See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/8563855

What Makes a Good Alibi? A Proposed Taxonomy

Article in Law and Human Behavior · April 2004

DOI: 10.1023/B:LAHU.0000022320.47112.d3 · Source: PubMed

CITATIONS READS

105 3,887

2 authors:

Elizabeth A. Olson Gary L. Wells

University of Wisconsin - Whitewater Iowa State University
17 PUBLICATIONS 1,418 CITATIONS 240 PUBLICATIONS 17,910 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Gary L. Wells on 20 October 2014.

The user has requested enhancement of the downloaded file.

P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

Law and Human Behavior, Vol. 28, No. 2, April 2004 (°

C 2004)

What Makes a Good Alibi? A Proposed Taxonomy1

Elizabeth A. Olson2,3 and Gary L. Wells2

A taxonomy of alibis is proposed involving two forms of supporting proof: physical

evidence and person evidence. Levels of physical evidence and person evidence were
combined to create 12 cells in the taxonomy. Participants (n = 252), who were asked
to assume the role of detectives, evaluated alibis representing these 12 cells. The believ-
ability of the alibis generally followed the taxonomy’s predicted pattern, but physical
evidence, when present, tended to overwhelm the person evidence more than had been
expected. In addition, alibi evaluators seemed to not consider the possibility that a
stranger who corroborated an alibi might be mistaken about the identity of the person.
Trait inferences regarding the alibi providers tended to follow the believability data,
even when the traits themselves were not relevant to believability of the alibi. We call for
the development of a literature on the psychology of alibis, recommend the taxonomy
as a framework, and suggest several avenues of inquiry.
KEY WORDS: alibi, alibi witness, corroborator, alibi provider, belief.

Alibi; n. A plea of having been at the time of the commission of an act elsewhere than at
the place of commission.

Suppose you discovered, much to your surprise, that you were considered a suspect
in a criminal investigation. You are innocent, so you simply have to account for your
whereabouts for the time of the crime. If you remember where you were at the time
of the crime, you have an alibi—an alibi is simply a claim that you were elsewhere.
However, having an alibi is the easy part; proving an alibi is another matter altogether.
We first developed an interest in alibis in part through examination of cases in
which innocent people were convicted by juries and their innocence was established
later using forensic DNA testing that was unavailable at the time of their conviction
(see Wells et al., 1998). Although the dominant evidence used against these innocent
people was eyewitness identification testimony, we noticed that “weak alibis” were
often exploited by prosecutors and used as incriminating evidence. Although the
1 Portions
of these data were presented at the 2001 Biennial Meeting of the Society for Applied Research
in Memory and Cognition.
2 Department of Psychology, Iowa State University, Ames, Iowa.
3 To whom correspondence should be addressed to Department of Psychology, Iowa State University,
W112 Lagomarcino Hall, Ames, Iowa 50011; e-mail: [email protected].

157
0147-7307/04/0400-0157/1 °
C 2004 American Psychology-Law Society/Division 41 of the American Psychology Association
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

158 Olson and Wells

individuals were innocent of the offense (and, therefore, their alibis were likely true),
their alibis were nevertheless ineffective.
These cases, in which innocent people’s alibis did not prevent them from being
erroneously convicted, led us to wonder, what is a weak alibi? Are there fundamental
properties of alibis that can be identified as strong or weak? This seemed to us to be
a natural interface between social psychology and law, and we expected to find an
empirical literature addressing the issue. Instead, we have been surprised and a bit
frustrated to discover that there is a dearth of empirical literature on alibis.
How should we approach the development of an empirical literature on the
psychology of alibis? The legal literature provides us with little guidance because it
focuses primarily on legal technicalities of the alibi defense. Two technicalities ac-
count for most of the legal writings on alibis, namely the prior notice rule and the
issue of instructions to juries. The prior notice rule simply states that the defense
must give pretrial notice that it intends to present an alibi defense. The legal rea-
soning behind this rule, which is applicable in most U.S. jurisdictions, is that an alibi
can be easily manufactured in the final hours of the defense, thereby preventing a
thorough investigation (Epstein, 1964). The instruction issue concerns the matter of
whether or not to inform the jury that the defendant is not required to prove an
alibi and that the prosecution has the burden to negate the alibi. The requirement
of such instructions varies widely across U.S. jurisdictions (Friedman, 1998). Other
legal writings allude to distinctions between interested alibi-corroborating witnesses
(usually, family members) and disinterested alibi-corroborating witnesses, but fail
to develop deeper understandings of this distinction (Gooderson, 1977; Sullivan,
1971).
Although an empirical literature on alibis has not yet been developed, there are
a few empirical studies that have used alibis to test hypotheses about other issues,
principally eyewitness identification issues. Leippe (1985) for example, used alibi
testimony to examine mock juror judgments of various forms of eyewitness identi-
fication and nonidentification evidence, but the alibi information was held constant,
not manipulated. McAllister and Bregman (1989) manipulated whether an alibi wit-
ness positively identified the defendant (thereby corroborating the alibi) or not and,
as expected, failure of the alibi witness to corroborate the defendant’s alibi led to
more guilty verdicts from mock jurors. Lindsay, Lim, Marando, and Cully (1986) ma-
nipulated whether or not an alibi witness was a relative of the defendant and found
that only the nonrelative alibi witness was able to reduce convictions when there was
an eyewitness who had identified the defendant as the culprit. Although all three
of these studies used alibi witnesses, the studies were focused primarily on issues of
eyewitness identification, not alibis. Alibis were used merely as tools to find out how
people think about eyewitness identification issues. The only empirical study that
we have been able to find that was devoted to alibis per se was one conducted by
Culhane and Hosch (in press). In their study, the alibi witness was either a neighbor
or a girlfriend and the witness was either certain or not certain in making either an
identification or nonidentification of the defendant as being at his home during the
time of the crime. Their results, like those of Lindsay et al., showed that mock jurors
were persuaded by the alibi only if the alibi corroborator had no relationship with
the defendant.
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

A Taxonomy of Alibis 159

Clearly, the relationship between the defendant and the alibi corroborator af-
fects the believability of the alibi. But this represents a small start to what seems to us
to be a potentially rich literature. It is our intent to facilitate the development of an
empirical and conceptual literature on the psychology of alibis, and accordingly, this
paper has two goals. First, we introduce some basic distinctions, concepts, and ideas
that could serve as a framework for the development of an empirical literature on the
psychology of alibis. In doing this, we hope to prime the interest of psychological re-
searchers in the fascinating applied and theoretical possibilities that can accrue from
the systematic study of alibis. Second, we propose and test a taxonomy of alibis. We
argue that these goals are closely related—a taxonomy of alibis is needed to develop
a coherent framework for hypotheses about alibis and to study alibis empirically.

