their hypothesis-generating and infornmtion-seeking behaviors. Due to time constraints, the present study utilized only one CAN vignette. Thus, caution must be taken in interpreting the results of this study, and replications must be attempted in order to assess the stability of the obtained pattern of findings, it is also plausible that, at some point, the advantages to the investigative process accrued by generating more hypotheses and requesting more information are offset by the risks incurred by delaying intervention. For example, if a sufticient amount of information that supports a particular allegation of abuse was presented, wotdd respondents who generated more hypotheses and requested more information avoid making a decision whose delay might be detrimental to the child'? Because the primary goal of such research is to suggest ways of optimizing the efficiency and effectiveness of overall case management, such possibilities also could be prolitably explored in future research.

SUMMARY Among a group of professionals involved in various aspects of child abuse and neglect cases, numbers of (a) unwarranted assumptions, (b) hypotheses generated, and (c) requests lk)r additional information concerning a hypothetical reporled case of child abuse and neglect significantly predicted level of agreement with a premature decision to remove the child from home. Lower levels of agreement were associated with (a) fewer unwarranted assumptions, (b) a greater number of hypotheses generated, and (c) a greater number requests for information concerning the case. Although no significant difference between professionals and undergraduates in mean level of agreement with the intervention was observed, a significantly greater percentage of professionals requested infornmtion and a significantly smaller percentage of professionals made unwarranted assumptions about the case. Length of professionals' experience in dealing with child abuse and neglect cases was not significantly correlated with their level of agreement. Our findings suggest that future research geared toward examining the structure, in addition to the content, of decision makers' reasoning in the area of child abuse and neglect case management would be valuable. Acknowh'dgenwnt--The authors thank Victoria Desroches for her valuable assistance in this study. We also thank three reviewers for their comments on an earlier draft of this paper.

Hypothesis generation and information seeking

1061

REFERENCES Ceci. S. J.. Leichtman, M. D.. & White, T. (in press). Interviewing pre~hoolers. In D. Peters (Ed.), The child wimess in ¢'o~nitive. social, and legal context. The Netherlands: Kluwer. Conte, J., & Berliner, L. (1981). Sexual abuse of children: Implications for practice. The Journal of Contemporary Social Work. 62. 6OI-64~6. Dent. H. R. (1982). The effects of interviewing strategies on the results of interviews with child witnes~s. In A. Trankell (Ed.). Reconstructing the past (pp. 279-297). Stockholm, Sweden: Norstedt &Soners. Failer. K. C. (1984). Is the child victim of sexual abuse telling the truth.'? Child Abuse & Neglect, 8. 471-481. Failer, K. C.. Bowden. M. L.. Jones, C. O., & Hildebrandt. H. M. (1981). In K. C, Failer (Ed.), Social work with abused and neglected children: A manual of interdisciplinary practice (pp. 13-31 ). New York: The Free Press. Hampton. R. L.. & Newberger. E. H (1985). Child abuse incidence and reporting by hospitals: Significance of severity, class, and race. American Journal of Public Health, 7 5 . 5 6 - 6 0 . Howe. A. C., Herzberger, S.. & Tennen, H. (1988), The influence of personal history of abuse and gender on clinicians" judgments of child abuse. Journal of Fami(v Violence, 3. 105- 119. Howell, D. C. (1987). Statistical methods for psychology (2nd ed.). Boston. MA: PSW-Kent. Hunter. R.. Yuille, J. C., & Harvey. W. (1990). A coordinated approach to interviewing in child sexual abuse investigations. Cana&t "s Mental Health, 38, 14-18. Kahneman. D., & Tversky. A. (1972). Subjective probability: A judgment of representativeness. Cagnitive Psychoh~gy, 3. 430-454. MacFarlane, K., & Bulkley. J. (1982). Treating child sexual abuse: An overview of current program models. Journal ~ Social Work and Haman Sexuality, !. 71-93. Meddin, B. J. (1984). Criteria for placement decisions in protective services. Child Welfitre. a3. 367-373. National Center on Child Abuse Prevention Research (1992). Carrent trends in chiM ahuse reporting and fatalities: The results of the 1991 annualfi]ty state survey. Chicago. IL: National Committee tilt Prevention of Child Abuse. O'T¢~Ie. R., Turlx, tt. P., & Nalepka, C. (1983). Theories. professional knowledge, and diagnosis of child abuse. In D. Finkelhor. R. J. Gelles. G. T. llotaling. & M. A. Straus (Eds.), The &irk sigh, ,~ffitmilies: Currentfitmily vioh'nce research (pp. 349-362). Beverly Ilills. CA: Sage Publicatilms. Raskin. D. C.. & Yuille. J. C. (1989). Problems in evaluating interviews of children in sexual abuse cases. In M. P. Toglia, S. J. Ceci, & D, F. Ross (Eds.), Adult per~'~y,tions oJ" chihl testimony (pp. 184-2(17). New York: SpringerVerlag. Rice, S. A. (1929). Interviewer bias as a contagion. American Journal ~Jf So~'ioh~gy, 35. 421-423. Schr¢~lcr, II. M., I)rivcr, M., & Streuferl. S. (1967), thunan infi~rmation pro,'essing. New York: llolt, Rinehart. & Winston. Straus. M. A.. & Kantor. (]. K (1987). Stress and child abuse. Iq R. E. Ilclfer & R. S. Kcmpe (Eds.), 7"he battered c'hihl (4th ed.. rev,. pp. 42-59). Chicago. I1,: University of Chicago Press. Sucdfcld. P.. & Rank, I). (1976). Revolutionary leaders: Long term success as a function of changes in conceptual complexity. Journal ~y" Personality and Social It~'y~'hologv. 34. 169-178. U.S. Advisory Board on Child Abuse and Neglect. (1990). Chihl abase and negh'~'t: Criti
R~,.sum~ll y a deux aspects importants Iorsqu'on poursuit une enqu~le au niveau des enfants mahrailes ou n~glig~s : ( I) engendrer autant d'hyl~th~ses que [x~ssible; et (2) il faut rechercher autant de renseignementsnon contamin~s que possibles. D'anlre part aborder un cas "fi partir de pr~somptions non d~sirahles risque de miner la prise de decision :,u tours de I'cnqu~te. Dans leur ~tude. les autcurs onl pr~senl~ une situation hypoth~tique "~des intervenants professionnels travaillant -',upr~s de I'enfance maltrait~e pour ddterminer si ( I ) le mnnbre de pr~somptions non valables. (2) les hypoth;zses idcntili,Ees, et (3) la poursuite de renseignements additionnels m~:ner.'fient les intervenants vers la m~Sme d,Scision, soit de retirer pnSmatur~ment I'enfant de son fipyer. Tel que pr~vu, on a not~ un faible consensus vis-

