Running head: EYEWITNESS IDENTIFICATION REFORMS 1. Eyewitness Identification Reforms: Are Suggestiveness-Induced Hits and Guesses True Hits?

Transcription

1 Running head: EYEWITNESS IDENTIFICATION REFORMS 1 Eyewitness Identification Reforms: Are Suggestiveness-Induced Hits and Guesses True Hits? Gary L. Wells Iowa State University Nancy K. Steblay Augsburg College Jennifer E. Dysart John Jay College of Criminal Justice Corresponding Author: Nancy K. Steblay Department of Psychology Augsburg College 2211 Riverside Avenue Minneapolis, MN USA steblay@augsburg.edu

2 Running head: EYEWITNESS IDENTIFICATION REFORMS 2 Abstract Research-based reforms for collecting eyewitness identification evidence (e.g., unbiased prelineup instructions, double-blind administration) have been proposed by psychologists and adopted in increasing numbers of jurisdictions across the United States. It is well known that reducing rates of mistaken identifications can also reduce accurate identification rates (hits). But the reforms are largely designed to reduce the suggestiveness of the procedures they are meant to replace. Accordingly, we argue that it is misleading to call any hits obtained because of suggestive procedures hits and then saddle reforms with the charge that they reduce the rate of these illegitimate hits. Eyewitness identification evidence should be based solely on the independent memory of the witness, not aided by biased instructions, cues from lineup administrators, or the use of lineup fillers who make the suspect stand out. Failure to call out these hits as being illegitimate can give solace to those who are motivated to preserve the status quo.

3 Running head: EYEWITNESS IDENTIFICATION REFORMS 3 Eyewitness Identification Reforms: Are Suggestiveness-Induced Hits and Guesses True Hits? Research-based reforms for collection of eyewitness identification evidence have been proposed by psychological scientists in recent years, and a large number of jurisdictions in the United States have adopted or are considering adopting these changes. The laboratory results that provide the foundation for the recommendations usually are reported in terms of witness response rates to culprit-present lineups (correct culprit identifications, incorrect filler picks, or incorrect no-picks) and culprit-absent lineups (correct rejections, incorrect filler picks, or incorrect false identifications of an innocent suspect). These measures provide direct evidence of the impact of the experimental manipulation and an evaluation of procedural effectiveness. It is understandable that the most salient measures from the lab hits (culprit identifications) and false alarms (mistaken identifications of an innocent suspect) might be narrowly equated with costs (correct identifications lost) and benefits (false identifications avoided) of a change in identification procedure. Such an analysis was presented by Clark (2012) in his positing of costs and benefits associated with lineup reforms. There are substantial limitations, however, in characterizing reforms in terms of a trade-off between hits and false alarms, and the deficiency inherent in a strict hit/false alarm analysis is revealed with a simple question: Why? For example, why do suspect identifications rise with one type of pre-lineup instruction compared to another? Why is the false alarm rate higher with one procedure versus another? These questions matter, in psychology and in the legal system, but a narrow hit/false alarm analysis is largely silent in response to this critical inquiry and cannot provide satisfactory answers to complex why questions.

4 Running head: EYEWITNESS IDENTIFICATION REFORMS 4 Accordingly, psychologists have brought much more to lineup reform than a list of specific recommendations predicated on a ratio of hits/false alarms. The recommendations are anchored in a deep research literature about human memory, social influence, and decisionmaking. Over 20 years ago, Wells and Luus (1990) captured an essential principle for lineup reforms in their analogy between police conducting a lineup and researchers conducting an experiment: All the things that can go wrong in an experiment can go wrong when police conduct a lineup, such as experimenter-expectancy effects, demand characteristics, absence of controls, and cross-subject contamination, to name only a few. Just as a rigorous experimental protocol employs procedures that will lead to more reliable scientific evidence, lineup reform has been based on analogous principles for securing reliable evidence of eyewitness memory. For its part, the legal system is very clear in its presumption that an identification decision from an eyewitness should be the result of an independent recollection by the witness and must not be influenced by administrator cues, suggestive procedures, or any form of external intrusion. In January of 2012, the U.S. Supreme Court revisited eyewitness identification issues for the first time since 1977 in Perry v. New Hampshire (2012). While acknowledging that suggestiveness in eyewitness identification procedures can derive from sources other than law enforcement, the Court strongly reiterated the view that unnecessary suggestiveness in police procedures is a constitutional issue involving due process that gives rise to judicial review and possible suppression of the evidence. This basic legal premise underlies the reforms for eyewitness evidence collection procedures. Clark (2012) concedes that there is a greater probability that an identification is accurate when reform procedures are used. In other words, there is no debate that the reforms not only reduce mistaken identifications but also improve the proportion of accurate identifications likely