Some Basic Concepts and Distinctions Regarding Alibis

We make a distinction between alibi in the colloquial sense and alibi as used in
this paper. Generally, alibi is used to refer to the story, or claim, that is proffered by a
suspect in a criminal investigation, and when people say “no alibi” they are referring
to an alibi that has no accompanying proof (e.g., many people would consider a
plea that the suspect was home alone to be “no alibi” at all). However, we use the
definition of alibi from the legal system: “a defense that places the defendant at
the relevant time of the crime in a different place than the scene involved and so
removed therefrom as to render it impossible for one to be the guilty party” (Nolan,
1990). Hence, being home alone is in fact an alibi because it is a claim that places the
person at a time and place that, if true, makes it impossible for that person to have
committed the crime. From a legal defense perspective, however, the claim itself is
only one component of the alibi because one must also have evidence, or proof, that
is offered in support of the alibi claim.
Alibis can vary greatly in their surface attributes—the story line of the claim.
By surface attributes, we mean the setting (e.g., home, work, public event), activities
(e.g., sleeping, eating, playing, working), and other characteristics of the story. But,
it is not the surface attributes themselves that make some alibis strong and others
weak. Instead, it is the level of proof that can be produced in support of the story.
An at-work alibi might be no more persuasive than an at-home alibi if there are no
work records or coworker corroborators to support the alibi. What distinguishes one
alibi from another is the level of proof that can be produced to support the alibi, not
the surface attributes of the alibi itself. Although we acknowledge that there are key
surface attributes to alibis (e.g., at-home setting vs. public event setting) that affect
the likelihood or difficulty of having proof, we assert that the important dimension
is proof, not the surface attributes themselves.
Defining alibi in terms of the proof offered in support of the alibi allows us
to discuss features of strong or weak alibis and compare various alibis with each
other regardless of their surface attributes. The surface attributes of alibi stories can
vary greatly, but the types of evidence offered to prove the alibi stories fall into one
of two broad categories: physical evidence and person evidence. Physical evidence
might include such items as receipts, videotape recordings, or other physical records
that indicate the person was at a particular location at a given time. Person evidence
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

160 Olson and Wells

might involve testimony from friends, relatives, or strangers who can corroborate the
claim that the alibi provider was at a particular place at a given time.4 An important
quality of any corroborating evidence is that it must solve a time/space problem. The
proof must speak to both the time of the crime and the space the alibi provider was
occupying at that time. In short, a mere claim that one was at a particular location
at a given time is likely to be insufficient; there must be some physical or person
evidence to support the claim for an alibi to be considered at all believable.
In calling for a literature on the psychology of alibis, one distinction that we think
should be made is between the alibi provider and the alibi itself. Undoubtedly, there
are characteristics of alibi providers that make their alibi more or less believable than
someone else who gives the same alibi with the same level of proof. The ex-convict
who produces an ATM receipt from a distant location from the crime for the time
the crime was committed is going to be less believable than the business owner with
no prior record who produces an ATM receipt. It is tempting to think that varying
the characteristics of alibi providers is the same as studying alibis, but the difference
between the ex-convict and the businessman has nothing to do with alibis per se.
Hence, it is important to make clear that an understanding of the believability of
alibis requires us to study the alibis themselves, not the characteristics of the alibi
provider.
The concept of alibi proof raises one of the most important questions that we can
imagine regarding alibis. Even if innocent people can reconstruct where they were
at a particular time, how often can they actually prove it? How often can we expect
innocent people to have physical evidence to corroborate their alibis? How often are
people alone, rendering them unable to provide person evidence to corroborate their
alibi? How often are people around others in public who could not later recognize
them so as to corroborate their alibi? How often are people with someone they know,
but that person cannot absolutely attest to the time/space issue or remembers the
time differently? These questions fall into a domain that we call “alibi generation.”

A Proposed Taxonomy of Alibis

We envision a research program on alibi generation in which innocent people
would have to account for their whereabouts for specific days and times in the re-
cent past. We are currently collecting such data, but the initial research reported in
the current paper necessarily begins at a more rudimentary level. This research is
concerned with the development of a taxonomy of alibis.
We argue that a taxonomy of alibis is needed to move on to other research
questions about alibis. For instance, suppose a researcher wanted to “score” alibis
from actual cases for descriptive purposes. How does one classify alibis for purposes
of scoring them? Or, suppose a researcher hypothesized that people with certain de-
mographic characteristics (e.g., unemployed, unmarried) are less likely to have good
alibi proof than are people with other demographic characteristics (e.g., employed,
4 The term alibi provider refers to the suspect or defendant who is being questioned regarding his or her
whereabouts at the time of the crime. We call a person whose statements are put forward to support the
alibi an alibi corroborator. Although an alibi corroborator is “providing an alibi” for the suspect in the
colloquial sense, we reserve the term alibi provider for the suspect him or herself.
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

A Taxonomy of Alibis 161

married). How does one score such alibis for “goodness” of proof or for qualitative
differences in types of alibis? There must be some way to organize the alibis that are
generated under varying conditions. Our proposed taxonomy is intended to function
as an organizing structure that can be put to such purposes.
We think that the most useful taxonomy is one that sorts alibis according to
the extent to which alibi evaluators will accept the alibi, that is, the extent to which
evaluators believe that the alibi provider could not have committed the crime. Thus,
the underlying psychological construct in our taxonomy is the “believability” of
the alibi. Believability seems to be the basic psychological dimension along which
all alibi evaluators, from detectives to prosecutors to judges to jurors, make their
judgments about the alibi. Although there is no way to have a taxonomy of the
“absolute level” of alibi believability (because alibis will interact with other details
of a particular case to affect final judgments about the alibi provider), we can examine
the relative believability of alibis within the taxonomy by holding nonalibi evidence
constant. Believability is relatively easy to measure on a continuous scale; most
people are accustomed to reporting such beliefs on a percentage scale. Accordingly,
the believability construct not only drove our initial construction of the taxonomy
but also served as the principal dependent measure for purposes of validating the
taxonomy.
Whereas the underlying psychological dimension for our taxonomy is believabil-
ity, the underlying structure of the taxonomy is based on the two forms of proof that
can be used with alibis, namely physical evidence and person evidence. We can think
of no other form of supporting evidence for an alibi other than a person who is brought
forward to corroborate the alibi or physical evidence to support the alibi. Because an
alibi provider can have physical evidence, person evidence, both physical and person
evidence, or neither physical nor person evidence, it seemed natural for us to create
a factorial taxonomy with physical evidence and person evidence as the factors.
Another underlying construct in our proposed taxonomy is “perceived ease of
fabrication.” By perceived ease of fabrication, we mean the evaluator’s perception of
how easily the alibi provider can manufacture or orchestrate the evidence supporting
the alibi. For example, an alibi provider who claims he was watching a television
program at the time of the crime might support his alibi by describing events in
the television program. However, it would be easy to have taped the program and
watched it later or downloaded a program summary from the Internet at a later time.
On the other hand, it would be difficult for an alibi provider whose image appears
on a dated and timed security video from a store across town to have fabricated
the video. Of course, it is possible to have tampered with the store’s tapes, but the
relative difficulty of fabricating the video should make the alibi proof very strong.
We apply the same underlying construct to person corroborators. It would be easier
to orchestrate alibi-corroborating testimony from a relative or friend than from a
stranger, for instance, because their relationship to the alibi provider could motivate
them to fabricate a story favoring the alibi provider. Sullivan (1971) foreshadows
this distinction when he classes alibi corroborators into two general categories: those
interested in the outcome of the case (such as relatives and friends) and those who
are not. He goes on to assert that disinterested witnesses are rare and valuable to an
alibi provider’s case.
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

