The Journal of Credibility Assessment and Witness Psychology
1997, Vol. 1, No. 1, 33-43
Published by the Department of Psychology of Boise State University
Lexical Decision and the Detection of Concealed Information
Lawrence Locker, Jr. and Marc
E. Pratarelli, Oklahoma State University
Copyright 1997 by the Department of Psychology of Boise State University and the Author. Permission for non-profit electronic dissemination of this article is granted. Reproduction in hardcopy/print format for educational purposes or by non-profit organizations such as libraries and schools is permitted. Any modification of this document is expressly forbidden. For all other uses of the this article, prior advance written notice is required. Send inquiries by hardcopy to: Charles R. Honts, Ph. D., Editor, The Journal of Credibility Assessment and Witness Psychology, Department of Psychology, Boise State University, 1910 University Drive, Boise, Idaho 83725, USA.
Abstract The present study utilized the lexical
decision task (LDT) to assess whether participants with concealed
information regarding a subset of the word stimuli performed differently
than those without. In Experiment 1, Experimental test
subjects believed they were concealing special knowledge of certain
word items from the Experimenter. In Experiment 2, Informed
test subjects did not attempt to conceal their knowledge of
the word items because they had received instructions from the
Experimenter to study the items overnight. The Informed group
responded more like Controls who possessed no special knowledge
of the stimuli in the LDT. However, the Experimentals, having
to conceal their knowledge of certain word stimuli, responded
slower than either of the other two groups on the test items as
well as on neutral word items and nonwords. The reaction time
effects for different classes of words and the nonwords were constant
across all three groups, an indication that prelexical processes
are not affected. However, deceit appears to negatively impact
word recognition speed by slowing all responses during postlexical
This study examines the efficacy of using a well known speeded reaction time paradigm, the lexical decision task (LDT), in the detection of deceit of individuals who possess concealed information. For this study, concealed information refers to privileged information that one group possesses and another group does not. The study represents a departure from previous research in that it does not make use of the traditional methods and paradigms of lie detection, such as the Guilty Knowledge Test or the Control Question Test coupled with the polygraph or Interrogative Polygraphy (Barland & Raskin, 1975; Farwell & Donchin, 1991; Lykken, 1959; Reid & Inbau, 1977). Instead, the present study involves the detection of a special set of acquired knowledge consisting of certain words that were afforded special status before the test in certain subjects. Thus, the study attempted to ascertain whether subjects who had acquired special knowledge would manifest a bias in their patterns of responding in the LDT. In addition, because emerging technologies like Interactive Polygraphy have methodological constraints which require timelocking to the onset of a stimulus, another purpose of the study was to explore new behavioral approaches needed to accommodate these technological concerns. Hence, the present report is the result of a pilot series aimed at developing task demands that would reveal deception and conform to the methodological constraints of collecting event-related brain potentials (ERP).
The conventional approach to lie detection utilizes the polygraph in conjunction with behavioral paradigms like the Guilty Knowledge Test (GKT) or the Control Question Test (CQT). The polygraph measures autonomic responses to questions involving critical events, e.g., of a crime, and other relevant and irrelevant questions in the hopes that the guilty knowledge items will elicit a stronger response from the guilty individual. Moreover, the physiological responses should also reject innocent individuals because they would not be physiologically aroused to the critical questions, by virtue of their not having specific knowledge of the event. The GKT involves a number of key items about the crime and a number of irrelevant items. A person with the knowledge of the key items and little response to the irrelevant items should have different physiological reactivity than a person with no knowledge of the key terms. Alternatively, the CQT involves questions related to the crime, neutral questions, and control questions that concern basic honesty, e.g., "Did you ever use a fake ID?" It is assumed that a guilty person would be more reactive to the relevant crime questions than an innocent person; the latter would be more concerned with the control questions.