1062

D. R. Mandel, D. R. Lehman, and J. C. Yuille

/,-vis I'intervention :~ exdcuter (I) Iorsqu'i] existait peu de presomptions non valables; (2) lorsqu'il y avait eu au pr6alable beucoup d'hypoth~ses; et (3) Iorsqu'on s'etait mis/, la recherche de beaucoup de renseignements additionnels. Les r e p o r t , s des intervenants ont 6t6 compardes .~ celles d'etudiants de premier cycle. Comparativement, un plus grant hombre de professionnels allaient chercher des renseignemenLs additionnels et un plus petit pourcentage de professionnels travaillaient h partir de presomptions non justifiables. !1 est interessant de noter, toutefois, que par rapport /~ la disposition qu'il fallait prendre pour prot~ger I'enfant, les professionnels et les dtudiants ont d~montr6 sensiblement peu de difference au niveau du consensus moyen. L'article propose de futures recherches ~ poursuivre. R e s u m e n - - D o s aspectos vitales en el proceso de investigaci6n de los casos de abuso y negligencia contra los nifios~'as) (CAN) son: (a) generar la mayor cantidad de hip6tesis posibles y (b) buscar t~ta la informaci6n no contaminada que se pueda. Por el contrario, suposiciones infundadas sobre la naturaleza de los casos de CAN pueden impedir la capacidad de decision de la investigaci6n. Examinamos si el n6mero de (a) suposiciones infundadas. (b) hip6tesis generadas, y (c~ solicitudes de informaci6n adicional en relaci6n al nivel de acuerdo pronosticado con un caso hipot6tico de reporte de CAN y una decisidn prematura de sacar un nifio~a) de su hogar entre un grupo de profesionales de CAN. Como era de esperarse, se asociaron niveles mas bajos de acuerdo con (a) suposiciones menos infundadas, (b) un mayor ndmero de hip6tesis generadas, y (c) mils solicitudes de informaci6n concernientes al caso. Comparados con un grupo de estudiantes de tcrmino, un poreiento signiticativamente mayor de profesionales de CAN requer{an informacidn y un porciento signiticativeamente menor de profesionales hacian suposiciones infundadas. Sin embargo, es interesante que no se observ6 diferencia significativa en el nivel de acuerdo con la intervencidn entre profesionales y estudiantes de tdrmino. Se discuten las direcciones para futuras investigaciones.