162 Olson and Wells

On the basis of the ease-of-fabrication construct, we created multiple levels of

person and physical evidence for the taxonomy. However, we felt that the ease-of-
fabrication construct did not capture all of what needed to be captured for the person
evidence dimension. A person who corroborates an alibi might be either lying or
mistaken. A complete stranger runs the risk of misidentifying the alibi provider, but
has no motive to lie. A person with a close relationship to the suspect (e.g., spouse),
on the other hand, is unlikely to misidentify the alibi provider, but has a potential
motive to lie for the suspect to protect him or her. Attribution theories show strong
support for the principle of discounting in which the presence of one explanation
leads observers to discount another explanation (Gilbert & Malone, 1995). Hence,
attribution theory might predict that the focal explanation for the alibi corroborator’s
claim (i.e., the corroborator actually saw the suspect at the critical time) would be
discounted by the possibility of mistaken identification in the stranger case and by
lying in the close relationship case. Perhaps the most believable corroborator is
one who is familiar with the alibi provider but not motivated to protect or lie for
the alibi provider—a nonmotivated familiar other. Hence, we thought it would be
important for the taxonomy to distinguish between three kinds of alibi corroborators:
a motivated familiar other (who is not likely to be mistaken but might lie), a complete
stranger (who is not motivated to lie but might be mistaken), and a nonmotivated
familiar other (who is not motivated to lie for the person and also is not likely to
have mistakenly identified the person).

Overview
We combined three levels of the factor physical evidence (none, easy to fabri-
cate, and difficult to fabricate) and four levels of the factor person evidence (none,
motivated familiar other, nonmotivated familiar other, nonmotivated stranger) into
the taxonomy in Fig. 1. The taxonomy is organized from the lowest level of predicted
believability (upper left of the factorial taxonomy) to the highest level of believability
(lower right of the factorial taxonomy).
We created several exemplar alibis for each cell of the taxonomy. Using multiple
exemplars for each cell of the taxonomy (e.g., mother, friend, brother representing
the motivated familiar other category) was important so as to not confuse the effect
of a category-level cell (e.g., motivated familiar other) with a particular instance
(e.g., mother) of that category cell (see Wells & Windschitl, 1999). Participants read
a brief description of a crime and then read the alibis of three suspects. Participants
were asked to evaluate each alibi as though they were detectives investigating this
criminal case. Each alibi was evaluated sequentially before reading the next alibi.
Our primary measure was the believability of the alibi, but participants also made
judgments of the likelihood of guilt. Participants also evaluated the alibi providers
on a number of personality trait dimensions, some of which were relevant to alibi
believability and some of which were not.
We expected each of the two factors in the taxonomy, that is, physical evidence
and person evidence, to have main effects on believability. We also expected an
interaction. In particular, although we expected to find differences across levels of
person evidence when physical evidence was nonexistent or easy to fabricate, we
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

A Taxonomy of Alibis 163

Fig. 1. A taxonomy of alibi believability.

expected levels of person evidence to not have an impact at the highest level of
physical evidence. We expected the likelihood-of-guilt and relevant trait measures
to follow the same general pattern as the believability data, but we did not expect
these measures to be as sensitive to the taxonomy levels as the believability measure.
We did not expect the traits that were irrelevant to believability to be affected by
levels of the taxonomy.

METHOD

Participants
Participants for the main study were 252 students from a large Midwestern
university recruited for an experiment titled “Police Detective Reasoning Skills.”
A separate sample of 46 participants were used to conduct a manipulation check
on the scenarios. Participants earned extra credit in psychology classes for their
participation. All participants were treated in accordance with the “Ethical Princi-
ples of Psychologists and Code of Conduct” (American Psychological Association,
1992).
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

164 Olson and Wells

Design
This experiment was a 4 (Person evidence) × 3 (Physical evidence) mixed fac-
torial design. Presentation of physical evidence was manipulated within participants,
and presentation of person evidence was manipulated between participants. Because
participants viewed three alibis within one level of person evidence, it was necessary
to create three exemplars for each category of person evidence to reduce suspicion
and to alleviate stimulus sampling concerns (Wells & Windschitl, 1999). For example,
the exemplars from the motivated familiar other category were mother, brother, and
best friend, whereas the nonmotivated familiar other category included a grocery
store cashier, a bookstore clerk, and a taxi driver. This multiple-exemplar approach
ensured that a given participant would not encounter the same exemplar of a person
corroborator in more than one alibi.
The order of the within-subjects physical evidence was counterbalanced so that
each level of physical evidence occurred equally often in the first, second, and third
alibis. For instance, one participant might see alibis with no physical evidence, easy-to-
fabricate physical evidence, and difficult-to-fabricate physical evidence, in that order.
Another participant might see alibis with difficult-to-fabricate physical evidence, no
physical evidence, and easy-to-fabricate physical evidence, in that order. Likewise,
the order of exemplars was counterbalanced so that each exemplar appeared equally
often in the first, second, and third alibis, as well as equally often with each level of
physical evidence.