In a newer and somewhat different approach, interrogative polygraphy uses ERPs as a lie detection tool (Farwell & Donchin, 1991; Rosenfeld, Angell, Johnson & Qian, 1991), again coupled with a behavioral paradigm that usually contrasts relevant and irrelevant information. This procedure calls for the analysis of highly visible brain waves elicited to isolated timelocked presentations of key words or pictures related to the crime. Stimulus items that subjects recognize as relevant should cause their brain to react with more activity than irrelevant items, even though a verbal or manual response may not be required. While the interactive polygraphy methods or the conventional polygraph approaches have met with varying degrees of success and acceptance, all of them continue to struggle with the difficult task of measuring the illusive psychological constructs of deceit and guilt. Moreover, none can claim to eliminate the number of false positives that continue to pose a problem in the practical application of these technologies. Presently, the aim of the study reported here is to examine the efficacy of another approach to measuring performance differences partially or wholly attributable to deceit or guilt.
The logic underlying the two experiments in the present study is based upon the distinction between automatic and controlled information processing. Automatic processes are not consciously controlled, i.e., they are normally outside a subject's awareness, and they are triggered very quickly at the mere presentation of the physical characteristics of the stimulus (Kihlstrom, 1987). In contrast, controlled processes are normally accompanied by awareness or access to consciousness. Controlled processes also require more processing time than automatic processes. The important theoretical issue for the present study is that the more cognitive demands/processes recruited during the task in a given individual, the more processing time they will need.
Concealing guilt or deceiving another individual involves a conscious
and controlled effort to avoid acknowledging that one possesses
a critical and relevant class of information. The present study
attempts to circumvent the controlled cognitive processes and
examines whether the automatic processes beyond a person's control
are differentially affected during the processing of concealed
information. In the LDT, the individual sees randomly presented
words or nonwords (e.g., GLUG), and responds as quickly as possible
by indicating whether the target item was one or the other. Because
a conscious effort is focused on the word-nonword decision required
by the task demands of the LDT, participants may not be able to
dedicate enough attentional resources to their reactions to different
classes of words nested within the word condition. Therefore,
an individual hoping to use their concealed information to their
advantage, may incidentally, generate side effects quite beyond
their control. In principle, this should produce a behavioral
marker of the possession of concealed relevant information.
The focus of Experiment 1 was to assess the performance differences
between individuals who have concealed information regarding certain
items that will appear in their tests, and a set of control subjects
who were not offered information. Therefore, the contrast was
between those who concealed their knowledge, and attempted to
use it to their advantage, and those who concealed nothing. The
experiment was not only aimed at detecting which group had the
relevant information, but also to evaluate which behavioral effect
was most sensitive for differentiating those who concealed their
knowledge of that information.
Forty-two participants (ages 18-24 years) recruited from introductory psychology classes were offered generous amounts of extra course credit for participation in the study. All participants gave their signed consent to participate on two consecutive days; they were screened for neurological history, normal or corrected to normal vision, reading disabilities, handedness, years of education, and whether English was their first and predominant language. Participants were assigned to one of two groups, referred to as Controls and Experimentals, in a quasi-random fashion. Each was told that their name appeared only on the consent form which would be given to their Instructor in order to record their extra credit. Afterward, the forms were discarded. Critical to the integrity of the study were instructions which stated that all information related to their performance was completely confidential, and could never be traced back to their name. Confidentiality was achieved by coding each participant's data with a randomly generated alpha-numeric code.
Four paragraphs were chosen from randomly selected topics. These included a paragraph about the planet Mars, one about chemistry, another dealing with fish, and another about the boiling point of liquids. For the stimuli in the LDT, 150 items/trials composed of 75 words and 75 pronounceable nonwords were generated in the following manner. Twenty-five words which appeared two or more times in the paragraphs, were selected as one category of word items. These words were designated the Relevant items. These words were matched to 50 additional words in the stimulus set on the basis or word length and word frequency of occurrence using the norms of Francis and Kucera (1982). Twenty-five of these latter 50 word items were designated Irrelevant items, i.e., they did not appear in the paragraphs and were not distributed in study lists. The last 25 word items were designated DOD-words because they comprised the concealed information. The DOD words also did not appear in the paragraphs, but certain participants believed these words would appear on the recognition/recall task given on the second day. The balance of the stimulus set consisted of 75 nonwords matched with the words for length and bigram frequency. All four stimulus categories were randomly ordered in the stimulus list. The list was presented in blocks of 50 trials.