5 Running head: EYEWITNESS IDENTIFICATION REFORMS 5 to reach a jury (i.e., the probative value). But, Clark notes that suppression of mistaken identifications and the better ratio of accurate to mistaken identifications (i.e., the diagnosticity ratio) from the reforms are achieved at a cost of lowering valid accurate identifications. He intimates that researchers have not been forthcoming about this cost. Are All Reductions in Hit Rates True Costs? Clark (2012) argues that there is a little or no cost presumption underlying the reform movement with respect to eyewitness identifications from lineups. 1 By cost, he means that the reforms that reduce mistaken identifications can also reduce the rate at which accurate identifications (hits) are made. We agree that there exists a little or no cost presumption about hits in some circles, especially within media and other secondary sources. And, the early scientific literature showed little or no reduction in hit rates resulting from the reform procedures. But, the empirical literature has clearly reported what happens to both hit rates and false alarm rates. In fact, Clark s quantitative analysis is of data directly drawn from these previously-published reports, which clearly demonstrates that nothing was hidden 2. Although we agree that the little or no cost characterization has been used by some advocates of reform, we believe that the meaning of that statement is very different from the way Clark (2012) depicts it. Specifically, to the extent that those who endorse the reforms tend to discount reductions in hits, it is because they do not consider the hits that are lost to have been legitimate hits in the first place. Accordingly, those who have used the phrase little or no cost should probably have said little or no cost to legitimate hits. We welcome the Clark article because it gives us the opportunity to clarify this nuanced but extremely important distinction between legitimate and illegitimate hits.

6 Running head: EYEWITNESS IDENTIFICATION REFORMS 6 Consider, for example, a suspect in a biased lineup who stands out as the only one fitting the description of the culprit, a situation that the literature calls a biased lineup composition. Clark s summary of the literature indicates that the reform, an unbiased lineup composition in which the suspect does not stand out, can produce fewer hits than the biased one if the suspect is in fact the culprit. But, is a hit that would not occur except for the suggestive lineup procedure really a hit? If not, then is elimination of a suggestiveness-induced hit a true cost? If we say that all hits are legitimate, we would be asked to lament the lower hit rate that comes from not simply telling witnesses which lineup member they should pick. There are legitimate hits and illegitimate hits and Clark does not appropriately acknowledge a difference between the two. Clark treats extra hits that are obtained from biased instructions, biased fillers, and non-blind procedures as if they were all legitimate instead of products of suggestion. The legal system has both an obligation and an interest in adopting reforms that can eliminate suggestive procedures even if those changes might lower hit rates over what would be obtained with suggestive procedures. Some reduction in culprit identifications with lineup reforms should come as no surprise because the reforms are predominantly strategies to reduce the suggestiveness that otherwise may boost suspect identification rates. In this sense, the reforms do indeed have a cost in reduced rates of identifying the culprit if he is in the lineup. However, whereas the legal system at times must struggle to determine whether an identification obtained under suggestive circumstance may nevertheless be reliable, in the laboratory we can clearly discern the impact of suggestive procedures on witness reliability: If the unbiased procedure makes X% of hits go away, then these X% hits were simply products of the suggestive procedure. Furthermore, to belabor these lost identifications as costs is a bit

7 Running head: EYEWITNESS IDENTIFICATION REFORMS 7 backward, as this assumes that the acceptable standard is a suggestive procedure, with costs incurred as suggestiveness is removed. Recent research in decision making has shown that decision queries posited from the default (current state; no action) produce differences in decision preferences due to reference dependence (Dinner, Johnson, Goldstein, & Liu, 2011). We believe that the focus on the default (no lineup reform and the presence of suggestiveness) leads some to construe any reduced hits from reforms as unmitigated losses. But, if one instead focuses on the reform procedures as the starting point, it becomes clearer that additional hits from suggestive procedures are ill-gotten. If the reform procedures were the default, the additional false alarms that would result from changing to suggestive procedures would likely make the suggestive procedures all the less plausible as an alternative. It is necessary to inform the justice system that reforms can reduce hit rates, but it is more important to stress that the reforms are meant to replace techniques that are less reliable indicators of eyewitness memory. Discussions between eyewitness identification researchers and policy makers in the justice system have involved not just hit and false alarm rates but also the legitimacy of eyewitness evidence obtained under a given procedure. Both science and law enforcement recognize that the evidence supporting a hypothesis/case must not be improperly facilitated by suggestive procedures. Clark (2012) acknowledges the importance of procedural justice and perceived legitimacy, but does not seem to recognize how these considerations undermine his argument about lost hits. In short, there is nothing in Clark's mathematical approach that represents the concept of legitimacy. We discuss a series of lineup reforms below. For each, we consider the rationale for the reform and its impact on eyewitness decisions 3. The Double-Blind Administration Reform