Materials
Stimulus Materials
The crime scenario was in the form of a police report consisting of two parts.
The initial report contained an account of a police officer’s investigation of an armed
robbery at a convenience store. A follow-up report (shown on the same page) ex-
plained that the investigation had reached a dead end with multiple suspects, and
it had become necessary to evaluate the suspects on the basis of their alibis. There
were 36 short alibi reports used in the experiment, three exemplars for each cell of
the 4 × 3 taxonomy.
Table 1 gives four examples of the 36 alibis. Example A is from the no person ev-
idence and no physical evidence cell, example B is from the motivated familiar-other
and easy-to-fabricate physical evidence cell, example C is from the nonmotivated fa-
miliar other and difficult-to-fabricate physical evidence cell, and example D is from
the nonmotivated stranger and difficult-to-fabricate physical evidence cell. Some
readers might be concerned that the differences between one alibi and another alibi
in Table 1 involve more than just the evidence offered because they also differ in
the basic story line (e.g., one was eating pizza, one was taking a walk, and so on).
However, it is crucial to note that a given story line (e.g., was in a grocery store at
the time) was used as the setting for no physical evidence, easy-to-fabricate physi-
cal evidence (e.g., cash receipt), and difficult-to-fabricate physical evidence (security
video) equally often across participants. Hence, although no individual participant
evaluated more than one alibi with the same alibi story line, alibi story lines were not
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

A Taxonomy of Alibis 165

Table 1. Alibi Examples

Type of alibi Example
No person evidence Suspect A. M. initially could not remember where he was between
No physical evidence 7:30 and 8:30 on the evening in question. Later in the interview
he claimed that he had been out for a walk in his neighborhood
on the east side of the city. A. M. has no record of gun ownership.
Motivated familiar other Suspect B. L. said he was entertaining his brother in his home on
Easy-to-fabricate the west side of the city on the evening in question. He said they
had ordered pizza and provided a pizza delivery receipt, paid in
cash, timed 8:07 pm. A statement from the brother was taken; the
brother claimed they had been in the home for the entire time
between 7:30 and 8:30 pm. Suspect does not own a gun.
Nonmotivated familiar other Suspect C. Z. claimed he had been at a Check-Into-Cash store in the
Difficult-to-fabricate central city between 8:00 and 8:20 pm on the evening in question.
He said that he regularly goes to that store to get cash. The teller
at the store recognized a picture of the suspect and agreed that
he is a regular customer. The teller also indicated that he was
there that night. Security camera video from the store showed
the suspect in the store between 8:03 and 8:18. Suspect has no
history of gun ownership.
Nonmotivated stranger Suspect C. Z. claimed that he had been at a grocery store in the
Difficult-to-fabricate city between 7:45 and 8:30 pm on the evening in question. He
said that he had been grocery shopping, and then stopped at the
ATM in the grocery store to withdraw cash. After seeing a pic-
ture of the suspect, a cashier at the customer service desk said
that she remembered seeing the suspect at the store that night at
approximately 8 pm. Camera video from the ATM showed the
suspect withdrawing money at 8:26 pm. Suspect has no history of
gun ownership.

confounded with evidence levels. All 36 alibis can be viewed at the following internet
address: http://www.psychology.iastate.edu/faculty/glwells/alibiscenarios.htm

Manipulation Checks on the Scenarios

It was intended that the three person corroborator conditions would differ along
two dimensions: the willingness of the corroborator to lie for the suspect and the
chances that the corroborator would be honestly mistaken. More specifically, it was
intended that the main difference between the motivated-other corroborator and the
nonmotivated familiar-other corroborator would primarily be one of willingness to lie
whereas the main difference between the nonmotivated familiar-other corroborator
and the stranger corroborator would primarily be in the chances of an honest mistake.
With regard to physical evidence, it was intended that the two physical evidence sce-
narios differ in terms of perceived ease of fabricating the evidence. A separate sample
of 46 participants were asked to evaluate the person corroborator scenarios for will-
ingness of the corroborator to lie for the suspect and for the chances that the corrob-
orator could have been honestly mistaken. These same 46 participants also evaluated
the easy and difficult-to-fabricate physical evidence scenarios for ease of fabrication.
As expected, the motivated other was perceived as much more likely to lie for the
suspect than was the nonmotivated familiar other, M = 7.11 and 3.63, respectively,
t(45) = 7.94, p < .001. Also as expected, the stranger corroborator was perceived
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

166 Olson and Wells

as more likely to be mistaken than was the nonmotivated familiar other, M = 6.24
and 5.34, respectively, t(45) = 2.09, p < .05. Finally, the difficult-to-fabricate physical
evidence was perceived as more difficult to fabricate than was the easy-to-fabricate
physical evidence, M = 7.50 and 4.76, respectively, t(45) = 6.00, p < .001.

Dependent Measures
For the primary dependent measure, participants rated each alibi on an 11-point
Likert-type scale of believability (0 = I do not believe him at all, 10 = I believe him
completely) immediately after reading each alibi (hereafter called the belief mea-
sure). Participants also rated on an 11-point Likert-type scale the probability that
the suspect is the one who committed the crime (0 = totally unlikely, 10 = he is
certainly the gunman) after reading each alibi (hereafter called the likelihood mea-
sure). This likelihood measure was not considered a primary measure of believability
because it does not focus on the alibi itself; rather, such a judgment could evoke a
broad range of considerations, such as the number of other suspects or the absence
of other evidence. Although not the primary measure, the likelihood-of-guilt judg-
ment was included to see if it followed the same pattern as the primary measure
(belief).
Participants also rated each alibi provider on 20 traits. Some of these traits we
considered to be relevant to the believability of an alibi provider (e.g., suspicious,
cunning, scheming, deceitful) and others we considered to be irrelevant to the believ-
ability of an alibi provider (e.g., shy, curious, funny, content). We used the resulting
correlations between traits and alibi believability to sort traits into the relevant and
irrelevant categories for purposes of analysis. Traits were rated on a 7-point Likert-
type scale (1 = does not describe this suspect at all, 7 = describes this suspect perfectly).
We expected the relevant traits to be affected by our manipulations of alibis and we
expected the irrelevant traits to not be affected by our manipulations.
After participants had read and rated each alibi individually, they were asked an
open-ended, exploratory question regarding what made them believe and disbelieve
each alibi. The open-ended question was not subjected to formal analyses, but was
included merely to assist the current authors’ development of future hypotheses
about how people evaluate alibis.

Procedure
The crime scenario, consisting of an initial crime report and a follow-up report,
was given to all participants upon arrival at the study. Participants were asked to
assume the role of a detective and evaluate the alibis of six suspects. Although par-
ticipants were told that they would evaluate six alibis, participants were given only
three alibis to evaluate; this was done to help prevent order effects. If participants
believed they were approaching the last of the possible suspects, they may have
injected some strategy (e.g., “I didn’t think it was any of the others, so it must be
this one”) into their decision, rather than evaluating the alibi itself. After partici-
pants finished evaluating their three alibis, they were fully debriefed, thanked, and
dismissed.
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

A Taxonomy of Alibis 167

Table 2. Mean Alibi Belief as a Function of Evidence

Physical evidence
Person evidence None Easy-to-fabricate Difficult-to-fabricate Overall mean
None 4.79 (1.88)a 6.44 (2.18)bd 6.97 (2.22)bd 6.07 (2.28)
Motivated
Familiar other 5.40 (2.20)ab 6.83 (2.28)cd 7.19 (2.21)cd 6.47 (2.35)
Nonmotivated
Familiar other 5.83 (2.29)ab 6.46 (2.39)bd 7.41 (2.23)cd 6.57 (2.38)
Nonmotivated
Stranger 6.63 (2.06)b 6.68 (2.43)bd 7.11 (2.66)bd 6.81 (2.39)
Overall mean 5.66 (2.20) 6.60 (2.31) 7.17 (2.33)
Note. Standard deviations are given in parentheses. Scale range is from 0 to 10. Higher numbers indicate
greater belief in the alibi. Means not sharing a common subscript within row or column differ at p < .05
(a Bonferroni correction was used to hold overall alpha at p < .05).