Each participant was informed that the experiment involved listening
to one experimenter reading the four paragraphs to them on the
first day, and that a recognition and recall task on a personal
computer would be conducted on the subsequent day. They were also
instructed that they would be working with two different experimenters
on each of the two days. On the first day, four paragraphs would
be read to them, and their instructions were to listen attentively
because certain items from the paragraphs would appear on the
task (LDT) the following day. The more items they recognized and
recalled on the second day, the better they performed, and the
more extra credit bonus points they would earn. The participants
were informed that they would receive two extra credit points
just for showing up on both days. However, they could also receive
up to three additional points, depending on their performance
on the second day. All participants were told during their debriefing
that they actually received all five points of extra credit regardless
of the quality of their performance, along with instructions not
to discuss the experiment. The Controls received no further instructions
except to return the next day at the appointment time. Those in
the Experimental group were offered a study list by a confederate
posing as another participant who had finished his second day
of the experiment in the adjoining room. The first experimenter
made an excuse to leave the lab for about five minutes while this
occurred. The confederate informed each participant in the Experimental
group that he had also been given the list the day before and
that it had been helpful in obtaining the maximum amount of extra
credit. He also asked them not to reveal to the experimenter that
they had been given the list. The confederate laid the list on
the desk and left the room. If the participant accepted and concealed
the study list, that individual was included in the Experimental
group. The list was offered to each Experimental participant in
order to create a situation in which they believed they were genuinely
concealing something from the second experimenter. In order to
create a valid blind condition and to protect the anonymity of
the participant, the confederate did not reveal to either of the
experimenters which of them were in the Control group or the Experimental
group. Furthermore, to maintain and ensure privacy and confidentiality,
the randomly generated participant codes had either an E or a
C added by the confederate to distinguish their data during analysis.
A two (Group) by four (Word Condition) analysis of variance was used to analyze group and trial-type differences in reaction time in the LDT. Table 1 lists the means and standard errors for both groups in each condition.
|Legend: REL-W = Related Words; DOD-W = Detection of Deception Words; IREL-W = Irrelevant Words; and NON-W = Nonwords. All means and standard errors are in milliseconds.|
The trial type differences are also illustrated in Figure 1(red/diamonds and blue/squares). The analysis of variance produced a significant main effect of Group (F[1,40] = 2.88, p < .045), and a main effect of Word Condition (F[3,120] = 32.52, p < .0005). The analysis did not show a significant interaction effect between Group and Condition.
A pooled average for all three Word Conditions versus the single
Nonword Condition revealed that both groups responded faster to
word targets than to nonword targets. Table 1 also lists the respective
means and standard errors. These results are consistent with scores
of previous studies using the LDT (cf., Forster & Taft, 1994;
Pratarelli & McIntyre, 1994). No analysis was performed on
this pooled data because the nonwords were nonconsequential to
the focus of the study. However, because the respective means
and standard errors do not overlap provides us with a reasonable
assurance that both groups were performing and attending to the
principal task of differentiating words and nonwords.
Figure 1 reveals two important features of the data. First, Controls and Experimentals reacted comparably to word versus nonword targets. The word-nonword effect is important because it verifies that subjects were not directing their attention to the different categories nested in the word trials. During debriefing, none gave any indication that they were aware of different word categories when they were performing the LDT. The failure to recognize different word conditions provides a second assurance that both groups focused on the task demands of differentiating words from nonwords. Therefore, because both groups produced comparable word-nonword effects, we can reasonably ascertain that both were focused in similar ways, i.e., on making a rapid word-nonword decision.