8 Running head: EYEWITNESS IDENTIFICATION REFORMS 8 We begin our discussion with the double-blind lineup procedure, a central reform for law enforcement that clearly exemplifies the danger of a myopic focus on hit and false alarm rates. First proposed by eyewitness researchers 24 years ago (Wells, 1988), the double-blind lineup-- which simply requires that the person who administers the lineup (and anyone else in the room) not know which person is the suspect and which are merely fillers is now standard in many major police departments (e.g., Boston, Dallas, Denver, and others), is the law for every police department in some states (e.g., Ohio, North Carolina, Texas, and Connecticut), is the recommendation of numerous important national legal associations (e.g., American Bar Association, International Association of Chiefs of Police) and research societies (e.g., the American Psychology-Law Society), and has been a staple recommendation of almost every state reform commission in the last decade. The double-blind lineup procedure derives from the lineups-as-experiments analogy noted above (Wells & Luus, 1990), specifically the danger of an experimenter-expectancy effect. The lineup identification task involves a close interaction between the lineup administrator and the witness, with conversation flowing in both directions. Standard practice in law enforcement has been for the case detective, who knows the identity of the suspected person, to administer the lineup. Hence, there are rich possibilities for the lineup administrator to influence the witness, even inadvertently. Laboratory research indicates that lineup administrators do in fact influence witnesses in line with administrators beliefs regarding which person is the suspect. A double-blind procedure thereby has the potential to reduce both mistaken and accurate identifications (e.g., Greathouse & Kovera, 2009). Therein Clark sees a cost incurred by double-blind lineup procedure. However, Clark misconstrues the problem: If non-blind lineups produce more hits,

9 Running head: EYEWITNESS IDENTIFICATION REFORMS 9 this means that the lineup administrator likely influenced the results, an influence that the legal system itself considers illegitimate. Bemoaning any reduction in hits from the use of doubleblind lineup procedures misrepresents the very nature of the legal system and its requirement that an eyewitness identification be the result of an independent recollection. To return to the analogous domain of science, imagine the outrage if a scientist were to argue against the use of double-blind research procedures with the claim that researchers would more often confirm their hypotheses if experiments were conducted using non-blind procedures. Our view is that the only proper answer to the question of how many identifications of the culprit might be lost by using double-blind lineup procedures? is that there would be no legitimate losses. A reliable witness who has a good memory of the culprit should be able to identify the culprit without the assistance of a lineup administrator who knows which person is the suspect and which are fillers. The Admonition Reform Clark (2012) additionally argues that there is a cost in terms of lower hit rates with the pre-lineup admonition to witnesses that the culprit may or may not be in the lineup. This admonition is commonly referred to as the unbiased instruction, and is used to explicitly alert a witness to the possibility that the lineup can be legitimately rejected, that the witness may accurately conclude he s not there. In contrast, biased instructions explicitly state, or imply, that the perpetrator is in the lineup and that it is the witness s job to pick him out (Clark, p. XX). We agree that the review of early studies (Steblay, 1997) showed no reduction in hit rates but that a later analysis by Clark (2005) revealed a 9% reduction in hits from unbiased instructions. 4 Despite this, use of the admonition has been largely noncontroversial, and

10 Running head: EYEWITNESS IDENTIFICATION REFORMS 10 jurisdictions across the country that have made lineup reforms now require this instruction first introduced 31 years ago by Malpass and Devine (1981). Clark s (2012) concern seems to be that policy makers have not labored more over the reduction in hits relative to the reduction in mistaken identifications. He notes that the important New Jersey Supreme Court case of State v. Henderson (2011) made mention of the effect of unbiased instructions in reducing mistaken identifications but made no mention of reduced accurate identifications. But, the New Jersey Supreme Court reviewed Clark s (2005) article, in which he made that argument, and nevertheless the Court insisted on the pre-lineup admonition as a critical element of any properly conducted lineup. How could this be, that the New Jersey Supreme Court and other jurisdictions have not been pre-occupied with a potential loss of hits from unbiased instructions? We believe that law enforcement and the courts have adopted the lineup admonition reform not because they are unaware of the possible reduction in hits, but because they understand that a biased lineup instruction that explicitly or implicitly states that the culprit is in the lineup is a highly suggestive procedure. In contrast, the unbiased lineup instruction serves to disabuse eyewitnesses of the false assumption that the culprit is always in the lineup. 5 No one knows how often police lineups contain the culprit but, to this day, there are no laws that require police to have any evidence beyond a mere hunch as a premise for conducting a lineup. Accordingly, the admonition the person who committed this crime may or may not be in the lineup is a factual statement and consistent with a presumption of innocence. Furthermore, every DNA exoneration case involving mistaken lineup identification teaches us that the correct answer was none of the above. 6 Policy makers adoption of the pre-lineup admonition rests on a basic premise that a reliable witness with a good memory of the culprit should be able to identify the culprit without the suggestiveness of a biased instruction.