RESULTS

We expected that participants would differentiate among types of physical ev-

idence and types of person evidence, rating some alibis as more believable than
others in ways reflected in the taxonomy as shown in Fig. 1. We also hypothesized
that difficult-to-fabricate physical evidence would be powerful evidence and would
overwhelm all person evidence with which it was combined, resulting in no effect
across levels of person evidence for difficult-to-fabricate physical evidence.

The Belief Measure

Table 2 shows means and standard deviations for the belief measure according
to condition, and Fig. 2 shows the overall pattern of means. For ease of viewing, we

Fig. 2. Mean believability rating as a function of corroborating evidence. Bars represent one standard
error.
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

168 Olson and Wells

switched the order of the nonmotivated familiar other and nonmotivated stranger
conditions to reflect the order of the obtained means rather than the order in the
taxonomy. A general 4 × 3 mixed ANOVA revealed a significant interaction between
physical evidence and person evidence on how believable evaluators rated alibis,
F(6, 494) = 2.49, MSE = 4.53, p < .05, Cohen’s f = 0.17. There was also a significant
main effect for person evidence, F(3, 247) = 2.76, MSE = 6.21, p < .05, Cohen’s
f = 0.18, as well as a significant main effect for physical evidence, F(2, 494) = 32.68,
MSE = 4.53, p < .05, Cohen’s f = 0.36.
Planned comparisons were then performed to clarify the relationships among
person and physical evidence. The first three simple effects were tests of whether
person evidence affected believability within each level of physical evidence. Results
indicated that type of person evidence only affected believability when there was
no physical evidence. There was no significant simple effect of person evidence at
the level of easy-to-fabricate evidence F(3, 247) = .39, MSE = 5.38, ns, Cohen’s
f = 0.07, or at the level of difficult-to-fabricate physical evidence F(3, 247) = .40,
MSE = 5.45, ns, Cohen’s f = 0.07. There was, however, a significant simple effect for
person evidence at the level of no physical evidence F(3, 248) = 8.48, MSE = 4.46,
p < .05, Cohen’s f = 0.32. Six single-degree-of-freedom contrasts were conducted
to examine differences among the four levels of person evidence under conditions in
which there was no physical evidence. We used the Bonferroni correction to hold the
overall alpha level at p < .05 by dividing the alpha level by 6, thereby requiring that
any single contrast be significant at p < .008. Three of these contrasts were significant.
The nonmotivated stranger was considered more believable than no person evidence,
t(251) = 4.90, p < .008, d = 0.62, and was also considered more believable than the
motivated other, t(251) = 3.29, p < .008, d = 0.42. Also, the nonmotivated familiar
other was considered more believable than no person evidence t(251) = 2.77, p <
.008, d = 0.35.
The next four simple effects were tests of whether physical evidence affected
believability within each level of person evidence. Results indicated that physical
evidence affected believability at all levels of person evidence except the nonmoti-
vated stranger level. When the alibis included no person evidence, physical evidence
made a significant difference in the ratings of alibis, with no physical evidence the
weakest and difficult-to-fabricate evidence the strongest, F(2, 124) = 19.92, p < .05,
Cohen’s f = 0.57. Likewise, there was a simple effect of physical evidence at the level
of motivated familiar-other person evidence, F(2, 124) = 11.72, p < .05, Cohen’s
f = 0.44, as well as at the level of nonmotivated familiar-other person evidence,
F(2, 124) = 8.66, p < .05, Cohen’s f = 0.38. However, there was no significant sim-
ple effect for physical evidence when a stranger was a corroborator, F(2, 122) = 1.08,
ns, Cohen’s f = 0.13.
Three single-degree of-freedom contrasts were conducted to examine differ-
ences among the three levels of physical evidence in the absence of person evidence.
We used the Bonferroni correction to hold the overall alpha level at p < .05 by
dividing the alpha level by 3, thereby requiring that each single contrast be signif-
icant at p < .017. Two of these three contrasts were significant. Easily fabricated
physical evidence was rated significantly more believable than no physical evidence,
t(251) = 5.06, p < .017, d = 0.80. Likewise, difficult-to-fabricate physical evidence
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

A Taxonomy of Alibis 169

Table 3. Mean Likelihood of Guilt Rating as a Function of Evidence

Physical evidence
Person evidence None Easy-to-fabricate Difficult-to-fabricate Overall mean
None 5.41 (1.77)a 3.98 (2.37)b 3.35 (2.20)bd 4.25 (2.29)
Motivated
Familiar other 4.59 (2.04)ab 3.32 (2.35)bc 3.22 (1.98)cd 3.71 (2.21)
Nonmotivated
Familiar other 4.49 (2.19)ab 3.60 (2.41)bc 2.66 (2.10)cd 3.59 (2.35)
Nonmotivated
Stranger 3.98 (2.16)b 3.37 (2.23)b 2.81 (2.05)bd 3.38 (2.19)
Overall mean 4.62 (2.10) 3.56 (2.34) 3.01 (2.09)
Note. Standard deviations are given in parentheses. Scale range is from 0 to 10. Higher numbers indicate
greater likelihood that the alibi provider is the gunman. Means not sharing a common subscript within
row or column differ at p < .05 (a Bonferroni correction was used to hold overall alpha at p < .05).

was significantly more believable than no physical evidence t(251) = 5.60, p < .017,
d = 0.71.

The Likelihood Measure

Table 3 shows the means and standard deviations for the likelihood question ac-
cording to condition, and Fig. 3 shows the overall pattern of means for the likelihood
question according to condition. A general 4 × 3 mixed ANOVA revealed no signif-
icant interaction between physical evidence and person evidence on judgments of
likelihood that the alibi provider was the gunman, F(6, 486) = 0.98, MSE = 3.99, ns,
Cohen’s f = 0.11. However, there was a significant main effect for physical evidence
F(2, 486) = 40.84, MSE = 3.99, p < .05, Cohen’s f = 0.41, as well as a significant

Fig. 3. Mean likelihood rating as a function of corroborating evidence. Note. Bars represent one stan-
dard error.
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