The second and more important feature of the data, illustrated in Figure 1, was a standard delay across all four conditions for the Experimental group as compared to the Control group. That is, the Experimental participants reacted slower overall across all four trial types. Because the delay occurred in all trial-type conditions, it can be attributed to an additive postlexical effect rather that one which differentially affected some conditions and not others. Clearly, participants in the Experimental group responded differently to all conditions, not just to the DOD-words which they presumed they were studying without the knowledge of the experimenters. Moreover, because there were no differential effects between groups in word-nonword processing, the results suggest that the prelexical processes were not affected by the treatment variable, i.e., that the Experimental group had concealed information about the DOD items. Therefore, previous exposure to a study list, however covert, did not alter the early stages of word recognition. Such effects, if they had occurred, would have to be considered prelexical; that is, they took place prior to recognizing the item. Effects which alter processing by delaying all responses (words and nonwords) must occur postlexically. These findings can be reconciled using Posner and Snyder's (1975) two-process model in which the first stage is characterized by automatic stimulus processing, while the second stage is affected by the later conscious controlled processes. Presently, because no effects were observed which could be attributed to the automatic (prelexical) stage, the standard delay for the Experimental group must be attributed to a late conscious controlled factor (postlexical). Such a factor would require some attentional resources.
The primary difference between the two groups was in terms of the awareness each had about the possibility that certain words would appear during the LDT. The Experimental group in particular believed that because they had concealed information, they would benefit from having studied the list acquired from the confederate. The Controls had no such belief because there was no concealed information. However, in order to benefit from the concealed information, the Experimental group had to focus more attention and effort on their accuracy at the expense of speed. The conventional framework for the speed-accuracy tradeoff is that depending on specific task instructions, task demands, or the influence of particular treatment variables, participants may sacrifice one at the expense of the other (Pachella, 1974). An accuracy analysis was not possible because both groups performed at ceiling levels. However, because the stimulus list was randomized, participants could not predict when a DOD word would appear. Therefore, participants in the Experimental group would have to commit more attentional resources and effort on their accuracy at the expense of reaction time. More attention would have to be diverted from the word-nonword discrimination task demands, and focused instead on the mere possibility that a word from their pool of concealed information would appear. The delay in mean reaction times for all four trial-types for the Experimental participants reflected their attention to accuracy, which in turn was the result of their attempts to utilize the concealed information from the acquired list.
Similar postlexical effects have been observed with social loafing (Pratarelli & McIntyre, 1994), heightened vigilance (Harkins & Petty, 1982), and when having to solve mazes (Jackson & Williams, 1985). Generally, altering any of the task demands which must compete for the limited attentional resources will impact reaction time postlexically. Another issue that should be considered is the self-efficacy of the Experimental participants (Bandura, 1988). It is reasonable to suppose that the Experimental participants assumed they would perform better due to their concealed knowledge. They focused more on the use of this knowledge instead of entirely devoting their attention to the task. In contrast, the Control participants assumed that their success was entirely dependent on the principal task demands, and their memory of the paragraphs heard the previous day. Therefore, Controls only took into account the instructions of the experimenters. Using this logic, Experimental participants, relying more on their belief that the concealed information would improve their performance, made inaccurate assumptions about the task. The effect of such cognitions was to draw on the limited attentional resources which, as seen in the better performance of the Controls, ought to be focused on the speeded word-nonword decision. The result is that Experimental participants were faced with having to implicitly divide their attention, while the Controls focused on a singular task.
In terms of alternate explanations for these results, it is generally accepted that concealing special knowledge creates a state of arousal in many individuals; this continues to be the underlying premise for the use of the polygraph. It is possible that a higher state of arousal could impact processing time and account for the differences observed between the two groups. However, the weakness and questioned reliability of the polygraph method, which hinges on higher arousal levels to detect deception, suggests arousal might have only minimal impact.
Another possibility is that participants could have been trying
to overanalyze the task on the second day. On the first day, those
who left with the list and a brief description of the next day's
task, left with the crude understanding that somehow the list
would improve their ability to recall those items. Unbeknownst
to them, however, the LDT is not a recall task. Therefore, participants
were faced with an unfamiliar task which was not consistent with
their expectations of how the information would be useful to them.
Yet another consideration was that the Experimental participants
may have processed the words on at least two levels as compared
to the Controls. These levels involve (1) the obligatory recognition
of the stimulus, and (2) the later processing component involving
the realization that it was or was not a word from their list.
However, if this had occurred, the DOD items would have been processed
differently than the other word conditions, but they were not.