11 Running head: EYEWITNESS IDENTIFICATION REFORMS 11 Lineup Composition One recommendation for a non-suggestive lineup is that the suspect should not stand out from the fillers as the only lineup member who fits the description of the culprit or in some other way draws attention (Wells et al, 1998). Clark (2012) argues that there is a cost to correct identifications from this recommendation. Clark cites the little cost argument from the original lineup composition experiment conducted 32 years ago (Lindsay & Wells, 1980), which in fact showed almost no reduction in hit rates, and from Lindsay and Pozzulo s (1999) report of a specific experiment that showed that a match-to-description filler selection procedure did not significantly influence the rate of correct identification, both of which are statistically-correct conclusions from those experiments. It seems to us to be a stretch of his thesis to pull a few words from two articles to argue that eyewitness researchers have advocated a no cost idea to lineup composition reform. In doing this, Clark ignores a larger body of work that clearly describes and strongly discourages making the lineup fillers too closely resemble the suspect because such a strategy creates gratuitous similarity, going well beyond what is needed to protect an innocent suspect and serving to hide a guilty suspect (Wells, Rydell, & Seelau, 1993, p. 837). 7 Again, the question is why correct identifications increase if the suspect stands out and why such an increase in hits rates is unacceptable to law enforcement and the courts. Suspect identifications can be easily facilitated with a suggestive lineup composition that makes alternatives to the suspect implausible (as happens when the suspect is the only one fitting the witness s description of the culprit, for instance). Clearly, any increase in hits from making the suspect stand out is the result of a dubious and non-legitimate test of the witnesses memory of the culprit. We can push this argument to an extreme (hypothetical, we hope): The system will

12 Running head: EYEWITNESS IDENTIFICATION REFORMS 12 lose hits by not placing an arrow over the head of the suspect in the lineup, but we anticipate that lost suspect identifications from failure to use this arrow technique will not concern policy makers. The basis for a legitimate identification is that a reliable witness with a good memory of the culprit should be able to identify the culprit from a fair lineup of persons who also fit the general description of the culprit. In other words, no rigging of the lineup should be required for a reliable eyewitness. Sequential (versus Simultaneous) Lineup Reform The reforms discussed thus far are characterized by law enforcement and by scientists as less suggestive alternatives to procedures that are clearly suggestive. There is profound face validity to the idea that these default procedures often used by law enforcement (biased composition, biased instructions, non-blind administration) are suggestive and that the alternative (reform) procedures achieve their effects by lessening suggestiveness. However, we make no claim that reduced rates of identification from all possible alternative procedures are always a function of reduced suggestiveness. Indeed, we wish to avoid a circular logic that assumes any procedure that reduces hit and false alarm rates is the result of reduced suggestiveness. We return to the core issue the question of why and how a procedure affects identification rates in our consideration of sequential lineups. The staple database shared with policy makers since the very first meta-analysis (Steblay, Dysart, Fulero, & Lindsay, 2001) has clearly laid out the fact that lab data show the sequential lineup to produce fewer hits as well as fewer false alarms. This meta-analysis has been made widely available to jurisdictions that have considered this particular reform, and law enforcement has unquestionably considered the reduction of correct identifications in its policy making (see e.g., Gaertner & Harrington, 2009). The more recent meta-analysis (Steblay, Dysart, & Wells,

13 Running head: EYEWITNESS IDENTIFICATION REFORMS ) reports the same pattern of results. The fact that so many jurisdictions have nevertheless switched to sequential lineups is not testimony to a lack of awareness regarding costs, but instead a statement about the value placed on being able to trust identifications that are made from the sequential procedure more than one can trust identifications made from the simultaneous procedure (despite fewer identifications overall, the proportion of identifications that are accurate is higher with the sequential procedure, which even Clark s 2012 analysis shows). The sequential lineup reform stands out from those discussed above because the literature has offered no per-se claim (that we are aware of) that a sequential procedure is less suggestive than the simultaneous procedure (although it might be, in that the same data pattern occurs for the sequential/simultaneous effect as for other effects that involve reduction of suggestiveness). The recommendation for use of a sequential lineup does not rest on the argument that it is less suggestive than the simultaneous. Instead we make three observations: (1) the legitimacy of some of the hits that come from simultaneous lineups can be questioned (Why wouldn t a reliable witness be able to identify the culprit from a sequential array?), (2) lost hits can be recovered by following the sequential lineup with a de facto simultaneous presentation (i.e., laps) as permitted in field applications, and (3) the cost/benefit analysis is much more complex than Clark s depiction, particularly in the treatment of filler identifications from culprit-absent lineups. We address these in turn. Why Are Some Witnesses Unable to Identify the Suspect From A Sequential Lineup? There are witnesses who do not make an identification of the suspect with the sequential procedure but do so with a simultaneous array. Presumably, these witnesses need to compare one lineup member to other lineup members in order to make a choice and are more likely to choose the lineup member who looks most like their memory of the perpetrator. Is that really