170 Olson and Wells

main effect for person evidence F(3, 243) = 4.29, MSE = 5.97, p < .05, Cohen’s
f = 0.23.
Planned comparisons were then performed to clarify the relationships among
person and physical evidence. The first three simple effects were tests of whether per-
son evidence affected likelihood judgments within each level of physical evidence.
Results indicated that type of person evidence affected likelihood judgments only
when there was no physical evidence. There was no significant simple effect of per-
son evidence at the level of easy-to-fabricate physical evidence F(3, 246) = 1.05,
MSE = 5.48, ns, Cohen’s f = 0.12, or at the level of difficult-to-fabricate evidence
F(3, 247) = 1.55, MSE = 4.34, ns, Cohen’s f = 0.14. There was a significant sim-
ple effect for person evidence at the level of no physical evidence F(3, 246) = 5.21,
MSE = 4.18, p < .05, Cohen’s f = 0.25. Six single-degree-of-freedom contrasts were
conducted to examine differences among the four levels of person evidence under
conditions in which there was no physical evidence. We used the Bonferroni cor-
rection to hold the overall alpha level at p < .05 by dividing the alpha level by 6,
thereby requiring that any single contrast be significant at p < .008. One of these
contrasts was significant: Alibi providers with a nonmotivated stranger corroborator
were considered less likely to be the gunman than were alibi providers with no person
corroborator t(250) = 3.87, p < .008, d = 0.49.
The next four simple effects were tests of whether physical evidence affected
likelihood judgments within each level of person evidence. Results indicated that
type of physical evidence affected likelihood judgments at all levels of person evi-
dence. When the alibi provider offered no person evidence, physical evidence made
a significant difference in the judgments of likelihood that the alibi provider was
the gunman, with no physical evidence the most likely and difficult-to-fabricate evi-
dence the least likely, F(2, 122) = 17.19, p < .05, Cohen’s f = 0.53. Likewise, there
was a similar simple effect of physical evidence when a motivated familiar other was
a corroborator, F(2, 124) = 8.19, p < .05, Cohen’s f = 0.36, when a nonmotivated
familiar other was a corroborator, F(2, 120) = 12.02, p < .05, Cohen’s f = 0.45, and
when a stranger was a corroborator, F(2, 120) = 6.14, p < .05, Cohen’s f = 0.32.
Three single-degree of-freedom contrasts were conducted to examine differ-
ences among the three levels of physical evidence in the absence of person evidence.
We used the Bonferroni correction to hold the overall alpha level at p < .05 by di-
viding the alpha level by 3, thereby requiring that each single contrast be significant
at p < .017. Two of these three contrasts were significant. Alibi providers with easily
fabricated physical evidence were rated significantly less likely to be the gunman
than those with no physical evidence, t(251) = 4.65, p < .017, d = 0.59. Likewise, al-
ibi providers with difficult-to-fabricate physical evidence were rated significantly less
likely to be the gunman than those with no physical evidence t(251) = 5.52, p < .017,
d = 0.70.

The Trait Measures

All trait ratings were scored so that higher scores bore a negative connotation.
Correlations between traits and alibi believability were then calculated to determine
which traits were relevant and which were irrelevant. The traits honest, deceitful,
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

A Taxonomy of Alibis 171

Table 4. Mean Relevant Traits Rating as a Function of Evidence

Physical evidence
Person evidence None Easy-to-fabricate Difficult-to-fabricate Overall mean
None 4.15 (0.71)a 3.68 (0.90)ab 3.58 (0.90)bd 3.80 (1.11)
Motivated
Familiar other 3.90 (0.82)ab 3.37 (0.95)bc 3.30 (0.90)cd 3.52 (1.10)
Nonmotivated
Familiar other 3.89 (0.74)ab 3.61 (0.90)bc 3.41 (0.74)cd 3.64 (1.10)
Nonmotivated
Stranger 3.60 (0.75)b 3.47 (0.79)b 3.40 (0.86)bd 3.49 (1.10)
Overall mean 3.89 (0.78) 3.53 (0.89) 3.42 (0.85)
Note. Standard deviations are given in parentheses. Scale range is from 0 to 10. Higher numbers indicate
greater ratings of the alibi provider as dishonest, suspicious, deceitful, untrustworthy, etc. Means not
sharing a common subscript within row or column differ at p < .05 (a Bonferroni correction was used to
hold overall alpha at p < .05).

suspicious, conniving, open, cunning, sincere, scheming, trustworthy, calculating, loyal,

and intelligent were significantly correlated with the belief measure (.13 < r > .35),
and the traits content, shrewd, friendly, caring, ambitious, curious, funny, shy were not
significantly correlated with the belief measure (.01 < r > .09). Thus, the former 12
traits were considered relevant and the latter eight traits were considered irrelevant
for purposes of the main analyses.5 Coefficient alpha for the 12 relevant traits was
.79 and for the eight irrelevant traits was .43.

Relevant Traits
The relevant trait ratings were averaged into a single score for each participant.
Table 4 shows means and standard deviations for the overall relevant traits rating
according to condition. A general 4 × 3 mixed ANOVA revealed no significant inter-
action between physical evidence and person evidence on the relevant trait ratings of
alibi providers F(6, 494) = 1.07, MSE = 0.67, ns, Cohen’s f = 0.12. However, there
was a significant main effect for physical evidence, F(2, 494) = 23.17, MSE = 0.67,
p < .05, Cohen’s f = 0.31, as well as a significant main effect for person evidence,
F(3, 247) = 2.93, MSE = 1.08, p < .05, Cohen’s f = 0.19.

Irrelevant Traits
The irrelevant trait ratings were averaged into a single score for each partici-
pant. Table 5 shows means and standard deviations for the overall irrelevant traits
rating according to condition. A general 4 × 3 mixed ANOVA revealed no signif-
icant interaction between physical and person evidence F(6, 492) = 1.40, MSE =
0.21, ns, Cohen’s f = 0.13. Likewise, there was no main effect for person evidence,
F(3, 246) = 2.29, MSE = 0.78, ns, Cohen’s f = 0.17. Interestingly, there was a sig-
nificant main effect for physical evidence F(2, 492) = 21.55, MSE = 0.21, p < .05,

5 To avoid overanalysis of the trait data, we did not perform tests of the simple effects for the relevant and
irrelevant trait measures.
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

172 Olson and Wells

Table 5. Mean Irrelevant Traits Rating as a Function of Evidence

Physical evidence
Person evidence None Easy-to-fabricate Difficult-to-fabricate Overall mean
None 4.39 (0.53)a 4.21 (0.57)a 4.24 (0.61)a 4.28 (0.96)
Motivated
Familiar other 4.29 (0.62)a 3.94 (0.60)a 4.00 (0.65)a 4.08 (0.96)
Nonmotivated
Familiar other 4.26 (0.62)a 4.23 (0.64)a 4.15 (0.54)a 4.21 (0.96)
Nonmotivated
Stranger 4.29 (0.60)a 4.10 (0.58)a 4.13 (0.59)a 4.18 (0.96)
Overall mean 4.31 (0.60) 4.12 (0.60) 4.13 (0.60)
Note. Standard deviations are given in parentheses. Scale range is from 0 to 10. Higher numbers indicate
greater ratings of the alibi provider as shy, uncaring, unfriendly, unfunny, etc. Means not sharing a common
subscript within row or column differ at p < .05 (a Bonferroni correction was used to hold overall alpha
at p < .05).