Indeed, the most reasonable explanation is that the metaknowledge
of having concealed information slowed all their responses, because
it could never be predicted that a known (DOD) word would appear
on any given trial. However, a question remains whether simply
having the privileged access to information, rather than the guilt
related to that knowledge, is the driving force behind the reaction
time delay. Therefore, we determined that further study was needed
to focus solely on the effects of privileged knowledge without
any guilt, concealment or deception.
The focus of the second experiment was to assess differences in
a new group of participants given the privileged knowledge without
any attempt to hide that knowledge. The purpose of this experiment,
therefore, was to assess whether poorer performance in the Experimental
group earlier was actually due to mere access to privileged information,
or to the fact that they believed they had concealed information.
If indeed the slower response times for the Experimental group
were due to their increased attention on the material they had
studied, then anyone having similar opportunities to study in
advance ought to manifest comparable delays.
Twenty-one additional students (ages 18-24 years) were recruited for the new treatment condition. These were referred to as Informed participants. Because we decided to compare them to the existing Control group, participants could not be randomly assigned. The new participants received the same instructions, screening, coding, and extra credit as the participants in Experiment 1.
Stimuli and Procedures
The same stimuli in the same randomized order were used in Experiment
2. The procedures differed in that the Experimenter gave all the
participants the study list after reading the four paragraphs
to them. The Experimenter added that the items on the study list
would appear in the task the next day, and they would have a better
chance of receiving all the extra credit points if they studied
the list items. No further instructions were given except to return
the next day at their appointment time to complete the computer
A two (Group) by four (Word Condition) analysis of variance was
utilized to analyze group and trial-type differences in reaction
time in the LDT. Table 2 lists the means and standard errors for
the Informed group along with the means and standard errors from
the Control group in Experiment 1.
|Legend: REL-W = Related Words; DOD-W = Detection of Deception Words; IREL-W = Irrelevant Words; and NON-W = Nonwords. All means and standard errors are in milliseconds.|
An analysis of variance did not reveal a significant main effect
of Group, but only a main effect of Word Condition (F[3,120]
= 34.11, p < .0005). The analysis also showed no significant
effect in the Group by Word Condition interaction. A pooled average
for all three word conditions/trials versus the nonword trials
again revealed that both groups responded faster to word targets
than to nonword targets. Table 2 lists the respective means and
standard errors compared to the Controls. As in Experiment 1,
these results provide a reasonable assurance that the focus of
the participants' attention was on differentiating words from
nonwords. The data for the Informed group were subsequently plotted
against those from Experiment 1 (Figure 1, green/stars).
There are two notable features in the present experiment. First, both Control and Informed participants responded comparably to word and nonword items. This effect is important in that it directs the participants' attention away from the word categories. The second notable feature of the new data is that the Informed participants did not respond significantly different than the Control group in Experiment 1. However, the data from Table 2, and as illustrated in Figure 1, reveal that there appears to be a small standard delay across all four Word Conditions. Although not statistically significant, this effect bears some resemblance to the pattern produced by the Experimental group earlier. The delay for the Informed group is different in that it more closely approximates the performance of the Controls than the Experimentals. A similar analysis of variance comparing the Informed group with the Experimental group was also not significant. Despite the small similarity, the difference in the standard delay between Informed and Experimental groups may be attributed to the manner in which their special knowledge was framed, i.e., nonconcealment versus concealed.
Both Experimental and Informed participants received the special knowledge for some of the stimulus items with the instructions that it would "make it easier to get" extra credit points. Because the Informed participants did not have to focus any effort on concealing their special knowledge from the Experimenter, one possible explanation for their slightly faster performance, when compared to Experimentals, is that the guilt and deceit related to the knowledge somehow usurped some of the attentional resources needed for responding quickly. A mechanism by which this happens remains to be explored. Using similar reasoning, the Informed participants focused more attention on the task demands because there was no concealment associated with their knowledge of the list. Although not significant, the small but uniform differences between the Informed and Control participants might be explained in the same way by saying that merely having special knowledge and the idea that the knowledge would improve performance, slowed their response times slightly, but not significantly. The Experimental group possessed the same idea but were further impacted by the need to conceal their having acquired the list. Using this rational, the conscious cognitive process of concealing privileged information draws on limited attentional resources to some degree.