14 Running head: EYEWITNESS IDENTIFICATION REFORMS 14 recognition memory? Is that a proper basis for incriminating a person? A common facet of human decision-making is that we frequently answer a different question than that asked and do so without recognizing the subtle change in the decision process (Kahneman, 2011). In this case, the witness is unable to answer the difficult question (Is this the culprit?) and instead shifts to an easier question (Which is closest?). The increase in choosing rate with a simultaneous lineup is presumed to reflect the fact that even witnesses who have a poor memory of the culprit can still make a selection of the person who is closest to their memory, a process of relative-judgment rather than recognition. Indeed, this relative-judgment process is insensitive to whether the culprit is in the lineup at all because there will always be someone who best matches the witness s memory. Consistent with this logic, we know that more witnesses who identify someone with the simultaneous procedure (but could not do so with the sequential lineup) select an innocent person (either a filler or an innocent suspect). Thus, the simultaneous lineup changes the nature of the eyewitness decision for some witnesses and leads to significantly higher rates of choosing and more misidentifications. Of course, along with this higher rate of choosing innocent people from lineups, the simultaneous also nets some guilty (hits). An argument has been made, however, that the higher rate of hits from the simultaneous lineup is actually just a result of lucky guesses stemming from a higher rate of choosing (e.g., Penrod, Garcia, & Robertson, 2005). Of course, the lucky hits emanating from a higher choosing rate are not pure luck in the same sense that a roll of the die produces a hit. The witness might use partial memory to eliminate some lineup members and then choose from among the remainder. Increased propensities to choose someone from a lineup will necessarily result in an increase in hits by chance (along with an increase in mistaken identifications), therefore we

15 Running head: EYEWITNESS IDENTIFICATION REFORMS 15 believe it is important to draw a sharp distinction between the reliability of an eyewitness and the accuracy of an identification. Consider, for instance, an eyewitness who has no memory for a culprit s face but nevertheless selects a member from the lineup and, by chance, picks the actual culprit. Although the choice was accurate, the witness is not reliable. Any lineup procedure that merely increases propensities to choose is likely to have some lucky hits by chance. But Clark s focus on hits and false alarms rather than the why of those decisions does not permit that level of analysis. A Sequential Can Always be Followed by A Simultaneous, But Not Vice Versa The sequential procedure does not easily permit relative judgment because the witness never views the lineup members side by side and does not know what the (unseen) remaining lineup members look like when making a decision on any single person. But, if detectives want to know which lineup member looks most like the witness s memory of the culprit, they can always follow the sequential lineup with a simultaneous viewing or permit the witness a second lap through the sequential. In fact, jurisdictions that have switched to sequential lineups (that we are aware of) grant a second lap of the sequential lineup upon a witness request. The second lap can be construed as a de facto simultaneous lineup in the sense that during the second lap (unlike the first) the witness knows what all lineup members look like and can more easily make a mere relative judgment. Of course, the second lap is a matter of record and should be considered in the evaluation of the witness reliability. A recent field study with actual eyewitnesses found that permitting a second lap for witnesses who explicitly requested it eliminated the small difference in suspect hit rates between simultaneous and sequential lineups while still producing significantly fewer filler identifications with the sequential technique (Wells, Steblay, & Dysart, 2011).

16 Running head: EYEWITNESS IDENTIFICATION REFORMS 16 Filler Identifications from Culprit-Absent Lineups are Not the Same as Correct Rejections We take strong issue with Clark s (2012) treatment of non-suspect filler picks as the equivalent of no identification. 8 In doing so, Clark ignores detailed observations made by Steblay et al. (2011), an article that Clark otherwise cites extensively. Fillers are, of course, known innocents who are in the lineup merely to fill it out. 9 A filler will not be charged with a crime and perhaps for this reason Clark feels free to treat filler picks from a culprit-absent lineup as unimportant. However, filler identifications are far from harmless errors. A filler pick is a wrong answer, with serious practical implications for the procedure that produced it as well as for the witness and the criminal investigation. Specifically, a filler identification spoils the witness. Consider a witness who views a culprit-absent lineup and picks a filler, an event that is more likely to occur with the simultaneous procedure. If the subsequent police investigation turns up the true culprit, it is now too late for the detective to bring that witness back for a new lineup; the witness has already positively identified a known-innocent person. In short, the witness is spoiled as unreliable, perhaps simply because the simultaneous procedure made choosing from the lineup an easier decision. Had the witness made no identification, as is more likely with the sequential procedure, the witness would still be viable for a later lineup. Far from being a rare occurrence, recent field data using actual witnesses showed that 42% of those making an identification with the simultaneous procedure picked a known-innocent filler compared to 31% with the sequential procedure (Wells et al., 2011). Steblay et al. (2011) noted the greater ability of the sequential procedure to hold down witness spoilage by suppressing filler identifications. This moderates concern about the gap between the simultaneous and sequential in hit rates. This particular sequential lineup advantage