Cohen’s f = 0.30. Thus, participants rated those alibi providers with stronger phys-
ical evidence as funnier, more caring, friendlier, etc.

DISCUSSION

Our proposed taxonomy of alibi believability received some support from par-
ticipants’ alibi evaluations. Main effects for person evidence and for physical evi-
dence indicate that participants in the role of detectives tended to make distinctions
among these alibis along the lines we had expected. In addition, the presence of
an interaction followed the general idea that difficult-to-fabricate physical evidence
overwhelms person evidence. Nevertheless, there were some aspects to the results
that were not expected. Our interaction prediction was that the difficult-to-fabricate
physical evidence would trump the person evidence, but we had not expected that
even the easy-to-fabricate physical evidence was sufficient to render the person ev-
idence irrelevant. Some caution is required in generalizing this conclusion because
the impact of a manipulated variable is sensitive to the particular ways in which it
is operationalized in a given study. Nevertheless, we were surprised that even easy-
to-fabricate physical evidence (e.g., a cash receipt) wiped out the effects of having a
person to corroborate the alibi even when that person had no apparent motivation
to lie for the suspect.
Another mild surprise was that participants did not find the alibi corroborators to
be more credible when they were nonmotivated familiar others than when they were
nonmotivated strangers. In both cases there was no motive for the alibi corroborator
to lie, but the stranger should be more likely than the familiar other to be honestly
mistaken. If anything, alibi corroboration from the nonmotivated stranger was seen
as more credible than corroboration from the nonmotivated familiar other (although
we remind the reader that this difference was not statistically significant). Recall that
our manipulation check participants perceived the stranger to be more likely than
the nonmotivated familiar other to have been honestly mistaken. However, the ma-
nipulation check participants were explicitly asked to consider the chances of an
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

A Taxonomy of Alibis 173

honest mistake by the corroborator and this might have made that dimension salient
for the manipulation check participants in a way that was not salient for the partic-
ipants who were asked to evaluate the believability of the alibi. This suggests to us
that our results might have shown the expected differences between the nonmoti-
vated familiar-other corroborator and the stranger corroborator if this had been a
trial setting in which a prosecutor cross-examined the corroborator along lines sug-
gesting that the stranger corroborator could have been mistaken. For this reason,
we advise maintaining the taxonomy distinction between stranger corroborators and
nonmotivated familiar-other corroborators.
As expected, the pattern of responses for the likelihood of guilt measure fol-
lowed the (inverse) pattern of the believability measure. Although the interaction
pattern between physical and person evidence was similar for the likelihood measure
and the believability measure, the interaction did not reach statistical significance for
the likelihood of guilt measure. We believe that this is because likelihood of guilt mea-
sures are not particularly sensitive measures of alibi believability. There are several
reasons why we believe that believability of the alibi rather than likelihood of guilt is
a more sensitive measure for studying alibis. First, questions about guilt are likely to
lead at least some participants to invoke a variety of other considerations such as the
presumption of innocence, the totality of evidence, the presentation of evidence by
both sides, and the need for affirmative evidence of guilt. These other considerations
that people use to make judgments of guilt can introduce noise into a paradigm that is
specifically attempting to focus on the believability of the alibi itself. In this sense, guilt
measures are, at best, indirect measures of alibi believability. Second, the absence of
proof to support one’s alibi might undermine the believability of the alibi, but it is
not actually evidence of guilt. Being alone at home at the time that an offense was
committed, but having no way to prove it, should not count as affirmative evidence
of guilt. Third, even if an alibi is proven wrong, it does not automatically mean the
suspect is guilty. For instance, false alibis can be given to cover up social indiscretions
(e.g., an affair). Our general point is that we were not surprised that the likelihood
of guilt measure was not as sensitive as the believability measure in response to our
manipulations. Our taxonomy is a taxonomy of the believability of alibis, not a tax-
onomy of the perceived likelihood of guilt of those using these alibis. Although the
believability of an alibi and the likelihood of guilt are related, guilt judgments invoke
a much broader range of considerations and a broader range of evidence than just the
alibi.
We found an interesting pattern of responses for the personality trait ratings of
the alibi provider. Although we did not have a strong prediction for the trait ratings,
we assumed that some traits, such as deceitful and conniving, would be related to
alibi believability. We expected that other traits, however, such as caring and funny,
should not be affected by alibi proof. A correlation analysis indicated that the 12
relevant traits were each significantly correlated with alibi believability, whereas none
of the eight irrelevant traits were significantly correlated with alibi believability. The
results showed that both relevant and irrelevant traits were affected by the taxonomy
manipulations. The effects on irrelevant traits were not as strong as the effects on
relevant traits, but we were surprised to find irrelevant traits to be affected at all by
alibi proof. For instance, why should a person with a weak alibi be rated as less funny
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

174 Olson and Wells

than would a person with a strong alibi? In hindsight, this seems not so surprising.
All traits can be sorted along a good–bad dimension, even those traits that seem
not related to alibis. A suspect whose alibi is evaluated negatively is perceived as
someone who is likely to have committed the crime in question and is, therefore,
likely to suffer from a number of negative inferences about his or her character.
Serendipitously, perhaps we have discovered an implicit, or at least indirect, method
for assessing alibi believability. As a proxy for directly asking people to evaluate the
believability of an alibi, they could be asked to make trait inferences about the alibi
provider.
We find it curious that the condition representing the strongest alibi, which
included not only a neutral person corroborating the alibi but also a dated, timed
security video capturing a clear image of the suspect, nevertheless yielded a mean be-
lievability score of only 7.4 (out of a possible 10.0). What kind of proof would it take
to get a believability score of an 8, 9, or 10? In an earlier draft of this paper, we sug-
gested that this speaks to the incredible skepticism with which people seem to greet
alibis. We suggested that the mere labeling of a statement as being an alibi evokes a
sense of disbelief and challenges people to create imaginative scenarios worthy of a
great crime novel as to how the person could nevertheless have committed the crime.
Reviewers of the earlier version of this paper objected to our interpretation, noting
that there are many reasons why participants might have avoided the high end of
the scale. We agree with the reviewers’ cautions, and so we offer the alibi-skepticism
idea as only a hypothesis for future research. We think it is quite plausible that the
term alibi is an extremely loaded term. Consider someone responding to the question
“Where were you last night?” versus responding to the question “What is your alibi
for where you were last night?” Asking for an alibi is more than just being inquisi-
tive; it is a form of accusation regarding a bad act for which there seems reason to
suspect the person being asked. In effect, giving an alibi is a form of denial of crimi-
nality and research on innuendo shows that denials of criminality can themselves be
incriminating. For example, research participants who read the headline “Andrew
Winters Not Connected to Bank Embezzlement” produced assessments of Andrew
Winters that were as negative as those coming from the headline “Andrew Winters
Connected to Bank Embezzlement” (Wegner, Wenzlaff, Kerker, & Beattie, 1981).
Research on anchoring and adjustment processes in human judgment illustrate the
power of “starting points” (Koehler, 1991). If alibi evaluators start with an assump-
tion that the alibi is likely to be false (an anchor), then adjustments toward believing
the alibi based on proffered proof are likely to be insufficient.
Our primary motivation for developing this taxonomy was to help launch a sys-
tematic literature on the psychology of alibis. We think that this taxonomy represents
a reasonable starting point for thinking about alibis. In developing this taxonomy, we
tried to maintain a degree of simplicity that would permit easy comprehension while
serving a heuristic role for generating new hypotheses. We invite others to refine this
taxonomy or to propose alternative taxonomies, but we think any useful taxonomy
will need to be based on the underlying concept of proof to support the alibi rather
than on the surface attributes of the alibi. This is not to say that the surface attributes
of the alibi story cannot affect alibi evaluations at all. Consider two men who claim to
P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