Given the slight but nonsignificant delay between the Controls and the Informed group shown in Figure 1, we conducted a post-hoc power analysis to estimate how many participants would have been required to achieve a significant effect with that degree of variability. The results suggest that approximately 150 participants would have been necessary to achieve significance. This is important because the interpretation of the present results would have to be altered somewhat to accommodate the mere effect of being Informed, versus being informed but having to conceal the knowledge. It is also important inasmuch as ERP studies normally have group sizes ranging from 10 to 30 participants; more than 30 is generally not feasible. However, many ERP studies have demonstrated that ERP effects can be very large even in the absence of significant behavioral effects.
On the whole, participants in the Informed and Experimental groups performed progressively slower as a function of the amount of resources each group may have dedicated to the process of consciously recognizing individual words during the LDT. Because nonwords were also affected, but were never candidates from the study list, the effect on the attentional system was to draw resources away from the task which in turn slowed the response to all target items. In terms of final conclusions, concealed information may alter certain cognitive processes which may differentiate performance on certain attention demanding tasks. Further study is required to examine the effects of other possible situational variables as well as individual differences. Moreover, the timebase of the effects of concealing one's knowledge of information can be further understood by adapting the LDT or similar paradigms with newer technologies like ERPs and interrogative polygraphy.
Bandura, A. (1988) Self efficacy: toward a unifying theory of behavioral change. Psychological Review, 84, 191-215.
Barland, G. H., & Raskin, D. C. (1975) An evaluation of field techniques in detection of deception. Psychophysiology, 12, 321-330.
Farwell, L. A., & Donchin, E. (1991) The truth will out: interrogative polygraphy ("Lie Detection") with event-related brain potentials. Psychophysiology, 28, 531-547.
Forster, K. I. & Taft, M. (1994) Bodies, antibodies, and neighborhood density effects in masked form priming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 455-464.
Francis, W. N., & Kucera, H. (1982) Frequency analysis of English usage: Lexicon and grammar. Boston: Houghton-Mifflin.
Harkins, S. G., & Petty, R. (1982) Effects of task difficulty and task uniqueness on social loafing. Journal of Personality and Social Psychology, 43, 1214-1229.
Jackson, J. M., & Williams, K. D. (1985) Equity in effort: an explanation of the social loafing effect. Journal of Personality and Social Psychology, 49, 937-942.
Kihlstrom, J. F. (1987) The cognitive unconscious. Science, 237, 1445-1452.
Lykken, D. T. (1959) The GSR in the detection of guilty knowledge. Journal of Applied Psychology, 43, 385-388.
Pachella, R. G. (1974). The interpretation of reaction time in information processing research. In B.H. Kantowitz (Ed.), Human information processing: Tutorials in performance and cognition (pp.41-82). New York: Wiley.
Posner, M. I., & Snyder, C. R. R. (1975) Attention and cognitive control. In R.L. Solso (Ed.), Information Processing in Cognition: The Loyola Symposium. Hillsdale, N.J.: Erlbaum.
Pratarelli, M. E., & McIntyre, J. A. (1994) Effects of social loafing on word recognition. Perceptual and Motor Skills, 78, 455-464.
Reid, J. E., & Inbau, F. E. (1977) Truth and Deception: The Polygraph Technique. Baltimore: Williams & Wilkins.
Rosenfeld, J. P., Angell, A., Johnson, M., & Qian, J. (1991)
An ERP-based, control-question lie detector analog: algorithms
for discrimination effects within individuals' average waveforms.
Psychophysiology, 28, 319-335.
We would like to thank John Locker and Alvin Sunni for their performance
as confederates, and Mary Devitt for her editorial suggestions
and statistical support. We also wish to thank all of our student
participants. Address correspondence to: Marc E. Pratarelli, Ph.D.,
215 North Murray, Department of Psychology, Oklahoma State University,
Stillwater, OK 74078.
Article Submitted: 21 September 1996
Article Accepted for Publication: 13 January 1997
Article Published: 5 February 1997
Number of page accesses since 9 February 1997:
Return to the JCAAWP Home Page