17 Running head: EYEWITNESS IDENTIFICATION REFORMS 17 is not modeled in lab studies (or in Clark s analysis) because lab studies end the experimental session when the witness makes an identification decision. But in actual criminal investigations, a witness who rejects a culprit-absent lineup might later go on to identify the true culprit in a subsequent lineup. Large numbers of law enforcement agencies have adopted the sequential procedure and have done so knowing that it can reduce identifications of the culprit. But they also know that the sequential procedure reduces mistaken identifications of innocent people, reduces the rate of filler picks, and improves the ratio of accurate to mistaken identifications. Moreover, law enforcement can pick up any lost hits by permitting a second lap. Final Remarks We welcome discussion of the potential costs and benefits of reforms to eyewitness identification procedures. As we have explained, however, a narrow focus on hits and false alarms dangerously impedes progress in the development of research-based procedures for collection of legitimate memory evidence. The double-blind lineup procedure provides a particularly compelling illustration of how any increase in hits that results from a suggestive procedure (a non-blind lineup) cannot be considered legitimate. This argument extends to biased instructions, the use of lineup procedures that make the suspect stand out, and practices that permit witness certainty to be contaminated by external information. A limited analysis of hits and false alarms also reveals a stunning lack of appreciation for why identification rates change with adjustments to lineup procedure. Our central purpose in this paper has been to clarify the rationale for revisions to identification procedures. However, our concern with a Clarkian analysis also extends to substantive issues beyond the eyewitness lineup decision. The analysis of costs and benefits for

18 Running head: EYEWITNESS IDENTIFICATION REFORMS 18 reforms in the real world also must consider the broader context, a point made clear as we considered the realities of how sequential lineups are employed in field practice and of the implications of filler identifications for ongoing criminal investigations. Four points highlight additional problems in the Clarkian position. First, we note that Clark (2012) tends to equate a failure to identify a culprit from a lineup with a criminal going free. But this neglects the fact that guilty people are frequently convicted at trial or plead guilty to the crime even when the eyewitness did not identify the culprit from the lineup because guilty people tend to have other evidence against them that is uncovered during an investigation. Moreover, even Clark s own writings show that multiple witness cases are extremely common and that if one witness does not identify the culprit, another witness might well do so (Clark & Wells, 2008). Second, it is useful to revisit an argument first made by Steblay et al. (2011) regarding the damages incurred by mistaken identifications. Specifically, Steblay et al. ask us to compare two equations: Mistaken identification = inculpate the innocent + culprit escapes identification Failure to identify the culprit = culprit escapes identification Note that a mistaken identification is always two errors, namely an identification of an innocent person and a failure to identify the culprit. A failure to identify the culprit, in contrast, is only one error, namely the culprit escapes identification. Accordingly, unless one places no negative value on inculpating the innocent, the two errors cannot be equally bad and the mistaken identification has to have more negative value. In addition, a mistaken identification often leads the investigation away from the actual culprit as charges move forward against the innocent

19 Running head: EYEWITNESS IDENTIFICATION REFORMS 19 suspect, whereas a mere failure to identify the culprit does not necessarily set him free it might simply lead to a more thorough search for other evidence against that person. Third, the potential implications of Clark s argument are troubling to us. Consider, for example, what would happen if we extended a Clarkian analysis of hits and false alarms to Loftus (1979) classic work on post-event information effects. How many more witnesses might correctly report a stop sign (a hit ) if an earlier question subtly inserted the words stop sign? As construed through the lens of Clark s analysis, failure to use this leading question will produce a cost to the investigation (a lower hit rate). But, this perspective seriously misses the mark of why the legal system should avoid leading questions. Leading questions are suggestive and they obscure the evaluation of what the witness independently remembers. A leading-question prompts witnesses to report more than they actually know and, in the case of misinformation (when no stop sign actually exists), to report details that never existed. Finally, we are concerned that a mere empirical count of lost accurate identifications can be used to justify the continued use of suggestive procedures. If one construes of suggestiveness-induced hits as legitimate, then suggestiveness will continue. Clark has provided some in the justice system with an argument against double-blind lineup procedures, against unbiased instructions, and against the use of fillers who do not let the suspect stand out. But even identifications of the guilty must be obtained using non-suggestive procedures. Calling that a cost of reforms is an interesting argument, but we believe it to be a misleading message to policy makers and we lament the increased resistance to reform (and continued use of suggestive procedures) that some jurisdictions might now exhibit in response to a myopic hit/false alarm perspective.