A Taxonomy of Alibis 175

have been alone at home at the time of the crime but one says he was smoking crack
at the time and the other says he was reading his Bible. Clearly, inferences about
the character of the two men based on the surface attributes of their alibi stories are
likely to lead to different evaluations. Still, we remind readers that these differences
are more at the level of differences in alibi providers, not differences in the alibis
themselves.
It would be misleading to suggest that our alibi taxonomy, or any taxonomy,
can capture all the dynamics that determine the believability of an alibi. Clearly,
the taxonomy assumes that other variables are held constant. These other variables
include such things as background characteristics of the alibi provider (e.g., prior
record, no prior record) and behavioral cues of the person giving the alibi story
(e.g., nervousness, pauses). We also think that alibi evaluators might be especially
suspicious of someone who is able to readily provide an alibi for a distant past event
when there seems to have been no reason for an innocent person to be able to recall
what she or he was doing that day. And, what if the person initially errs in recall (“I
was playing golf with friends that afternoon”) only to discover later that she recalled
the wrong day and has to change the alibi story? These dynamic features are not
captured in the taxonomy and we do not think that the taxonomy itself can be easily
modified to capture them. Nevertheless, we think that the taxonomy represents a
useful structure from which to study these other variables. For example, we suspect
that these other variables (e.g., prior record, nervousness, change of alibi story) will
affect alibi credibility primarily when the level of proof is weak (upper left cells
of the taxonomy) and will matter less when levels of proof are strong (lower right
cells of the taxonomy). Hence, the taxonomy should be useful for researchers even
when their hypotheses concern alibi variables that are not part of the taxonomy
itself.
A big door that we hope we have opened on the study of alibis concerns the
ability of innocent people to generate and prove alibis. Clearly, some innocent people
simply will not be able to reconstruct where they were at a particular time in the past.
Others might mistakenly recall where they were and then later have to change their
alibi story, a behavior that is likely to be met with considerable incredulity. But, even
when innocent people can accurately recall where they were, how often could they
provide proof at one of the higher levels of our taxonomy? How often would people
be stuck in the upper-left cell of the taxonomy (i.e., no person or physical evidence to
support the alibi)? We suspect that physical evidence is rather rare and that person
evidence, when it exists, will usually come from a motivated familiar other (friend
or relative), which is the second-lowest level of believability in the taxonomy. At the
same time, we suspect that people overestimate their ability to prove their alibi. For
example, they might think that their girlfriend or cousin will somehow be a powerful
form of corroboration or that their corroborator will always recall the time frame
the same way that they do. In theory, it seems that the alibi defense ought to work
with some regularity for innocent people; in practice we worry that it will rarely work
because of the difficulty of proving the alibi and the incredible skepticism that seems
to accompany alibi evaluations. We challenge psychological researchers to take up
some of these important questions.
 P1: IZO
Law and Human Behavior [lahu] pp1162-lahu-484199 March 10, 2004 13:47 Style file version June 4th, 2002

176 Olson and Wells

ACKNOWLEDGMENT

This research is based on the first author’s Master’s thesis.

REFERENCES

American Psychological Association. (1992). Ethical principles of psychologists and code of conduct.
American Psychologist, 47, 1597–1611.
Culhane, S. E., & Hosch, H. M. (in press). An alibi witness’s influence on jurors’ decision making. Journal
of Applied Social Psychology.
Epstein, D. M. (1964). Advance notice of alibi. The Journal of Criminal Law, Criminology, and Police
Science, 55, 29–38.
Friedman, J. P. (1998). Alibi instructions and due process of law. Western New England Law Review, 20,
343–386.
Gilbert, D. T., & Malone, D. S. (1995). The correspondence bias. Psychological Bulletin, 117, 21–38.
Gooderson, R. N. (1977). Alibi. London, UK: Heinemann Educational.
Koehler, D. J. (1991). Explanation, imagination, and confidence in judgment. Psychological Bulletin, 110,
499–519.
Leippe, M. R. (1985). The influence of eyewitness nonidentifications on mock jurors’ judgments of a court
case. Journal of Applied Social Psychology, 15, 656–672.
Lindsay, R. C. L., Lim, R., Marando, L., & Cully, D. (1986). Mock-juror evaluations of eyewitness testi-
mony: A test of metamemory hypotheses. Journal of Applied Social Psychology, 15, 447–459.
Nolan, J. R. (1990). Black’s law dictionary (6th ed.). St. Paul, MN: West Publishing.
McAllister, H. A., & Bregman, N. J. (1989). Juror underutilization of eyewitness nonidentifications: A test
of the disconfirmed expectancy explanation. Journal of Applied Social Psychology, 19, 20–29.
Sullivan, T. P. (1971). Presentation of the defense. The Journal of Criminal Law, Criminology, and Police
Science, 62, 153–172.
Wegner, D. M., Wenzlaff, R., Kerker, R. M., & Beattie, A. E. (1981). Incrimination through innuendo: Can
media questions become public answers? Journal of Personality and Social Psychology, 40, 822–832.
Wells, G. L., Small, M., Penrod, S., Malpass, R. S., Fulero, S. M., & Brimacombe, C. A. E. (1998). Eyewit-
ness identification procedures: Recommendations for lineups and photospreads. Law and Human
Behavior, 22, 603–647.
Wells, G. L., & Windschitl, P. D. (1999). Stimulus sampling in social psychological experimentation. Per-
sonality and Social Psychology Bulletin, 25, 1115–1125.

View publication stats