20 Running head: EYEWITNESS IDENTIFICATION REFORMS 20 References Clark, S. E. (2005). A re-examination of the effects of biased lineup instructions in eyewitness identification. Law and Human Behavior, 29, doi: /s Clark, S. E. (2012). Costs and benefits of eyewitness identification reform: Psychological science and public policy. Perspectives on Psychological Science. Clark, S. E., & Godfrey, R.D. (2009). Eyewitness identification evidence and innocence risk. Psychonomic Bulletin & Review, 16, doi: /PBR Clark, S. E., & Wells, G. L. (2008). On the diagnosticity of multiple-witness identifications. Law and Human Behavior, 32, doi: /s Dinner, I., Johnson, E. J. Goldstein, D. G. & Liu, K. (2011). Partitioning default effects: Why people choose not to choose. Journal of Experimental Psychology: Applied, 17, doi: /a Douglass, A. B., & Steblay, N. (2006). Memory distortion in eyewitnesses: A meta-analysis of The post-identification feedback effect. Applied Cognitive Psychology, 20, doi: /acp.1237 Dysart, J. E., Lawson, V. Z., & Rainey, A. (in press). Blind lineup administration as a prophylactic against the postidentification feedback effect. Law and Human Behavior. doi: /h Gaertner, S., & Harrington, J. (2009). Successful eyewitness identification reform: Ramsey County s blind sequential lineup protocol. The Police Chief, 76 (4), Greathouse, S., & Kovera, M. (2009). Instruction bias and lineup presentation moderate the effects of administrator knowledge on eyewitness identification. Law and Human Behavior, 33, doi: /s x

21 Running head: EYEWITNESS IDENTIFICATION REFORMS 21 Kahneman, D. (2011). Thinking, fast and slow. New York: Farrar, Straus, & Giroux. Lindsay, R. C. L., & Pozzulo, J. D. (1999). Sources of eyewitness identification error. International Journal of Law and Psychiatry, 22, doi: /s (99)00014-x Lindsay, R. C. L., & Wells, G. L. (1980). What price justice? Exploring the relationship between lineup fairness and identification accuracy. Law and Human Behavior, 4, Loftus, E. F. (1979). Eyewitness testimony. Cambridge, MA: Harvard University Press. Malpass, R. S., & Devine, P. G. (1981). Eyewitness identification: Lineup instructions and the absence of the offender. Journal of Applied Psychology, 66, doi: // Penrod, S., Garcia, L., & Robertson, R. (July 2005). Assessing the impact of eyewitness guessing and lineup bias on eyewitness performance. 29th International Congress on Law and Mental Health, Paris. Perry v. New Hampshire, 565 U.S. (2012). State v. Henderson, A. 3d -, 2011 WL (N.J. Aug. 24, 2011). Steblay, N. M. (1997). Social influence in eyewitness recall: A meta-analytic review of lineup instruction effects. Law and Human Behavior, 21, Steblay, N. K. (in press). Lineup instructions. In B.L. Cutler (Ed.), Reform of Eyewitness Identification Procedures. Washington, D.C.:APA Press. Steblay, N., Dysart, J. E., Fulero, S., & Lindsay, R. C. L. (2001). Eyewitness accuracy rates in sequential and simultaneous lineup presentations: A meta-analytic review. Law and Human Behavior, 25, doi: /a:

22 Running head: EYEWITNESS IDENTIFICATION REFORMS 22 Steblay, N., Dysart, J. E., Fulero, S., & Lindsay, R. C. L. (2003).Eyewitness accuracy rates in police showup and lineup presentations: A meta-analytic comparison. Law and Human Behavior, 27, Steblay, N. K., Dysart, J. E., & Wells, G. L. (2011). Seventy-two tests of the sequential lineup superiority effect: A meta-analysis and policy discussion. Psychology, Public Policy, and Law, 17, doi: /a Wells, G. L. (1988). Eyewitness identification: A system handbook. Toronto: Carswell Legal Publications. Wells, G. L., & Bradfield, A. L. (1998). Good, you identified the suspect : Feedback to eyewitnesses distorts their reports of the witnessing experience. Journal of Applied Psychology, 83, doi: / Wells, G. L., & Luus, C. E. (1990). Police lineups as experiments: Social methodology as a framework for properly conducted lineups. Personality and Social Psychology Bulletin, 16, Wells, G. L., Rydell, S. M., & Seelau, E. P. (1993). The selection of distractors for eyewitness lineups. Journal of Applied Psychology, 78, Wells, G. L., Small, M., Penrod, S., Malpass, R. S., Fulero, S. M., & Brimacombe, C. A. E. (1998). Eyewitness identification procedures: Recommendations for lineups and photospreads. Law and Human Behavior, 22, doi: /A: Wells, G. L., Steblay, N. M., & Dysart, J. E. (2011). A test of the simultaneous vs. sequential lineup methods: An initial report of the AJS national eyewitness identification field studies. Des Moines, IA: American Judicature Society.

23 Running head: EYEWITNESS IDENTIFICATION REFORMS 23 Wells, G. L., & Turtle, J. W. (1986). Eyewitness identification: The importance of lineup models. Psychological Bulletin, 99, doi: /

24 Running head: EYEWITNESS IDENTIFICATION REFORMS 24 Footnotes 1. Clark later shortens the little or no cost presumption to a No Cost presumption, which we believe could be misleading to readers and is inconsistent with the few articles he quotes. 2. Our focus in this article is primarily on the idea that reductions of illegitimate hits that result from reform procedures are not properly construed as costs if the hits were the result of suggestive procedures. But we also have issues with Clark s summary of the data from the extant literature. For example, other recent meta-analyses have yielded different numbers for biased and unbiased lineup instructions (Steblay, in press) and for simultaneous versus sequential lineups (Steblay et al., 2011). Hence, our focus on the issue of the illegitimacy of hits deriving from suggestive procedures in this article should not be construed as agreeing with Clark s methods of representing and calculating summary statistics on these procedures. In addition, we question his conclusion that a base rate of 60% for the police suspect being guilty is fairly low. We know of no evidence or source of information for this conclusion, which is critical and central to his conclusions regarding the effects of certain procedures on lost hits. 3. Clark was somewhat selective in his choice of reforms for discussion. Other reforms, for example, include the single-suspect lineup model (Wells & Turtle, 1986) that prevents placing more than one suspect in a lineup. Another major procedural reform based on the empirical literature is the certainty reform, which involves securing a certainty statement at the time of the identification (i.e., obtaining this statement prior to other events, in particular post-identification feedback, that can contaminate the certainty of the witness, see Wells & Bradfield, 1998; Douglass & Steblay, 2006; Dysart, Lawson, & Rainey, in press).

25 Running head: EYEWITNESS IDENTIFICATION REFORMS A more recent pre-identification instruction bias meta-analysis (Steblay, in press) examined the impact of the specific recommended may or may not admonition (earlier reviews included variations of suggestive versus non-suggestive instructions), showing a 5% reduction in hits, not 9%. 5. The fact that the pre-lineup admonition, which is a true statement, reduces choosing rates relative to no instruction is evidence that witnesses presume that the culprit is in the lineup unless specifically disabused of that presumption. 6. Two known exceptions are the DNA exoneration cases of John Jerome White and Marvin Anderson. Although witnesses in both cases mistakenly chose the innocent suspect who was later exonerated, the actual culprits had been present in the lineups and were not chosen. 7. In fact, making fillers highly resemble the suspect is not the recommended reform; as explained in the American Psychology-Law Society position paper in eyewitness identification procedures selecting distractors so as to make them resemble the suspect can create undue homogeneity and interfere with recognition of the actual culprit. Using the fit-to-description criterion, on the other hand, preserves sufficient variability across lineup members and yet does not make the suspect stand out (Wells, et al., 1998, p.25). This statement and its reasoning, taken from the most widely distributed and adopted reform recommendations article, clearly illustrates the holes in Clark s thesis that the literature has made recommendations without regard to what happens to hit rates. 8. Clark s estimates of mistaken identification rates rely on a presumption that, in the absence of an a-priori innocent suspect in a culprit-absent lineup, the rate of mistaken identification of an innocent suspect is the filler identification rate times 1/N, where N is the number of lineup members (Clark & Godfrey, 2009,). But, this assumes that a lineup is perfectly fair and that

26 Running head: EYEWITNESS IDENTIFICATION REFORMS 26 the chance of picking an innocent suspect is no different from the chance that any given filler would be picked. Real-world estimates for this are not available, but even lab studies that attempt to use perfectly fair lineup procedures tend to show higher rates for identifying an innocent suspect than for identifying a filler. In the controlled lab studies that compared show-ups to lineups, for example, the rate of choosing the innocent suspect from a culpritabsent lineup is over three times the rate of the average filler from absent lineups (see metaanalysis by Steblay, Dysart, Fulero, & Lindsay, 2003). Thus, Clark s false identification rates for the simultaneous procedure are likely to be underestimates. 9. Somewhat ironically, Clark fails to acknowledge that the use of only one suspect embedded among known-innocent fillers (called the single-suspect lineup model) is a reform that came from psychology (Wells & Turtle, 1986). And, lest one think that this need for knowninnocent fillers was already obvious to law enforcement, we note that the initial photo lineup from which Jennifer Thompson identified Ronald Cotton in 1984, a central example of a mistaken identification featured in Clark s article, was actually an all-suspect lineup. Hence, Clark s hypothetical about Jennifer identifying the person next to Cotton as being a mere filler identification does not apply to the photo-lineup that Jennifer saw - everyone in that initial photo lineup was a suspect, there were no fillers, and anyone identified by her would likely have been charged with the crime.